Impacts of Electric Vehicles Charging in Low-Voltage Distribution Networks: A Case Study in Malta

Abstract

A high penetration of electric vehicle (EV) charging in low voltage (LV) networks can challenge grid stability due to voltage variations and limited feeder capacity. This research paper examines the integration of electric vehicle (EV) charging in real-life residential low voltage (LV) networks in Malta. The study utilizes smart metering data and presents a methodology framework and tools to analyze the impacts of EV charging on grid stability. The likelihood of challenges in the LV network is assessed by conducting simulations and deriving cumulative distribution functions (CDFs). The study also evaluates the impact of EV charging on the occurrence of network challenges and identifies predominant issues through multi-feeder analyses. Additionally, a regression analysis tool is developed to predict the impacts based on feeder characteristics. The results show strong relationships between feeder characteristics and EV charging processes, offering valuable insights for network planning and operations. However, it should be noted that the current EV charging penetration in the Maltese grid is below 1% in any LV feeder, suggesting the absence of significant technological hurdles at present.

1. Introduction

The decarbonization of the energy system relies on the electrification of transportation due to the significant contribution of the transportation sector to global greenhouse gas emissions. To promote clean energy and mobility, many countries have set national goals for electric vehicle (EV) deployment. By using electricity more efficiently, road transportation can become more environmentally friendly, reducing noise and local contaminants. This research aims to assess the statistical impacts of EV penetration, specifically the number of households with EVs, on a real low voltage (LV) network using actual smart meter profiles. High EV penetration can have adverse effects on LV distribution networks, which are further complicated by the dynamic nature of EV charging. To ensure the successful integration of EVs into distribution networks, a deeper understanding of EV charging patterns is necessary.

Both system dependability and quality must be ensured by the utility providers. The extent to which increased EV adoption affects electric networks heavily relies on the technology and charging methods employed. Typically, EV charging takes place in residences and places of business; therefore, significant EV adoption is anticipated to have an influence on LV distribution networks in residential and commercial regions initially, much as PV integration issues [1]. In addition, EVs are not stationary loads as conventional loads are, therefore, a deeper comprehension of EV charging habits is required to ensure proper integration of these vehicles into distribution networks.

Considering an electric vehicle as a basic load, it may use up a lot of electricity, and, as a result, voltage issues in overloaded grids are possible, particularly when peak demands coincide with EV charging times. Therefore, it is critical to establish load management measures to prevent serious technical issues and their effects, which might undermine the advantages of EV integration. Understanding how much EV integration affects the stability of LV grids in terms of voltage profile and feeder loads is the first step. Power flow analysis is used in this process, and Monte Carlo simulation is used to account for uncertainty. The research in [2] examined the consequences of extensive EV deployment, taking into account the owners’ diverse travel needs, such as distance, road conditions, vehicle type, etc. According to the study, EV charging causes voltage violations and overloading in rural distribution networks.

A multistage time-variant EV modeling technique was suggested by the study [3] to investigate the impacts of EV charging on the distribution network. The voltage profile and energy losses on the IEEE 69-bus distribution test network were evaluated by the authors. The findings confirm that the load models have a significant influence on energy losses and terminal voltages.

According to compliance with standards (such as EN50160, ANSI, etc.), a number of studies evaluated the effects of EV integration on voltage profile [4,5,6,7,8,9,10,11,12]. Regarding compliance with voltage standards, all cited studies revealed voltage violations. The majority of studies, however, only examined general or representative networks [6,7,8], which would not necessarily have the same capacity for hosting EV integration as actual ones.

This paper, an extended paper from [9], is set up as follows. The methodology-based framework of the study is described in Section 2. The implementation of the methodological and framework is then shown in Section 3 through an effect study on a real-world Maltese LV network originating from a single 11/0.4 kV substation with seven LV feeders. Section 4 uses the extracted cumulative distribution function (CDF) from the Monte Carlo approaches to discuss the likelihood of LV network challenges. After that, Section 5 presents and discusses the detailed real-life Maltese LV multi-feeder impact study results. The regression analysis tool and its findings are then highlighted in Section 6. The primary findings and a crucial discovery that differs from other research on the insufficient measure for non-linear regression are presented in Section 7, which concludes the article.

2. The Methodology-Based Framework

A three-phase unbalanced time-series power flow is performed by OpenDSS [13]. The methodology for evaluating PV/EV penetration on LV networks based on the size and behavior uncertainty [1] is shown in Figure 1. Smart meter load and EV profiles are loaded after creating computer models for LV feeders and EV profiles. For a given feeder, a Monte Carlo simulation is performed.

For each EV penetration, various power flows (from 0% to 100% in 10% stages) are assessed. A random load and EV profile are chosen for each power flow study from a pool of profiles. The effect analysis measure for each simulation (the utilization factor and percentage of users experiencing voltage problems) is saved. One concludes by comparing a large number of feeders to identify feeder characteristics that might clarify technical issues.

2.1. Data Energy Profiles

Adopting time-series profiles is crucial to providing an accurate assessment of the effects of PV and EV integration. The widespread installation of EV charging units necessitates a clear requirement for long-term knowledge of LV networks. By 2014, 90% of Malta’s energy operator Enemalta plc’s customers had access to smart meters, making it the first country in the world to do so [14]. In this investigation, residential loads and EV profiles were employed. Based on data provided by smart meters, a probabilistic methodology is used in this study from 5000 residential loads and 2000 EV profiles generated from the My Electric Avenue (MEA) initiative [15].

This section continues to give a general overview of converting monitoring device data into OpenDSS format and building EV profiles using that data. The recorded data of each electric car’s charging activity is referred to as CARWINGS data. Every time the EV is charged, a sample of this data is taken, making each charge a representative charge. Start charging time, beginning state of charge, and ultimate state of charge are among the variables that are recorded for each sample. Nissan Leaf’s charge level is represented as 12 units, each containing 2 kWh. Data from CARWINGS have been studied to better understand the behaviors of EV users, such as when they are most likely to charge their vehicles, as well as their beginning and ultimate charging states.

The probability distribution functions (PDFs) of the charging behavior comparing various factors taken from the CARWINGS data are shown in Figure 2. These numbers indicate that the majority of customers charge their batteries entirely and that the peak time for charging EVs is about 18:00.

The PDFs described in Figure 2, which represent the start charging time, starting iSOC, and ultimate fSOC, are used to predict how EVs would behave. In order to comprehend the potential effects of EV charging, it is crucial to take into account the battery’s power, charge level, and charging demand. The amount of time the battery must be connected to the power source is determined by these factors.

It is important to note that the charging process is regarded as continuous, which means that once it begins, it will not end until the user disconnects the car or the battery is entirely charged. It is, therefore, feasible to construct EV profiles to be used in subsequent research while taking the PDFs and EV into consideration.

The EV utilized in the simulation is based on the Nissan Leaf, a genuine electric vehicle with a battery capacity of 24 kWh and 3.3 kW maximum charging power. Therefore, EV has to be plugged into a power source for roughly 8 h in order to completely charge the battery. Given that the SOC data is divided into 12 units, one unit corresponds to 40 min. This type of vehicle and slow charging facilities represent a realistic scenario on LV feeders in Malta that may feature in the future, given low range requirements and limited power capacities. Taking the aforementioned into account, EV profiles may be developed and used in future impact assessments. The profile-creation procedure entails:

Step 1: A random connection time is chosen based on the related PDF.

Step 2: A random beginning state of charge is chosen based on the associated PDF.

Step 3: The ultimate state of charge is chosen at random based on the related PDF.

Step 4: Based on the number of units that must be charged, the amount of time needed to get from the starting state of charge to the end state of charge is computed.

According to the information above, billing takes place between the connection and disconnecting times. It is feasible to construct thousands of EV profiles by using the methods described in the preceding chapter. Figure 3 displays five independently chosen individual EV profiles to illustrate this. The averaged profile may be made from separate profiles. This might help you comprehend the peak demand for EVs. The aggregated load, EV, and total demand profiles are shown in Figure 4. The typical peak demand is at 1.08 kW. Additionally, Figure 4 shows that when all homes incorporate EVs, the maximum average demand of residential load increases from 0.5 kW to 1.55 kW. This provides a crucial perspective for evaluating the effects of EV integration.

2.2. Impact Assessment

The present study’s methodology utilizes Monte Carlo simulations to address the uncertainties associated with the size and behavior of electric vehicles (EVs). In order to analyze two EV allocation scenarios, a total of 100 simulations were carried out. These simulations were conducted across 11 penetration levels, which ranged from 0% to 100% in increments of 10%. In the context of power distribution, EVs are assigned from the substation to the ultimate consumer, which is commonly referred to as downstream. EVs are allocated from the final consumer to the transformer in the upstream direction.

The utilization of allocation scenarios is implemented to address unfavorable circumstances in order to ascertain the limits of low voltage (LV) networks. It is imperative to acknowledge that the technical implications of EV charging consumption are subject to variation throughout the day due to their temporal fluctuations. Hence, it is imperative to consider technological considerations while quantifying. The impact evaluation incorporates a pair of dice for this purpose.

Voltage-related concerns: the present index employs the computation of voltage for individual customers to ascertain its conformity with the European Standard EN50160 [16], which stipulates that the voltage magnitude must fall within the range of 230 V +/− 10%. The violation is detected at a voltage level of either below 207 V or above 253 V, denoted as in [17]. The present study employs the terminology “voltage issues/problems” or “voltage challenges” to denote instances of voltage violation characterized by voltage magnitude, primarily overvoltage resulting from reverse power flows triggered by the substantial integration of photovoltaic systems. A non-compliant behavior exhibited by a consumer is indicative of a problem. The percentage denoting the overall count of consumers encountering voltage issues is ascertained. Adhering to the EN50160 standard, any deviation from the prescribed values in the profiles, which have a resolution of fifteen minutes, would be considered a violation.

The utilization of a feeder can be determined by dividing the maximum current by the ampacity head of the feeder. The present index serves to demonstrate the diverse applications of the LV feeder across different degrees of penetration.

3. Application of the Methodology Framework

3.1. Real-Life Maltese LV Network

The methodology framework was applied to a real-life Maltese LV network originating from an 11/0.4 kV substation with seven three-phase LV feeders, as depicted in Figure 5, in order to emphasize the impact analysis.

Table 1. Main characteristics of LV Networks [1].

Feeder	Total Length (m)	No of Loads	Phase Connectivity
			1	2	3
1	1706.3	121	0.31	0.39	0.3
2	461.9	30	0.33	0.33	0.33
3	1558.1	128	0.294	0.319	0.387
4	1391.4	146	0.28	0.372	0.348
5	1015.6	83	0.351	0.378	0.271
6	778.1	71	0.354	0.384	0.262
7	565.2	50	0.25	0.375	0.375

presents the primary attributes of the LV network, which were determined through Geographic Information System (GIS) modeling. This modeling took into account various factors, including conductor characteristics, consumer location, phase connectivity, and network topology. LV Feeder 4 was selected to emphasize the per-feeder outcomes, as it is considered to be the 160th most heavily loaded feeder in this instance [1].

3.2. Summary of Results for Real-Life Maltese LV Network

Table 2. Summary of penetration levels and technical challenges thresholds.

Feeder	Technical Challenges Due to EV Charging Penetration Levels
	Voltage Issues Downstream	Voltage Issues Upstream	Utilization Factor > 70% Downstream	Utilization Factor > 70% Upstream
1	20%	20%	20%	20%
2	-	-	-	-
3	10%	10%	20%	20%
4	30%	10%	10%	10%
5	80%	50%	30%	30%
6	80%	30%	40%	50%
7	-	-	90%	100%

presents a summary of the outcomes for the seven (7) feeders. The findings illustrate the initial instances of technical obstacles resulting from the proliferation of EV charging. A utilization capacity threshold of 70% was chosen to signify potential constraints in headroom capacity for operational duties.

3.3. Voltage Issues

Figure 6a depicts the impact of varying electric vehicle (EV) penetrations on consumers experiencing voltage issues, including the mean value and standard deviation of these occurrences. The observable differences in impacts are evident when comparing allocation scenarios between downstream and upstream. As an illustration, the proportion of consumers experiencing voltage issues is less than 20% in the downstream scenario, while it immediately represents some issues to a small percentage of affected customers from the first EV integration in the upstream scenario at a very low electric vehicle adoption rate. The aforementioned data indicates that voltage problems typically arise at an approximate EV penetration rate of 10% and 30% for upstream and downstream allocation scenarios, respectively.

3.4. Utilization

The utilization factor mean and standard deviation at the feeder’s head for both allocation scenarios are depicted in Figure 6b. As anticipated, there exists a minor discrepancy between situations involving downstream and upstream factors. On average, the loading level at the outset is 70%. The magnitude of the aforementioned phenomenon exhibits an upward trend in direct proportion to the number of electric vehicle units that are interconnected. The utilization factor experiences an upsurge subsequent to the penetration of Evs, reaching overloading levels at around 30% penetration levels.

3.5. Conclusion on Single LV Network Analysis

The observations made suggest that feeders with minimal loads and shorter lengths do not exhibit immediate technical complications, albeit the results are confined to a single LV network. Conversely, feeders that are heavier and have greater lengths are prone to encountering technical difficulties when reaching a certain level of penetration. Notably, voltage and utilization issues were mostly not observed until the penetration reached 10%. Nevertheless, in order to thoroughly examine the aforementioned technical issues across various LV feeders, a comprehensive evaluation of multiple feeders is conducted in Section 4.

4. Likelihood of LV Network Challenges

The preceding sections have presented an impact analysis that can be further extended due to its stochastic nature. This can be achieved by extracting the cumulative distribution function (CDF) to showcase the likelihood of encountering voltage or utilization capacity problems. This instrument is utilized to determine the acceptability of a specific electric vehicle charging penetration level that may result in technical difficulties. The quantification of this probability can aid utility companies in determining the feasibility of accepting penetration levels that represent low probabilities of technical challenges instead of investing in infrastructure.

Based on the results obtained from 100 simulations, an empirical cumulative distribution function (CDF) has been selected; this bears a resemblance to the cumulative distribution function (CDF). Both of these are probabilistic models utilized for data analysis. Empirical cumulative distribution function (ECDF) models are based on observed data, while the cumulative distribution function (CDF) is a theoretical distribution model. The empirical cumulative distribution function (ECDF) allocates a probability of 1/n to each data point and subsequently computes the cumulative sum of these probabilities up to and including each datum. The outcome is a discontinuous function that exhibits a staircase-like pattern, with increments of 1/n at each step.

Let X denote the proportion of customers experiencing voltage violations. The probability of encountering voltage issues in downstream and upstream allocation scenarios can be represented by the cumulative distribution functions (CDFs) depicted in Figure 7. These functions indicate the likelihood of having a minimum percentage of consumers experiencing voltage problems. The findings indicate that in the downstream allocation scenario, the probability of encountering voltage issues among over 20% of consumers is estimated to be approximately 0.15, 0.42, and 0.8 for 50%, 60%, and 70% of penetration, respectively, Figure 7a. In contrast, the upstream allocation scenario, Figure 7b, suggests a probability of 0.95 and 1 for 50% and 60% of electric vehicle (EV) penetration. It can be observed that in both downstream and upstream scenarios, voltage challenges are experienced by over 20% of consumers following 60% and 90% penetration.

The primary hindrance to integrating electric vehicles is the overloading issue. According to the data presented in Figure 8a, it can be inferred that the likelihood of observing a utilization factor exceeding 100% is 0.2, 0.5, and 0.9 when the penetration rates are 20%, 30%, and 40%, respectively, in the downstream allocation scenario. According to the data presented in Figure 8b, the likelihood of encountering issues related to overloading is 0.4 and 0.8 when the penetration levels are at 30% and 40%, respectively, in the context of the upstream allocation scenario. Furthermore, it is noteworthy that there will invariably arise issues of overloading subsequent to a penetration level of 50% in both upstream and downstream scenarios.

Regarding the acknowledgment of a potential technical problem, utility companies have the discretion to determine whether an issue is deemed significant if the likelihood of experiencing voltage-related complications among their consumers surpasses a specific threshold, such as 1%. Therefore, if the utility fails to respond to potential technical difficulties, the aforementioned parameter may be assigned a value of 0. This suggests that voltage problems are expeditiously detected upon emergence in any of the simulations. Conversely, in the event that the utility is amenable to accepting a certain degree of potential technical complications, it may elect to institute a more elevated threshold.

Figure 9a depicts the mean percentage of customers encountering voltage issues, accompanied by the standard deviation, concerning the assimilation of electric vehicles on Feeder 4. Furthermore, the probability of encountering more than 1% of customers facing voltage problems per level of penetration is illustrated. Therefore, if the utility system fails to accommodate all possible voltage fluctuations, issues may arise when the penetration level reaches 30%, with a likelihood of 0.1. On the other hand, the upstream situation illustrated in Figure 9b gives rise to voltage complications in a prompt manner, exhibiting a likelihood of 0.3 when the penetration level is as low as 10%. The main obstacle to the incorporation of electric vehicles is the concern about overloading. The probability of immediate issues is depicted in Figure 10, which reveals a linear increase in utilization factor for both downstream (a) and upstream (b) scenarios at a mere 10% penetration level.

5. Multi-Feeder EV Charging Impact Analysis

This section summarizes and discusses the results obtained from the first-ever comprehensive real-life Maltese LV multi-feeder stochastic impact analysis on EV charging following a similar methodology shown in [1].

In the context of power distribution systems, the identification of a voltage violation occurs when the likelihood of a feeder having a proportion greater than 1% of its consumers experiencing voltage-related problems, denoted as +X, exceeds a predetermined threshold denoted as α. Specifically, a voltage violation is flagged when the probability of +X being greater than or equal to 1, represented as P(X ≥ 1), is equal to or surpasses the threshold α.

The utility company establishes the threshold based on its assessment of the potential technical issues that may arise. The term “overloaded feeder” refers to a scenario where the likelihood of a utilization factor exceeding the rated feeder capacity by 100% is equal to or greater than the value of α, represented by the equation P(Y > 100) ≥ α. Thus, in the event that the utility set α is set to zero, technical issues will be promptly documented subsequent to the occurrence of a flagged case in one of the simulations. Conversely, when α is assigned a value of 0.05, the technical complications are documented, provided that a minimum of 5% of the simulations exhibit identified issues.

Table 3. Percentage of feeders with technical problems.

Case	α = 0		α = 0.05
	Voltage Issues	Overloading	Voltage Issues	Overloading
Downstream	63.1%	76.2%	58.3%	71.4%
Upstream	65.5%	76.2%	58.3%	69.1%

presents a summary of the multi-feeder analysis pertaining to the proportion of feeders experiencing voltage and utilization capacity issues across two thresholds, namely, a conservative α value of 0 and α value of 0.05. The latter threshold is widely deemed acceptable by utility companies. In the conservative scenario, it was observed that approximately 64% and 76% of the feeders experienced technical issues related to voltage and utilization capacity at any given simulation and penetration level. The primary technical challenges that surfaced were related to capacity utilization rather than voltage issued.

5.1. First Occurrence of LV Network Challenges

The initial instance of LV network challenges entails a more comprehensive exploration of the assessment of the impact of EV charging penetration. Section 5.1 presents histograms depicting the penetration level at which feeders encounter technical difficulties. The calculation of the penetration level is performed using Equations (1) and (2).

p₁ ≡ {min(pi) ∈ Q|P(X(pi) ≥ 1) ≥ α}

(1)

p₂ = {min(pi) ∈ Q|P(Y(pi) > 100) ≥ α}

(2)

where Q is the set of penetration levels (0% to 100%), and pi is the penetration level i. Therefore, p₁ and p₂ reflect the initial penetration level when voltage or overloading difficulties occur [1].

The outcome of the initial instance of LV network difficulties, encompassing voltage and utilization capacity technical problems at any simulation and penetration level, is listed in

Table 4. Standard error of regression for each parameter.

Parameter	Downstream	Upstream
Feeder length	44.4	42.9
Number of customers	24.8	28.6
Total path resistance	23.1	22.2
Initial loading	17.6	25.9
Main path	54.7	50.1
Main path resistance	56.5	52.8
Total resistance	47.9	47.4

.

6. A Tool for Predicting Impacts through Regression Analysis

The utilization of regression analysis is a highly effective technique within the realm of network planning and operation. The regression analysis findings can provide valuable insights to utility companies regarding the hosting capacity of low-voltage networks in Malta. Consequently, the utility can evaluate and anticipate the technical obstacles that may arise from increased electric vehicle (EV) charging penetration levels. Additionally, it can determine the limitations of EV hosting capacity on particular low voltage (LV) network characteristics without needing power flow analysis. The plotted data represents the parameters that exhibit the most optimal fit. In this study, the standard error of the regression, also referred to as the standard error of the estimate, is utilized as a more favorable indicator of the goodness-of-fit compared to the coefficient of determination, R². This approach is because the standard error of the regression, denoted as S, is applicable for both linear and non-linear models, whereas R² is not a valid measure for non-linear models [18].

6.1. Defining Characteristics of LV Feeders

The seven feeder characteristics that were examined are explicitly defined as follows:

i.

The feeder length refers to the complete length of the feeder, encompassing both underground and aerial cables;

ii.

The number of customers is the amount of individual utility service connections on the grid for a feeder;

iii.

The total path resistance (TPR) refers to the cumulative resistance values that exist between the busbar and individual consumers within an electrical circuit as defined in [1];

iv.

The initial utilization factor pertains to the initial usage of a resource or system. The average utilization factor is determined by dividing the maximum current by its corresponding ampacity at the feeder’s head based on 100 simulations that do not incorporate EV integration;

v.

The main path is the measurement of the distance separating the busbar from the farthest consumer;

vi.

Main path resistance (MPR) is the cumulative resistance along the primary path from the substation to the final consumer;

vii.

Total resistance encompasses all feeder resistances, including both underground and overhead cables.

It is noteworthy that the computation of complex impedance is not undertaken, and solely the determination of resistance is taken into account, as it represents a more direct and cost-effective method for the utility to execute.

6.2. Regression Analysis Tool Methodology

Regression analysis based on the methodology in [1] examined the integration of electric vehicles (EVs) charging infrastructure along LV networks. The investigation involved seven feeder characteristics plotted against the penetration level at which technical issues arise. Two thresholds were considered for potential technical problems, namely α = 0 and α = 0.05. Therefore, the fourth calculation determines the minimum level of penetration at which feeders encounter technical difficulties.

Equation (3) is a mathematical expression for determining the minimum penetration at which feeders experience technical challenges in relation to two variables, namely p₁ and p₂, corresponding to voltage and utilization capacity issues, respectively:

p_min = min {p₁,p₂}.

(3)

Subsequently, a regression analysis is conducted to determine the optimal fit by utilizing the standard error of the regression (S) as a measure.

Given that over 20% of the feeders exhibited no complications across all simulations for penetration levels up to 100%, further penetration levels are being explored. Hence, a total of 31 penetration levels have been analyzed, ranging from 0% to 300%, with an interval of 10%. This approach implies that a residential dwelling or utility customer has the capability to incorporate numerous electric vehicle units concurrently. The evaluation of penetration levels exceeding 100% can provide significant insights that may be incorporated into regression analysis, as it facilitates the identification of the penetration level that initiates a technical obstacle in the feeder. Consequently, this will yield more precise approximations.

6.3. Regression Tool Analysis

The regression analysis is scattered data points representing the results of each studied LV feeder plotted. The scattered data points are plotted on the EV penetration levels at which the impact assessment records the first occurrence of LV network challenge against the characteristics of LV Feeders.

Table 4 summarizes the regression analysis results in the standard regression error for each parameter for downstream and upstream allocation scenarios. The results for the corresponding S is shown for α = 0, the most conservative threshold. α = 0.05 results provided similar results; the smaller S, the stronger the relationship. This table suggests that the parameters with the strongest relationships are the number of customers, total path resistance, and initial loading. The total path resistance embeds the overall resistance of the feeders. Hence, there is a good relationship with the possible voltage drop issues in the case of EV Charging. On the other hand, the initial loading provides information on how assets are utilized at present, that is, 0% penetration (no EV case), and hence can predict the occurrence of overloading problems.

To illustrate an example of regression analysis, Figure 11 and Figure 12 present the observed penetration values at which problems appear vs. the initial utilization factor and the regression line considering downstream and upstream allocation scenarios, respectively, for EV Charging integration. By looking at Figure 11 and Figure 12, the reduction in sparseness can be observed from PV integration [1] to EV integration, resulting in an improvement of S from 39.1 and 45.1 for the number of customers in PV allocation to 17.6 and 25.9 for downstream and upstream allocation scenarios respectively in this case for EV integration. In addition, given that S is smaller for total path resistance than that of the number of consumers, the penetration has smaller variations, thus, more accurate estimates.

Therefore, to adopt this tool, a utility must compromise between more accurate estimates but require more effort to calculate, for example, initial utilization factor and total path resistance and lower accuracy but having a specific parameter for feeder length and the number of consumers. However, in some cases, the parameters such as feeder length and the number of consumers present a good fit, for example, EV downstream for the number of consumers. Appendix A presents the regression analysis plots for the six key feeder characteristics.

7. Conclusions

This article presents the first-of-its-kind study conducted in Malta, focusing on the analysis of the multi-feeder effects and regression analysis tool for electric vehicles (EV) charging in a real-world low voltage (LV) network. As EVs continue, the transport electrification and have a significant impact on the electrical sector, Malta is positioning itself as an electric mobility priority nation. This study stands out as the pioneering effort to assess the technical effects of integrating high levels of electric vehicle (EV) charging on a large scale in the low voltage (LV) distribution networks, taking into account real-world scenarios and smart meter data.

The methodology employed in this study involves utilizing OpenDSS for power flow analyses and implementing the Monte Carlo technique on a real three-phase LV distribution network with seven feeders. Cumulative distribution functions (CDFs) are then used to assess the likelihood of voltage and utilization issues. The findings highlight that the utilization of LV feeders is the main bottleneck for EV integration which relates to feeders overloading. Additionally, a multi-feeder impact evaluation reveals patterns in the early detection of network issues, leading to the development of a regression analysis tool to investigate relationships between various parameters and the occurrence of technical issues.

Similar to other PV integration studies, EV integration in LV distribution networks is limited, with only a few incorporating regression analysis. During these impact studies, it was determined that the commonly used metric for non-linear regression, the coefficient of determination (R-square), is not suitable. In fact, this work makes use of a more appropriate metric, the standard error of the regression (S), to evaluate the goodness-of-fit of the non-linear model [1].

At present, significant technical hurdles are not evident in the Maltese LV network, as EV penetration levels remain below 1% in any LV feeder. The results demonstrate that feeder overloading concerns, with limited undervoltage, are encountered, particularly at high penetration levels. These overloading concerns may be mitigated through vehicle-to-grid (V2G), vehicle-to-home (V2H), and vehicle-to-load (V2L) approaches in terms of load management and providing ancillary services [8,9,10]. The study evaluates worst-case scenarios to define network boundaries and reveals fewer EV integration problems compared to earlier studies, attributed to the existing infrastructure.