Research on Intelligent Scheduling Strategy for Electric Heavy Trucks Considering Photovoltaic Outputs

Abstract

Due to the extensive use of fossil fuels, energy conservation and sustainable transportation have become hot topics. Electric vehicles (EVs), renowned for their clean and eco-friendly attributes, have garnered considerable global attention and are progressively being embraced worldwide. However, disorganized EV charging not only reduces charging station efficiency but also threatens power grid stability. In this low-carbon era, photovoltaic storage charging stations offer a solution that accommodates future EV growth. However, due to the significant instability in both the charging load and photovoltaic power generation within charging stations, it is critical to maximize local photovoltaic power consumption and minimize the impact of disorganized EV charging on the power grid. This paper formulates an intelligent scheduling strategy for electric heavy trucks within charging stations based on typical photovoltaic output data. The study focuses on a photovoltaic storage charging station in an industrial zone in Xinjiang. While considering the electricity procurement cost of the charging station, the aim is to minimize fluctuations in the electricity procurement load. A simulation analysis was conducted using MATLAB 2021a software, and the results indicated that, compared to an uncoordinated charging strategy for electric heavy trucks, the proposed strategy reduced electricity procurement costs by CNY 1348.25, decreased load fluctuations by 169.45, and improved the utilization efficiency of photovoltaic energy by 30%. A statistical analysis was also used to support the reduction in electricity procurement costs and load variations. Finally, a sensitivity analysis of the weight factors in the objective function was performed, proving that the proposed strategy effectively reduces electricity procurement costs and improves the utilization efficiency of photovoltaic energy.

1. Introduction

In 2023, global fossil energy use accounted for 82% of the primary energy used [1]. China is the world’s largest energy producer and consumer, accounting for 32% of global carbon emissions in 2023, ranking 1st in the world [2]. Fuel vehicles are the main source of carbon dioxide emissions in transportation. The wide spread of new energy electric vehicles (EVs) can effectively reduce carbon dioxide emissions, carbon pollution, and the operating costs of automobiles [3]. The electrification of automobiles is undoubtedly the main way to achieve the goal of “carbon peak, carbon neutral” in the field of transportation, especially in China [4]. The ease of charging electric vehicles is a key factor in their widespread adoption. However, due to the high level of randomness and uncertainty associated with EV charging loads, charging activities that are not properly managed and controlled may pose numerous challenges to the power network [5]. The uncoordinated charging behavior of a large number of electric vehicles affects the safe and stable operation of the grid. The formation of power consumption peaks affects power quality [6]. Currently, it is recognized that the most promising solution for their application is an energy replenishment infrastructure for electric vehicles—optical storage charging stations [7]. At the same time, solar photovoltaic power generation, the main energy source of optical storage charging stations, is also rapidly developing [8]. By integrating photovoltaic power generation with energy storage technology, charging stations can provide charging for electric vehicles directly during the day using solar energy. At the same time, the additional electricity can be saved in an energy storage facility. Such an approach not only enables the charging station to achieve a state of energy self-supply, but also helps to balance the load on the grid and adjust power demands, effectively reducing the pressure on the distribution grid [9]. In addition, the electricity market’s structure and energy transactions play a crucial role in influencing the operation and economic performance of charging stations. The electricity market typically includes multiple segments such as generation, transmission, and distribution. Energy transactions can occur through various forms, including real-time markets, day-ahead markets, and long-term contracts. Understanding and utilizing these market mechanisms are of great importance for optimizing the energy scheduling strategies of charging stations, reducing their operational costs, and enhancing economic benefits.

In recent years, the optimization of electric vehicle (EV) charging strategies has garnered extensive attention. As the adoption of electric vehicles and the rapid development of renewable energy progress, the effective management and control of EV charging behavior have become crucial issues. Below is a review of some significant research outcomes in recent years. Yan Jun et al. successfully smoothed the total charging load curve by utilizing time-of-use pricing and vehicle-to-grid (V2G) interactions, as well as limiting charging power, significantly reducing peak-to-off-peak differentials [10]. Wei Dajun et al. developed a multi-objective optimization model that optimized charging and discharging schedules based on a time-sharing tariff system, achieving the goal of reducing grid load fluctuations and user costs [11]. Wang Ruijuan et al. proposed a multi-time-scale hierarchical joint optimization scheduling method suitable for charging stations with distributed PVs. This method effectively utilizes photovoltaic power generation, achieving optimal energy allocation [12]. Ruixue Li et al. constructed a multi-scenario day-ahead optimization model considering the stochastic nature of PV output, implementing a rolling optimization strategy for the optimal scheduling of energy storage systems [13]. Cao Lingjie analyzed the energy management on the supply side of the system, particularly focusing on the mismatch between peak load and power output at charging stations, and proposed corresponding management measures [14]. Lina Zhang et al. developed an optimization model aimed at minimizing the total operating cost of the system, successfully solving it using a particle swarm optimization algorithm [15]. Chen Gang et al. proposed a distributed cooperative control strategy for energy storage units based on a multi-agent consensus algorithm, considering the life-loss cost of energy storage batteries. This strategy ensured system stability while extending battery life [16]. Luo et al. used stochastic dynamic programming and greedy algorithms to calculate adequate charging prices, ensuring the economic viability of charging stations [17]. Hao Yue et al. employed genetic algorithms (GAs) to optimize a scheduling model with the goal of maximizing single-day profits, providing important references for the economic operation of charging stations [18]. Hongling Jiang et al. considered future power purchase costs and the operating revenue of an energy storage system but did not address demand-side charging planning, indicating further research potential in this area [19]. Recently, some researchers have proposed methods that integrate various optimization strategies to enhance the overall efficiency of EV charging systems. For example, Li Xiaolong et al. proposed a comprehensive optimization model that combines time-of-use pricing, demand response, and distributed energy storage and is capable of reducing grid load fluctuations while increasing user satisfaction [20]. Additionally, Zhao Xuefeng et al. explored charging optimization strategies based on artificial intelligence and machine learning, achieving more precise charging scheduling by predicting user charging behavior and power demand [21]. Cheng Hao et al. proposed a deep learning-based EV charging load forecasting method, significantly improving the accuracy of load forecasting and providing reliable data support for charging scheduling optimization [22]. Wang Wei studied multi-objective optimization algorithms to optimize EV charging and discharging schedules, effectively smoothing the grid load curve [23]. Liu Yang explored the integration of distributed energy and smart grids, proposing a blockchain-based distributed energy trading platform, significantly improving energy use efficiency and transaction transparency [24]. Yang Ming studied the collaborative optimization control strategies of PV–storage systems and smart grids, effectively reducing grid peak-to-valley differences and enhancing system stability [25]. Karmali, L. P., Gholami, A., and Nezamoddini, N. studied the integrated optimization of production planning and the charging and discharging scheduling of electric trucks, proposing an optimization approach to enhance system sustainability and efficiency [26]. Çiçek, A., and Erdinç, O. proposed a risk-averse optimal bidding strategy considering a bi-level approach for a renewable energy portfolio manager, including EV parking lots for imbalance mitigation [27]. The aforementioned literature demonstrates that the optimization of EV charging strategies primarily focuses on achieving load balance and cost minimization through time-of-use pricing, demand response strategies, the integration of distributed PV and energy storage technologies, supply side energy management, and various optimization algorithms. These studies provide valuable theoretical foundations and practical experiences for the construction and operation of future EV charging stations. However, with the continuous growth of EV numbers and the ongoing advancement of new energy technologies, further optimization of charging strategies to enhance grid stability and economic efficiency remains an essential area of ongoing exploration.

The existing literature on EV charging optimization has made some progress but still has shortcomings. Most studies focus on ordinary EVs and do not pay attention to the high-power demands of electric heavy trucks. Additionally, many studies use static scheduling methods that do not fully consider the variability and real-time nature of PV power generation and lack dynamic adjustment mechanisms. Cost considerations are also incomplete, primarily focusing on grid load balancing while neglecting the electricity purchase costs of charging stations. Traditional scheduling strategies are often rule-based, lacking intelligence and flexibility. Although some studies address load peak–valley optimization, they do not provide clear measures to reduce grid load fluctuations, failing to fully resolve their impact on grid stability. Furthermore, many methods are primarily theoretical explorations, lacking practical application validation, and thus exhibit shortcomings in their practicality and operability.

We propose an intelligent scheduling strategy with the following attributes: it specifically targets the high-power charging needs of electric heavy trucks, filling a gap in the research; it dynamically adjusts charging and electricity purchases based on typical photovoltaic (PV) output curves, maximizing the use of PV generation and reducing dependence on the grid; it comprehensively considers electricity purchase costs, thereby improving the economic and operational efficiency of charging stations; and it dynamically adjusts charging behavior, significantly reducing the impact on grid stability and ensuring safe grid operation. We conducted a detailed model comparison between this intelligent scheduling strategy and traditional scheduling strategies and also incorporated a deep reinforcement learning model into our analysis. The results show that this intelligent scheduling strategy outperforms traditional strategies in terms of economic efficiency, operational efficiency, and grid stability, and the deep reinforcement learning model further enhances the prediction accuracy and optimization effects of this strategy.

The sections of this paper are organized as follows:

The Section 1 introduces the research background and its significance, summarizes previous work, and highlights the novelty of this study. The Section 2 explains the principles of the photovoltaic storage charging station and provides specific values for important parameters. The Section 3 analyzes the charging behavior of electric heavy trucks. The Section 4 introduces the problem and methods, presenting the objective function and constraints of this study. The Section 5 completes the simulation modeling and result analysis, demonstrates the effectiveness of the scheduling strategy through a statistical analysis, and conducts a sensitivity analysis of the weighting coefficients to further prove the effectiveness of the scheduling. The Section 6 presents the results and discussion, summarizing the work and findings of this paper.

2. Photovoltaic Storage Charging Station, PSCS

An integrated photovoltaic storage charging station combines photovoltaic power generation, energy storage facilities, and electric heavy-duty truck charging functions, forming a microgrid system with both island and grid-connected modes, as shown as in Figure 1. In island mode, PV power generation covers the charging demand, with excess electricity stored in the energy storage device to ensure self-sufficiency. In grid-connected mode, the system can exchange electricity with the grid to ensure stable operation; surplus PV electricity can be sold to the grid to generate revenue. The combination of PV and energy storage enhances system efficiency, reduces the impact on the grid, supports local renewable energy consumption, and enhances economic viability and effectiveness.

The following are the main components of the photovoltaic storage charging station:

(1)

Energy storage system

The energy storage system of the photovoltaic storage charging station (PSCS) mainly includes electrochemical batteries and bi-directional converter technology equipment. The primary function of the electrochemical batteries is to store electrical energy. When solar photovoltaic (PV) generation exceeds immediate demand, these batteries convert excess electrical energy into chemical energy for storage. Conversely, when PV generation is insufficient to meet demand, the system releases stored energy to ensure the continuous operation of the charging facilities [28]. The key parameters of the storage battery include:

Capacity: 200 kWh;

Charge/discharge efficiency: 90%;

Maximum charging power: 100 kW;

Maximum discharging power: 100 kW.

(2)

Photovoltaic power generation system

The photovoltaic panel is the core of a photovoltaic system, operating on the principle of converting solar energy into electrical energy through the photovoltaic effect of materials such as monocrystalline silicon. This electrical energy is processed by internal electronics and transmitted from the collector lines to the charging facility to power electric vehicles or drive other equipment. In this study, the photovoltaic system has an installed capacity of 700 kW. The electricity generated is used directly to charge electric heavy-duty trucks. When the generated power exceeds demand, the surplus electricity is stored in the energy storage system; conversely, when the generated power is insufficient, the storage system supplements the power supply to ensure the energy supply of the charging stations. Therefore, photovoltaic systems are crucial to the energy supply.

(3)

Charging system

The charging system of the photovoltaic storage charging station (PSCS) is responsible for transferring the electrical energy generated by the storage batteries and photovoltaic system to electric heavy trucks (EHTs). The key components of the charging system include charging posts and the power management system. The key parameters are as follows:

Number: 5;

Maximum Charging Power per Post: 50 kW;

Efficiency: 95%.

The research subject of this paper is a solar energy storage and charging station located in an industrial zone in Xinjiang. Both its photovoltaic output and electric heavy truck charging exhibit an amount of randomness. Without proper guidance for electric heavy trucks, their disorganized charging behavior will reduce the efficiency of the photovoltaic output’s utilization and have adverse effects on the power distribution network in which the charging station is located. Therefore, this paper first introduces the basic composition of the solar energy storage and charging station, focusing mainly on the analysis of the photovoltaic power generation system and the energy storage system. This lays the groundwork for determining the charging optimization goals of the charging station in the subsequent sections and designing intelligent scheduling strategies.

3. Analysis of Charging Behavior of Electric Heavy-Duty Trucks

The use of electric heavy-duty trucks (EHDTs) in industrial zones is becoming increasingly widespread. Their charging behavior usually follows relatively fixed patterns in terms of time and route, reflected in the start and end times of charging and the state of charge (SOC) of their battery. Due to the limited publicly available data on EHDT charging, it is challenging to construct a daily charging demand curve based on existing data. Therefore, this study conducted on-site data collection in industrial zones to obtain relevant data.

Reference [29] analyzed the commuting patterns of private vehicles in Beijing and found that peak commuting times are typically from 7:00 to 9:00 a.m. and from 5:00 to 7:00 p.m. Considering the variations in arrival and departure times among different regions and commuters, this study assumes that the arrival and departure times of EHDTs at photovoltaic storage charging stations follow a normal distribution. Additionally, reference [30] analyzed the 2009 nationwide household vehicle travel data released in the United States and concluded that the initial SOC of commuting vehicles at the start of charging follows a normal distribution. Therefore, this study assumes that the initial SOC of EHDTs at the start of their charging follows a normal distribution.

For the parameter settings, the actual arrival time range is from 00:00 to 24:00, with a mean value set at 10:00 based on industrial zone data and a standard deviation of 1. The actual departure time parameter is set at 19:00, with a standard deviation of 1. The SOC follows a normal distribution with parameters (0.45, 0.12), ranging from 0 to 1 with a step size of 0.01. Based on the above model and parameter settings, we calculated the probability density of EHDTs arriving at and departing from the charging station, as well as the battery’s SOC probability density. The results are shown in Figure 2.

Through the above model and analysis, we gained a deeper understanding of the charging habits of EHDTs in industrial zones. These results provide significant support for the scheduling of charging stations, helping to improve charging efficiency and the utilization of clean energy, thereby promoting the green development of the electric heavy-duty truck transportation industry.

4. An Intelligent Dispatch Strategy for Electric Heavy-Duty Trucks

In summary, the charging behavior of electric heavy-duty trucks exhibits a certain degree of regularity and randomness. For photovoltaic energy storage charging stations within industrial parks, failing to correctly and effectively regulate the large-scale charging behavior of electric heavy-duty trucks will lead to an increase in the peak of the load curve on the distribution side, overloading infrastructure such as transformers and adversely affecting the safe and stable operation of the power grid. Therefore, this section considers the electricity purchase cost of photovoltaic charging stations and proposes a corresponding intelligent energy scheduling strategy to reduce the volatility of the electricity purchase curve. This strategy involves dynamic energy scheduling based on the power generation information of the photovoltaic system to regulate the charging behavior of electric vehicles. The following section elaborates on the intelligent energy scheduling model from the aspects of its objective function and constraints.

4.1. Objective Function

Overall Objective Function:

$$C_{total}=\alpha \cdot C_g+\beta \cdot {\varphi }_g$$

(1)

Electricity Purchase Cost Function:

$$C_g={\displaystyle \sum _{t=1}^{T_s}{q}_g\left(t\right)}\cdot p_g\left(t\right)\cdot T$$

(2)

where Ts is the total number of time slots, q_g(t) is the price of electricity during time period t, p_g(t) is the power received by the photovoltaic storage charging station during time period t, and T is the length of each time period.

Load Variance Function:

$${\varphi }_g=\sqrt{{\displaystyle \sum _{t=1}^{T_S}{\left(p_g\left(t\right)-p_{avg}\right)}^2}}$$

(3)

where p_avg is the average power received.

In summary, these formulas are based on our original design. The aim is to optimize the charging schedule for electric heavy-duty trucks while balancing electricity costs and load variance. To support the validity of this design, several relevant studies in this field have been cited [31,32].

4.2. Restrictive Condition

Total Charging Power Limit:

$${\displaystyle \sum _{t=1}^n{p}_i}(t)\le p_g\left(t\right)+p_v\left(t\right),\forall t\in \left[1,T_s\right]$$

(4)

This constraint ensures that, at any time period t, the total charging power of all electric heavy-duty trucks does not exceed the sum of the photovoltaic power p_v(t) and the grid power supply to the charging station p_g(t). This prevents the charging demand from exceeding the available power supply, ensuring system stability.

Full Charging Constraint:

$${\displaystyle \sum _{i=1}^{T_s}{p}_i\left(t\right)=p_i*\left(1-SO{C}_i\right)},\forall i\in \left[1,n\right]$$

(5)

This constraint requires that each electric heavy-duty truck must be fully charged before leaving the charging station, meaning that the total amount of charge each truck receives must meet its charging needs. Here, SOC_i represents the state of charge of truck i at the beginning of the charging process.

Charging Power Range:

$$0\le p_i(t)\le p_r,\forall i\in \left[1,n\right],t\in \left[1,T_s\right]$$

(6)

This constraint stipulates that the charging power of each electric heavy-duty truck at any time period t must be between 0 and the rated power of the charging pile p_r. This ensures that the charging power of each truck does not exceed the capacity of the charging pile and prevents negative power situations.

Non-negative Photovoltaic Power:

$$p_g\left(t\right)\ge 0,\forall t\in \left[1,T_s\right]$$

(7)

This constraint ensures that the photovoltaic power p_g(t) is non-negative at any time period t. This reflects physical reality, as photovoltaic power cannot be negative.

Charging Time Window Constraint:

$$p_i\left(t\right)=0,\forall i\in \left[1,n\right],t\notin \left[T_{start,i},T_{end,i}\right]$$

(8)

This constraint specifies that each electric heavy-duty truck is only allowed to charge within its designated charging time window [T_start,i, T_end,i]. This ensures that each truck charges during its specified time period, avoiding charging time conflicts.

5. Arithmetic Simulation

5.1. Example Data

This example simulation takes a charging station in an industrial zone in Xinjiang as its field application scenario, builds a simulation model based on simulation software, and simulates and analyzes the proposed scheduling strategy for electric heavy trucks.

The charging conversion efficiency of the electric heavy trucks is 95% and the charging piles in the zone are rated at 50 kW. We set the scheduling calculation strategy to 10 min/time. In addition, this section also sets the real-time electricity price of each period according to the time-of-sale price system of general industrial and commercial enterprises in Urumqi, Xinjiang. It should be noted that the time-of-sale price is the same as the photovoltaic power price, and its specific values are shown in Figure 3.

Based on the results of the field study, the actual number of electric heavy-duty trucks in use in the zone is 40, with a battery rating of 200 kW. As analyzed above, the arrival and departure times at which electric trucks start charging and stop charging, as well as their battery load at the beginning of charging, follow a normal distribution, and their distribution characteristics are shown in Figure 4.

The typical output curve of the photovoltaic system is shown in Figure 5.

5.2. Algorithm Analysis

To demonstrate the effectiveness of the proposed strategy, this paper first simulates the unordered charging behavior of electric heavy-duty trucks at a photovoltaic (PV) charging station. Unordered charging behavior refers to the scenario in which the PV charging station does not regulate the charging process of the electric heavy-duty trucks. The trucks start charging as soon as they arrive at the station and continue until their batteries are fully charged, with the charging power remaining constant throughout the process.

The simulation results shown in Figure 6 reveal the behavioral characteristics of electric heavy-duty trucks in an uncoordinated charging mode. In this mode, the charging demand of electric heavy-duty trucks is primarily concentrated between 8 a.m. and 11 a.m., coinciding with the start of photovoltaic system generation. However, at this time, the capacity of photovoltaic generation is insufficient to fully meet the intensive charging demand, resulting in the charging stations having to rely on the grid for additional power supply. As time progresses, by noon, the majority of electric vehicles have completed charging, leading to a sudden drop in the demand to zero, while the photovoltaic system is nearing or reaching its peak generation. Since the charging consumption of electric heavy-duty trucks is far from large enough to absorb the surplus photovoltaic generation, the excess electrical energy is fed back to the grid through bidirectional inverters. During the peak charging period, the utilization rate of photovoltaic generation can reach 100%; however, as the charging demand decreases, the utilization efficiency of photovoltaic generation correspondingly decreases. This simulation not only highlights the challenges posed by the charging patterns of electric heavy-duty trucks to photovoltaic charging stations and the grid in the absence of effective scheduling but also underscores the importance of optimizing charging scheduling strategies.

In this simulation analysis, the PV charging station purchased as much as 4372.68 kW of electricity from the grid, which cost CNY 2069.78, showing the relatively high cost of operating the charging station. Although the photovoltaic system produced 6033.41 kW of energy, only 3627.32 kW was actually used for charging electric heavy trucks, i.e., 60.06% of the total amount of photovoltaic electricity generated. Further, the electricity provided by PV generation only meets 45.34% of the total charging demand of electric heavy trucks, revealing that there is still much more room for clean energy to support the charging of electric heavy-duty trucks. In addition, the charging process of electric heavy-duty trucks shows significant load volatility, with a variance as high as 187.53, which poses an additional challenge to the stable operation of the power grid.

Figure 7 shows the simulation results of the proposed scheduling strategy under the same scenario. As can be seen from the figure, before 8:00, the output power of the photovoltaic (PV) system is zero, so the only power source for the PV charging station is the grid, which is significantly constrained. Consequently, the charging power of the electric vehicles (EVs) under this regulation is relatively low, and the EV charging curve overlaps with the grid power purchasing curve. The charging power of the electric heavy-duty trucks starts to gradually increase from 7:00, with a trend similar to the PV power output but with a higher magnitude. It peaks between 12:00 and 15:00 and then gradually decreases, dropping to zero around 19:00. Since most of the charging power of electric heavy-duty trucks comes from the PV power output, the power purchased from the grid is relatively low throughout the day. There is a peak in grid power purchasing at noon, which then gradually decreases, reaching zero around 21:00. Throughout the process, the PV output curve overlaps with the output curve of the PV system supplying power to the electric vehicles. Clean energy, as the main power source of the PV charging station, is fully utilized, achieving the goal of the local consumption of clean energy.

During the scheduling process, the grid provides 1996.58 kW of power to electric vehicles at a cost of CNY 721.53, significantly reducing its high operating cost. The total output of the PV system is 6033.41 kW, all of which is used for charging electric vehicles. Furthermore, the PV system’s output accounts for 75.41% of the total energy used for charging electric vehicles, indicating a high proportion of clean energy used. The variance of the electric vehicle charging compliance curve is 18.08, showing that its volatility has a low impact on the grid.

To demonstrate the effectiveness and significance of this scheduling strategy, this paper uses statistical analysis to support its claim of reduced power purchase cost and load variance. First, we calculated the mean and standard deviation of power purchase costs under the disorder model and the scheduling model to initially present the distribution of the two sets of data. For the scheduling model, the mean is 724.46 and the standard deviation is 54.06. For the disorder model, the mean is 2069.33 and the standard deviation is 155.31.

Subsequently, we performed an independent sample t-test on the two sets of data, with the results as follows: the t-value is −81.37 and the p-value is −143.

Regarding the t-value, this indicates how much the mean difference between the two sets of data is relative to the overall standard deviation of the data. The larger the t-value (in absolute terms), the more significant the mean difference between the two sets of data.

Regarding the p-value, this represents the probability of observing the results under the null hypothesis (i.e., no mean difference between the two sets of data). The smaller the p-value, the more likely the mean difference between the two sets of data is significant. Typically, a p-value less than 0.05 is considered significant.

In our results, the p-value is far less than 0.05, indicating that the mean difference between the two sets of data is very significant. This further demonstrates the significance and effectiveness of the scheduling model in reducing power purchase costs compared to the disorder model.

Similarly, the load variance data of the disorder model and the ordered scheduling model have means of 188.27 and 18.46, respectively. Their standard deviations are 12.82 and 1.432, respectively. The independent sample t-test results for the two sets of data are t-value = 113.92 and p-value = −48.17. This proves that the scheduling model has significant effectiveness in reducing load variance compared to the disorder model.

Figure 8 and Figure 9 show the comparisons of the statistical analysis of the power purchase cost and load variance of the models, respectively.

5.3. Weighting Factor Sensitivity Analysis

A sensitivity analysis is a systematic method for studying and evaluating the impact of changes in model parameters on model output results. Its primary purpose is to identify which parameters significantly influence model outcomes, thereby helping us understand model behavior, improve model accuracy, and guide parameter selection in practical applications. In optimization problems, a sensitivity analysis is particularly important as it reveals the interrelationships and trade-offs between parameters, guiding decision-makers to make more rational decisions in uncertain environments. In the context of electric truck charging scheduling optimization, the weighting coefficients α and β, respectively, determine the relative importance of cost minimization and load balancing in the objective function. Different combinations of weighting coefficients can lead to significant variations in optimization results, making a sensitivity analysis essential.

Figure 10 shows the trends of C_g and ϕ_g with varying α and β. The red curve represents the changes in C_g, while the blue curve represents the changes in ϕ_g. Observing the chart, it is evident that the global optimum is achieved around α = 0.6 and β = 0.4.

Figure 10 and Figure 11 show the trends of C_g and ϕ_g with varying α and β. The darker the color, the higher the values of C_g and ϕ_g. It is also evident that the global optimum for C_g and ϕ_g is achieved around α = 0.6 and β = 0.4.

In summary, when α = 0.6 and β = 0.4, C_g = 721.53 and ϕ_g = 18.08, resulting in the minimum total cost C_total = 440.15.

5.4. Model Comparison

To comprehensively evaluate the performance of the proposed scheduling model, this paper selects the deep reinforcement learning (DRL) model as a comparative model. The DRL model optimizes the charging strategy of electric vehicles through deep neural networks and reinforcement learning algorithms. Specifically, the DRL model is suitable for power grid environments that use renewable energy sources (such as photovoltaic systems). By learning the variations in charging demand, the DRL model reduces peak grid loads and improves overall system efficiency.

The DRL model leverages interactions between the agent and the environment through reinforcement learning to continuously optimize its charging strategy. Specifically, the core components of the DRL model include the following:

(1)

State: the current state of the system, such as the remaining battery level of the electric vehicle, photovoltaic power generation, and the grid load.

(2)

Action: the actions taken by the system, such as starting charging, stopping charging, or adjusting the charging power.

(3)

Reward: the feedback received after taking a certain action, such as reducing grid load or decreasing electricity purchase costs.

(4)

Policy: the mapping function from states to actions, implemented through deep neural networks.

During the training process, the DRL model continuously explores and learns, gradually optimizing the charging strategy to maximize cumulative rewards [33,34,35].

To comprehensively evaluate the performance of different models, we selected the following key metrics: photovoltaic power utilization, load variance, and electricity purchase cost. Figure 12 presents a comparison of the three models in terms of electricity purchase cost, load variance, and photovoltaic power utilization using box plots and stacked bar charts.

By analyzing the three charts, it is clear that the scheduling model demonstrates significant superiority in all key performance metrics. As shown in Figure 12a, the photovoltaic power utilization of the scheduling model is significantly higher than that of the other two models, ranging from 70% to 85%. Its median is approximately 75%, compared to 50% for the disordered model and 60% for the comparison model. The scheduling model has fewer outliers, indicating its greater stability and effectiveness in photovoltaic power utilization. In summary, the scheduling model significantly improves photovoltaic power utilization, maximizing the use of renewable energy and enhancing overall system efficiency. As shown in Figure 12b, the electricity purchase cost of the scheduling model is significantly lower than that of the other two models, around CNY 723. The disordered model consistently has the highest electricity purchase cost, approximately CNY 2060, while the comparison model’s cost is around CNY 1020. The scheduling model effectively reduces electricity purchase costs by charging during low electricity price periods, thereby achieving higher economic benefits. As shown in Figure 12c, the load variance of the scheduling model is the lowest, around 20. The disordered model has the highest load variance, around 192, while the comparison model’s is around 51. In conclusion, the scheduling model has a significant advantage in balancing the charging load, effectively reducing load fluctuations during the charging process and ensuring grid stability.

In summary, the scheduling model demonstrates significant superiority in its photovoltaic power utilization, electricity purchase cost, and load variance, proving its effectiveness and feasibility in optimizing electric vehicle charging strategies.

6. Conclusions

This paper investigates a solar energy storage and charging station in an industrial zone in Xinjiang, focusing on the charging behavior of commuting electric heavy trucks. It analyzes the impact of the disorganized charging behavior of electric heavy trucks on the stability of the power grid within this photovoltaic charging station. Building upon this analysis, this paper proposes an intelligent energy scheduling strategy, aiming to minimize the variance of the electricity purchase load curve while ensuring solar energy storage and lowering charging station’s electricity purchase cost. This strategy, combined with typical photovoltaic output power, regulates the charging behavior of electric heavy trucks, thereby achieving the intelligent scheduling of a solar energy storage and charging station for electric heavy trucks. Finally, a comparative analysis with a deep reinforcement learning model was conducted to demonstrate the advantages of the proposed strategy.

• During the implementation of the proposed organized charging strategy, the electricity purchase cost decreased by CNY 162.08, leading to a reduction in the overall cost of the charging station. The proportion of electricity supplied by photovoltaic power generation increased by 38.52%, thereby enhancing the utilization of clean energy. Additionally, the variance of the electricity purchase load decreased by 567.55, demonstrating that the proposed scheduling strategy effectively mitigated the impact of electric heavy truck charging on the power distribution network. In the simulation results comparing the deep reinforcement learning model with the scheduling strategy proposed in this paper, the following points were observed in terms of the performance of the deep reinforcement learning model: electricity purchase cost—approximately CNY 1020, Load variance—51, photovoltaic power utilization rate—approximately 60%. Although the deep reinforcement learning model does not perform as well as the scheduling strategy proposed in this paper in these metrics, it still outperforms the unordered scheduling model. These results highlight the superiority of the scheduling strategy proposed in this paper, which achieves lower electricity purchase costs, a smaller load variance, and higher photovoltaic power utilization rates for electric heavy-duty truck charging. In summary, the simulation comparison clearly demonstrates the significant advantages of the scheduling strategy proposed in this paper for optimizing electric heavy-duty truck charging.
• A statistical analysis demonstrated the effectiveness and significance of the proposed scheduling strategy in reducing power purchase costs and load variance. In terms of power purchase costs, the scheduling model had a mean of 724.46 and a standard deviation of 54.06, while the disorganized model had a mean of 2069.33 and a standard deviation of 155.31. An independent sample t-test yielded a t-value of −81.37 and a p-value of −143, indicating a significant difference. In terms of load variance, the scheduling model had a mean of 18.46 and a standard deviation of 1.432, compared to the disorganized model’s mean of 188.27 and standard deviation of 12.82, with t-test results of t = 113.92 and p = −48.17. These results highlight the scheduling model’s significant effectiveness in reducing both power purchase costs and load variance compared to the disorganized model.

Through the research presented in this paper, not only have a deeper understanding of the charging behavior of electric heavy trucks within photovoltaic storage charging stations and their impact on the power grid been gained, but effective strategies and technical support have also been provided for achieving more efficient, economical, and environmentally friendly energy management at charging stations.