Feasible set deduction based method and system for simulating full-time state decision of vehicle and pile road network

Abstract

This invention provides a method and system for real-time decision simulation of vehicle-pile-road network based on feasible set derivation. The method uses a pre-constructed multi-model fusion architecture to perform parallel load data prediction based on the spatiotemporal feature set of the acquired vehicle-pile-road network system. Based on the load data prediction results, a fused load prediction result containing confidence intervals is output through a dynamic weighted fusion strategy. Then, based on the fused load prediction results, a feasible decision set for each electric vehicle is constructed by comprehensively considering multi-dimensional factors. Based on this feasible decision set, under the premise of satisfying the distribution network safety constraints, a production simulation strategy containing the openable capacity of transformer substations and vehicle access instructions is obtained using a pre-constructed openable capacity assessment model. This solves the problems of large load fluctuations, complex time-series characteristics, inaccurate transformer substation capacity assessment, and difficulty in effectively fusing multi-model prediction results caused by large-scale electric vehicle access to the distribution network.

Description

Technical Field

[0001] This invention relates to the fields of power systems and intelligent transportation, specifically to a method and system for real-time decision simulation of vehicle-pile-road networks based on feasible set derivation. Background Technology

[0002] With rapid global economic development and continuously rising energy consumption, the over-reliance on fossil fuels such as coal and oil has exacerbated resource scarcity and environmental pollution. It is projected that by 2050, global energy consumption will increase by nearly 50% compared to current levels, further highlighting the imbalance between energy supply and demand. Transportation is a major source of energy consumption and carbon emissions, accounting for over 60% of oil consumption and nearly 20% of CO2 emissions. As a major global automobile producer, China has over 400 million vehicles, and its oil consumption and vehicle emissions continue to climb, making the transportation sector a key challenge for my country's energy security and environmental governance.

[0003] Electric vehicles (EVs), as a crucial pathway for green transportation, offer significant advantages in energy conservation and emission reduction, leading to their rapid global development in recent years. Policy support, technological advancements, and market forces have collectively fueled the rapid growth of China's EV industry. Large-scale EV integration will have a profound impact on power distribution network operation. Their concentrated charging behavior will widen the peak-to-valley difference in the power grid, causing voltage fluctuations and localized line overloads, thus posing pressure on the safe operation of the distribution network. Simultaneously, as mobile energy storage units, EVs possess considerable regulation potential. Through intelligent charging control technology, they can not only effectively smooth load curves but also participate in ancillary services such as grid frequency regulation, enhancing the operational flexibility of the power system. Currently, there is a need to establish accurate load forecasting models and optimized scheduling methods to fully leverage the energy storage characteristics of EVs while ensuring the safe operation of the power grid, achieving coordinated optimization between charging load and grid operation.

[0004] Therefore, it is of great significance to solve the problems of large load fluctuations, complex time-series characteristics, inaccurate assessment of transformer capacity, and difficulty in effectively integrating the prediction results of multiple models caused by the large-scale access of electric vehicles to the power distribution network. Summary of the Invention

[0005] To address the problems of large load fluctuations, complex time-series characteristics, inaccurate transformer capacity assessment, and difficulty in effectively integrating prediction results from multiple models caused by the large-scale integration of electric vehicles into the power distribution network in existing technologies, this invention proposes a real-time decision simulation method and system for vehicle-pile-road networks based on feasible set derivation.

[0006] Firstly, a method for real-time decision simulation of vehicle-pile-road networks based on feasible set derivation is provided, including: Based on the spatiotemporal feature set of the acquired vehicle-pile-road network system, a pre-constructed multi-model fusion architecture is used to perform parallel load data prediction. Based on the load data prediction results, a fused load prediction result containing confidence intervals is output through a dynamic weighted fusion strategy. The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model. Based on the fusion load prediction results, a feasible decision set for each electric vehicle is constructed by comprehensively considering multi-dimensional factors. Based on the feasible decision set, and under the premise of satisfying the distribution network security constraints, a production simulation strategy including the openable capacity of transformer substations and vehicle access instructions is obtained by using a pre-constructed openable capacity assessment model.

[0007] Secondly, a real-time decision-making simulation system for vehicle-pile-road networks based on feasible set derivation is provided, including: The prediction module is used to perform parallel load data prediction based on the spatiotemporal feature set of the acquired vehicle-pile-road network system using a pre-built multi-model fusion architecture, and output a fused load prediction result containing confidence intervals based on the load data prediction result through a dynamic weighted fusion strategy. The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model. The module is used to construct a feasible decision set for each electric vehicle based on the fused load prediction results and by comprehensively considering multi-dimensional factors. The generation module is used to obtain a production simulation strategy that includes the openable capacity of the transformer substation and vehicle access instructions based on the feasible decision set and under the premise of satisfying the distribution network security constraints, using a pre-built openable capacity assessment model.

[0008] In another aspect, this application also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a real-time decision simulation method for vehicle-pile-road network based on feasible set derivation, as described above, is implemented.

[0009] Furthermore, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the above-described method for full-time decision simulation of vehicle-pile-road network based on feasible set derivation.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method and system for real-time decision simulation of vehicle-pile-road network based on feasible set derivation. The method uses a pre-constructed multi-model fusion architecture to perform parallel load data prediction based on the spatiotemporal feature set of the acquired vehicle-pile-road network system. Based on the load data prediction results, a fused load prediction result containing confidence intervals is output through a dynamic weighted fusion strategy. Then, based on the fused load prediction results, a feasible decision set for each electric vehicle is constructed by comprehensively considering multi-dimensional factors. Based on this feasible decision set, under the premise of satisfying the distribution network safety constraints, a production simulation strategy containing the openable capacity of transformer substations and vehicle access instructions is obtained using a pre-constructed openable capacity assessment model. This solves the problems of large load fluctuations, complex time-series characteristics, inaccurate transformer substation capacity assessment, and difficulty in effectively fusing multi-model prediction results caused by large-scale electric vehicle access to the distribution network.

[0011] Specifically, this invention addresses the challenges of distribution network operation and management in the context of large-scale electric vehicle integration. It proposes an intelligent analysis method that integrates multi-model prediction and dynamic capacity assessment. By fully combining historical load data, environmental characteristics, and vehicle behavior information, and employing multi-model collaboration, probabilistic analysis, and dynamic optimization, it improves the accuracy of load forecasting, the rationality of capacity estimation, and the overall system adaptability. Compared to traditional static assessment methods or single-algorithm models, this method exhibits higher predictive stability, adaptability, and deployability, effectively supporting the safe operation and scheduling decisions of distribution systems in complex scenarios. It possesses significant engineering value and promising prospects for widespread application. Attached Figure Description

[0012] Figure 1 The flowchart is a process for the real-time decision simulation method of vehicle-pile road network based on feasible set derivation of the present invention. Figure 2 This is a schematic diagram illustrating the dispatchability of electric vehicles based on the real-time decision-making simulation method for vehicle-pile-road network based on feasible set derivation of the present invention. Figure 3 This is a schematic diagram illustrating the information flow and energy flow interaction between electric vehicles and distribution stations in the real-time decision-making simulation method for vehicle-pile-road networks based on feasible set derivation of the present invention. Figure 4 This is a schematic diagram of the large-scale electric vehicle scheduling framework of the vehicle-pile-road network all-time decision simulation method based on feasible set derivation of the present invention; Figure 5 This is a schematic diagram of the electric vehicle load prediction process of a single model of the vehicle-pile-road network all-time decision simulation method based on feasible set derivation of the present invention. Figure 6 This is a schematic diagram of the electric vehicle load prediction framework based on the LSTM-ARIMA-RF combined model of the vehicle-pile-road network all-time decision simulation method based on feasible set derivation of the present invention. Figure 7This is a schematic diagram of the improved IEEE-33 node system structure of the vehicle-pile-road network all-time decision simulation method based on feasible set derivation of the present invention. Figure 8 This is a schematic diagram of the structure of the vehicle-pile-road network all-time decision simulation system based on feasible set derivation of the present invention; Figure 9 This is a schematic diagram of an electronic device structure according to the present invention. Detailed Implementation

[0013] This invention aims to address the problems of inaccurate transformer capacity assessment and insufficient load forecasting accuracy during the large-scale integration of electric vehicles (EVs) into power distribution networks. Addressing the technical challenges of large load fluctuations, complex time-series characteristics, and difficulties in effectively fusing prediction results from multiple models, this invention proposes a method for assessing the dispatchability of EVs, optimizing transformer capacity, and accurately forecasting loads based on multi-feature time-series modeling and robust dynamic weighted fusion. This method deeply mines the energy storage characteristics and user behavior patterns of EVs, achieving high-precision load forecasting while quantitatively assessing the dispatchability potential of EV clusters. Figure 2 and Figure 3 As shown, an integrated technical system of "prediction-assessment-scheduling" is established. Specifically, this invention innovatively integrates the assessment of the dispatchability of electric vehicles into the traditional capacity assessment framework. By constructing a multi-objective optimization model that considers battery characteristics, user acceptance, and grid constraints, it achieves a technological leap from passive prediction to active scheduling, providing quantitative basis and decision support for electric vehicles to participate in grid peak shaving and frequency regulation and promote the inclusion of new energy sources.

[0014] By applying the method of this invention, the adaptability and management level of the power distribution network to electric vehicle access can be significantly enhanced, ensuring the safe and stable operation of the power grid, optimizing the energy structure, promoting the large-scale promotion of new energy vehicles and the development of green transportation, thereby achieving the goals of energy conservation, emission reduction and environmental protection, and promoting the coordinated development of smart grids and the energy internet.

[0015] To better understand the present invention, the following description, in conjunction with the accompanying drawings and embodiments, will further illustrate the content of the present invention.

[0016] Example 1: A real-time decision simulation method for vehicle-pile-road network based on feasible set derivation, such as Figure 1 As shown, it includes: Step 1: Based on the spatiotemporal feature set of the acquired vehicle-pile-road network system, load data is predicted in parallel using a pre-built multi-model fusion architecture. Based on the load data prediction results, a fused load prediction result containing confidence intervals is output through a dynamic weighted fusion strategy. Step 2: Based on the fused load prediction results, construct a feasible decision set for each electric vehicle by comprehensively considering multi-dimensional constraints; Step 3: Based on the feasible decision set, under the premise of satisfying the distribution network security constraints, use the pre-built openable capacity assessment model to obtain a production simulation strategy that includes the openable capacity of the transformer area and vehicle access instructions.

[0017] The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model.

[0018] Specifically, this invention addresses the problems of increased load uncertainty, difficulty in capacity assessment, and low utilization of scheduling resources caused by the large-scale integration of electric vehicles into distribution transformer areas. It proposes a method for transformer area capacity assessment and load forecasting that integrates multi-model collaborative prediction and a robust weighting mechanism. This method establishes a multi-source feature-driven time series prediction model, integrating algorithms such as LSTM, ARIMA, and random forest to improve adaptability to fluctuating scenarios while maintaining prediction accuracy. Furthermore, it constructs a capacity assessment mechanism based on the fused prediction results, considering network constraints and scheduling feasibility. Figure 2 As shown, this invention ultimately achieves scientific access and efficient management of electric vehicles in urban power distribution networks. The method of this invention has advantages such as clear structure, accurate calculation, and strong robustness, and is applicable to scenarios such as intelligent power distribution systems, areas with concentrated new energy vehicles, and collaborative scheduling of multiple charging stations.

[0019] In this embodiment, before performing parallel prediction of load data using a pre-built multi-model fusion architecture in step 1, the data of the vehicle-pile-road network system needs to be pre-processed to adapt to the requirements of different model structures. Specifically, this includes: Data cleaning was performed on the multi-source heterogeneous data of the vehicle-pile-road network system, and a probabilistic statistical analysis mechanism was introduced to summarize and analyze the marginal distribution, variation characteristics and joint relationships of the cleaned multi-source heterogeneous data. Probabilistic statistical analysis was performed on the cleaned multi-source heterogeneous data to extract statistics including mean, standard deviation, and skewness. Combined with time period characteristics, the load probability distribution changes in the temporal and spatial dimensions of morning and evening peak hours, weekdays and holidays were identified to obtain a spatiotemporal feature set.

[0020] Among them, multi-source heterogeneous data includes one or more of the following: electric vehicle status data, charging facility resource data, road traffic network data, or power distribution network operation parameters.

[0021] In one specific embodiment, such as Figure 4As shown, this invention constructs an intelligent assessment system for distribution substations. The overall structure includes a data preprocessing module, a load forecasting module, a model fusion module, and a capacity assessment module. The system operates at a 15-minute time granularity, supports multi-site and multi-type data input, and integrates multi-dimensional information such as time, space, and environment. Input data includes, but is not limited to: historical charging load, grid operating parameters, meteorological data (such as temperature and humidity), vehicle charging records, holiday markers, and site numbers. To improve the accuracy and generalization ability of model training, the system first performs data cleaning, including time series alignment, outlier removal, and missing value imputation; subsequently, it performs standardization processing and transforms categorical variables to adapt to different model structure requirements.

[0022] like Figure 5 As shown, during data preprocessing and modeling preparation, the system further introduces a probabilistic statistical analysis mechanism to summarize and analyze the marginal distribution, variation characteristics, and joint relationships of various types of raw data, identifying the core variables and interaction characteristics affecting load fluctuations. By extracting statistical measures of historical variable behavior (such as mean, standard deviation, skewness, etc.) and combining them with the probability distribution changes of load data in different time periods, the system establishes a feature profile of charging behavior in the temporal and spatial dimensions, revealing the behavioral patterns and probabilistic characteristics of electric vehicle charging load during typical periods such as morning and evening peak hours, weekdays, and holidays.

[0023] Furthermore, the system evaluates the overall correlation structure and trends among input variables to assist in feature selection, error analysis, and strategy adjustment for the prediction model. Through long-term statistical accumulation and pattern extraction, the system can form a simplified representation of high-dimensional input data, laying a high-quality data foundation for subsequent multi-model load forecasting and capacity assessment. It can flexibly connect to various distribution network operation platforms and electric vehicle cluster control systems, demonstrating good adaptability, portability, and engineering value.

[0024] In this embodiment, after preprocessing the acquired multi-source heterogeneous data to obtain a spatiotemporal feature set based on the aforementioned steps, parallel prediction of load data can be performed using a pre-built multi-model fusion architecture based on this spatiotemporal feature set. Through the pre-built joint framework of fusing multi-model prediction and capacity assessment, accurate prediction of electric vehicle load changes and dynamic estimation of the available capacity of transformer substations can be achieved, forming an intelligent analysis system with adaptive adjustment capabilities, specifically including: The time-series load data in the spatiotemporal feature set is input into the long short-term memory neural network model. Nonlinear temporal dependencies are captured through forget gate, input gate, and output gate mechanisms, and deep prediction components are output. The stationary load sequence in the spatiotemporal feature set is input into the differential autoregressive moving average model for linear trend fitting, and the trend prediction component is output. The high-dimensional features in the spatiotemporal feature set are input into the random forest model, a decision tree is constructed based on the random sub-feature set and the training sample set, and the feature interaction prediction component is output. The depth prediction component, trend prediction component, and feature interaction prediction component are weighted and fused using a dynamic weighted fusion strategy to output a fused load prediction result containing confidence intervals. The dynamic weighted fusion strategy comprehensively considers the error variance of each model in the current prediction window, the performance stability of the past multiple periods, and the volatility of the prediction results to dynamically calculate the fusion weights, and introduces a dynamic adjustment factor to automatically optimize the weights based on the previous prediction errors.

[0025] In one specific embodiment, such as Figure 6 As shown, this invention designs a multi-model collaborative prediction architecture to address the highly time-varying and nonlinear characteristics of electric vehicle loads. The system establishes a deep time-series prediction model based on Long Short-Term Memory Neural Network (LSTM), a linear trend modeling module based on ARIMA, and a feature nonlinear learning model based on Random Forest, respectively, to fully exploit the temporal trends, periodicity, and high-dimensional nonlinear correlations in the load data. Each model is independently trained on a unified training and validation set and outputs prediction results. The system retains historical prediction error records and confidence interval estimates for each model.

[0026] In one embodiment, the LSTM model takes time-series load data as input. Its recursive relation is:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032] in, The input vector at the current time step, This is the hidden state from the previous moment. Output for the forget gate. For input gate output, Candidate cell state, This represents the current state of the cell. This represents the cell state at the previous moment. For output gate output, Hide the current state. , , , These represent the weight matrices, , , , These represent the bias vectors, respectively. For the sigmoid function, It is a hyperbolic tangent curve.

[0033] Model output , This is the output layer weight matrix. This is the output layer bias term, used to form the rolling prediction value. This information is then fed back to the feasible set and capacity calculation module.

[0034] Assume the load sequence after stabilization is , The model is in the following form:

[0035] Where B is the lag operator, For the AR part polynomial, For the MA part polynomial, This is white noise error. The differential operator is used to fit the short-term load change trend.

[0036] The RF model makes predictions by training multiple decision trees and integrating the results. Let the input features be... The goal is Its model is:

[0037] in, These are the predicted values ​​from the random forest. For the number of decision trees, For the input feature vector, Indicates the first Each tree is trained using a random subset of features and a subset of training samples to form features including, but not limited to: time information, holiday identifiers, ambient temperature, humidity, historical load, charging station ID, etc. Each tree is split based on the CART algorithm, and the optimal splitting features are determined by minimizing the MSE loss function.

[0038] in, For the sample size, Let i be the true value of the i-th sample. Let be the predicted value for the i-th sample. Building upon this, the present invention proposes a robust dynamic weighted fusion strategy that weights and combines the outputs of each model. The fusion strategy comprehensively considers the error variance of each model within the current prediction window, its performance stability over multiple past periods, and the volatility of the prediction results, dynamically calculating the fusion weights. This mechanism can automatically reduce the weight of performance fluctuation models when the system encounters sudden load changes during holidays, weather disturbances, or charging surges, thereby improving the overall stability and robustness of the prediction. The fused output has confidence upper / lower bound estimation capabilities, providing upper and lower boundary conditions for subsequent capacity assessment.

[0039] In one embodiment, to fully utilize the advantages of each model, the present invention employs a weighted fusion strategy, which combines the prediction results of the three models according to their weights, resulting in the following final prediction value:

[0040] in, To integrate load forecast results, These are the weight coefficients of the Long Short-Term Memory (LSTM) neural network model. These are the weighting coefficients of the differential autoregressive moving average model. These are the weight coefficients of the random forest model. This represents the load prediction output of the Long Short-Term Memory (LSTM) neural network model at time t. This represents the load forecast output of the differential autoregressive moving average model at time t. This is the load prediction output of the random forest model at time point t; Weights can be determined through cross-validation, weighted least squares, or robust optimization. In addition, the system also introduces a dynamic adjustment factor to automatically optimize the weights based on previous prediction errors.

[0041] in, Let be the fusion weight of the k-th prediction model at time t. Let j be the historical average absolute error of the k-th model up to time point t-1, and j be the summation index. Let be the average absolute error of the j-th prediction model before time point t-1.

[0042] In this embodiment, after obtaining the fused load prediction result containing confidence intervals through parallel prediction of load data using a pre-built multi-model fusion architecture in step 1, a feasible decision set for each electric vehicle can be constructed based on the fused load prediction result and multi-dimensional factor constraints. By considering the dynamic evolution method of the feasible set with multiple subjects and multiple constraints, vehicle driving characteristics, path accessibility, charging pile availability, and substation access capability are integrated to construct a time-rolling feasible decision domain and realize dynamic updating and simulation throughout the entire process. Specifically, this includes: The vehicle's drivable distance is calculated based on the current battery charge state and drivable distance of the electric vehicle. Path reachability is determined based on the vehicle's drivable distance and the actual path distance from the electric vehicle to the target charging station, and a path reachability set is generated. A set of physical access points for charging piles is generated based on their idle status, rated power, and interface type compatibility. The set of site resource constraints is obtained by filtering based on the number of available charging piles, the status of charging piles, and the optional queuing waiting time threshold for each charging station. A feasible set of distribution area capacity is generated based on the total capacity of the distribution area, the currently occupied load, and the predicted results of the integrated load. The feasible set of distribution area capacity takes into account the power flow equation of the distribution network, voltage deviation limit, and line current carrying capacity limit. Based on the path reachability set, the charging pile physical access set, the site resource constraint set, and the distribution area capacity feasibility set, and taking into account vehicle state constraints, path reachability constraints, charging station capacity availability constraints, and distribution area constraints, a feasible decision set for the dynamic evolution of each electric vehicle over time is constructed.

[0043] In one specific implementation, suppose the first... i The electric vehicle's state of charge (SOC) at time t is SOC. i,t The driving distance is:

[0044] in, The unit of SOC is the driving range (km / kWh). The vehicle's rated battery capacity (kWh). Let be the maximum distance that the i-th electric vehicle can travel at time t based on its current battery state. Let be the minimum allowed remaining battery charge percentage for the i-th electric vehicle. Let be the rated capacity of the battery of the i-th electric vehicle. This formula is used to characterize the feasible range of movement for different vehicles in the current state and is the basis for constructing the transportation-energy feasible set.

[0045] Electric vehicle from current position x iTo a certain charging station j The actual path distance is d i,j If satisfied and

[0046] Then Include the current set of feasible charging stations for vehicles. , This is the set of available charging stations. This set constitutes a class of dynamic edges in the vehicle-to-charging-station connection graph.

[0047] Furthermore, considering vehicle state constraints, path reachability, charging station capacity availability, and distribution area constraints, a comprehensive feasible decision set for the i-th vehicle at time t is constructed. :

[0048] in, For the set of path reachability, For the physical access set of charging stations, This is a set of site resource constraints (such as the number of idle stakes). For the set of feasibility constraints on the capacity of the transformer area, Let represent the possible scheduling decisions that vehicle i might make at time t.

[0049] In this embodiment, after constructing the feasible decision set for each electric vehicle based on step 2 above, a production simulation strategy can be obtained based on this feasible decision set, taking into account the security constraints of the distribution network, using a pre-constructed open capacity assessment model. This strategy specifically includes: A multi-objective collaborative optimization scheduling model is established with the objective functions of minimizing the load fluctuation rate of the distribution substation area and maximizing the electric vehicle charging demand satisfaction rate. The feasible decision set and the openable capacity of the transformer area obtained through the pre-constructed openable capacity assessment model are used as the boundary constraints of the multi-objective collaborative optimization scheduling model. The multi-objective collaborative optimization scheduling model is solved by a rolling time-domain optimization algorithm, which outputs the optimal charging power allocation instruction in the future scheduling cycle, and generates a vehicle access instruction in all time mode by combining the available capacity of the substation. A production simulation strategy is derived based on the available capacity of the transformer area and the vehicle access command.

[0050] In one specific embodiment, after completing the short-term electric vehicle load forecast, such as Figure 7As shown, this invention further constructs a distribution area-level accessibility capacity assessment model to dynamically estimate the scale of electric vehicles that can be safely accessed in future time periods. The capacity assessment module, based on fused prediction results, comprehensively considers factors such as distribution network power flow models, voltage compliance, cable / switch capacity limitations, and node accessibility constraints to establish an optimized model containing continuous variables and binary scheduling variables. The model is updated using a rolling prediction mechanism to adapt to intraday scheduling and day-ahead planning scenarios.

[0051] In one specific implementation, assume the total capacity of the current transformer area is... The occupied load is Lt, and the current predicted access load is... Then the remaining open capacity for:

[0052] in, This is a safety margin factor used to reserve system redundancy capacity. This indicator is used to construct the feasible power supply domain between vehicles and transformer substations.

[0053] To enhance the model's practical scheduling feasibility, this invention incorporates vehicle path constraint modeling, maximum scheduling distance limits, and site availability factors under multi-site access scenarios during the evaluation process, guiding scheduling strategies to more closely align with actual operational logic. The model output includes key operational indicators such as the maximum available capacity curve for each transformer area at 15-minute granularity, recommended access loads for each site, and redundancy capacity. The system also supports historical load fluctuation analysis and future capacity trend simulation, providing fundamental data support for electric vehicle access decisions and grid peak-shaving ancillary services, demonstrating promising application prospects and engineering promotion value.

[0054] The present invention has the following beneficial effects: (1) Improve the accuracy and stability of load forecasting: This invention constructs a multi-model forecasting architecture including LSTM, ARIMA and random forest, which integrates the advantages of deep learning and statistical modeling, and dynamically adjusts the model weights based on historical errors and real-time performance. Compared with the traditional single-model forecasting method, this fusion mechanism can more comprehensively capture the nonlinear trend, periodic changes and random disturbances of electric vehicle load, and achieve more stable and accurate short-term forecasting results.

[0055] (2) Enhance the robustness of the system in dealing with uncertain scenarios: Introduce probability statistical analysis and behavior pattern extraction mechanism to classify and model charging behavior under different time periods and environmental conditions, and introduce a feedback mechanism for data distribution changes in the model training and fusion process to enable the system to have stronger anomaly recognition and adaptive capabilities.

[0056] (3) Achieving dynamic capacity assessment under constrained conditions: Combining the fusion prediction results, a capacity assessment model is constructed that considers multiple constraints such as distribution network voltage limits, line capacity, and current protection. This model can dynamically output the electric vehicle capacity that can be opened for access in each distribution area at each time period. The model introduces multi-factor modeling such as site distribution, vehicle travel distance, and charging priority to achieve distributed capacity scheduling under multi-site and multi-path conditions. This provides a quantitative basis for the distribution system to scientifically accept charging loads and improves the accuracy and practicality of access planning.

[0057] (4) Support for multi-site collaborative scheduling and strategy optimization: Supports data sharing and collaborative modeling among multiple charging stations, and can dynamically generate the optimal access scheme based on grid status, load forecast and site distribution. The system is compatible with the path planning constraints and site access access access of electric vehicles. By introducing a flexible scheduling strategy framework, it achieves optimal matching between vehicles and sites, improves the utilization rate of charging facilities, alleviates the problem of load concentration, and helps to build a more flexible and distributed energy management pattern.

[0058] (5) Good practical adaptability and deployment flexibility: The method proposed in this invention adopts a modular design, which has good system compatibility and flexible deployment capabilities, and is suitable for general energy management needs at the distribution substation level. The system can be deployed and run on conventional computing platforms or substation control terminals, with low requirements for hardware and data communication conditions, making it easy to integrate and implement on existing infrastructure. At the same time, the method itself has strong fault tolerance for the structure and quantity of input data, and can adapt to operating data in different regions and at different times, possessing certain engineering feasibility and promotion potential, and is suitable for application in typical scenarios such as electric vehicle demonstration areas, urban distribution substations, or small and medium-sized charging stations with certain load forecasting needs.

[0059] Example 2: Based on the same inventive concept, this invention also provides a real-time decision-making simulation system for vehicle-pile-road networks based on feasible set derivation, such as... Figure 8 As shown, it includes: The prediction module is used to perform parallel load data prediction based on the spatiotemporal feature set of the acquired vehicle-pile-road network system using a pre-built multi-model fusion architecture, and output a fused load prediction result containing confidence intervals based on the load data prediction result through a dynamic weighted fusion strategy. The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model. The module is used to construct a feasible decision set for each electric vehicle based on the fused load prediction results and by comprehensively considering multi-dimensional factors. The generation module is used to obtain a production simulation strategy that includes the openable capacity of the transformer substation and vehicle access instructions based on the feasible decision set and under the premise of satisfying the distribution network security constraints, using a pre-built openable capacity assessment model.

[0060] Preferably, the prediction module is further configured to: The time-series load data in the spatiotemporal feature set is input into the long short-term memory neural network model. Nonlinear temporal dependencies are captured through forget gate, input gate, and output gate mechanisms, and deep prediction components are output. The stationary load sequence in the spatiotemporal feature set is input into the differential autoregressive moving average model for linear trend fitting, and the trend prediction component is output. The high-dimensional features in the spatiotemporal feature set are input into the random forest model, a decision tree is constructed based on the random sub-feature set and the training sample set, and the feature interaction prediction component is output. The depth prediction component, trend prediction component, and feature interaction prediction component are weighted and fused using a dynamic weighted fusion strategy to output a fused load prediction result containing confidence intervals. The dynamic weighted fusion strategy comprehensively considers the error variance of each model in the current prediction window, the performance stability of the past multiple periods, and the volatility of the prediction results to dynamically calculate the fusion weights, and introduces a dynamic adjustment factor to automatically optimize the weights based on the previous prediction errors.

[0061] Preferably, the weighted fusion in the prediction module is calculated using the following formula:

[0062]

[0063] in, To integrate load forecast results, These are the weight coefficients of the Long Short-Term Memory (LSTM) neural network model. These are the weighting coefficients of the differential autoregressive moving average model. These are the weight coefficients of the random forest model. This represents the load prediction output of the Long Short-Term Memory (LSTM) neural network model at time t. This represents the load forecast output of the differential autoregressive moving average model at time t. This is the load prediction output of the random forest model at time point t. Let be the fusion weight of the k-th prediction model at time t. Let j be the historical average absolute error of the k-th model before time point t-1, and j be the summation index.

[0064] Preferably, the building module is further configured to: The vehicle's drivable distance is calculated based on the current battery charge state and drivable distance of the electric vehicle. Path reachability is determined based on the vehicle's drivable distance and the actual path distance from the electric vehicle to the target charging station, and a path reachability set is generated. A set of physical access points for charging piles is generated based on their idle status, rated power, and interface type compatibility. The set of site resource constraints is obtained by filtering based on the number of available charging piles, the status of charging piles, and the optional queuing waiting time threshold for each charging station. A feasible set of distribution area capacity is generated based on the total capacity of the distribution area, the currently occupied load, and the predicted results of the integrated load. The feasible set of distribution area capacity takes into account the power flow equation of the distribution network, voltage deviation limit, and line current carrying capacity limit. Based on the path reachability set, the charging pile physical access set, the site resource constraint set, and the distribution area capacity feasibility set, and taking into account vehicle state constraints, path reachability constraints, charging station capacity availability constraints, and distribution area constraints, a feasible decision set for the dynamic evolution of each electric vehicle over time is constructed.

[0065] Preferably, the drivable distance of the vehicle in the construction module is obtained by the following calculation formula:

[0066] in, Let be the maximum distance that the i-th electric vehicle can travel at time t based on its current battery state. Let be the distance that the i-th electric vehicle can travel with a single unit of battery charge. Let be the percentage of remaining battery charge of the i-th electric vehicle at time t. Let be the minimum allowed remaining battery charge percentage for the i-th electric vehicle. Let be the rated capacity of the battery of the i-th electric vehicle.

[0067] Preferably, the generation module is further configured to: A multi-objective collaborative optimization scheduling model is established with the objective functions of minimizing the load fluctuation rate of the distribution substation area and maximizing the electric vehicle charging demand satisfaction rate. The feasible decision set and the openable capacity of the transformer area obtained through the pre-constructed openable capacity assessment model are used as the boundary constraints of the multi-objective collaborative optimization scheduling model. The multi-objective collaborative optimization scheduling model is solved by a rolling time-domain optimization algorithm, which outputs the optimal charging power allocation instruction in the future scheduling cycle, and generates a vehicle access instruction in all time mode by combining the available capacity of the substation. A production simulation strategy is derived based on the available capacity of the transformer area and the vehicle access command.

[0068] Preferably, the acquisition of the spatiotemporal feature set in the prediction module includes: Data cleaning is performed on the multi-source heterogeneous data of the vehicle-charging-road network system, and a probabilistic statistical analysis mechanism is introduced to summarize and analyze the marginal distribution, variation characteristics and joint relationships of the cleaned multi-source heterogeneous data. The multi-source heterogeneous data includes one or more of the following: electric vehicle status data, charging facility resource data, road traffic network data or power distribution network operation parameters. Probabilistic statistical analysis was performed on the cleaned multi-source heterogeneous data to extract statistics including mean, standard deviation, and skewness. Combined with time period characteristics, the load probability distribution changes in the temporal and spatial dimensions of morning and evening peak hours, weekdays and holidays were identified to obtain a spatiotemporal feature set.

[0069] Example 3 like Figure 9 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0070] The processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to realize the corresponding method flow or corresponding function, so as to realize the steps of the vehicle-pile-road network full-time decision simulation method based on feasible set in the above embodiment.

[0071] Example 4 Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of the vehicle-pile-road network full-time decision simulation method based on feasible set derivation in the above embodiments.

[0072] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A real-time decision simulation method for vehicle-pile-road network based on feasible set derivation, characterized in that, include: Based on the spatiotemporal feature set of the acquired vehicle-pile-road network system, a pre-constructed multi-model fusion architecture is used to perform parallel load data prediction. Based on the load data prediction results, a fused load prediction result containing confidence intervals is output through a dynamic weighted fusion strategy. The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model. Based on the fusion load prediction results, a feasible decision set for each electric vehicle is constructed by comprehensively considering multi-dimensional factors. Based on the feasible decision set, and under the premise of satisfying the distribution network security constraints, a production simulation strategy including the openable capacity of transformer substations and vehicle access instructions is obtained by using a pre-constructed openable capacity assessment model.

2. The method according to claim 1, characterized in that, The method of performing parallel load data prediction based on the spatiotemporal feature set using a pre-built multi-model fusion architecture, and outputting a fused load prediction result containing confidence intervals based on the load data prediction results through a dynamic weighted fusion strategy, includes: The time-series load data in the spatiotemporal feature set is input into the long short-term memory neural network model. Nonlinear temporal dependencies are captured through forget gate, input gate, and output gate mechanisms, and deep prediction components are output. The stationary load sequence in the spatiotemporal feature set is input into the differential autoregressive moving average model for linear trend fitting, and the trend prediction component is output. The high-dimensional features in the spatiotemporal feature set are input into the random forest model, a decision tree is constructed based on the random sub-feature set and the training sample set, and the feature interaction prediction component is output. The depth prediction component, trend prediction component, and feature interaction prediction component are weighted and fused using a dynamic weighted fusion strategy to output a fused load prediction result containing confidence intervals. The dynamic weighted fusion strategy comprehensively considers the error variance of each model in the current prediction window, the performance stability of the past multiple periods, and the volatility of the prediction results to dynamically calculate the fusion weights, and introduces a dynamic adjustment factor to automatically optimize the weights based on the previous prediction errors.

3. The method according to claim 2, characterized in that, The weighted fusion is calculated using the following formula: in, To integrate load forecast results, These are the weight coefficients of the Long Short-Term Memory (LSTM) neural network model. These are the weighting coefficients of the differential autoregressive moving average model. These are the weight coefficients of the random forest model. This represents the load prediction output of the Long Short-Term Memory (LSTM) neural network model at time t. This represents the load forecast output of the differential autoregressive moving average model at time t. This is the load prediction output of the random forest model at time point t. Let be the fusion weight of the k-th prediction model at time t. Let j be the historical average absolute error of the k-th model before time point t-1, and j be the summation index.

4. The method according to claim 1, characterized in that, The construction of feasible decision sets for each electric vehicle based on the fused load prediction results and the multi-source heterogeneous data, taking into account multiple dimensions of constraints, includes: The vehicle's drivable distance is calculated based on the current battery charge state and drivable distance of the electric vehicle. Path reachability is determined based on the vehicle's drivable distance and the actual path distance from the electric vehicle to the target charging station, and a path reachability set is generated. A set of physical access points for charging piles is generated based on their idle status, rated power, and interface type compatibility. The set of site resource constraints is obtained by filtering based on the number of available charging piles, the status of charging piles, and the optional queuing waiting time threshold for each charging station. A feasible set of distribution area capacity is generated based on the total capacity of the distribution area, the currently occupied load, and the predicted results of the integrated load. The feasible set of distribution area capacity takes into account the power flow equation of the distribution network, voltage deviation limit, and line current carrying capacity limit. Based on the path reachability set, the charging pile physical access set, the site resource constraint set, and the distribution area capacity feasibility set, and taking into account vehicle state constraints, path reachability constraints, charging station capacity availability constraints, and distribution area constraints, a feasible decision set for the dynamic evolution of each electric vehicle over time is constructed.

5. The method according to claim 4, characterized in that, The vehicle's drivable distance is calculated using the following formula: in, Let be the maximum distance that the i-th electric vehicle can travel at time t based on the current battery state. Let be the distance that the i-th electric vehicle can travel with a single unit of battery charge. Let be the percentage of remaining battery charge of the i-th electric vehicle at time t. Let be the minimum allowed remaining battery charge percentage for the i-th electric vehicle. Let be the rated capacity of the battery of the i-th electric vehicle.

6. The method according to claim 1, characterized in that, The production simulation strategy, based on the feasible decision set and under the premise of satisfying the distribution network security constraints, utilizes a pre-constructed openable capacity assessment model to obtain a model that includes the openable capacity of transformer substations and vehicle access commands, including: A multi-objective collaborative optimization scheduling model is established with the objective functions of minimizing the load fluctuation rate of the distribution substation area and maximizing the electric vehicle charging demand satisfaction rate. The feasible decision set and the openable capacity of the transformer area obtained through the pre-constructed openable capacity assessment model are used as the boundary constraints of the multi-objective collaborative optimization scheduling model. The multi-objective collaborative optimization scheduling model is solved by a rolling time-domain optimization algorithm, which outputs the optimal charging power allocation instruction in the future scheduling cycle, and generates a vehicle access instruction in all time mode by combining the available capacity of the substation. A production simulation strategy is derived based on the available capacity of the transformer area and the vehicle access command.

7. The method according to claim 1, characterized in that, The acquisition of the spatiotemporal feature set includes: Data cleaning is performed on the multi-source heterogeneous data of the vehicle-charging-road network system, and a probabilistic statistical analysis mechanism is introduced to summarize and analyze the marginal distribution, variation characteristics and joint relationships of the cleaned multi-source heterogeneous data. The multi-source heterogeneous data includes one or more of the following: electric vehicle status data, charging facility resource data, road traffic network data or power distribution network operation parameters. Probabilistic statistical analysis was performed on the cleaned multi-source heterogeneous data to extract statistics including mean, standard deviation, and skewness. Combined with time period characteristics, the load probability distribution changes in the temporal and spatial dimensions of morning and evening peak hours, weekdays and holidays were identified to obtain a spatiotemporal feature set.

8. A real-time decision-making simulation system for vehicle-pile-road network based on feasible set derivation, characterized in that, include: The prediction module is used to perform parallel load data prediction based on the spatiotemporal feature set of the acquired vehicle-pile-road network system using a pre-built multi-model fusion architecture, and output a fused load prediction result containing confidence intervals based on the load data prediction result through a dynamic weighted fusion strategy. The multi-model fusion architecture includes a long short-term memory neural network model, a differential autoregressive moving average model, and a random forest model. The module is used to construct a feasible decision set for each electric vehicle based on the fused load prediction results and by comprehensively considering multi-dimensional factors. The generation module is used to obtain a production simulation strategy that includes the openable capacity of the transformer substation and vehicle access instructions based on the feasible decision set and under the premise of satisfying the distribution network security constraints, using a pre-built openable capacity assessment model.

9. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method for full-time decision simulation of vehicle-pile-road network based on feasible set derivation as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium, characterized in that, It contains an execution program, which, when executed, implements the vehicle-pile-road network full-time decision simulation method based on feasible set deduction as described in any one of claims 1 to 7.

Images