Charging pile virtual and real operation simulation verification system based on digital twinning

Abstract

The invention discloses a charging pile virtual and real operation simulation verification system based on digital twinning, and the system comprises the following modules: a data collection and standardization module which is used for collecting and standardizing an original charging multi-mode data set; the causal relationship analysis module is used for performing causal relationship test; the information aggregation module is used for carrying out local neighborhood information aggregation through the Grapher network; the time sequence prediction module is used for performing time sequence prediction through an improved DeepAR model; the virtual-reality feedback model construction module is used for generating a virtual-reality feedback model by using a virtual-reality dynamic feedback mechanism; the resource scheduling generation module is used for carrying out strategy distribution on charging pile resource scheduling; and the operation scheme generation module is used for initializing the virtual simulation model and performing dynamic simulation reasoning and optimization. According to the invention, through the improved DeepAR model, the resource scheduling and operation efficiency of the charging pile is improved.

Description

Technical Field

[0001] This invention relates to the field of simulation verification technology, and in particular to a virtual-real operation simulation verification system for charging piles based on digital twins. Background Technology

[0002] With the increasing popularity of electric vehicles (EVs), charging stations, as a crucial infrastructure for EVs, have become increasingly important in terms of construction and management. However, traditional charging station resource scheduling and operation management systems often face numerous challenges. In existing technologies, most charging station resource scheduling and operation optimization rely on static models or simple rule engines, lacking dynamic adaptability. Traditional charging station management systems typically employ rule-based scheduling methods, using preset scheduling strategies for different user needs and charging station loads. However, this approach has significant limitations in practical applications. As the scale of the charging station network expands, the system cannot effectively adjust in real time to cope with complex load fluctuations, environmental changes, and diversified user needs, leading to low resource utilization efficiency and even overload or idleness of some charging stations, thus affecting charging efficiency and user experience.

[0003] Furthermore, existing virtual simulation systems for charging piles typically fail to accurately reflect the dynamic operating status of charging piles in real-world environments. Traditional virtual simulation systems are mostly based on static models, making it difficult to provide real-time and effective support in changing charging environments. While some systems attempt to optimize by simulating the charging behavior of electric vehicles and the operating status of charging piles, the lack of in-depth analysis of spatiotemporal causal relationships often prevents them from accurately capturing the underlying patterns in charging pile operation. This results in significant discrepancies between simulation results and actual operation, impacting the effectiveness of scheduling and optimization strategies. Therefore, improving the consistency between virtual simulation and actual operating status of charging piles, and optimizing charging pile resource scheduling schemes in real time based on actual data, have become urgent technical problems to be solved. Most systems rely on basic sensor data collection but do not deeply integrate and standardize charging pile operating data from different data sources. Since the operating status of charging piles is affected by various factors, such as voltage, current, ambient temperature, and user behavior, these factors exhibit different dependencies across different time and spatial dimensions. Existing technologies cannot fully grasp the spatiotemporal characteristics of charging piles through a single data source or a single analysis model, thus failing to accurately predict future charging demand and leading to lags and failures in resource allocation and scheduling strategies.

[0004] Therefore, how to provide a simulation and verification system for the virtual and real operation of charging piles based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a digital twin-based virtual-real operation simulation verification system for charging piles. This invention combines digital twin technology with a virtual-reality dynamic feedback mechanism, utilizing spatiotemporal causal relationship analysis and an improved DeepAR time-series prediction model to optimize charging pile resource scheduling schemes in real time. This system can dynamically adjust the virtual simulation model to ensure consistency between the virtual and actual operating states, thereby improving charging pile resource utilization efficiency, load balancing, and fault prediction accuracy, significantly enhancing the operating efficiency and user experience of the charging pile system.

[0006] A virtual-real operation simulation verification system for charging piles based on digital twins according to an embodiment of the present invention includes the following modules:

[0007] The data acquisition and standardization module is used to collect and standardize the raw charging multimodal dataset from charging piles to obtain a standardized charging multimodal dataset.

[0008] The causal relationship analysis module is used to perform causal relationship testing on the charging multimodal dataset, determine the spatiotemporal dependencies between various factors during the operation of the charging pile, and generate a charging spatiotemporal causal relationship graph.

[0009] The information aggregation module is used to input the charging spatiotemporal causal relationship graph into the Graphormer network to perform local neighborhood information aggregation and obtain a charging embedding feature vector set.

[0010] The time-series prediction module is used to input the charging embedded feature vector set into the improved DeepAR model, and perform time-series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status.

[0011] The virtual-reality feedback model construction module is used to generate a virtual-reality feedback model based on the prediction result dataset of the charging pile operation status using a virtual-reality dynamic feedback mechanism.

[0012] The resource scheduling generation module is used to allocate strategies for charging pile resource scheduling based on the virtual-reality feedback model, and obtain a charging pile resource scheduling scheme.

[0013] The operation plan generation module is used to initialize a virtual simulation model based on the charging pile resource scheduling plan, and perform dynamic simulation reasoning and optimization to obtain an optimized charging pile operation plan.

[0014] A method for simulating and verifying the virtual and real operation of a charging pile based on digital twins according to an embodiment of the present invention includes the following steps:

[0015] Step 1: Collect the original charging multimodal dataset from the charging piles, and standardize the original charging multimodal dataset to obtain a standardized charging multimodal dataset.

[0016] Step 2: Perform causal relationship testing on the charging multimodal dataset to determine the spatiotemporal dependencies between various factors during the operation of the charging pile and generate a charging spatiotemporal causal relationship graph;

[0017] Step 3: Input the charging spatiotemporal causal relationship graph into the Graphormer network to aggregate local neighborhood information and obtain the charging embedding feature vector set;

[0018] Step 4: Input the charging embedded feature vector set into the improved DeepAR model, and perform time series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status;

[0019] Step 5: Based on the predicted data set of charging pile operating status, a virtual-reality dynamic feedback mechanism is used to generate a virtual-reality feedback model;

[0020] Step 6: Allocate charging pile resources according to the virtual-reality feedback model to obtain a charging pile resource scheduling scheme;

[0021] Step 7: Initialize the virtual simulation model based on the charging pile resource scheduling scheme, and perform dynamic simulation reasoning and optimization to obtain the optimized charging pile operation scheme.

[0022] Optionally, the original charging multimodal dataset includes charging pile operation status data, charging pile external environment data, and user behavior data; the standardization processing steps include outlier removal, missing value imputation, timestamp alignment, and normalization processing for different types of data in the original charging multimodal dataset to obtain a standardized charging multimodal dataset.

[0023] Optionally, step two specifically involves:

[0024] Time series of equal size are selected from the standardized charging multimodal dataset, and each selected time series is transformed into a stationary time series through differencing.

[0025] The lag order of a stationary time series is determined using the Bayesian information criterion, where the lag order is the time difference between the current observation and previous observations.

[0026] Set the null and alternative hypotheses, specifically including:

[0027] The null hypothesis is that one time series has no causal relationship with another time series;

[0028] The alternative hypothesis is that one time series has a causal relationship with another time series.

[0029] The p-value is used to determine whether a previous time series can predict a subsequent time series given a lag number. If the p-value is less than a preset significance level, the null hypothesis is rejected, indicating that a causal relationship exists.

[0030] Based on the results of causal relationship testing, the spatiotemporal dependencies between various factors during the operation of the charging pile are determined, and a spatiotemporal causal relationship diagram of the charging pile is constructed. The construction steps include:

[0031] Each factor in the operation of the charging pile is treated as a node, resulting in a node set;

[0032] The causal relationship test results are used to mark the causal relationship directions between different time series, thus obtaining a set of connection edges;

[0033] By connecting different nodes in the node set according to the connection relationship in the edge set, a charging spatiotemporal causal relationship graph is obtained.

[0034] Optionally, step three specifically includes:

[0035] The charging spatiotemporal causal relationship graph is input into the input layer of the Graphormer network. The input layer performs one-hot encoding on the nodes and edges in the charging spatiotemporal causal relationship graph, converting each node and edge into a numerical vector representation, thereby obtaining the initial embedding vector set of the charging spatiotemporal causal relationship graph.

[0036] Through graph convolution operations, local neighborhood information aggregation is performed on the initial embedding vector of each node. The local neighborhood information aggregation is achieved by using the edge information in the charging spatiotemporal causal relationship graph to propagate the relationship information between adjacent nodes to the current node for weighted averaging, and updating the initial embedding vector of each node to obtain the graph convolution feature vector of the node.

[0037] The graph convolutional feature vector is input into the temporal information fusion layer. Combined with the temporal information of the nodes, the temporal features and graph structure features are fused through the self-attention mechanism to obtain the temporal feature embedding vector of each node.

[0038] The self-attention mechanism layer aggregates information by utilizing the causal relationships between nodes in the charging spatiotemporal causal relationship graph.

[0039] The temporal feature embedding vector is input into the fully connected layer and then through the multilayer perceptron layer to generate the final charging embedding feature vector set. The charging embedding feature vector set contains the embedding representation of each node, representing the spatiotemporal causal relationship and dynamic characteristics of various factors in the charging pile.

[0040] Optionally, the improved DeepAR model is specifically as follows:

[0041] The charging embedded feature vector set is input into the time series modeling module in the order of time series, and the time series data of the charging pile is combined with the embedded features to create a time series model of the charging pile's operating status.

[0042] The time series model uses the autoregressive properties of time series data to model the changing patterns of the charging pile's operating status by learning from historical data, and generates historical time series feature representations, which represent the past operating status of the charging pile.

[0043] The historical time series features are input into the periodic modeling module. By establishing a periodic feature modeling mechanism, the periodic changes in the load and current variables of the charging pile over time can be captured.

[0044] The extracted periodic variation features are added to the time series model to obtain a periodic feature vector, which includes periodic fluctuation information in the charging pile's operating status.

[0045] The periodic feature vector is input into the residual correction module to calculate the difference between the historical predicted value and the actual observed value of the charging pile's operating status, thereby obtaining the prediction error for each time step. The historical predicted value is the mean value of the historical time series feature representation.

[0046] For time series data with a prediction error greater than a preset error threshold, the future prediction value is adjusted by learning the error pattern through a support vector machine to obtain a corrected prediction feature vector. The adjustment involves adding the prediction error back to the original prediction value.

[0047] In the prediction module, the historical time series feature representation from the time series modeling module, the periodic feature vector from the periodic modeling module, and the prediction feature vector from the residual correction module are combined and weighted according to preset weights to generate a dataset of prediction results for the charging pile's operating status.

[0048] Optionally, step five specifically includes:

[0049] The actual operating data of the charging piles and the predicted results of the charging pile operating status are transmitted to the virtual simulation system for real-time synchronization.

[0050] Calculate the deviation between the virtual simulation results and the actual charging pile operation data;

[0051] When the deviation between the virtual simulation results and the actual operating status of the charging pile exceeds a preset threshold, an adjustment mechanism is triggered.

[0052] When the deviation exceeds a preset threshold, the simulation value of the corresponding simulation parameter in the virtual system is automatically optimized through a feedback mechanism until the virtual simulation matches the actual behavior, thus obtaining a virtual-reality feedback model. The simulation parameters are used to correct the error in the virtual simulation system.

[0053] Optionally, step six specifically includes:

[0054] Resource demand information for each charging pile is extracted from the actual operation data of the charging pile and the predicted data of the charging pile obtained from the virtual-reality feedback model. The resource demand information includes the charging load demand, power demand, charging duration, current idle status of the charging pile and expected user demand.

[0055] Based on the resource demand information, the charging piles are sorted according to the preset priority rules to generate a resource allocation priority list for each charging pile.

[0056] Based on the resource allocation priority list and charging pile resource demand information, the available resources of the charging piles are strategically allocated to obtain a charging pile resource scheduling scheme.

[0057] Optionally, step seven specifically includes:

[0058] The virtual simulation model is initialized according to the charging pile resource scheduling scheme. The initialization involves constructing the initial operating state of the charging pile based on its resource requirements, scheduling priority, and load allocation; and setting the charging pile operating parameters and resource configuration of the virtual simulation model in advance.

[0059] The operation status of charging piles is simulated in real time through a dynamic simulation reasoning process based on a virtual simulation model. The dynamic simulation reasoning process simulates the operation status of charging piles under different loads, environments and user needs according to the resource scheduling scheme of the charging piles. In the dynamic simulation reasoning process, the virtual simulation model dynamically adjusts the simulation results according to real-time input data, predicts future operation trends, and obtains the simulation results of the charging pile operation status.

[0060] Based on the simulation results of the charging pile's operating status, the resource allocation and scheduling strategy is optimized according to the set minimum energy consumption, resulting in an optimized charging pile operation scheme.

[0061] The beneficial effects of this invention are:

[0062] The digital twin-based virtual-real operation simulation verification system for charging piles proposed in this invention effectively solves a series of problems in existing charging pile resource scheduling and operation management systems. Traditional charging pile resource scheduling methods mostly rely on static preset rules, lacking the ability to respond dynamically to the real-time operating status of charging piles. This invention, by introducing spatiotemporal causal relationship analysis and dynamic simulation reasoning, can not only accurately capture the potential patterns in charging pile operation but also adaptively optimize resource scheduling strategies based on real-time data, significantly improving resource utilization efficiency and load balancing. With the support of a virtual-real feedback mechanism, the system can continuously correct the simulation model during charging pile operation, ensuring consistency between virtual simulation and actual charging pile behavior, thereby improving the accuracy of system predictions and the reliability of decision-making. Furthermore, this invention standardizes and performs causal relationship analysis on multimodal data of charging piles, effectively integrating relationships between different data sources, overcoming the shortcomings of existing technologies that rely on only a single data source or simple rules. The time-series prediction module, combined with an improved DeepAR model, can accurately predict the operating status of charging piles through periodic modeling and residual correction mechanisms, identifying potential fluctuations in charging demand and fault risks in advance, providing accurate reference data for subsequent resource scheduling. This innovation effectively overcomes the problems of traditional methods, such as the inability to accurately predict fluctuations in charging demand and the lag in resource allocation. By combining a virtual-reality dynamic feedback mechanism, this invention enables the charging pile system to dynamically adjust the virtual model based on real-time operational data, promptly identify and correct deviations in the system, and ensure the consistency between the virtual simulation and the actual operating state of the charging pile. Ultimately, this real-time feedback capability provides strong support for optimizing charging pile resource scheduling schemes, ensuring that the system can make optimal resource allocation decisions under different loads, environments, and user demand changes, thereby improving the operating efficiency of the charging pile system and the user experience. Overall, this invention not only fills the gap in existing technology but also greatly improves the accuracy, flexibility, and adaptability of charging pile resource scheduling, and has broad application prospects. Attached Figure Description

[0063] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0064] Figure 1 This is a schematic diagram of the structure of a charging pile virtual-real operation simulation verification system based on digital twin proposed in this invention;

[0065] Figure 2 This is an overall flowchart of a digital twin-based virtual-real operation simulation verification method for charging piles proposed in this invention. Detailed Implementation

[0066] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0067] refer to Figure 1 A digital twin-based simulation and verification system for the virtual and real operation of charging piles includes the following modules:

[0068] The data acquisition and standardization module is used to collect and standardize the raw charging multimodal dataset from charging piles to obtain a standardized charging multimodal dataset.

[0069] The causal relationship analysis module is used to perform causal relationship testing on the charging multimodal dataset, determine the spatiotemporal dependencies between various factors during the operation of the charging pile, and generate a charging spatiotemporal causal relationship graph.

[0070] The information aggregation module is used to input the charging spatiotemporal causal relationship graph into the Graphormer network to perform local neighborhood information aggregation and obtain a charging embedding feature vector set.

[0071] The time-series prediction module is used to input the charging embedded feature vector set into the improved DeepAR model, and perform time-series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status.

[0072] The virtual-reality feedback model construction module is used to generate a virtual-reality feedback model based on the prediction result dataset of the charging pile operation status using a virtual-reality dynamic feedback mechanism.

[0073] The resource scheduling generation module is used to allocate strategies for charging pile resource scheduling based on the virtual-reality feedback model, and obtain a charging pile resource scheduling scheme.

[0074] The operation plan generation module is used to initialize a virtual simulation model based on the charging pile resource scheduling plan, and perform dynamic simulation reasoning and optimization to obtain an optimized charging pile operation plan.

[0075] refer to Figure 2 A simulation verification method for the virtual and real operation of charging piles based on digital twins includes the following steps:

[0076] Step 1: Collect the original charging multimodal dataset from the charging piles, and standardize the original charging multimodal dataset to obtain a standardized charging multimodal dataset.

[0077] Step 2: Perform causal relationship testing on the charging multimodal dataset to determine the spatiotemporal dependencies between various factors during the operation of the charging pile and generate a charging spatiotemporal causal relationship graph;

[0078] Step 3: Input the charging spatiotemporal causal relationship graph into the Graphormer network to aggregate local neighborhood information and obtain the charging embedding feature vector set;

[0079] Step 4: Input the charging embedded feature vector set into the improved DeepAR model, and perform time series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status;

[0080] Step 5: Based on the predicted data set of charging pile operating status, a virtual-reality dynamic feedback mechanism is used to generate a virtual-reality feedback model;

[0081] Step 6: Allocate charging pile resources according to the virtual-reality feedback model to obtain a charging pile resource scheduling scheme;

[0082] Step 7: Initialize the virtual simulation model based on the charging pile resource scheduling scheme, and perform dynamic simulation reasoning and optimization to obtain the optimized charging pile operation scheme.

[0083] In this embodiment, the original charging multimodal dataset contains multiple different types of data from the charging pile system. This data helps to effectively analyze and optimize the operation of the charging piles. Specifically, the original dataset includes the following parts:

[0084] Charging pile operation status data: This type of data mainly records all important parameters generated during the operation of the charging pile, such as charging current, voltage, charging power, charging time, temperature, and humidity. This data reflects the actual working status of the charging pile and helps monitor power consumption and equipment health during the charging process.

[0085] External environmental data for charging stations: This section includes information related to the environment in which the charging station is located, such as outdoor temperature, humidity, atmospheric pressure, and weather conditions. Changes in the external environment can directly affect the operating efficiency and equipment stability of the charging station; therefore, this data is crucial for simulation and analysis.

[0086] User behavior data: This type of data includes user behavior information about charging stations, such as users' charging habits, charging frequency, charging time periods, and charging power requirements. This behavioral data can not only help predict charging demand but also be used to optimize the resource allocation of charging stations and the user experience.

[0087] The data mentioned above comes from different sensors and monitoring systems, and typically has different formats, units, and acquisition cycles. It is crucial to ensure that all data can be formatted uniformly for subsequent analysis. Standardization processing steps include:

[0088] Outlier removal: Identifying and removing outliers from the raw data to prevent these anomalies from affecting subsequent analysis and model training. Outliers are usually caused by sensor malfunctions, data acquisition errors, or other abnormal situations, which may increase the error of model predictions.

[0089] Missing value imputation: During data acquisition, some data may be missing due to various reasons (such as sensor malfunction, network problems, etc.). To avoid inaccurate analysis caused by missing data, mean imputation is used to fill in the missing values ​​and ensure the integrity of the dataset.

[0090] Timestamp alignment: Since different sensors may have different data acquisition cycles, the timestamps of the raw data may be inconsistent. By aligning the timestamps, we ensure that all data are synchronized within the same time frame and transform them into a unified time standard, so that subsequent data analysis and model training can be processed based on a consistent time series.

[0091] Normalization: To enable data with different dimensions to be processed in the same model, data normalization is necessary. Normalization scales the data to a uniform range (e.g., between 0 and 1), eliminating the influence of differences in dimensions, thereby improving the model's convergence speed and prediction accuracy.

[0092] In this embodiment, step two specifically includes:

[0093] Time series of equal size were selected from a standardized multimodal charging dataset. These time series data contain multiple parameters at different time points during the operation of charging piles, such as charging load, current, voltage, environmental factors, and user behavior. When selecting time series, it is necessary to ensure that the data volume of each time series is the same to avoid bias in the analysis results due to inconsistent data volume. Then, by differencing each selected time series, non-stationary time series are transformed into stationary time series. The characteristic of stationary time series is that their statistical properties (such as mean and variance) do not change over time, which can meet the requirements of causal relationship analysis.

[0094] For stationary time series, the Bayesian Information Criterion (BIC) is used to determine the lag order. The lag order refers to the time difference between the current observation and previous observations, reflecting the predictive power of previous data on the current data. The Bayesian Information Criterion helps determine the optimal lag number, avoiding the negative impact of too many or too few lag orders on the results of causal relationship analysis. By selecting the most appropriate lag order, it is ensured that causal relationship tests can effectively capture the dependencies between time series.

[0095] After determining the lag order, the next step is to test the causality relationship. This involves setting the null hypothesis and the alternative hypothesis. The null hypothesis states that one time series has no causal relationship with another time series; that is, the data from the preceding time series cannot effectively predict the changes in the following time series. The alternative hypothesis states that one time series has a causal relationship with another time series; that is, the preceding time series can effectively predict the changes in the following time series. The p-value is used to determine the significance level. If, given the lag order, the p-value is less than the preset significance level (e.g., 0.05), the null hypothesis is rejected, indicating that the preceding time series has causal predictive power over the following time series.

[0096] Based on the causal relationship test results, the system can further analyze the spatiotemporal dependencies between various factors during the operation of the charging pile. According to the causal relationship test results, a spatiotemporal causal relationship graph for charging is constructed. This causal relationship graph contains the causal dependencies between various factors in the charging pile system (such as load, current, voltage, environmental factors, etc.). Specific steps include treating each factor in the charging pile operation process as a node to generate a node set; and marking the causal relationship directions between different time series according to the causal relationship test results to obtain a set of connection edges. Finally, based on the connection relationships in the connection edge set, different nodes in the node set are connected to construct a complete spatiotemporal causal relationship graph for charging.

[0097] This spatiotemporal causal relationship diagram for charging provides an accurate causal reasoning basis for subsequent simulation verification, resource scheduling and system optimization. It can comprehensively reflect the interdependencies between various factors in the charging pile system, greatly improving the accuracy and decision-making ability of the simulation system.

[0098] In this embodiment, step three specifically includes:

[0099] The charging spatiotemporal causal relationship graph is input into the input layer of the Graphormer network. The input layer performs one-hot encoding on the nodes and edges in the charging spatiotemporal causal relationship graph, converting each node and edge into a numerical vector representation, thereby obtaining the initial embedding vector set of the charging spatiotemporal causal relationship graph.

[0100] Through graph convolution operations, local neighborhood information aggregation is performed on the initial embedding vector of each node. The local neighborhood information aggregation is achieved by using the edge information in the charging spatiotemporal causal relationship graph to propagate the relationship information between adjacent nodes to the current node for weighted averaging, and updating the initial embedding vector of each node to obtain the graph convolution feature vector of the node.

[0101] The graph convolutional feature vector is input into the temporal information fusion layer. Combined with the temporal information of the nodes, the temporal features and graph structure features are fused through the self-attention mechanism to obtain the temporal feature embedding vector of each node.

[0102] The self-attention mechanism layer aggregates information by utilizing the causal relationships between nodes in the charging spatiotemporal causal relationship graph.

[0103] The temporal feature embedding vector is input into the fully connected layer and then through the multilayer perceptron layer to generate the final charging embedding feature vector set. The charging embedding feature vector set contains the embedding representation of each node, representing the spatiotemporal causal relationship and dynamic characteristics of various factors in the charging pile.

[0104] In this embodiment, the improved DeepAR model is specifically as follows:

[0105] The charging embedded feature vector set is input into the time series modeling module in the order of time series, and the time series data of the charging pile is combined with the embedded features to create a time series model of the charging pile's operating status.

[0106] The time series model uses the autoregressive properties of time series data to model the changing patterns of the charging pile's operating status by learning from historical data, and generates historical time series feature representations, which represent the past operating status of the charging pile.

[0107] The historical time series features are input into the periodic modeling module. By establishing a periodic feature modeling mechanism, the periodic changes in the load and current variables of the charging pile over time can be captured.

[0108] The extracted periodic variation features are added to the time series model to obtain a periodic feature vector, which includes periodic fluctuation information in the charging pile's operating status.

[0109] The periodic feature vector is input into the residual correction module to calculate the difference between the historical predicted value and the actual observed value of the charging pile's operating status, thereby obtaining the prediction error for each time step. The historical predicted value is the mean value of the historical time series feature representation.

[0110] For time series data with a prediction error greater than a preset error threshold, the future prediction value is adjusted by learning the error pattern through a support vector machine to obtain a corrected prediction feature vector. The adjustment involves adding the prediction error back to the original prediction value.

[0111] In the prediction module, the historical time series feature representation from the time series modeling module, the periodic feature vector from the periodic modeling module, and the prediction feature vector from the residual correction module are combined and weighted according to preset weights to generate a dataset of prediction results for the charging pile's operating status.

[0112] This step, by combining time-series modeling, periodic modeling, and residual correction modules, achieves accurate prediction of the charging pile's operational status. Utilizing the autoregressive properties and periodic variation characteristics of time series data, the system can capture the operating patterns of charging piles across different time periods, ensuring that fluctuations in key parameters such as load and current are accurately reflected. The residual correction module further improves prediction accuracy and reduces deviations caused by prediction errors. When the prediction error exceeds a preset threshold, a support vector machine is used for error pattern learning to effectively adjust the predicted values, ensuring more accurate future predictions. The final generated prediction dataset provides high-quality data support for charging pile resource scheduling and operational optimization, improving the system's resource utilization efficiency and charging pile operational stability.

[0113] In this embodiment, step five specifically includes:

[0114] The actual operating data and predicted results of the charging piles need to be transmitted to the virtual simulation system for real-time synchronization. The actual operating data includes operating parameters such as the charging pile's load, current, and voltage, as well as external environmental data and user behavior data. This data is collected in real-time by the charging pile system and transmitted to the virtual simulation system to ensure that the charging pile status reflected in the virtual system is synchronized with the status in the actual system. Through real-time synchronization, the virtual simulation system can promptly obtain the latest data corresponding to the actual operation of the charging piles, thereby enabling effective simulation and adjustments.

[0115] The virtual simulation system calculates the deviation between the simulation results and the actual operating data of the charging pile. This calculation process is crucial for verifying whether the virtual simulation matches the actual charging pile system. The deviation is primarily calculated based on the errors between the charging pile's operating parameters (such as load, current, and voltage). By comparing the simulation results with the actual data, the system can quantify the difference between the two, providing a basis for subsequent adjustments and optimizations.

[0116] When the deviation exceeds a preset threshold, the system will trigger an adjustment mechanism. This threshold is set based on the operating characteristics of the charging pile system and the accuracy requirements of the virtual simulation system. If the deviation between the virtual simulation results and the actual operating state exceeds the set threshold, the system will automatically enter adjustment mode to ensure that the virtual simulation results more accurately reflect the actual operating conditions of the charging pile.

[0117] Once the adjustment mechanism is triggered, the feedback mechanism will begin to work, automatically optimizing the simulated values ​​of the simulation parameters in the virtual simulation system. The core of this optimization process lies in adjusting key parameters in the simulation system through feedback, making the behavior of the virtual simulation system more closely resemble that of the actual charging pile system. For example, if there is a significant voltage or load deviation between the simulation results and the actual operating data, the system will automatically adjust the voltage, current, or load models in the simulation until the deviation between the simulation results and the actual data is reduced to an acceptable range.

[0118] Through this series of feedback and adjustments, the system will obtain a virtual-reality feedback model. This model can dynamically correct errors in the virtual simulation system, ensuring that the virtual simulation behaves consistently with the actual charging pile. The obtained simulation parameters will be continuously used to correct errors in the virtual system during subsequent simulations, enabling the system to achieve precise resource scheduling and operational optimization under different loads, environments, and user requirements.

[0119] Through this process, this implementation method effectively eliminates the gap between virtual simulation and actual charging pile operation status, improves the accuracy of charging pile resource scheduling and operation optimization, and ensures the reliability and practicality of the simulation system.

[0120] In this embodiment, step six specifically includes:

[0121] Resource demand information for each charging station is extracted from actual operational data and predicted data obtained from a virtual-reality feedback model. This information includes several key parameters such as charging load demand, power demand, charging duration, current idle status, and projected user demand. This data, sourced from a real-time monitoring system and a virtual simulation model, reflects the current and future resource demands of the charging stations. By integrating this information, the system can gain a comprehensive understanding of the load status of each charging station and provide data support for subsequent resource scheduling.

[0122] Based on the extracted resource demand information, the system sorts the charging piles according to preset priority rules. These priority rules consider multiple factors, including the charging load of the charging pile, battery charging demand, the urgency of the charging time period, and the availability status of the charging pile (e.g., idle or busy). These factors determine the resource allocation priority of the charging piles. For example, when multiple charging piles have high load demand during the same time period, the system can prioritize allocating resources to the charging piles with higher loads. Through priority sorting, the system can ensure that limited resources are allocated to the charging piles with the most urgent needs.

[0123] Based on the generated resource allocation priority list and the resource demand information of the charging piles, the system strategically allocates the available resources of the charging piles. This process rationally allocates power resources and the number of charging piles in the system based on the real-time demand and priority list of the charging piles. Resource allocation not only considers the immediate demand of the charging piles but also needs to predict future demand changes. Therefore, through a virtual-reality feedback mechanism, resource configuration can be dynamically adjusted to avoid situations of excessive or insufficient system resources. The result of this strategic allocation process is the charging pile resource scheduling scheme.

[0124] The charging pile resource scheduling scheme is an optimized resource allocation scheme that meets real-time requirements. It effectively solves the problem of uneven distribution or improper use of charging pile resources under high load and high demand conditions. Through this step, the charging pile system can automatically adjust resource allocation under different charging periods, different load conditions, and changes in user demand, improving the operating efficiency of the charging pile network and avoiding overload or resource idleness. This implementation method obtains real-time resource demand information from actual operating data and virtual-reality feedback models, sorts it according to priority rules, and then allocates charging pile resources through intelligent strategies, ultimately generating an optimized resource scheduling scheme, thereby improving the resource utilization, scheduling efficiency, and overall performance of the charging pile system.

[0125] In this embodiment, step seven specifically includes:

[0126] The virtual simulation model is initialized based on the charging pile resource scheduling scheme. This initialization process includes constructing the initial operating state of the charging piles according to their resource requirements, scheduling priorities, and load allocation. In this step, the operating state parameters of the charging piles (such as current, voltage, charging load, etc.) and resource configurations (such as the number of charging piles, available power, etc.) are set to initial values ​​according to actual needs. The virtual simulation model needs to accurately reflect the operating state of the charging piles under different conditions; therefore, this initialization step is crucial. By setting the initial operating state of the virtual simulation model, it is ensured that subsequent simulation results can accurately reflect the actual operating behavior of the charging pile system.

[0127] Based on the initialized virtual simulation model, the operating status of charging piles is simulated in real time through a dynamic simulation inference process. The core of this dynamic simulation inference process lies in simulating the operation of charging piles under different loads, environments, and user demands according to the resource scheduling scheme of the charging piles. In actual operation, the load of the charging piles, environmental changes (such as temperature changes and power fluctuations), and user demands change constantly, and these factors have a profound impact on the operating status of the charging piles. Through the virtual simulation model, the system can adjust the operating status of the charging piles in real time within the simulated environment and dynamically update the simulation results based on real-time input data, thereby predicting the operating trend of the charging piles in the future. The simulation results of the charging pile operating status output by this process provide an important basis for subsequent optimization of resource scheduling and decision-making.

[0128] Based on the simulation results of the charging pile's operating status, the system will undergo further optimization. In this stage, the system further optimizes the operating scheme by adjusting the charging pile resource allocation and scheduling strategies according to the set goal of minimizing energy consumption. The optimization objectives include reducing energy consumption, balancing the charging pile load, and improving system operating efficiency. Through the optimization algorithm in the simulation model, the system can rationally allocate resources and schedule charging piles while ensuring user needs are met, thereby maximizing the overall system performance and resource utilization efficiency. Ultimately, the optimized scheme will become the optimal operating scheme for the charging pile, ensuring the most efficient and stable operation of the charging pile under different charging loads and user demands. This implementation method, through dynamic simulation reasoning and real-time optimization mechanisms, can accurately predict the operating status of the charging pile and dynamically optimize resource scheduling accordingly, ensuring the efficiency and stability of the charging pile in actual operation and solving the problem that traditional charging pile resource scheduling methods cannot cope with real-time load fluctuations and changes in user demand.

[0129] Example 1:

[0130] To verify the feasibility of this invention in practice, it was applied to a smart charging pile network in a large city, deploying a digital twin-based virtual-real operation simulation verification system for charging piles. This system covers charging piles in multiple areas and can monitor and schedule their operational status in real time to ensure optimal resource allocation and utilization. In this case, the operation of charging piles is affected by multiple factors, including electricity demand, external environmental conditions (such as weather and temperature), and user charging needs, and these factors have complex spatiotemporal dependencies. Traditional charging pile scheduling systems often struggle to handle the dynamic changes of these factors, resulting in problems such as excessive resource waste and uneven charging pile load.

[0131] To optimize the resource scheduling and operational efficiency of charging piles, we adopted a digital twin-based system. This system addresses the shortcomings of traditional methods, such as insufficient real-time prediction and lagging resource scheduling, by performing time-series modeling, periodic modeling, and residual correction on real-time data from each charging pile. In practical application, the system uses a data acquisition and standardization module to collect real-time operational status data, current, voltage, ambient temperature, and user behavior data from each charging pile. After standardization, this data forms a standardized multimodal dataset for each charging pile. This data reflects key operational indicators such as load demand, power demand, and charging duration for each charging pile. Then, the system uses a causal relationship analysis module to perform causal relationship testing on this data, generating a spatiotemporal causal relationship graph for charging piles. This causal relationship graph helps us deeply understand the operational patterns of charging piles and the spatiotemporal dependencies between various factors, laying the foundation for subsequent time-series prediction and resource scheduling.

[0132] By aggregating local neighborhood information from the spatiotemporal causal relationship graph using a Grapher network, a charging embedding feature vector set was obtained. This feature vector set helps the system accurately capture the operating modes and trends of charging piles. Next, the system inputs these feature vectors into an improved DeepAR time-series prediction model. Through the time-series modeling module, periodic modeling module, and residual correction module in the model, the system successfully predicted the future load demand and power consumption of the charging piles. This prediction result was applied to a virtual-reality dynamic feedback mechanism, achieving synchronization and correction between virtual simulation and the actual operating state of the charging piles. To verify the effectiveness of this method, the system compared the charging pile operating effects before and after optimization. Before optimization, charging pile resource scheduling was based on simple rules, resulting in frequent overload during peak hours and idle resources during low-demand periods, leading to low system resource utilization efficiency. After optimization, the load and resource scheduling of the charging piles were dynamically adjusted, achieving balanced load during peak hours to avoid overload, and avoiding resource waste during low-demand periods.

[0133] In terms of specific data, the system monitored the operational status of the same batch of charging piles before and after optimization, and compared key indicators such as load balancing, resource utilization, and energy consumption. Through optimization, the load balancing rate of the charging piles increased from 65% to 88%, resource utilization improved by 20%, and energy consumption decreased by 10%. The following is a comparison table of data before and after optimization:

[0134] Table 1 Comparison of Charging Pile Operation Status Data

[0135] index Data before optimization Optimized data Improvement range Charging pile load balance 65% 88% +23% resource utilization rate 70% 90% +20% Energy consumption (kWh) 5000kWh 4500kWh -10% Number of overload events 15 times / day 2 times / day -13 times / day Percentage of charging station idle time 30% 10% -20%

[0136] As shown in Table 1, the load balancing of charging piles has significantly improved through system optimization, increasing from 65% to 88%. This indicates a more rational allocation of charging pile resources, with the load no longer concentrated on a few piles, thus avoiding resource waste. Resource utilization has also increased by 20%, demonstrating the system's ability to utilize charging pile resources more effectively. Regarding energy consumption, the optimized system consumes only 10% less energy, proving the effectiveness of the optimized scheduling scheme in reducing energy consumption. Furthermore, overload events at charging piles have decreased by 13 times per day, indicating that the system effectively prevents overload, avoiding equipment damage and user dissatisfaction. Most importantly, the idle time of charging piles has decreased by 20%, meaning that the utilization rate of charging piles has greatly improved after system optimization, and resources are not idle. This practical application fully demonstrates that the present invention can effectively improve the resource utilization efficiency and load balancing capability of charging piles, reduce energy consumption, and improve the overall performance and stability of the charging pile system.

[0137] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A digital twin-based simulation and verification system for the virtual and real operation of charging piles, characterized in that, Includes the following modules: The data acquisition and standardization module is used to collect and standardize the raw charging multimodal dataset from charging piles to obtain a standardized charging multimodal dataset. The causal relationship analysis module is used to perform causal relationship testing on the charging multimodal dataset, determine the spatiotemporal dependencies between various factors during the operation of the charging pile, and generate a charging spatiotemporal causal relationship graph. The information aggregation module is used to input the charging spatiotemporal causal relationship graph into the Graphormer network to perform local neighborhood information aggregation and obtain a charging embedding feature vector set. The time-series prediction module is used to input the charging embedded feature vector set into the improved DeepAR model, and perform time-series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status. The virtual-reality feedback model construction module is used to generate a virtual-reality feedback model based on the prediction result dataset of the charging pile operation status using a virtual-reality dynamic feedback mechanism. The resource scheduling generation module is used to allocate strategies for charging pile resource scheduling based on the virtual-reality feedback model, and obtain a charging pile resource scheduling scheme. The operation plan generation module is used to initialize a virtual simulation model based on the charging pile resource scheduling plan, and perform dynamic simulation reasoning and optimization to obtain an optimized charging pile operation plan.

2. The charging pile virtual-real operation simulation verification system based on digital twin as described in claim 1, characterized in that, The modules are connected in the following way: Step 1: Collect the original charging multimodal dataset from the charging piles, and standardize the original charging multimodal dataset to obtain a standardized charging multimodal dataset. Step 2: Perform causal relationship testing on the charging multimodal dataset to determine the spatiotemporal dependencies between various factors during the operation of the charging pile and generate a charging spatiotemporal causal relationship graph; Step 3: Input the charging spatiotemporal causal relationship graph into the Graphormer network to aggregate local neighborhood information and obtain the charging embedding feature vector set; Step 4: Input the charging embedded feature vector set into the improved DeepAR model, and perform time series prediction through the time series modeling module, periodic modeling module, residual correction module and prediction module to obtain the prediction result dataset of the charging pile operation status; Step 5: Based on the predicted data set of charging pile operating status, a virtual-reality dynamic feedback mechanism is used to generate a virtual-reality feedback model; Step 6: Allocate charging pile resources according to the virtual-reality feedback model to obtain a charging pile resource scheduling scheme; Step 7: Initialize the virtual simulation model based on the charging pile resource scheduling scheme, and perform dynamic simulation reasoning and optimization to obtain the optimized charging pile operation scheme.

3. The charging pile virtual-real operation simulation verification system based on digital twin as described in claim 2, characterized in that, The original charging multimodal dataset includes charging pile operation status data, charging pile external environment data, and user behavior data; the standardization processing steps include outlier removal, missing value imputation, timestamp alignment, and normalization processing of different types of data in the original charging multimodal dataset to obtain a standardized charging multimodal dataset.

4. The charging pile virtual-real operation simulation verification system based on digital twin as described in claim 2, characterized in that, Step two specifically involves: Time series of equal size are selected from the standardized charging multimodal dataset, and each selected time series is transformed into a stationary time series through differencing. The lag order of a stationary time series is determined using the Bayesian information criterion, where the lag order is the time difference between the current observation and previous observations. Set the null and alternative hypotheses, specifically including: The null hypothesis is that one time series has no causal relationship with another time series; The alternative hypothesis is that one time series has a causal relationship with another time series. The p-value is used to determine whether a previous time series can predict a subsequent time series given a lag number. If the p-value is less than a preset significance level, the null hypothesis is rejected, indicating that a causal relationship exists. Based on the results of causal relationship testing, the spatiotemporal dependencies between various factors during the operation of the charging pile are determined, and a spatiotemporal causal relationship diagram of the charging pile is constructed. The construction steps include: Each factor in the operation of the charging pile is treated as a node, resulting in a node set; The causal relationship test results are used to mark the causal relationship directions between different time series, thus obtaining a set of connection edges; By connecting different nodes in the node set according to the connection relationship in the edge set, a charging spatiotemporal causal relationship graph is obtained.

5. The charging pile virtual-real operation simulation verification system based on digital twin as described in claim 2, characterized in that, Step three specifically involves: The charging spatiotemporal causal relationship graph is input into the input layer of the Graphormer network. The input layer performs one-hot encoding on the nodes and edges in the charging spatiotemporal causal relationship graph, converting each node and edge into a numerical vector representation, to obtain the initial embedding vector set of the charging spatiotemporal causal relationship graph. By performing graph convolution operations, local neighborhood information aggregation is performed on the initial embedding vector of each node. The local neighborhood information aggregation is achieved by using the edge information in the charging spatiotemporal causal relationship graph to propagate the relationship information between adjacent nodes to the current node for weighted averaging, and updating the initial embedding vector of each node to obtain the graph convolution feature vector of the node. The graph convolutional feature vector is input into the temporal information fusion layer. Combined with the temporal information of the nodes, the temporal features and graph structure features are fused through the self-attention mechanism to obtain the temporal feature embedding vector of each node. The self-attention mechanism layer aggregates information by utilizing the causal relationships between nodes in the charging spatiotemporal causal relationship graph. The temporal feature embedding vector is input into the fully connected layer and then through the multilayer perceptron layer to generate the final charging embedding feature vector set. The charging embedding feature vector set contains the embedding representation of each node, representing the spatiotemporal causal relationship and dynamic characteristics of various factors in the charging pile.

6. The charging pile virtual-real operation simulation verification system based on digital twin as described in claim 2, characterized in that, The improved DeepAR model is specifically as follows: The charging embedded feature vector set is input into the time series modeling module in the order of time series, and the time series data of the charging pile is combined with the embedded features to create a time series model of the charging pile's operating status. The time series model uses the autoregressive properties of time series data to model the changing patterns of the charging pile's operating status by learning from historical data, and generates historical time series feature representations, which represent the past operating status of the charging pile. The historical time series features are input into the periodic modeling module. By establishing a periodic feature modeling mechanism, the periodic changes in the load and current variables of the charging pile over time can be captured. The extracted periodic variation features are added to the time series model to obtain a periodic feature vector, which includes periodic fluctuation information in the charging pile's operating status. The periodic feature vector is input into the residual correction module to calculate the difference between the historical predicted value and the actual observed value of the charging pile's operating status, thereby obtaining the prediction error for each time step. The historical predicted value is the mean value of the historical time series feature representation. For time series data with a prediction error greater than a preset error threshold, the future prediction value is adjusted by learning the error pattern through a support vector machine to obtain a corrected prediction feature vector. The adjustment involves adding the prediction error back to the original prediction value. In the prediction module, the historical time series feature representation from the time series modeling module, the periodic feature vector from the periodic modeling module, and the prediction feature vector from the residual correction module are combined and weighted according to preset weights to generate a dataset of prediction results for the charging pile operation status.

7. The charging pile virtual-real operation simulation verification system based on digital twin according to claim 2, characterized in that, Step five specifically involves: The actual operating data of the charging piles and the predicted results of the charging pile operating status are transmitted to the virtual simulation system for real-time synchronization. Calculate the deviation between the virtual simulation results and the actual charging pile operation data; When the deviation between the virtual simulation results and the actual operating status of the charging pile exceeds a preset threshold, an adjustment mechanism is triggered. When the deviation exceeds a preset threshold, the simulation value of the corresponding simulation parameter in the virtual system is automatically optimized through a feedback mechanism until the virtual simulation matches the actual behavior, thus obtaining a virtual-reality feedback model. The simulation parameters are used to correct the error in the virtual simulation system.

8. The charging pile virtual-real operation simulation verification system based on digital twin according to claim 2, characterized in that, Step six specifically involves: Resource demand information for each charging pile is extracted from the actual operation data of the charging pile and the predicted data of the charging pile obtained from the virtual-reality feedback model. The resource demand information includes the charging load demand, power demand, charging duration, current idle status of the charging pile and expected user demand. Based on the resource demand information, the charging piles are sorted according to the preset priority rules to generate a resource allocation priority list for each charging pile. Based on the resource allocation priority list and charging pile resource demand information, the available resources of the charging piles are strategically allocated to obtain a charging pile resource scheduling scheme.

9. A charging pile virtual-real operation simulation verification system based on digital twin as described in claim 2, characterized in that, Step seven specifically involves: The virtual simulation model is initialized according to the charging pile resource scheduling scheme. The initialization involves constructing the initial operating state of the charging pile based on its resource requirements, scheduling priority, and load allocation; and setting the charging pile operating parameters and resource configuration of the virtual simulation model in advance. The operating status of charging piles is simulated in real time through a dynamic simulation reasoning process based on a virtual simulation model. The dynamic simulation reasoning process involves simulating the operating status of charging piles under different loads, environments, and user demands based on the resource scheduling scheme of the charging piles. During the dynamic simulation reasoning process, the virtual simulation model dynamically adjusts the simulation results based on real-time input data, predicts future operating trends, and obtains the simulation results of the charging pile operating status. Based on the simulation results of the charging pile's operating status, the resource allocation and scheduling strategy is optimized according to the set minimum energy consumption, resulting in an optimized charging pile operation scheme.

Images