1. Introduction
The growing energy demand, the environmental pollution, and the attractive renewable policies lead consumers to adopt renewable energy sources. The emission of greenhouse gases is reduced by prioritizing the utilization of available renewable resource energies rather an the utility grid. Even though the reduction in the emission of greenhouse gases is a major advantage, the interruption of Distributed Generations (DGs) makes the reliability of the grid vulnerable. In the Smart Grid context, DSM technique is extremely promising technique to make the grid stable and dependable, limit consumer energy usage and decrease power generation. The schematic approach of DSM evolves around DR Demand Response programs, energy efficiency programs, and of all energy conservation programs.
It is essential for the DSM framework to have the quality of mitigating certain criteria such as electricity cost minimization, user comfort maximization, and peak demand reduction. This can be achieved using various optimization algorithms. The literature section discusses different optimization algorithms that have been used to implement DSM. Various optimization algorithms such as CS, GWO, PSO, SA, HS, DEA, ACO and GA have been employed. All bio-inspired algorithms have their advantages and limitations. The primitive goal of proposing hybrid forms of algorithms is to achieve better and most promising results by combining appropriate conventional algorithms. BPSOGSA is a proposed hybrid algorithm proposed which is calibrated as both BPSO and GSA for better outcomes. The algorithm offers a decent improvement in the resolution of the intended purpose to minimize electricity costs. The search process in the hybrid BPSOGSA algorithm is supported by agents that replicate the characteristics of BGSA in the research phase and BPSO in the utilization phase. The ultimate aim of hybridizing efficient GSA and PSO is to incorporate the rationale of PSO and the exploration of GSA. Followed by that, distinguished performance of optimization algorithms entirely relies on the tuning algorithm with respect to the system. Generally, it can be achieved by adopting various upgrading methods such as changing the search space, modifying convergence parameters, and lastly, hybridizing the individual algorithms. From the mentioned means of improvisation, hybridizing strengths of individual algorithms results in better performance.
In this extensive study, the hybrid BPSOGSA algorithm is employed in shifting and scheduling the institutional loads, which are classified into two classes, namely, SLs (Shiftable Loads) and NSLs (Non-Shiftable Loads). The hybrid BPSOGSA algorithm optimizes the operating period of each type of load by shifting the load to a period with a lower electricity price. An integrated model that schedules loads using the traditional Ant Colony Optimization (ACO) methodology to minimize overall journey length and unpredictable demand has been used to estimate delivery routes and truck loads [
1]. Four heuristic-based strategies for optimum DG location and sizing were proposed by using GSA, PSO, CSA, and GA based on loss minimization criteria [
2]. A path for planning and navigation has emerged using GSA, PSO, and Simulated Annealing (SA) algorithms involving autonomous mobile robots [
3]. A unique methodology by combining PSO with adaptive GSA for improving FACTS devices to ensure voltage stability in power transmission networks has been established [
4]. A unique GSA (Gravitational Search Algorithm) has been evaluated as it is useful in multiple objective problems [
5].
In a smart grid, a heuristic search-type controller for energy management has been suggested for the implementation of DSM [
6]. Their work examines five different heuristic algorithms, including GA, BFOA, WDO, BPSO, BFOA strategy, and concluded with a suggested hybrid genetic wind-driven (GWD) approach. A familiar Supply Chain Management (SCM) model Vendor Managed Inventory (VMI) model made of a PSO-based algorithm has been developed to find the near optimum [
7]. In [
7], for defuzzification, the centroid defuzzification method has been used to discover ideal retailer order amounts to minimize overall inventory and transportation costs. A control scheme called PS-GSA has been developed for a double induction wind turbine generator by means of Fuzzy’s sliding mode control [
8]. A hybrid PSOGSA-based model has been developed to analyze High-Dimensional (HD) data to provide the solution to the convergence problem [
9].
A Binary Grey Wolf Optimisation Algorithm (BGWO)-based DSM has been developed to solve the non-linear objective of cost minimization, concerning the electricity tariff of a real-time residential load [
10]. Their paper proposes strategic shifting and scheduling of the residential loads optimally deploying the swarm intelligence-based BGWO algorithm. A reinforced DSM approach involving enhanced BGWO has been designed and developed to schedule pre-categorized classes of loads in an educational institution with an aim of deducing peak demand and thereby indirectly minimizing the electricity tariff [
11]. A DSM technique based on the PSO algorithm has been developed to diminish the electricity price and schedule different loads in residential, commercial, and industrial areas [
12]. A multi-objective stochastic optimization algorithm for scheduling both reserves and energy by allowing simultaneous participation of loads with large wind energy penetration has been proposed [
13]. A combination of both the heuristic and randomization algorithms considering both electricity cost and user convenience changes the non-convex method into a convex method [
14]. For distributed generation sources, a two-stage optimal scheduler has been formulated [
15] and it employs MILP to deduce the bi-objective problem of minimizing net annualized cost TAC and net emission TAE to a single-objective problem. The BPSOGSA algorithm has been utilized in optimally scheduling the available hybrid renewable energy resources for minimizing LCOE and also the probability of power supply failure [
16].
The clustering problem has been solved by hybridizing GSA with one of the promising heuristic methods, which is done to improve results gained from GSA [
17]. In [
18], the convergence speed of GSA is improved by employing position-based learning GSA. In [
19], the convergence of GSA is improved by a modified Immune Gravitation Optimization Algorithm (IGOA). The combined quality of both social thinking and individual thinking of PSO is inherited in hybrid PSOGSA which continues the capability of problem solving as GSA aids in parameter identification of hydraulic turbine governing systems [
20].
In [
21], an algorithm GSA was developed by adopting the concepts of Newton’s laws of motion and gravity that form interaction among the search agents [
21]. Similarly, the binary version of GSA, namely, BGSA (Binary Gravitational Search Algorithm) was suggested [
22]. A DBHS algorithm is developed utilizing pitch variation problems [
23]. Whereas, harmony search and pitch adjustments deployed in MBDE have been proposed to solve distinct optimization problems [
24]. The discrete binary version of PSO has been developed to deal with discrete space [
25] while hybrid binary version of PSOGSA, known as BPSOGSA, has been developed to achieve a better performance with a regard of both exploitation and exploration utilizing positive aspects of both PSO and GSA [
26]. A hybrid PSOGSA algorithm suggested to train feedforward neural networks for solving a local convergence problem [
27].
A fully distributed and interactive learning approach has been proposed for integrated charging regulators of electric vehicles, to regulate three socio-technical aspects: dependability, discomfort, and fairness [
28]. The daily load profiles of residential loads have been assessed by incorporating clustering k-means and self-organizing map algorithms and taking into account the social traits and lifestyles of real-time communities [
29]. An IoT-based DSM is proposed using hardware to dynamically optimize the load pattern in actual commercial buildings [
30]. The smart socket was designed in real time using the mobile Blynk application for remote monitoring of Power and connected loads management [
31]. The socket may be accessed remotely over wireless media. This can be further upgraded by introducing WPC (wireless power communication) for ensuring effective and efficient communication to facilitate user end [
32]. Three modes of communication among various access points, co-located power station, and wireless user enhances reliable and fastest means of interaction [
33]. For accelerating and promoting user preferences, using an IoT-enabled recommended system is promising [
34]. Following this, introducing remote monitoring and control options for data physical fusion technology is essential. Mobile IoT interlinked with appropriate applications paves the way for betterment in performance and also adds research significance [
35]. In this case, investigating the smart grid without creating an adverse impact on resiliency and efficiency is the requirement. Hence, IoT and cloud-based detection systems were found to be feasible with integrated AI approaches [
36]. All that is necessary for AI is the quality of data for processing. The real-time data retrieved are stored in the cloud/local storage and undergo AI treatment based on the nature of labeling data [
37]. The data involved in the Smart grid are of various parameters addressing power quality, operating status, fault detection, and resulting optimal solutions [
38].
1.1. Research Gap
From the literature review conducted so far, it can be stated that optimization algorithms were used for different nonlinear problems, in different domains to solve various real-time optimization objectives. Among the previous works, the DSM approaches and techniques were hardly deployed using a real-time educational load profile. This research offers a unique BPSOGSA hybrid optimization algorithm-based demand-side management for KCET—Kamaraj College of Engineering and Technology, an academic higher education institution situated in Southern India.
KCET’s infrastructure spans 47.92 acres of land and is scheduled to work from morning 9:00:00 to evening 16:00:00. The institution offers 16 undergraduate and postgraduate programs with 58 highly equipped laboratories of the latest machinery. KCET has a maximum demand of 1000 kVA and is facilitated by a 32 kW solar PV system with battery storage and converter system along with three diesel generator sets, one DG of 500 kVA, and the rest two of 250 kVA. These spacial and temporal factors make the load profile of educational institutions differ from one another. On the whole, the consumption trend of any institutional building completely relies on their working time, the nature of load types used, and the shifting of load operating patterns. The placement of DSM techniques and approaches in educational institutions helps in minimizing the electricity consumption price considerably.
Thus, the multi-objective SDSM system proposed in KCET for reconciling and highlighting following objectives
Minimize the cost involved in electricity consumption;
Categorize a wide range of loads for allocating suitable time of operation;
Accountable reduction in PAR and peak demand;
To shift and schedule load optimally by considering all the possible constraints;
Addressing user comfort index (UCI)-based optimization.
1.2. Contributions
The following contributions to the effective implementation of DSM in KCET, an Indian educational institution, have been proposed:
Internet of Things (IoT)-enabled Smart Demand-Side Management System (SDSMS) has been implemented in real time for energy monitoring and connects the solar power line for the laboratory loads during peak periods;
The load classification of KCET was obtained based on load shifting with and without interruption;
The institutional load profile is optimally scheduled by applying the hybrid BPSOGSA algorithm using MATLAB software;
The load profile of KCET is shifted and scheduled a day ahead effectively to improve the degree of user comfort by utilizing the renewable energy (RE) solar source set up at the institution;
Next, the real-time hardware implementation of SDSMS was applied to increase the degree of user comfort (DUC) during the peak hours powered by the Solar PV (SPV) resource, and all the determining real-time electrical parameters are monitored with the aid of the Blynk application.
1.3. Novelty of Proposed Work
In this research work, an IoT-based smart demand-side management system is developed using a hybrid BPSOGSA algorithm to achieve the main goal of lowering the cost involved in electricity consumption. It can be made possible by optimally scheduling the different classes of institutional loads considering the operational constraints associated with each class of load. The outcomes of the hybrid BPSOGSA algorithm highlight its performance over other two algorithms—BPSO and BGSA—in electricity tariff reduction, avoiding reach of peak demand, and maintaining PAR. Additionally, the statistical performance of all the three algorithm BPSOGSA, BPSO, and BGSA, has been determined using statistical tests. Holm, Hommel, and Holland tests are considered for this study. Next, the degree of user comfort has been improved by utilizing the solar PV generation located at the institution. Finally, real-time hardware implementation of SDSMS has been deployed at the Renewable Energy lab to improve the degree of user comfort, and the electrical parameters obtained are continuously monitored using the Blynk application.
The paper is organized as follows: The methodology adopted is interpreted in
Section 2. Optimization algorithms are outlined in
Section 3. The BPSOGSA algorithm is formulated in
Section 4, while
Section 5 emphases on results and discussions.
Section 6 concludes the work with key findings.
6. Conclusions
This paper deploys an Internet of Things (IoT)-based Smart Demand Side Management System (SDSMS) that connects the solar power line for the laboratory loads during peak hours to increase the degree of user comfort (DUC) at an educational institution KCET located at Madurai, Tamilnadu, India. In this paper, DSM for the institutional loads is first implemented using a hybrid BPSOGSA algorithm and the same is integrated as hardware in real time. The KCET loads are categorized into shiftable loads and non-shiftable loads (NSLs) loads. The overall reduction of electricity consumption cost of the institution is achieved by deploying day-ahead load shifting and scheduling in the shiftable loads (SLs). The performance of the hybrid BPSOGSA algorithm has been evaluated in terms of parameters such as PAR and peak demand reduction, and electricity cost minimization by comparing it with individual optimization algorithms BPSO and BGSA using MATLAB. A overall cost reduction is achieved by 35.27% using the proposed hybrid BPSOGSA algorithm, while 27.72% and 24.502% reduction in electricity cost by BPSO, and BGSA respectively. PAR without DSM is 4.12 whereas using BGSA, BPSO, and BPSOGSA, PAR is reduced to 4.01, 3.73, and 3.26 respectively. Furthermore, peak demand without DSM is 1855.47 kW, whereas using BGSA, BPSO, and BPSOGSA, PAR is reduced to 1756.23 kW, 1704.62 kW, and 1502.24 kW, respectively. Hence, BPSOGSA gives the best performance for the minimization of electricity cost, reduction in PAR, and reduction in peak demand. The major highlight of using the hybrid BPSOGSA algorithm-based DSM is that it is effective in eliminating the demerits of both BPSO and GSA algorithms. Further, the degree of user comfort (DUC) has been improved by deploying hybrid BPSOGSA incorporating the Solar PV (SPV) generation at the premises of KCET. The ratio of the number of retained operational hours to the period of operation is directly proportional to the DUC. The use of SPV in the hybrid BPSOGSA algorithm raises the DUC level without ratio dependency. Next, the hardware has been implemented in real time at the Renewable energy laboratory to improve DUC for different load types by optimal scheduling and it reduces the electricity cost during the peak period for the institution, and the corresponding electrical parameters are monitored in real time using the Blynk application.