A photovoltaic-energy storage-electric vehicle coordination v2g optimal scheduling method suitable for building scene

Abstract

The application discloses a photovoltaic-energy storage-electric vehicle coordinated V2G optimization scheduling method suitable for building scenes, relates to the technical field of building energy management and vehicle-network interaction, and comprises the following steps: obtaining a data generator through training; outputting sample data of new energy output and building load scenes through the data generator, and generating a plurality of typical scenes by using a clustering algorithm and a Nash equilibrium model; planning vehicle-network interaction facility configuration under each typical scene, and coordinately scheduling sources, loads, storages and vehicles; selecting participating vehicles, and scheduling the participating vehicles in combination with battery loss and compensation mechanisms; and comprehensively evaluating the economic efficiency and operation quality of the coordinately scheduled results of the sources, loads, storages and vehicles, the planning results of the vehicle-network interaction facility configuration and the scheduling results of the participating vehicles. The application takes uncertain scenes as input, realizes cost minimization through a two-stage model, and verifies the economic efficiency and feasibility of the scheme through evaluation.

Description

Technical Field

[0001] This invention relates to the field of building energy management and vehicle-to-grid (V2G) interaction technology, specifically to a photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios. Background Technology

[0002] Buildings, as a major energy-consuming sector, account for over 30% of total social energy consumption. The coordinated management of flexible load resources has become a key issue in the integrated development of smart grids and green buildings. Electric vehicles (EVs), as mobile loads with energy storage capabilities, can utilize V2G (vehicle-to-grid) technology (electric vehicles supplying power to the grid) to discharge during peak building load periods and charge during off-peak periods. This achieves peak shaving and valley filling, improves the absorption rate of renewable energy and system economics, and is an important path for energy transformation in the building sector. However, existing V2G facility planning is mostly based on a single economic objective, failing to optimize charging pile layout and capacity according to the temporal characteristics of building loads, resulting in low facility utilization and difficulty in adapting to the needs of different building scenarios.

[0003] Furthermore, current V2G technology research suffers from significant limitations in scenario adaptability: existing research primarily focuses on industrial parks or industrial settings, where load characteristics are relatively stable and equipment adjustment has ample room for maneuver. However, the loads of office buildings (e.g., concentrated load from 9:00 to 18:00), commercial buildings (e.g., surge in customer traffic during holidays), and residential communities (e.g., peak electricity consumption from 18:00 to 22:00) exhibit strong temporal fluctuations. Simultaneously, the behavior of electric vehicle users varies significantly across building scenarios—office workers typically have a fixed "9-to-5" schedule, commercial visitors stay randomly, and residential users' charging time is concentrated at night. This makes it difficult for existing scheduling strategies to match the peak-valley patterns of building loads with the spatiotemporal distribution characteristics of EVs, while also ignoring the rigid constraints of unadjustable building loads (e.g., air conditioning base load). This makes it difficult to balance the fluctuations in photovoltaic output with the charging and discharging demands of EVs. Moreover, the scenario generation methods do not construct typical scenarios for seasonal differences in building scenarios (e.g., surge in air conditioning load in summer) and weekday / non-weekday characteristics (e.g., sudden drop in commercial building load on weekends), resulting in insufficient robustness of the scheduling solutions. Therefore, constructing a V2G interactive scheduling method that adapts to the characteristics of building scenarios is of great significance for improving building energy efficiency and the practical application of V2G technology.

[0004] Overall, current V2G optimization scheduling still has the following shortcomings:

[0005] 1. Most existing V2G dispatch strategies focus on the "campus level" or "distribution network level," neglecting the diverse electricity consumption characteristics within buildings (such as air conditioning, lighting, and cooling) and failing to systematically model the building's own energy resources (such as photovoltaics and energy storage). Especially in office and commercial buildings, factors such as parking shortages, staggered traffic peaks, and distinct daily cyclical patterns in building loads all necessitate a restructuring of V2G strategies on the building side.

[0006] 2. Current V2G scheduling models neglect the impact of user behavior and battery health. In reality, whether an electric vehicle can participate in V2G is limited by user travel time (e.g., expected departure time), target SOC demand, and user sensitivity to battery cycle life (degradation cost). Simultaneously, building load exhibits peak-valley time-varying characteristics, significantly affected by temperature, lighting, and working hours. Therefore, if the scheduling model cannot simultaneously consider user adjustable travel windows, battery degradation cost functions, and dynamic building load characteristics, it will severely restrict the model's practical usability.

[0007] 3. The V2G dispatching process is affected by multiple uncertainties, including the power generation capacity of rooftop photovoltaic systems, the parking status of electric vehicles, and fluctuations in electricity prices (time-of-use pricing or real-time pricing). If strategies are formulated solely based on static parameters, it will be difficult to cope with sudden changes in actual operation, and may even lead to load mismatch or revenue loss.

[0008] 4. Most current studies assume that electric vehicles can accept "on-demand discharge," neglecting users' perceived economic compensation factors, such as: whether the discharge is profitable enough, whether battery depreciation is considered, and whether there are tiered incentives. If the incentives are insufficient or the calculation model lacks a reward design, users will be unwilling to participate in discharge, making scheduling execution difficult to achieve.

[0009] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0010] To address the problems in related technologies, this invention proposes a photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0011] Therefore, the specific technical solution adopted by the present invention is as follows:

[0012] A photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios includes:

[0013] Based on renewable energy output and building load data, a data generator is trained using a stable training adversarial network. The data generator outputs sample data of renewable energy output and building load scenarios, and uses clustering algorithms and Nash equilibrium models to generate several typical scenarios.

[0014] In each typical scenario, the configuration of vehicle-to-grid (V2G) interactive facilities is planned based on power balance constraints, AC power flow constraints, and budget constraints, with the goal of minimizing the total investment and operating costs. The coordinated scheduling of power sources, loads, storage, and vehicles is carried out in combination with preset time granularity, with the goal of minimizing the total lifecycle operating costs.

[0015] Based on the behavioral data of electric vehicles, participating vehicles are selected, and in conjunction with battery loss and compensation mechanisms, participating vehicles are scheduled.

[0016] The system comprehensively evaluates the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-grid (V2G) interactive facilities configuration, and the scheduling results of participating vehicles, and outputs the evaluation results to provide boundary conditions for scheduling decisions.

[0017] Furthermore, based on renewable energy output and building load data, and trained using a stable, trainable adversarial network, a data generator is obtained, including:

[0018] Initialize the generator and discriminator of the stable training adversarial network; input time-series samples of the building's renewable energy output and building load data into the stable training adversarial network;

[0019] The generator and discriminator are alternately optimized and iterated until convergence is achieved to obtain the data generator.

[0020] Furthermore, the data generator outputs sample data on new energy output and building load scenarios, including:

[0021] Input a new random noise vector that is consistent with the noise distribution during the training phase into the data generator, and output a simulated scene sequence that is consistent with the temporal characteristics of the original samples;

[0022] Repeatedly run the data generator to obtain the scenarios of new energy output and building load, and retain sample data that conforms to physical constraints.

[0023] Furthermore, using clustering algorithms and Nash equilibrium models, several typical scenarios are generated, including:

[0024] Using clustering algorithms and combining sample data of new energy output and building load scenarios, the new energy output and building load scenarios are clustered into several typical scenarios.

[0025] A scenario number structure is constructed, and a Nash equilibrium model is used to assign equilibrium weights to each scenario to reflect the probability distribution after the interests of each subject are balanced.

[0026] Furthermore, the Nash equilibrium model is used to assign equilibrium weights to each scenario to reflect the probability distribution after balancing the interests of the various stakeholders, including:

[0027] Identify the various entities, including power grid companies, operators, and users, and obtain the initial weighting scheme for each entity in the scenario;

[0028] Utility functions related to peak-valley difference, net income, and total compensation are constructed respectively. Then, using the Nash equilibrium model, the equilibrium scheme is obtained when each subject can change its own weight but cannot improve its own utility. The equilibrium weights are then assigned to the scenario.

[0029] Furthermore, with the goal of minimizing the total investment and operating costs, and based on power balance constraints, AC power flow constraints, and budget constraints, the planning of vehicle-to-grid (V2G) infrastructure configuration includes:

[0030] Obtain candidate nodes for building vehicle-to-grid (V2G) charging piles and set the facility capacity of the candidate nodes;

[0031] Establish node power balance constraints between node injected power and load demand, network loss, and tributary transmission power.

[0032] Based on the AC power flow model, voltage and current constraints are established, and non-convex terms are treated by second-order cone relaxation to ensure that the power, voltage, and current of nodes satisfy physical laws; the investment amount of vehicle-to-grid (V2G) interactive facilities is limited by a given cost.

[0033] A first-stage objective function is constructed, which includes facility investment costs, network loss costs, and expected operating costs. The function is then solved to obtain the planning results for the configuration of vehicle-to-grid (V2G) interactive facilities.

[0034] Furthermore, with the goal of minimizing the total lifecycle operating cost, and in conjunction with preset time granularity, the coordinated scheduling of sources, loads, storage, and vehicles includes:

[0035] Construct scheduling time series and model sources, loads, storage, and vehicles; based on historical data, extract the time series patterns of unadjustable loads and generate a fixed profile of unadjustable loads;

[0036] By introducing AC power flow constraints and combining the modeling results of sources, loads, storage, and vehicles with the fixed profile of unadjustable loads, a second-stage objective function is constructed, which includes electricity purchase costs, gas costs, curtailment costs, and battery life loss costs. The vehicle-to-grid economic compensation mechanism is used to subsidize the discharge of electric vehicles at a preset price.

[0037] Solving the objective function for the second stage yields the coordinated scheduling results of the source, load, storage, and vehicle.

[0038] Furthermore, by utilizing the vehicle-to-grid economic compensation mechanism, subsidies will be provided for the discharge volume of electric vehicles at a preset price, including:

[0039] Collect historical data of electric vehicle users, including discharge duration, discharge amount, participation frequency, SOC threshold preference, and sensitivity to subsidies;

[0040] Electric vehicle users are classified using clustering algorithms and based on historical data.

[0041] Based on the classification results of electric vehicle users, and using a deep learning model to train a discharge intention prediction model, the probability of electric vehicle users' discharge intention under the preset subsidy price is output.

[0042] A basic subsidy benchmark is determined, and subsidies are dynamically adjusted for different electric vehicle user groups based on the probability of electric vehicle users' willingness to discharge at a preset subsidy price.

[0043] Furthermore, a comprehensive evaluation of the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-grid (V2G) infrastructure configuration, and the scheduling results of participating vehicles will be conducted, including:

[0044] Following the coordinated scheduling of energy sources, loads, storage, and vehicles, the planning of vehicle-grid interaction facilities, and participatory vehicle scheduling, a comprehensive evaluation objective function is constructed, including annual operating costs, vehicle-to-grid compensation and battery degradation costs, and curtailment and network loss costs, and the output indicators are determined.

[0045] The impact of isolated variable disturbances on economic and technical indicators is quantified, and the sensitivity of parameter fluctuations to the overall evaluation results is obtained.

[0046] Furthermore, before quantifying the impact of isolated variable disturbances on economic and technical indicators and obtaining the sensitivity of parameter fluctuations to the overall evaluation results, the following steps are also included:

[0047] Select key variables and set perturbation gradients; perturb the target variables one by one and solve the evaluation indicators for each scenario; calculate the rate of change of the evaluation indicators with respect to the target variables and identify the critical threshold.

[0048] The beneficial effects of this invention are as follows:

[0049] 1. This invention, by comprehensively modeling the photovoltaic output, load forecast results, and electric vehicle entry and exit times within a building, can predict the daily load curve in advance, actively guide electric vehicle charging during low electricity price periods, and selectively discharge to the building's internal load or sell to the grid during high electricity price periods, significantly reducing the building's electricity costs, effectively improving energy efficiency, and achieving a balance between economic and environmental benefits.

[0050] 2. This invention utilizes the energy storage characteristics of electric vehicle batteries and integrates SOC status monitoring, vehicle driving pattern learning, and user participation intention classification models to dynamically identify the set of vehicles that can participate in V2G scheduling. It also automatically adjusts the charging and discharging strategies in conjunction with electricity price signals and building load changes, thereby improving the flexibility and intelligence of the scheduling strategy and enhancing the system's ability to cope with uncertainties.

[0051] 3. This invention constructs a 15-minute scheduling cycle based on a mixed integer second-order cone programming model, and unifies the modeling of V2G systems with various controllable devices in buildings such as photovoltaic, energy storage systems, and CHP (cogeneration technology). This enables collaborative optimization scheduling of the source-load-storage-vehicle system, effectively mitigating the risk of local transformer overload in buildings and improving system safety and operational stability.

[0052] 4. By setting a minimum SOC threshold for electric vehicles and quantifying and monetizing the depth of discharge and battery life degradation, this invention establishes an incentive and constraint mechanism that can effectively balance user participation willingness and system scheduling needs, enhance the enthusiasm and sustainability of electric vehicles participating in building V2G systems, and improve the scalability and replicability of the system.

[0053] 5. This invention introduces a generative adversarial network based on Wasserstein distance combined with the K-medoids clustering method, which can generate representative typical photovoltaic power output and building load scenarios based on historical data, improve the model's generalization ability under uncertain environments, and realize more robust building V2G facility configuration and scheduling strategy formulation.

[0054] 6. This invention integrates a dynamic compensation mechanism, a user participation probability assessment model, and an operating cost-benefit assessment index system during the scheduling optimization process. It can improve the renewable energy absorption rate and building peak shaving capacity while ensuring the system's economy, providing a reliable intelligent scheduling solution for new building energy management systems. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to an embodiment of the present invention. Detailed Implementation

[0057] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0058] According to embodiments of the present invention, a photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios is provided. From the perspective of building systems, a V2G scheduling modeling method uniformly considers five types of elements: vehicles, photovoltaics, energy storage, building load, and grid interaction.

[0059] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, the photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to an embodiment of the present invention includes:

[0060] S1. Based on the data of new energy power output and building load, a data generator is obtained by training a stable training adversarial network; the data generator outputs sample data of new energy power output and building load scenarios, and generates several typical scenarios by using clustering algorithms and Nash equilibrium models.

[0061] S2. In each typical scenario, with the goal of minimizing the total investment and operating costs, and based on power balance constraints, AC power flow constraints, and budget constraints, the configuration of vehicle-to-grid (V2G) interactive facilities is planned; with the goal of minimizing the full lifecycle operating costs, and combined with preset time granularity, the source, load, storage, and vehicle are coordinated and scheduled.

[0062] S3. Based on the behavior data of electric vehicles, select participating vehicles and schedule them in conjunction with battery loss and compensation mechanisms.

[0063] S4 conducts a comprehensive evaluation of the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-everything (V2X) infrastructure configuration, and the scheduling results of participating vehicles, and outputs the evaluation results to provide boundary conditions for scheduling decisions.

[0064] In one embodiment, a data generator is obtained by training a stable, trainable adversarial network based on new energy output and building load data.

[0065] Initialize the generator and discriminator of the stable training adversarial network; input time-series samples of the building's renewable energy output and building load data into the stable training adversarial network; alternately optimize the generator and discriminator and iterate until convergence to obtain the data generator.

[0066] In one embodiment, the data generator outputs sample data on renewable energy output and building load scenarios, including:

[0067] Input a new random noise vector that is consistent with the noise distribution during the training phase into the data generator, and output a simulated scene sequence that is consistent with the temporal characteristics of the original samples; repeatedly run the data generator to obtain the new energy output and building load scenarios, and retain sample data that conforms to physical constraints.

[0068] In one embodiment, several typical scenarios are generated using clustering algorithms and Nash equilibrium models, including:

[0069] Using clustering algorithms and combining sample data of new energy output and building load scenarios, the new energy output and building load scenarios are clustered into several typical scenarios; a scenario number structure is constructed, and a Nash equilibrium model is used to assign equilibrium weights to each scenario to reflect the probability distribution after the balance of interests of each subject.

[0070] In one embodiment, the Nash equilibrium model is used to assign equilibrium weights to each scenario to reflect the probability distribution after balancing the interests of the various stakeholders, including:

[0071] The process involves identifying various stakeholders, including the power grid company, operators, and users, and obtaining initial weighting schemes for each staker within a given scenario. Utility functions related to peak-to-valley difference, net revenue, and total compensation are constructed. Using a Nash equilibrium model, these functions are iteratively adjusted until individual stakers can no longer improve their utility by changing their weights, resulting in an equilibrium scheme. The equilibrium weights are then assigned to the scenario. Utility functions:

[0072] ;

[0073] ;

[0074] ;

[0075] in, For power grid companies in scenario weighting The effect of the following This represents the power grid company's focus on peak-valley differences. For the scene The weight, / Scene Peak / valley load (kW) This represents the weighting coefficient of the power grid company's electricity sales revenue. For the scene China Power Grid Corporation's electricity sales revenue; For operators in scene weight The effect of the following For V2G peak-valley arbitrage profits in scenario s, For the scene Revenue from ancillary services For the scene Operating costs For the scene Depreciation costs of equipment; For users in scene weight The effect of the following For the scene Total compensation received by users This is a risk aversion factor for users regarding battery degradation. For the scene Battery wear and tear costs for mid-range users.

[0076] In one embodiment, the planning of vehicle-to-grid (V2G) infrastructure configuration, with the objective of minimizing the total investment and operating costs and based on power balance constraints, AC power flow constraints, and budget constraints, includes:

[0077] Candidate nodes for constructing vehicle-to-grid (V2G) charging piles are identified, and their facility capacities are set. Node power balance constraints are established between node injected power and load demand, network loss, and branch transmission power. Voltage and current constraints are established based on an AC power flow model, and non-convex terms are handled using second-order cone relaxation to ensure that node power, voltage, and current satisfy physical laws. The investment amount for V2G facilities is limited by a given cost. A first-stage objective function is constructed, including facility investment cost, network loss cost, and expected operating cost, and solved to obtain the planning results for V2G facility configuration.

[0078] In one embodiment, the coordinated scheduling of sources, loads, storage, and vehicles, with the goal of minimizing total lifecycle operating costs and incorporating a preset time granularity, includes:

[0079] A scheduling time series is constructed, and sources, loads, storage, and vehicles are modeled. Based on historical data, the temporal patterns of unadjustable loads are extracted, and a fixed profile of unadjustable loads is generated. AC power flow constraints are introduced, and combined with the modeling results of sources, loads, storage, and vehicles and the fixed profile of unadjustable loads, a second-stage objective function is constructed, which includes electricity purchase costs, gas costs, curtailment costs, and battery life loss costs. The vehicle-to-grid economic compensation mechanism is used to subsidize the discharge of electric vehicles at a preset price. The second-stage objective function is solved to obtain the coordinated scheduling results of sources, loads, storage, and vehicles.

[0080] In one embodiment, subsidizing the discharge volume of electric vehicles at a preset price using a vehicle-to-grid economic compensation mechanism includes:

[0081] Historical data of electric vehicle users is collected, including discharge duration, discharge volume, participation frequency, SOC threshold preference, and sensitivity to subsidies. A clustering algorithm is used to classify electric vehicle users based on this historical data. Based on the classification results, a deep learning model is used to train a discharge intention prediction model, outputting the probability of a user's discharge intention at a preset subsidy price. A basic subsidy benchmark is determined, and subsidies are dynamically adjusted for different electric vehicle user groups based on the probability of their discharge intention at the preset subsidy price.

[0082] In one embodiment, a comprehensive evaluation of the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-everything (V2X) infrastructure configuration, and the scheduling results of participating vehicles includes:

[0083] Following the coordinated scheduling of energy sources, loads, storage, and vehicles, the planning of vehicle-to-grid interaction facilities, and participatory vehicle scheduling, a comprehensive evaluation objective function is constructed, including annual operating costs, vehicle-to-grid compensation and battery degradation costs, and curtailment and network loss costs, and output indicators are determined; the impact of isolated variable disturbances on economic and technical indicators is quantified, and the sensitivity of parameter fluctuations to the comprehensive evaluation results is obtained.

[0084] Select a single isolated variable (such as V2G unit construction cost, peak-valley electricity price difference, user participation rate, etc.) and apply gradient perturbations of ±10%, ±20%, ±30% to it. With other parameters fixed, substitute the perturbed values ​​into the collaborative scheduling and planning model to solve it again, and obtain the corresponding economic indicators (such as annual total operating cost, internal rate of return, etc.) and technical indicators (such as curtailed solar power, renewable energy absorption rate, etc.). Quantify the impact by calculating the sensitivity coefficient (the ratio of the rate of change of the indicator to the rate of perturbation of the variable). The larger the absolute value of the sensitivity coefficient, the more significant the impact of the variable. Finally, identify the key variables by sorting them by coefficient.

[0085] In one embodiment, before quantifying the impact of isolated variable disturbances on economic and technical indicators and obtaining the sensitivity of parameter fluctuations to the overall evaluation results, the following steps are included:

[0086] Select key variables and set perturbation gradients; perturb the target variables one by one and solve the evaluation indicators for each scenario; calculate the rate of change of the evaluation indicators with respect to the target variables and identify the critical threshold.

[0087] To facilitate understanding of the above technical solutions of the present invention, the working principle of the present invention in actual process will be described in detail below.

[0088] Step 1: V2G Facility Planning (Phase 1 Decision). Objective: To determine the node locations and capacity for V2G facility construction, minimizing the total investment and operating costs. Specifically, this includes:

[0089] Step 1.1: Construct the distribution network topology: Identify candidate nodes within the park where V2G charging piles can be built; Step 1.2: Define decision variables: Define the V2G facility capacity of each candidate node as a continuous variable or a 0 / 1 construction state; Step 1.3: Establish power balance constraints: Ensure that the injected power of each node = demand + loss; Step 1.4: Establish the following constraints based on the AC power flow model: Node power balance: The active / reactive power injected by each node is equal to the sum of load demand, network loss, and branch transmission power. Voltage constraint: The square of the node voltage amplitude must be within the safe range of 0.95-1.05 pu. Current constraint: The square of the branch current amplitude does not exceed the rated value. Power-voltage-current relationship: Non-convex terms are handled through second-order cone relaxation to ensure that the three satisfy physical laws. Step 1.5: Set budget constraints: Facility investment does not exceed the given cost limit; Step 1.6: Construct the objective function:

[0090] ;

[0091] node The scale of V2G charging pile construction, The initial investment costs associated with the scale of V2G infrastructure construction, The scenarios refer to uncertainties such as different new energy outputs, load demands, and electric vehicle behavior. For the scene The probability of occurrence The operating costs for the second-stage coordinated scheduling include facility investment costs and network loss costs; the expected operating costs from the second stage are introduced (the expected value of the operating costs of the second-stage coordinated scheduling over the entire cycle). Step 1.7: Solve the first-stage problem: Use Gurobi to optimize the mixed integer problem and output the optimal facility configuration scheme.

[0092] Step Two: Source-Load-Storage-Vehicle Coordinated Scheduling (Second Phase Scheduling); Objective: To coordinate the operation of adjustable equipment based on a 15-minute time granularity, minimizing the overall lifecycle operating cost. Specifically, this includes:

[0093] Step 2.1: Construct the scheduling time series: Set 96 time periods / day, each time period is 15 minutes; Step 2.2: Introduce the equipment operation model, photovoltaic output is an uncertain input; CHP considers thermoelectric coupling and minimum operating power; energy storage modeling includes charge and discharge status, power balance, and efficiency; electric vehicle modeling includes SOC changes and available time windows; Step 2.3: Set the non-adjustable load profile: such as lighting and air conditioning base load. Classify non-adjustable loads: clearly defined as emergency lighting, office lighting, air conditioning base load, elevators, etc.; Principle: Non-adjustable loads cannot be scheduled due to functional or equipment limitations. It is necessary to extract time series patterns based on historical data (such as seasonal and weekday differences) to generate a fixed profile that fits reality, providing accurate rigid constraints for the second-stage scheduling model and ensuring accurate power balance calculations; Step 2.4: Introduce AC power flow constraints: ensure voltage, current, and safety boundaries between nodes; Step 2.5: Define the objective function:

[0094] ;

[0095] in, The electricity price is the cost of purchasing electricity (including the cost of purchasing electricity and the revenue from selling electricity). For gas costs, To cover the cost of light curtailment, Cost of battery life loss (estimated using DoD / SoH).

[0096] Step 2.6: Introduce a V2G economic compensation mechanism: subsidize EV discharge at a specific price. Specifically, this includes:

[0097] Step 2.6.1: User Behavior Data Collection and Profile Construction: Collect historical data of electric vehicle users, including discharge duration, discharge volume, participation frequency, SOC threshold preference, and sensitivity to subsidies (such as changes in participation rate under different subsidy prices); classify users based on the K-means clustering algorithm. Step 2.6.2: Discharge Intention Prediction Model Training: Output user profile labels, real-time electricity price, weather (affecting travel demand), historical subsidy response rate, etc.; train the prediction model using a deep learning model (such as an LSTM neural network): output the probability (0-100%) of a user's discharge intention under a specific subsidy price. Step 2.6.3: Set Basic Subsidy Benchmark: Dynamically adjust subsidies for different user groups based on the discharge intention output by the prediction model. Step 2.6.4: Real-time Collection of User Response Data to Dynamic Subsidies (Deviation between Actual Discharge Volume and Predicted Value).

[0098] Principle: Different users have significantly different motivations for participating in V2G (e.g., some users are more concerned about battery life, while others are price-sensitive). Uniform subsidies are difficult to maximize the overall participation rate, while personalized strategies can accurately match user needs. By capturing the temporal characteristics of user behavior (e.g., weekday / weekend discharge patterns) through LSTM, the accuracy of discharge intention prediction is improved, avoiding cost waste caused by blindly increasing subsidies. Based on the principle of "maximizing marginal benefits", limited subsidy resources are allocated to price-sensitive users, while the strategy is continuously optimized through feedback iteration, balancing economic efficiency and user participation.

[0099] Step 2.7 Solve the scheduling optimization model: Use Gurobi to solve the scheduling problem based on the MISOCP format.

[0100] Step 3: Uncertainty Scenario Generation; Objective: To construct typical renewable energy output and park load scenarios to provide input data for decision optimization of V2G facility planning and coordinated scheduling in a two-stage stochastic programming model. Specifically, this includes:

[0101] Step 3.1: Collect historical data: Obtain actual photovoltaic output and load data of the building in recent years; Step 3.2: Train the data generator using WGAN (i.e., Stable-Trained Adversarial Network): Input raw time-series samples such as photovoltaic output and building load of the park, initialize the generator (to generate simulated data from random noise) and the discriminator (to judge the authenticity of the data); alternately optimize both—the generator attempts to generate more realistic data to deceive the discriminator, and the discriminator optimizes its ability to distinguish between real and generated data through Wasserstein distance; iterate until convergence to obtain a generator that can generate high-fidelity scenes (the principle of this step is to use the game mechanism of generative adversarial networks, and replace the traditional data with Wasserstein distance). The method employs cross-entropy loss to address training instability and ensures that the generated data distribution closely matches the probability distribution of the original samples, thereby generating typical scenarios that reflect the uncertainty of the source load. Step 3.3: Generate a large amount of sample data: Input a new random noise vector into the trained WGAN generator, consistent with the noise distribution during the training phase; The generator simulates a sequence of scenarios (15-minute time resolution) based on the noise output and the temporal characteristics of the original samples (photovoltaic output, building load); Repeat the above process to generate a large number of simulated scenarios, retaining samples that meet physical constraints (such as non-negative photovoltaic output and matching load time period characteristics); WGAN is a Wasserstein generative adversarial network, an improved generative adversarial network model.

[0102] Principle: After being trained by WGAN, the generator has learned the probability distribution of the original samples. Inputting random noise can map out new samples that follow this distribution. The batch-generated samples not only retain the uncertainty characteristics of the source load, but also expand the sample coverage through the generalization ability of the generator, providing a sufficient data foundation for subsequent scene clustering.

[0103] Step 3.4: Use K-medoids clustering: The daily data of photovoltaic output and building electricity load at 15-minute resolution throughout the year for a microgrid in a certain park are clustered using K-medoids into 8 typical daily samples for each of the four seasons (spring, summer, autumn, and winter), including both working and non-working days. Step 3.5: Output a typical scenario set: Each category contains 96 15-minute data points for model input. Step 3.6: Define a scenario tree structure in the model, assigning weights to each scenario to represent probability: Identify the power grid company (focusing on load smoothing), the operator (focusing on revenue), and the user (focusing on compensation), defining utility functions related to peak-valley difference, net revenue, and total compensation respectively; using a Nash equilibrium model, each entity proposes an initial weight scheme, iteratively adjusting until any entity can no longer improve its own utility by changing its weight alone, obtaining an equilibrium scheme; assign the equilibrium weights to the scenarios, ensuring the sum is 1, reflecting the probability distribution after balancing the interests of all parties.

[0104] Step 4: Electric Vehicle Behavior Modeling and Scheduling Mechanism Design; Objective: To accurately describe EV entry and exit times and SOC evolution, and to formulate V2G scheduling strategies. Specifically, this includes:

[0105] Step 4.1: Establish an EV travel mode library: Set commuting time periods and parking duration distribution based on building features; Step 4.2: Perform Monte Carlo simulation: Simulate daily EV entry and exit times and dwell time; allocate initial SOC to different vehicles; Step 4.3: Classify vehicle users: Set a portion as V2G participating vehicles, and the rest as charging vehicles; participating vehicles are allowed to schedule SOC and adjust time. Step 4.4: Define SOC constraints, including a lower limit: ensure minimum travel demand is met when leaving; upper limit: do not exceed the maximum SOC. Step 4.5: Introduce battery life loss modeling: use depth of discharge (DoD) as input; convert the corresponding life loss into monetary cost; Step 4.6: Set a V2G compensation function: set a compensation price based on kWh discharge amount; if life loss > revenue, the vehicle automatically exits the scheduling; Step 4.7: Centralized scheduling control mechanism: operators uniformly optimize the discharge time and power of participating vehicles.

[0106] Step 5: Optimize the solution and construct the evaluation index system; Objective: To comprehensively evaluate the model's effectiveness, including its economic efficiency and system operation quality. Specifically, this includes:

[0107] Step 5.1: Combine objective function terms: annual operating cost (electricity price, gas, operation and maintenance); V2G compensation and battery degradation cost; curtailment and network loss cost; Step 5.2: Define output indicators: annual operating cost (ten thousand yuan); renewable energy curtailment (kWh) and utilization rate (%); investment payback period (years); internal rate of return (IRR); Step 5.3: Operational sensitivity analysis: Select participation rate (30%-90%, step size 10%), facility cost (50,000-150,000 yuan / unit), and electricity price structure (peak-valley difference and peak period adjustment) as key variables, and set the perturbation gradient; keep other parameters constant, perturb the objective variables one by one, and substitute them into the model to solve for the IRR under each scenario. Step 5.4: Output scheduling result charts: 15-minute granular EV, energy storage, and CHP charge / discharge curves; grid power purchase curve and load response results; Step 5.5: Simulation verification and case summary: using real parameters from a certain industrial park; output comparison results of multiple typical daily operations.

[0108] In summary, this approach uses an uncertain scenario as input and a two-stage model (facility planning → coordinated scheduling) to minimize costs. Finally, the economic viability and feasibility of the solution are verified through evaluation. Each step supports the others, forming a process of "data input → decision optimization → result evaluation".

[0109] When this invention is applied:

[0110] 1. V2G Facility Planning: Evaluate parking lots in commercial complexes and select areas with high pedestrian traffic and convenient power access for V2G charging station construction. Calculate and determine the scale of charging station construction to meet user needs while controlling investment costs and improving facility utilization.

[0111] 2. Establishment of a Source-Load-Storage-Vehicle Coordinated Dispatch System: Construct an intelligent energy management system that integrates photovoltaic power generation, energy storage device status, real-time building load, and electric vehicle charging and discharging information. During peak daytime electricity consumption, photovoltaic power is prioritized, with energy storage supplementing any shortfall. When appropriate, electric vehicles can be used for V2G discharge to reduce grid power purchase costs.

[0112] 3. Generation and Application of Typical New Energy Output and Park Load Scenarios: Based on the operational characteristics of commercial complexes, typical energy scenarios for different seasons, weekdays, and non-working days are generated. For example, during the high temperatures of summer, data is used to predict changes in photovoltaic output and building load, allowing for proactive optimization of scheduling strategies and improved accuracy in energy allocation.

[0113] 4. Electric Vehicle Behavior Modeling and Scheduling: Data such as user charging habits and dwell time are collected to create models. For users with long dwell times, charging is utilized during off-peak electricity pricing periods; for users with short dwell times, basic electricity needs are prioritized, achieving efficient coordination with building energy dispatching.

[0114] Comprehensive assessment and optimization: Regularly evaluate the implementation effect, calculate annual operating costs, pay attention to indicators such as renewable energy utilization rate and investment payback period, and monitor the power grid purchase curve. Optimize dispatch parameters based on the results to improve effectiveness.

[0115] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios, characterized in that, include: A data generator is obtained by training a stable training adversarial network based on new energy output and building load data. The data generator outputs sample data of new energy output and building load scenarios, and uses clustering algorithms and Nash equilibrium models to generate several typical scenarios. In each typical scenario, the configuration of vehicle-to-grid (V2G) interactive facilities is planned based on power balance constraints, AC power flow constraints, and budget constraints, with the goal of minimizing the total investment and operating costs. The coordinated scheduling of power sources, loads, storage, and vehicles is carried out in combination with preset time granularity, with the goal of minimizing the total lifecycle operating costs. Based on the behavioral data of electric vehicles, participating vehicles are selected, and in conjunction with battery loss and compensation mechanisms, participating vehicles are scheduled. The system comprehensively evaluates the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-grid (V2G) interactive facilities configuration, and the scheduling results of participating vehicles, and outputs the evaluation results to provide boundary conditions for scheduling decisions. The comprehensive evaluation of the economic efficiency and operational quality of the coordinated scheduling results of sources, loads, storage, and vehicles, the planning results of vehicle-to-grid (V2G) infrastructure configuration, and the scheduling results of participating vehicles includes: Following the coordinated scheduling of energy sources, loads, storage, and vehicles, the planning of vehicle-grid interaction facilities, and participatory vehicle scheduling, a comprehensive evaluation objective function is constructed, including annual operating costs, vehicle-to-grid compensation and battery degradation costs, and curtailment and network loss costs, and the output indicators are determined. The impact of isolated variable disturbances on economic and technical indicators is quantified, and the sensitivity of parameter fluctuations to the overall evaluation results is obtained.

2. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, The data generator, which is based on new energy output and building load data and trained using a stable, trainable adversarial network, includes: Initialize the generator and discriminator of the stable training adversarial network; input time-series samples of the building's renewable energy output and building load data into the stable training adversarial network; The generator and discriminator are alternately optimized and iterated until convergence is achieved to obtain the data generator.

3. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, The sample data on new energy output and building load scenarios output by the data generator includes: Input a new random noise vector that is consistent with the noise distribution during the training phase into the data generator, and output a simulated scene sequence that is consistent with the temporal characteristics of the original samples; Repeatedly run the data generator to obtain the scenarios of new energy output and building load, and retain sample data that conforms to physical constraints.

4. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, Using clustering algorithms and Nash equilibrium models, several typical scenarios are generated, including: Using clustering algorithms and combining sample data of new energy output and building load scenarios, the new energy output and building load scenarios are clustered into several typical scenarios. A scenario number structure is constructed, and a Nash equilibrium model is used to assign equilibrium weights to each scenario to reflect the probability distribution after the interests of each subject are balanced.

5. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 4, characterized in that, The method of assigning equilibrium weights to each scenario using the Nash equilibrium model to reflect the probability distribution after balancing the interests of each party includes: Identify the various entities, including power grid companies, operators, and users, and obtain the initial weighting scheme for each entity in the scenario; Utility functions related to peak-valley difference, net income, and total compensation are constructed respectively. Then, using the Nash equilibrium model, the equilibrium scheme is obtained when each subject can change its own weight but cannot improve its own utility. The equilibrium weights are then assigned to the scenario.

6. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, The planning of vehicle-to-grid (V2G) infrastructure configuration, with the goal of minimizing the total investment and operating costs and based on power balance constraints, AC power flow constraints, and budget constraints, includes: Obtain candidate nodes for building vehicle-to-grid (V2G) charging piles and set the facility capacity of the candidate nodes; Establish node power balance constraints between node injected power and load demand, network loss, and tributary transmission power. Based on the AC power flow model, voltage and current constraints are established, and non-convex terms are treated by second-order cone relaxation to ensure that the power, voltage, and current of nodes satisfy physical laws; the investment amount of vehicle-to-grid (V2G) interactive facilities is limited by a given cost. A first-stage objective function is constructed, which includes facility investment costs, network loss costs, and expected operating costs. The function is then solved to obtain the planning results for the configuration of vehicle-to-grid (V2G) interactive facilities.

7. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, The goal of minimizing the overall lifecycle operating cost, combined with a preset time granularity, includes the coordinated scheduling of sources, loads, storage, and vehicles, which includes: Construct scheduling time series and model sources, loads, storage, and vehicles; based on historical data, extract the time series patterns of unadjustable loads and generate a fixed profile of unadjustable loads; By introducing AC power flow constraints and combining the modeling results of sources, loads, storage, and vehicles with the fixed profile of unadjustable loads, a second-stage objective function is constructed, which includes electricity purchase costs, gas costs, curtailment costs, and battery life loss costs. The vehicle-to-grid economic compensation mechanism is used to subsidize the discharge of electric vehicles at a preset price. Solving the objective function for the second stage yields the coordinated scheduling results of the source, load, storage, and vehicle.

8. The photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 7, characterized in that, The aforementioned mechanism for subsidizing the discharge of electric vehicles at a preset price using a vehicle-to-grid economic compensation method includes: Collect historical data of electric vehicle users, including discharge duration, discharge amount, participation frequency, SOC threshold preference, and sensitivity to subsidies; Electric vehicle users are classified using clustering algorithms and based on historical data. Based on the classification results of electric vehicle users, and using a deep learning model to train a discharge intention prediction model, the probability of electric vehicle users' discharge intention under the preset subsidy price is output. A basic subsidy benchmark is determined, and subsidies are dynamically adjusted for different electric vehicle user groups based on the probability of electric vehicle users' willingness to discharge at a preset subsidy price.

9. A photovoltaic-energy storage-electric vehicle collaborative V2G optimization scheduling method adapted to building scenarios according to claim 1, characterized in that, Before quantifying the impact of isolated variable disturbances on economic and technical indicators and obtaining the sensitivity of parameter fluctuations to the comprehensive evaluation results, the following steps are also included: Select key variables and set perturbation gradients; perturb the target variables one by one and solve the evaluation indicators for each scenario; calculate the rate of change of the evaluation indicators with respect to the target variables and identify the critical threshold.

Images