Demand Side Management in Industrial, Commercial, and Residential Sectors: A Review of Constraints and Considerations

Abstract

Demand Side Management (DSM) is a cost-effective approach to managing electricity networks, aimed at reducing capacity requirements and costs, increasing the penetration of renewable generation, and reducing power system emissions. This review article explores the distinctive characteristics of electricity demand in the industrial, commercial, and residential sectors, and their relationship to successful implementation of DSM. The constraints and considerations for DSM are characterized as technical, economic, and behavioral factors, such as process requirements, business operation constraints, and consumer decisions, respectively. By considering all three types of factors and their impacts in each sector, this review contributes novel insights that can inform the future implementation of DSM. DSM in the industrial and commercial sectors is found to be primarily constrained by technical considerations, while DSM in the commercial sector is also subject to economic constraints. Conversely, residential demand is found to be primarily constrained by human behavior and outcomes, highly variable, and the largest contributor to peak demand. This review identifies sector-specific opportunities to enhance DSM uptake. Industrial DSM uptake will benefit from technological and process improvements; commercial DSM uptake can benefit from enhanced economic incentivization; and residential DSM uptake can benefit from improved understanding of the interactions between human behavior, human outcomes, and energy use. Finally, this review investigates behavioral models and concludes that agent-based models are best suited for integrating these interactions into energy models, thereby driving the uptake of DSM, particularly in the important residential sector.

1. Introduction

Concerns about global warming and greenhouse gas (GHG) emissions are leading many countries to increase electricity generation from renewable sources [1]. However, electricity generation from renewable sources, such as wind, solar, and hydroelectric, can be intermittent and unpredictable [2,3]. Unmanaged, this intermittency can cause discrepancies between electricity supply and demand, leading to blackouts and other electricity system failures [4,5].

Demand Side Management (DSM) consists of acts modifying the level or pattern of electricity consumption [6]. One form of DSM is Demand Response (DR), where electricity consumption is intentionally shifted to suit supply or other constraints, either by decreasing, increasing, or shifting demand, known as peak shaving, valley filling, and load shifting, respectively [7,8]. By increasing electricity demand to coincide with times of excess supply, DSM can increase the utilization of electricity from renewable resources [9,10]. Thus, DSM can help resolve issues of intermittency and unpredictability, facilitate increased installation of renewable generation, and reduce GHG emissions from electricity generation. DSM can also reduce peak electricity demand to ensure that it remains within the limits of transmission and distribution infrastructure, extending the lifetime of components and reducing electricity distribution costs for businesses and consumers [11,12].

Electricity demand is typically categorized according to the end user, with three broad, main categories: industrial, commercial, and residential [13,14]. Each sector accounts for approximately one third of electricity consumption in developed countries [15,16,17]. Figure 1 shows the breakdown of electricity consumption by sector in New Zealand and the United States of America.

In all sectors, the implementation of DSM can be subject to technical and economic constraints [18] and is fundamentally constrained by the human needs that the energy use is serving. On the technical side, electric appliances can have different degrees of adjustability in terms of usage time and power consumption. For example, many industrial processes have high start-up costs and/or sensitive processes requiring a stable power supply. On the economic side, some electricity demand cannot be shifted without financial loss, such as for commercial loads required for successfully serving customers. Equally, some loads, such as electricity demand from equipment for critical medical applications, cannot be shifted without compromising human outcomes. Thus, understanding technical and economic constraints in each sector is required to successfully implement DSM and can inform policy and technical efforts to best address the barriers to DSM uptake.

DSM implementation can also be constrained by human behavior [19]. For example, charging an electric vehicle is constrained by driving behavior, because the vehicle can only be charged when connected to a power source and the energy required is dependent on travel requirements. Additionally, DSM can, in turn, impact human outcomes, as actions such as restricting electricity consumption during winter can excessively reduce indoor temperatures, increasing the risk of respiratory illness [20,21,22]. Ultimately, human outcomes are the drivers for electricity demand, as people do not desire electricity, but rather the many objects that electricity affords them, such as travel and thermal comfort.

However, people do not always behave as rational economic actors [23], and thus may not respond to DSM programs as predicted by standard economic theory. For example, changing electricity price has less impact on the energy consumption of high-income households than low-income households [24], as energy consumption behavior is the result of a complex process involving psychological, environmental, and economic factors [25], and thus has non-linear price elasticity. Thus, successful DSM requires understanding human outcomes, and the behaviors driven by the desire to obtain those outcomes, as well as understanding the more rational and directly measurable economic and technical constraints. In short, successful DSM requires understanding the rational and human behavior-driven constraints.

1.1. Recent Reviews of DSM

DSM in heavy industries is reviewed in [26], with a focus on the technical and economic constraints in cement manufacturing plants, aluminum smelting plants, and oil refineries. Zhang and Grossmann [27] review the application of industrial DSM using the concept of enterprise-wide optimization, where manufacturing, supply, and distribution are optimized to reduce inventories and costs [28]. In [29], the application of energy storage for industrial DSM is reviewed, and the technical and economic performance of different storage options is compared, including magnetic, mechanical, electrochemical, and thermal storage. A survey of industrial DSM is conducted in [30], which focuses on the implementation of Demand Response (DR) programs providing services to utility companies while accounting for the technical constraints of industrial processes. A review of industrial and commercial DSM is conducted in [31], which focuses on the importance of technical constraints and the impact on DSM of economics in electricity markets.

DR programs in “smart grids”, which are advanced electricity networks with interconnected communication systems able to react to changing supply and demand, have also been reviewed. A review of mathematical models and approaches for simulating DR in smart grids is conducted in [8]. Vardakas et al. [32] provide a review of pricing methods and optimization strategies employed to model DR programs in smart grids, with a specific focus on market-based implementation. Aghaei et al. [33] review DR programs in smart grids with renewable energy sources and include an economic assessment and comparison of DR programs.

Residential DSM programs have also been reviewed. Panda et al. [7] review the optimization methods used in residential DSM modeling, and the operational constraints affecting the implementation of residential DSM. Gyamfi et al. [19] review the behavioral factors influencing the implementation of residential DSM, including issues of infrastructure cost, price unresponsiveness, and equity. Additionally, a meta-analysis of DR programs in [34] analyzes the effects of program design and socioeconomic conditions on the effectiveness of DR programs.

However, all these reviews primarily focus on technical and economic constraints influencing the implementation of DSM, which are more measurable and quantifiable than behavioral constraints. The only review explicitly addressing behavioral constraints, [19], does not include industrial or commercial electricity consumption. Additionally, this review does not consider technical or economic constraints, which can interact with behavioral constraints in important ways. Thus, there has been no review of the combined effects on DSM of technical, economic, and behavioral constraints, and how the importance and interaction of these constraints can vary between the different sectors of electricity consumption.

1.2. Contributions

This paper reviews the current state of research in DSM for industrial, commercial, and residential users. This review has two novel features distinguishing it from previous work: (1) review of DSM in all three sectors (industrial, commercial, and residential); and (2) inclusion of technical, economic, and behavioral constraints influencing the uptake of DSM.

We hypothesize that the impacts of human behavior on industrial and commercial DSM will be limited, as these sectors are less exposed to human behavior and/or do not tolerate variability in consumption.

In particular, industrial DSM typically revolves around solutions tailored to a specific process and factory, which typically run more efficiently at constant speed without stoppages [26,27,29,35]. The limited flexibility of commercial loads arises from operators’ prioritization of occupant comfort and safety, and maintaining business operation, over energy savings, which in any event are often a smaller part of their total costs in the overall financial sustainability of their businesses [31,36]. Hence, the interaction of technical and economic constraints with human behavior is hypothesized to have the greatest impact on DSM implementation and success in the residential sector, which represents ~33% of electricity consumption (Figure 1).

1.3. Organization

This paper is organized as follows: Section 2 defines key terms and concepts. Section 3, Section 4 and Section 5 review industrial, commercial, and residential DSM, respectively. Section 6 discusses the current state and future of behavioral-based DSM, Section 7 outlines key considerations and practical implications, and Section 8 concludes and gives recommendations.

2. Problem Definitions

Strategies for implementing DSM can be classified into two categories: energy efficiency and Demand Response (DR) [11]. Energy efficiency refers to enhancements in the efficiency of electrical appliances or processes, which reduce users’ electricity demand, while DR refers to intentional shifts in electricity demand to suit supply or limits on transmission and distribution infrastructure, such as distribution transformer capacity [37]. There are three main methods of DR: (1) peak shaving, where consumption is reduced during times of high demand [38,39,40,41]; (2) valley filling, where consumption is increased during times of low demand [42,43,44,45]; and (3) load shifting, where consumption is shifted from times of high demand to times of low demand [46,47,48,49,50]. Figure 2 schematically illustrates these three types of DR.

DR programs can be classified into price-based and incentive-based programs [51,52], based on how they encourage consumer participation. Price-based DR programs, also known as time-based DR programs, use time-varying electricity prices designed to shift demand. The two main pricing methods are time-of-use (ToU) pricing, where retail electricity prices increase by a pre-defined amount during specified times of the day, and real-time pricing (RTP), where retail electricity prices vary in real time according to transmission and distribution charges and the wholesale cost of electricity [53,54,55,56,57,58]. Reports from DR trials indicate price-based DR programs are more appealing to consumers than incentive-based programs [59].

Wholesale electricity costs generally increase during times of high demand and decrease during times of excess supply, but can also depend on power plant utilization, which is governed by constraints such as fuel cost and resource availability [60,61,62,63]. Incentive-based DR programs offer incentives for customers to shift demand, such as reduced electricity bills or direct payments. Incentives can be offered for demand curtailment during times of high demand and demand increases during times of high supply, and for the provision of ancillary services such as frequency and voltage control [64,65,66,67]. Price-based DR is most common for residential and commercial consumers, while large industrial users’ bespoke agreements with utility companies mean that they typically engage in incentive-based DR, increasing and/or curtailing their consumption according to direct requests [68,69,70].

The capacity of an appliance, process, or user to change electricity consumption patterns is known as its “demand flexibility” [71,72,73]. High demand flexibility indicates that a load has a high capacity to participate in DR, while low demand flexibility indicates the opposite. For example, residential hot water cylinders have high flexibility, as their thermal storage means electricity demand can be decoupled from hot water demand and they can be used for DR without compromising demand fulfillment [74]. Conversely, appliances which consume electricity only while being used, such as washing machines and dishwashers, have little-to-no flexibility because electricity demand cannot be altered without also changing the times of appliance use.

However, having little flexibility does not mean a load cannot be used for DR, as consumers can also participate in DR programs by changing their use of non-flexible loads. The degree to which a consumer responds to changes in price is known as their “demand elasticity”, with a high demand elasticity meaning a great willingness to change electricity demand according to an incentive or a change in electricity price. For example, a consumer with high demand elasticity would be willing to use their high-flexibility loads, such as hot water cylinders, for DR, and may also be willing to change how, or when, they use their low-flexibility loads, such as washing machines and dishwashers, to increase their participation in DR.

Demand elasticity is related to other economic constraints, which include any monetary considerations influencing the willingness or ability to participate in a DR program, such as the start-up cost of a process and the direct and indirect costs of shifting demand. Conversely, technical constraints are the physical attributes influencing flexibility, such as the runtime of a process or the amount of thermal storage available in a hot water cylinder. In many cases, economic and technical constraints are related. For example, a process may be unsuitable for DR because it is at a critical stage in a production line (a technical constraint) and the factory manager does not want to lose money by stopping the production line (an economic constraint).

In contrast to readily measurable technical and economic constraints, human behavioral constraints are related to human preferences and desires to obtain certain outcomes. These constraints can be directly or indirectly related to consumer interactions with an appliance or process. For example, the temperature a consumer sets a heater thermostat is a direct interaction with the appliance, whereas whether the consumer is home, and thus whether they may use the heater, is an indirect interaction with the appliance. Both types of behavioral constraints are important considerations for the successful implementation of DR. Equally, both are highly variable choices across consumers and their wide range of socioeconomic and cultural behaviors, and are not as rationally explicable as technical and economic constraints.

In some cases, participation in DR programs may be limited for small consumers, as electricity utility companies may consider their demand too small to be of concern. These smaller consumers can still participate in DR programs through “demand aggregation”, where groups of smaller consumers work with a “demand aggregator” (typically a third party) to coordinate their electricity demand according to DR requests from the utility [75]. Thus, for DR purposes, the utility will treat this group of small consumers, whose individual demand may not be of concern, as one large consumer whose demand is of concern and is thus eligible to participate in DR programs.

These definitions and explanations are used throughout this review to set the context of the discussion. They also delineate rational, typically quantifiable and readily explicable choices, from what might appear to be more variable or irrational choices based on the confluence of a consumer’s direct, economic, and social/cultural preferences.

3. Industrial DSM

Electricity is used in many processes across multiple industries, including metal production, oil refining, chemical and cement production, and pulp and paper production [76]. The breakdown of electricity consumption by industry for New Zealand, where industrial electricity consumption comprises 37% of the national total, is shown in Figure 3. In many cases, electricity consumption is a major contributor to industrial operating costs [77,78]. Additionally, industrial processes, such as process heat, are being increasingly electrified, so industrial electricity consumption and electricity costs are expected to increase [79,80,81,82,83,84].

The breakdown in Figure 3 shows that the vast majority of industrial electricity in New Zealand is consumed by the basic metals, food processing, and wood, pulp, paper, and printing industries. Thus, implementation of DSM in these industries is likely to have more impact on overall electricity consumption than DSM in other industries in New Zealand. The high share of electricity consumption from basic metals reflects the large quantities of electricity required for the electrolytic reduction of alumina feedstock, iron smelting, and steel making [85].

3.1. Economic and Technical Constraints

Technical constraints for electricity demand can vary greatly between different industries, and between different consumers within the same industry, depending on several factors, such as processes, equipment, and schedules. Thus, industrial DR programs are typically tailored to the requirements of each consumer [27,29,30]. While flexibility is highly variable between different industrial consumers, many industrial processes are uninterruptible without advance notice, and others cannot be interrupted at all. Thus, the flexibility of industrial electricity demand is generally recognized as being considerably lower, with greater immediate negative costs for stoppages or reductions, than for other sectors [26,31].

The electricity demand of industrial consumers can be high enough to warrant a bespoke agreement with electricity utilities, and in some cases even their own connection to high-capacity electricity transmission infrastructure [16]. These agreements sometimes include industrial DR programs, the vast majority of which are incentive-based [86]. Many industrial processes are difficult and/or costly to interrupt, and electricity tends to be a small part of total costs for industrial consumers [87,88,89,90,91]. Thus, incentives for DR participation tend to be higher for industrial consumers than for residential or commercial consumers, to encourage participation [26,92]. However, even with these incentives, many consumers are unable to participate due to the inflexible nature of their processes [26].

3.2. Behavioral Constraints

During normal operation, industrial electricity demand is typically defined purely by the technical and economic constraints of the industrial process [26]. Thus, human behavior has little effect. Behavioral constraints, such as employees’ capacity to work extra hours, can be important considerations for operational scheduling, and thus for when the plant may require electricity [35]. However, these behavioral considerations tend to have a minimal effect on overall electricity consumption and participation in DR, and the negative costs of stoppages or reductions may point towards hiring extra staff as being more cost-effective.

3.3. Review of Industrial Demand Response

The majority of industrial DR is carried out by large electricity consumers, as their high consumption increases their bargaining power for bespoke agreements with utility companies [26]. Industrial electricity consumption can contribute a large portion of total costs for industrial operators [77,78] and a large portion of total demand for utility companies [15,16]. Thus, management of industrial electricity demand has the potential to mutually benefit both suppliers and consumers.

However, industrial demand flexibility is often limited due to technical constraints. Nevertheless, some industries are better suited than others to participate in DR. In general, the industrial consumers with the highest flexibility, and thus the highest DR potential, are pulp and paper mills, metalworking industries with furnaces, cold storage warehouses, cement manufacturers, and electrolysis plants [30,35].

Pulp and paper mills convert wood chips or other fibers into fiberboards, which are then processed into paper. The inventory of pulp tends to be higher than for paper, which enables flexibility in the scheduling of pulp production [35]. Helin et al. [93] assess the potential for, and profitability of, DR for a Finnish pulp and paper mill active on the Nordic electricity market. The large on-site storage, particularly of pulp, allows shifting of the use of grinding machines and paper machines to times of lower electricity prices, offering demand flexibility “for most hours of the year”. However, the authors indicate that participating in DR can put the mill at risk of failing to fulfill obligations to customers, so a risk premium should be added for sensitive processes, which could limit the mill’s DR capacity.

Metalworking industries have many uses for furnaces, including blast furnaces for the production of iron and heat treatment furnaces for tempering. While electric furnaces can be used for demand flexibility due to their very high thermal storage, metalworking plants are recognized as difficult environments in which to implement DR, due to their complex processes and tight temperature tolerances [94]. The potential for DR from optimal scheduling of furnaces and rolling mills in a steel plant in Germany is assessed in [95], where the proposed optimal schedule is shown to reduce electricity costs by 12–52% from the non-optimized costs, depending on capacity utilization.

Cold storage warehouses use large fridges or freezers to store temperature-sensitive products. Most products have a temperature range within which they can be stored, such as frozen meat products requiring temperatures between −24 °C and −18 °C [96]. Thus, fridges/freezers do not consume electricity power constantly, which allows flexibility of electricity demand through heat storage. In [86], the flexibility potential of price-based DR for cold storage is assessed for such a warehouse in Ireland. By regularly updating their operations schedule in response to changing electricity prices, the warehouse shifted electricity demand to off-peak times, reducing monthly electricity costs between 5% and 15% and increasing the utilization of electricity from wind generation.

Cement manufacturers make cement powder from stone and other raw materials through grinding, mixing, and milling processes. Cement manufacturers typically have large inventories, which allow flexible scheduling of mills and grinders. Additionally, many cement manufacturers have mills fitted with variable speed drives, or could be retrofitted with such drives, which allow the speed, and thus the electricity consumption, of the mills to be increased/decreased, which can be used for DR [97]. In [98], operations in a cement plant in the United Kingdom are optimally scheduled according to price-based DR, which utilizes the flexibility of these processes to reduce electricity costs by 4.2%.

Electrolysis is the process of using electrical current to drive a chemical reaction. Industrial electrolysis, such as that used for the extraction of aluminum from aluminum ore, takes place in electric smelting pots, where input voltage can be directly controlled [26]. The effects of large-scale DR of aluminum electrolysis on the German electricity system are investigated in [99]. Flexibility of the extraction process is shown to yield national DR load-shifting potential of up to 13 GWh; however, this industrial DR potential is shown to be lower than for non-industrial DR strategies, such as controlled charging of EVs or thermal storage in district heating networks.

While industrial DR is typically implemented by large electricity consumers, flexibility in industrial processes can also be utilized by smaller consumers. The potential for DR in a textile factory in Ethiopia is assessed in [100], where a model of the factory’s oil-powered boiler is used to simulate the integration of solar-assisted heating. By using DR strategies to align the factory’s operation schedule with solar production, the utilization of solar heat for the boiler is shown to increase by up to 10% per month. The authors also identify the requirement for “sector-specific solution(s)” to increase the use of solar power for low- and medium-heat industries whose participation in large-scale DR programs is limited.

3.4. Future Work

Utilizing the flexibility of industrial electricity consumers has been shown to be an effective method of DR, particularly for large consumers. However, a number of technical and economic challenges remain before the full potential of industrial DR can be realized [30]. While industrial DR has been analyzed in some industries, the flexibility potential of many others remains largely unstudied and unknown. For example, the focus on industry-specific DR programs may overlook flexibility opportunities existing across multiple industries, such as industrial compressed air systems [101,102].

Thus, the first challenge to increasing the potential for industrial DR is to assess the flexibility of processes in a wider range of industries and increase this flexibility where possible [27]. This assessment would involve modeling to understand the technical constraints of the electricity-dependent processes in these industries, which would facilitate greater understanding of how to increase their flexibility. Second, industrial DR remains largely available only to those consumers who have knowledge both of their plant’s processes and of scheduling according to DR incentives. Hence, those without these two skillsets are unable to participate [26,30].

Economic constraints tend not to be challenges to participation in DR for large industrial consumers, as these consumers typically have bespoke DR programs tailored to their process requirements and economics. However, smaller consumers can be overlooked, with their lower demand and resulting lower priority for electricity utilities [26]. Thus, the third challenge for industrial DR is to extend DR programs to smaller consumers. This extension could either be implemented by a utility extending bespoke DR agreements to these consumers, or through industrial demand aggregation.

3.5. Summary

Industrial electricity consumption can contribute a large portion of operating costs for consumers and a large portion of demand for electricity suppliers and distributors. Thus, industrial DR is beneficial both for consumers and for electricity utility companies. DR is highly utilized in industries such as pulp and paper mills, metalworking with electric furnaces, cold storage warehouses, cement manufacturers, and electrolysis plants. In general, industries with large inventory buffers and/or thermal storage in their processes, and those who have longer operational shifts, are best suited for industrial DR.

However, there exist several barriers to further industrial DR uptake: (1) small consumers are often overlooked, given their lower DSM potential than larger consumers; (2) while the knowledge of DR incentives and detailed knowledge of an industry’s processes are both required to adopt DR, both skillsets are not often present in the same organization; and (3) DR programs tend to be customer-specific, so industry-wide process-specific applications with high DSM potential, such as compressed-air systems, are often overlooked.

Industrial DR is highly dependent on the types of electricity-dependent processes and the nature of the industry, and as such tends to be through individual agreements between consumers and utility companies where the consumer’s process requirements are acknowledged. Thus, overall human behavioral constraints are typically of little concern for industrial DR.

4. Commercial DSM

Electricity is used in the commercial sector for a wide range of applications, such as space heating and cooling, lighting, water heating, and electronic devices. Figure 4 shows the breakdown of commercial electricity consumption by end use in New Zealand, where commercial consumption comprises 24% of the national total. It should be noted that electricity costs tend to contribute a small proportion of commercial operating costs, particularly for smaller businesses such as retail and office buildings [103,104,105].

Figure 4 shows the overall end use pattern for electricity use, yet this is likely to change depending on the type of premise. A high share of electricity use for space conditioning (25–30%) is characteristic in office, service, and wholesale trade buildings. However, retail buildings are characterized by having a predominant share of electricity use for refrigeration (35–40%) [104].

4.1. Economic and Technical Constraints

While commercial electricity demand is less process-focused than industrial demand, most is still subject to technical constraints. Much of the electricity demand of commercial customers is subject to regulation, such as for lighting, refrigeration, and space heating. For example, workplaces have recommended task-specific illuminance levels to meet while businesses are operational [107], which are set by standards such as AS/NZS 1680 [108]. Thus, the flexibility of electricity demand for lighting is limited. These constraints mean much of the electricity demand during business operating hours has limited flexibility, and there is little consumption in many businesses outside of these hours. Additionally, electricity demand may not make up a dominant part of operational costs, compared to building costs, employee wages, and materials, and the financial motivation for DSM uptake is thus of a lower priority than other financial considerations.

Commercial electricity demand is also subject to economic constraints relating to business operation. Most commercial electricity loads are required for business operation, so they have low demand elasticity. For example, a restaurant using electric cooking appliances is unlikely to shift their appliance use based on changing electricity prices, as doing so would compromise their ability to serve customers and generate revenue.

Therefore, the majority of commercial DR is carried out by appliances with thermal storage, such as water heaters and space heaters [109]. The electricity loads of these appliances are more flexible than most other commercial loads, as their thermal storage allows electricity demand to be decoupled from heating/cooling demand. However, these loads are still subject to technical constraints, such as the building’s thermal insulation, and behavioral constraints, such as employee and customer thermal comfort.

4.2. Behavioral Constraints

As the majority of commercial electricity demand is defined by technical and economic constraints, human behavioral constraints tend to be unimportant. Most commercial electricity demand is for appliances required either by regulations (such as lighting standards) or by business operations (such as the ability to serve customers), so human behavior is typically not a significant consideration for the majority of demand.

However, human behavior is a consideration for applications catering to thermal comfort. Preferences for thermal comfort can vary between individuals, which can influence electricity demand for space heating and cooling. While not all commercial consumers have the wherewithal or desire to adjust electricity consumption based on these differing preferences, when accounted for, these preferences can influence electricity demand [110,111].

4.3. Review of Commercial Demand Response

Commercial electricity demand has fewer opportunities for flexibility than industrial demand, so commercial DR is less well-studied. The majority of research into commercial DR investigates applications with thermal storage, such as space heating/cooling and refrigeration, or other forms of storage, such as electric vehicle batteries.

Office buildings have been shown to be effective participants in DR programs [36]. Many offices have Heating, Ventilation, and Air Conditioning (HVAC) systems with adjustable temperature setpoints, which can perform DR using the thermal storage of the rooms they service. However, DR programs with space heating/cooling appliances must balance DR participation with the thermal comfort of occupants. In [112], DR using space heating and other electrical loads is implemented in a small office in Portugal. The DR program aligns electricity demand with generation from the office’s rooftop solar PV panels, increasing the utilization of renewable electricity, and does so without compromising occupant comfort. In [113], the use of building thermal capacitance and behavior adaptation are proposed to adjust the HVAC system operation and reduce energy demand during shoulder (before and after winter) months, when electricity supply can be constrained.

In addition, pilot studies of DR programs in offices are described in [36], where in an administration building and a library in China, the flexibility of space heating and other appliance loads is used to respond to direct DR requests from the electricity utility company. These offices are shown to be capable of load shedding 15–20% of their electricity demand.

Supermarkets are also good candidates for commercial DR. Supermarkets have many electricity loads, such as lighting, cooking, and space heating/cooling, but their main contribution to DR programs is from the thermal storage in their large refrigerators [114]. In a pilot study in a Finnish supermarket [115], refrigerators are used in a DR program to reduce the supermarket’s greenhouse gas emissions. Leveraging the refrigerators’ thermal storage and aligning electricity with generation from renewables reduced greenhouse gas emissions and increased energy efficiency. However, the authors highlight that current practices in the Finnish electricity market make it infeasible for small electricity consumers to participate in DR, and this pilot study was only possible because the supermarket was operated by S-Group, which is Finland’s largest supermarket chain and the country’s largest non-industrial energy consumer.

An emerging category of commercial DR is controlled charging of electric vehicles (EVs), as many businesses are increasingly adopting EVs for their vehicle fleets. EVs are often connected to a charger for longer than it takes them to fully charge, so their batteries can offer demand flexibility to the electricity grid [116]. Optimal scheduling of commercial EV charging has been shown to be beneficial for both EV owners, by shifting charging to lower-priced times and reducing electricity costs, and for electricity utility companies, by reducing peak demand and providing ancillary services [117,118,119].

4.4. Future Work

Commercial DR programs are currently limited, so opportunities for improvement exist. Most commercial demand is for business-critical applications, so improvement in commercial DR should come from better understanding of existing flexibility. Simulation of thermal storage in buildings can be made more accurate using building information modeling (BIM), which is a three-dimensional modeling technique closely approximating the thermal characteristics of real buildings [120]. Further improvements can be made in DR programs accounting for thermal comfort, by using more accurate models of building occupancy and occupant behavior [121]. More accurate behavioral models would allow better quantification of occupants’ preferences and comfort, allowing DR programs to better optimize electricity demand without compromising thermal comfort.

Additionally, as with industrial DR, commercial DR tends to be limited to large consumers who can make bespoke agreements with electricity utility companies [115]. Thus, the DR potential of smaller consumers is often overlooked. To expand DR programs to smaller commercial consumers and utilize their demand flexibility, utility companies should extend offers for bespoke DR programs to smaller consumers or encourage DR through demand aggregators.

4.5. Summary

Electricity costs tend to contribute a small proportion of commercial operating costs compared to employee wages, material costs, and facility costs. Thus, the management of electricity demand tends to be of less importance for commercial consumers than for industrial consumers. The majority of commercial electricity demand is subject to technical and/or economic constraints limiting flexibility, such as regulations for workplace lighting and requirements to serve customers. Thus, commercial DR programs focus on applications with inherent storage, such as thermal storage in refrigerators and battery storage in EVs.

However, in comparison to other sectors, research in commercial DSM is limited, which may either indicate the sector has little potential for DSM, potentially due to the lower share of electricity consumption to their overall costs, or there is a significant research gap in this sector. Opportunities for improvements in commercial DR include a better understanding of demand flexibility for consumers, better modeling of building occupant preferences and comfort levels, and the extension of DR programs to smaller consumers. As human behavior is typically not important in defining commercial electricity demand, improvements in commercial DR should focus primarily on technical and economic constraints.

5. Residential DSM

Electricity is used in the residential sector for a number of applications, including cooking, lighting, and heating. Figure 5 shows a breakdown of energy consumption by end use in the residential sector in New Zealand, where residential electricity consumption comprises 33% of the national total.

As shown in Figure 5, most residential electricity consumption in New Zealand is for water heating, space heating/cooling, and refrigeration, all of which have some thermal storage [111]. Thus, residential DSM programs with thermal storage are likely to have an outsized impact on total electricity demand. Note that household energy consumption varies seasonally, with increased heating (up to 25% of total household energy use) during the winter months [122].

5.1. Economic and Technical Constraints

Unlike in the industrial and commercial sectors, residential electricity demand is not constrained by process requirements or workplace regulations. Instead, technical constraints for residential electricity demand are the physical properties of appliances and buildings, such as the power capacity of a heater and the thermal insulation of the room it is heating [123,124].

Electricity can contribute a large portion of household expenses [125]. Thus, managing electricity demand is a method used by many consumers to reduce their expenses [126]. However, for most consumers, this management consists solely of curtailing demand [127], such as reducing electricity use by heating homes to a lower temperature during colder months. Further, demand elasticity is higher for consumers in lower-income households, for whom economic constraints are a salient issue [24]. Thus, the adverse effects of demand curtailment due to economic constraints, such as lower indoor temperatures, residential dampness, and an increased risk of respiratory illness, are experienced most by populations characterized by low income, tenancy, and residential mobility [128].

5.2. Behavioral Constraints

Unlike in the commercial and industrial sectors, where electricity is consumed to fulfill process and business requirements, electricity in the residential sector is consumed primarily to fulfill human desires, such as the desire for thermal comfort, entertainment, and transportation. Thus, behavioral constraints are typically more important than technical and economic constraints in defining residential electricity demand [129]. While technical and economic constraints define the limits and elasticity of residential demand, respectively, participation in DSM depends primarily on the willingness of residential consumers to change their demand patterns, which may require them to compromise on the commodities for which they use electricity.

Additionally, behavioral constraints mean that the residential sector is the main contributor to peak electricity demand [130]. Residential electricity demand is highly dependent on time of day, and on external conditions such as weather [131]. In New Zealand, national power demand is typically highest between 7 a.m. and 9 a.m., when people ready themselves to leave home, and between 5 p.m. and 9 p.m., when people return home [132,133,134]. Demand is also highest during hot or cold weather events, such as heat waves and cold spells, as consumers use more electricity for cooling or heating, respectively. This increased demand is typically representative of a large portion of residential consumers, so this aggregate behavior can affect the entire electricity system.

Surveys of residential consumers show an overall positive attitude towards DR [130,135]. Residential consumers are generally sympathetic to price-based DR programs and respond to these programs by shifting electricity use away from higher-priced times [130]. However, price is not the only factor influencing the willingness to participate in DR programs, as many consumers are also concerned about increasing electricity supply security to avoid blackouts and reducing the GHG emissions of electricity generation [135]. Electricity consumption depends on both structural and motivational factors, so the receptiveness for DR schemes may also depend on internalized norms and self-expectations [136].

5.3. Review of Residential Demand Response

Residential electricity loads can be classified into two categories, based on whether they can be shifted without altering consumer behavior [73,137,138,139]. “Flexible loads” can be shifted without requiring major behavioral changes [140], such as water heating and electric vehicle (EV) charging. Conversely, “inflexible loads” or “fixed loads” cannot be shifted without major behavioral changes, such as computers and electric stoves.

Residential Demand Response (DR) research is typically aimed at assessing the DR potential of flexible loads [141]. In particular, most of the research focuses on EVs, space heating and cooling, water heating, battery storage, and flexible white appliances, such as washing machines and dishwashers [7,8,32,33].

If unmanaged, EV charging can increase peak electricity demand and reduce the feasibility of a transition to EVs [142]. However, managed EV charging can provide storage and ancillary services to the electricity grid [143,144], reduce GHG emissions from electricity generation [144,145,146], reduce peak electricity demand [145,147,148,149], reduce electricity costs for consumers [147,149,150], and increase the feasibility of a transition to EVs [151]. EV charging is one of the simplest ways to implement residential DR, because EVs’ battery storage provides inherent flexibility, so electricity demand can be shifted to times of lower electricity prices. Additionally, EV batteries can be used to feed electricity back into the grid when required for ancillary services. This bidirectional electricity flow to/from EV batteries is known as “Vehicle to Grid” (V2G), which can provide benefits both for electricity utilities and for consumers [152,153,154,155]. However, the use of EVs for V2G can increase the rate of battery degradation, as batteries are meant to fulfill either autonomy requirements or ancillary services, and fulfilling both conditions has been shown to compromise battery lifespan [152,156,157,158,159,160]. Thus, consumers may be less willing to participate in DR programs involving V2G if not properly compensated for this degradation.

Flexibility of space heating and cooling loads, and their implied potential for DR applications, is related to the thermal storage potential of the space in which they operate [111,161,162]. Well-insulated buildings can store thermal energy for longer periods, providing greater demand flexibility. This thermal storage means electricity demand can be decoupled from heating/cooling demand, allowing space heaters/coolers to be used for DR. Air conditioning units can be used for DR to reduce consumer electricity bills and peak power demand without compromising occupants’ thermal comfort [163,164]. The use of combined Heating, Ventilation, and Air Conditioning (HVAC) systems for DR can reduce total electricity demand [165], match demand with supply [166,167], and reduce consumer electricity costs by shifting electricity consumption to lower-priced off-peak times [167]. As in the commercial sector, the success of residential DR programs involving space heating/cooling loads can depend on occupants’ temperature preferences, and their sensitivity to deviations from their preferred temperature [168,169,170,171]. Thus, the success of DR programs involving space heating/cooling loads depends both on technical constraints such as thermal insulation and on behavioral constraints such as thermal comfort preferences, which can vary greatly.

Electric hot water cylinders (HWCs) can store energy in their tanks, allowing electricity demand to be decoupled from domestic hot water (DHW) demand and facilitating their use for DR [111,172]. The temperature in HWCs is typically regulated using setpoint controllers, which aim to keep water at a constant temperature. By predicting DHW time-of-use using stochastic models [173,174] or machine learning techniques [175,176,177], “smart controllers” can increase the DR potential of HWCs [74], reduce consumer electricity costs [173,176,178,179,180,181], overall electricity consumption [175,182,183], and peak demand [174], and increase the utilization of electricity generated from renewable sources [184].

The use of smart controllers is typically simulated to assess their accuracy before installation in HWCs, which requires a model of HWC temperature. These models can be separated into three levels of complexity: (i) fully mixed temperature models, which assume a constant cylinder temperature [178,179,180]; (ii) stratified temperature models, which use two variables to describe temperature of the upper (hotter) and lower (colder) halves of the cylinder [185]; and (iii) multi-nodal temperature models, which split the cylinder into a large number of nodes and calculate the temperature of each node [178]. However, increasing the DR potential of HWCs can increase the likelihood of failing to fulfill DHW demand [74], so consumers’ willingness to experience unmet demand may limit DR potential.

Residential DR potential can also be increased by using battery storage [186,187]. The utilization of electricity from “distributed generation”, electricity generated at homes or businesses [188,189], can be increased with battery storage [190,191,192]. Batteries can shift electricity demand to coincide with supply, without shifting appliance use [193]. Batteries can also be used for DR without the presence of distributed generation, by charging when supply is high and discharging when supply is low, which can provide ancillary services to the grid [194,195]. Further, other storage methods can be used for residential DR. In [196], residential hydrogen storage is used to reduce power system operational costs and increase the utilization of wind generation. However, these added storage capabilities come with capital costs too high for many to participate in.

Many white appliances, such as dishwashers, washing machines, and clothes dryers, can schedule their loads according to user preferences and can thus be used in DR programs [197,198]. By scheduling runtimes to coincide with DR requests or low electricity prices, white appliances can participate in load shifting [199]. DR scheduling for white appliances can be more complicated than for loads with inherent storage, such as HWCs or batteries, as white goods are subject to technical constraints, such as the length of time required for the appliance to complete its task (“runtime”), and behavioral constraints, such as the time a consumer requires the appliance to be finished [195,200,201]. For example, a dishwasher may take one hour to complete its washing cycle and may be required to provide clean dishes by 7 p.m.; if the dishwasher is turned on at 3 pm, its electricity demand can only be delayed by up to three hours, as its cycle must start no later than 6 pm. Within these constraints, white appliances can successfully participate in DR programs [202] and be used to increase the utilization of electricity from renewable sources and reduce peak electricity demand [203,204].

5.4. Summary

Residential Demand Response (DR) can be beneficial for both consumers and electricity utilities, as electricity can contribute a large portion of household expense and residential demand is the main contributor to peak total electricity demand. Therefore, as the main driver of peak demand, residential DSM is the key to maintaining network functionality, reducing network costs, and effectively integrating renewable energies. While technical and economic constraints can limit the feasibility of residential DR, behavioral constraints can have a far greater impact in this sector. Thus, residential DR programs typically focus on flexible loads, which can be shifted with minimal impact on occupant behavior.

Of these flexible loads, those with inherent storage are most used for DR, such as electric vehicles (EVs), hot water cylinders (HWCs), and space heating/cooling. Storage devices such as batteries and hydrogen storage are also used for residential DR but come at high capital costs for owners. Additionally, certain flexible loads, such as white appliances (e.g., dishwashers and washing machines), lack built-in storage. However, they can still contribute to shifting electricity loads by adjusting their operating schedules. By delaying their start times, these appliances can align their electricity consumption with periods of higher renewable energy generation or lower electricity demand, all while meeting consumers’ needs for clean dishes and clothes.

Unlike in the commercial and industrial sectors, human behavior is a primary consideration for the success of residential DR. In particular, residential electricity demand is primarily aligned with human needs rather than process or business requirements. Thus, an accurate understanding of human behavior, and the impact this behavior has on electricity demand and willingness to participate in DR programs, is crucial to the successful implementation of residential DR.

6. Current State and Potential Future of Behavior-Based DSM

An understanding of behavioral constraints is most important for DR in the residential sector, as opposed to the commercial and industrial sectors. Residential consumption is also a major contributor to variability in electricity demand, and thus a key source of demand exceeding supply in certain underlying cases [205,206]. Thus, understanding the behavior driven by consumer desires is crucial to understanding residential electricity demand and the potential efficacy and impact of Residential Demand Response (DR). It is also important to understand both the real and perceived impacts that DR programs may have on consumers, as they affect consumer willingness to participate in DR programs, and thus DR program success.

6.1. Types of Behavior Models

Methods for modeling human behavior can be broadly grouped into three categories, based on their method of modeling variations in actions and interactions [207,208,209,210]: (i) deterministic models, which assume behavior to be fixed; (ii) stochastic models, which randomly generate behavior from statistical distributions; and (iii) agent-based models (ABMs), which calculate actions based on simple models of behavior.

Deterministic models simulate behavior without any component of randomness. Typically, deterministic models use behavioral data which are either directly measured or calculated from measured data. For example, deterministic models of electric vehicle (EV) charging typically calculate charging profiles based on travel data directly measured from travel surveys [146,151] or calculated from average travel statistics [150]. Deterministic models can also be constructed using measured data for building occupancy [164,165], appliance use [193,195,200,201,211], and electricity from distributed generation [200,212,213]. In addition, deterministic behavioral profiles can be calculated from average, typically regional or national, data. For example, in [150], electricity demand for a house in Spain is calculated using the average profiles of appliance use and EV charging, and in [214], elasticity of electricity demand is assessed using demand profiles generated based on average appliance usage in the Nordic electricity market.

Stochastic models incorporate randomness by generating behavioral profiles from distributions of typical behavior and are thus best suited to modeling applications with well-known behavioral distributions. Stochastic models generate demand from statistical distributions, so their results can differ between simulations, and they are typically run multiple times to capture this variation and provide a distribution of expected results and likelihoods. In simulations of EV charging, stochastic models are used to calculate driving behavior from the distributions of typical start and end times of vehicle trips [143,144,149]. Stochastic models are also typically used to calculate temporal electricity demand from water heating [173,174] and appliance use [36,196,215,216].

Unlike standard stochastic and deterministic models, agent-based models (ABMs) do not operate from a “top-down” approach. Instead, ABMs use a “bottom-up” approach to simulate the actions and interactions of individual actors (“agents”) based on simple behavioral rules. This approach allows ABMs to capture the emergent behavior of groups arising from the sum of each individual agent’s decision making and their interactions with other agents. Because of their ability to capture emergent phenomena, which are difficult to model using top-down approaches, ABMs are considered the best tool to simulate behavior in financial markets [217,218,219], including electricity markets [220]. In models of electricity markets, ABMs are typically used to simulate either participation in electricity markets [97,147,148,163,221,222,223] or optimization of household electricity demand according to defined goals [137,194,224,225,226].

6.2. Variable Behavior in DSM Models

As the electricity demand profiles they produce are fixed, deterministic models are poorly suited for simulating variable electricity demand behavior, such as can arise from participation in DSM programs. However, in some cases, scenario analyses can be used to model the effect of behavioral variations. Scenario analyses can allow the simulation of different deterministic behavior profiles, such as the effects of a transition to EVs with different scenarios of vehicle charging [142].

Similarly, the random nature of stochastic demand profiles means stochastic models are limited in their ability to simulate behavioral changes resulting from DSM programs and other changes to electricity use. Statistical methods, such as Monte Carlo analysis, where stochastic simulations are run multiple times to assess the impact of random variables, can be used to address uncertainty in stochastic models of electricity demand [227,228,229]. However, these statistical methods do not involve the simulation of changes to the distributions from which behavioral profiles are calculated, so they do not address the effect or variability of the root behaviors driving energy use. Thus, stochastic models are poorly suited to studying the potential resulting behavioral changes from participation in DSM programs, such as the “rebound effect”, where behavioral changes decrease the gains from improvements in efficiency or reductions in price, because consumers increase their usage according to the changes [230,231,232,233].

On the other hand, ABMs are well suited for studying changing behavior, particularly where these changes occur as a result of interactions with other agents [234]. Where ABMs are used in simulations of electricity demand, the agents are typically not human consumers but rather computer programs, such as the controller of a smart home system, aiming to optimize the electricity consumption of building occupants [137,147,224,225]. However, ABMs can also be used to simulate human behavior [235,236,237], including behavior related to energy consumption [235,238,239]. Thus, ABMs of electricity demand could incorporate models of human actions and interactions, which would enable them to better assess the effects of variable behavior on the success of DSM programs.

As well as assessing the effects of variable behavior on DSM programs, it is important to assess the effects of DSM programs on human outcomes. ABMs allow the incorporation of human decision making, needs, and wants, to create realistic electricity demand profiles. Thus, ABMs can allow the inclusion of behavioral considerations in simulations of DSM and facilitate the assessment of human-based outcomes of energy decisions, such as whether increased electricity prices and price-based DR programs are likely to result in colder homes and the increased risk of respiratory illness and subsequent, related personal and societal costs [20,21,22,128].

However, behavioral simulations only produce results as good as the models used in the simulations. Thus, it is insufficient to simply produce an ABM of electricity demand with a behavioral component. Rather, behavioral models should be empirically tuned, based on actual human decision making. For example, while some ABMs assume all agents are perfectly rational and self-interested [221,240,241], findings from the field of behavioral economics indicate that this is not true [23,242,243].

Decision making about electricity use in households is not always rational and can be affected by habits [244] and human psychology aspects [245]. Therefore, future models of electricity demand should include ABMs informed by social science and behavioral economics. These behavioral ABMs will allow researchers to incorporate variable behavior and assess the impacts of DSM programs on human outcomes.

However, these bottom-up models may be unsuitable in some situations, as simulation of the actions and interactions of large groups of agents could cause them to be too computationally intensive. Furthermore, models with a high level of agent cognition need strong justification if they are to be considered valid [246]. As such, researchers should exercise professional care in their use to ensure that they are best able to capture aggregate behavior.

6.3. Summary and Future Work

Of the three main categories of models for simulating human behavior, ABMs are best suited to capture variations in electricity consumption behavior. However, some types of ABMs are better suited than others for this task. In particular, ABMs offer the potential to capture ranges of human behavioral responses across a group of agents, thus simulating populations. More specifically, in applying these agents, the following recommendations should be considered.

First, simple agents are better than complex agents, as simple decision-making rules are easier to justify and can perform just as well as more complex models [246]. Second, agents whose decision-making rules are grounded in social science and behavioral economics are more representative of real human actors than agents whose behaviors are assumed to be perfectly rational and self-interested [23,242,243]. As such, it is better to run several simpler models to see aggregate behavioral effects than attempt to construct one overly complex model perfectly simulating all behaviors. Thus, ABMs offer the potential to create complex interactions and behaviors built from combinations of known, simple behavior responses.

Overall, behavioral models are engines of prediction providing distributions of possible outcomes, rather than perfect forecasting mechanisms for the future [247]. Thus, ABMs with agents whose decisions are informed by appropriate models of human behavior should be considered the best tools for informing the effective uptake of DSM in areas where human behavior is an important driver of electricity consumption, such as the residential sector. However, these ABMs should be used to assess the possible distributions of human behavior, and how these may affect demand and human outcomes, rather than being used to provide some ideal expectation of a perfectly accurate prediction of the future.

7. Key Considerations and Practical Implications

7.1. Key Considerations

Table 1. Summary of technical, economic, and behavioral constraints and considerations for industrial, commercial, and residential DSM.

	Industrial	Commercial	Residential
Technical	Process constraints, such as manufacturing lead times [35] and requirements for steady electricity supply to equipment [26].	Regulatory constraints, such as workplace lighting standards [108]. Customer requirements, such as order fulfillment. Preservation of products with temperature and/or humidity sensitivities, such as food products.	Physical parameters of appliances, such as tank size for hot water cylinders [74], and of buildings, such as insulation level [111].
Economic	Equipment costs, such as start-up/shut-down costs, and customer requirements, such as order fulfillment dates [30].	Low contribution of electricity towards typical commercial operating costs [103].	Consumers’ price elasticity of electricity demand [135]. Socioeconomic factors, such as access to cost-effective heating equipment and contribution of electricity to total expenses [125].
Behavioral	Human behavior has minimal impact on industrial DSM.	DSM with space heating/cooling loads must account for occupant comfort [110,111]. Otherwise, minimal.	Residential DSM heavily depends on behavior, such as travel schedules [152] and hot water use [181]. DR programs can influence, and be influenced by, consumer behavior.

shows a summary of key constraints and considerations for DSM in the industrial, commercial, and residential sectors, and how considerations vary for consumers in the three sectors.

7.2. Practical Implications

Table 2. Positive indicators of DSM potential in the industrial, commercial, and residential sectors.

Industrial	Commercial	Residential
Mechanical processes using electric motors with storage or inventory buffers, such as milling and grinding. Variable speed drives (VSDs) increase the suitability for DSM. Heating and cooling processes with thermal storage, such as electric furnaces, electric heating, and refrigeration. Metal refining with electrolysis.	Electric space heating and cooling, such as HVAC systems and electric water heating. Refrigeration, such as cold storage. EV fleet charging.	Schedulable consumer appliances, such as dishwashers, washing machines, and dryers. Electric space heating and cooling, and electric water heating. EV charging.

shows a summary of factors indicating the potential for industrial, commercial, and residential DSM. These indicators can be used to assess whether a process, load, or consumer is well-suited to the implementation of DSM, allowing simple identification of the sectors and businesses where DSM implementation can be increased.

Based on the indicators in Table 2 and the sectoral consumption breakdowns in Figure 3, Figure 4 and Figure 5, the following areas appear to be good candidates for DSM in countries such as New Zealand:

• Metal refining, which comprises 46% of industrial electricity consumption (Figure 3) and typically involves electrolysis and/or electric furnaces [26];
• Wood, pulp, paper, and printing, which comprise 11% of industrial electricity consumption (Figure 3) and typically involve variable inventory buffers and mechanical processes with VSDs [35];
• Electric space heating/cooling and refrigeration (28% and 16% of commercial electricity consumption, respectively, as shown in Figure 4), which offer inherent thermal storage [111];
• Water heating, space heating/cooling, and refrigeration, which comprise 61% of residential electricity consumption (Figure 5), with all offering inherent thermal storage [7,111].

8. Conclusions and Recommendations

Demand Side Management (DSM) is an effective means of reducing customer electricity costs and greenhouse gas emissions of electricity consumption and providing services such as peak shaving and ancillary services to electricity utility companies. DSM is employed to varying capacities in the industrial, commercial, and residential sectors, and is subject to various technical, economic, and behavioral constraints. Understanding and addressing these constraints can increase the uptake of DSM and provide an effective means of integrating greater shares of renewable energies and effectively managing power networks.

Industrial DSM is limited by process constraints, such as the need for steady electricity supply to sensitive processes. However, electricity consumption can contribute a large portion of industrial operation costs. Thus, the economic benefit of Industrial DSM, where technically feasible, is high. The uptake of DSM is highest in industries with large inventory buffers and/or thermal storage in their processes, and for those who have longer operational shifts, as these characteristics increase the flexibility potential and can be used to target industries to increase DSM uptake.

Commercial DSM is limited by technical constraints such as workplace regulations and customer requirements. Additionally, unlike in the industrial sector, the economic benefits of DSM in the commercial sector are low, relative to operational costs. Overcoming knowledge barriers for consumers, improving human integrations in modeling, and extending DR incentives to smaller consumers are all opportunities for increasing DSM uptake in the commercial sector.

Unlike in the commercial and industrial sectors, electricity demand in the residential sector is primarily constrained and driven by human behavior, leading to large variability and peaks. Many households are sensitive to changes in electricity costs and many residential loads can be shifted with little impact on human outcomes. Therefore, the potential impact of residential DSM is large, yet requires a detailed understanding of behavior.

Three broad categories of behavioral models are used: deterministic models, stochastic models, and agent-based models (ABMs). Of these three types, ABMs are best suited to simulating variable electricity consumption behavior and understanding the connections between human behavior, human outcomes, and DSM. Future improvements in ABMs, such as incorporating decision-making models informed by social science and behavioral economics, will better capture the complex interactions between humans and electricity, and therefore lead to more effective DSM integration, particularly in the crucial residential sector.