People-vehicle-pile station-network multi-source data fusion method, medium and system

Abstract

The invention discloses a man-vehicle-pile station-network multi-source data fusion method, medium and system, and the method comprises the steps: S1, constructing a time-varying window function based on the voltage of a vehicle BMS battery, the current of a charging pile and the voltage fluctuation data of a power grid, and applying vehicle-pile synchronization constraint; s2, inputting the time-varying window function into the frequency spectrum estimation model, solving an objective function, and outputting an aligned vehicle-pile-network frequency spectrum characteristic vector; s3, inputting the characteristic vector, the user behavior time sequence data, the vehicle SOC sequence signal, the charging pile power and the power grid load rate into a four-branch neural network, calculating a characteristic weight, and outputting a fusion characteristic; s4, calculating confidence coefficients of four sources of human, vehicle, pile and network; and S5, according to the confidence coefficient and the online state indication function, carrying out weighted fusion on the fusion features according to the weight, and outputting charging load prediction and power grid dispatching instructions in a layered manner. The method has the advantages of remarkably improving the multi-source data fusion precision, effectively supporting the intelligent decision of the charging network and the like.

Description

Technical Field

[0001] This invention mainly relates to the field of electric vehicle technology, specifically to a method, medium, and system for multi-source data fusion of people, vehicles, charging stations, and networks. Background Technology

[0002] In the context of large-scale electric vehicle application, multi-source data fusion involving people, vehicles, charging stations, and the grid is a key technology for achieving intelligent charging scheduling and grid-based collaborative optimization. Its main function is to integrate user behavior, vehicle status, charging station operation, and grid load data to accurately predict charging demand, effectively mitigate peak-valley differences in the grid, and improve the capacity for renewable energy absorption. Currently, mainstream fusion methods mainly include centralized Kalman filtering, static feature splicing models, and fixed-weight decision mechanisms, among which attention-based neural networks are widely used due to their feature selection capabilities. However, due to differences in the spatiotemporal characteristics of the data (such as the mismatch between vehicle mobility and fixed grid monitoring points) and modal heterogeneity (discrete behavioral data versus continuous waveform data), the fusion accuracy of existing methods is limited, resulting in charging load prediction errors often exceeding 15%, affecting the economic efficiency of grid operation and control.

[0003] Current multi-source data fusion architectures are mainly divided into two categories: centralized fusion and hierarchical fusion. Centralized fusion involves uploading all raw data to a cloud platform for unified processing. Its advantages include high theoretical completeness and the ability to establish a global optimization model. However, it suffers from low processing efficiency, making it difficult to meet the millisecond-level response requirements of vehicle-to-everything (V2X) networks, and its system robustness is poor when vehicles frequently go offline. Hierarchical fusion adopts an edge computing + cloud-collaborative architecture. Its characteristic is that vehicle-charging pile data preprocessing is completed at the charging pile side before being uploaded to the power grid dispatch center. It can adapt to high-concurrency scenarios, but the limited computing power of edge nodes leads to insufficient feature extraction, and the lack of cross-modal alignment mechanisms for human-vehicle-charging pile-network data results in significant semantic gaps between different levels of data.

[0004] In existing technologies, hierarchical fusion generally uses Temporal Convolutional Networks (TCNs) as the core algorithm. This method extracts local features based on sliding windows and improves representation capabilities by increasing network depth. However, this method does not adequately consider two key issues: first, the asynchronicity of vehicle-charging pile signals (the vehicle BMS sampling period of 100ms is not synchronized with the charging pile's 200ms period), leading to phase deviations in spectrum analysis; and second, the dynamic decay of data source confidence (e.g., rain and fog can cause a 30% distortion rate in charging pile voltage monitoring), yet traditional methods still assign fixed weights to these data sources. The multi-source data fusion method proposed in this invention, involving people, vehicles, charging piles (stations), and the network, solves the aforementioned spatiotemporal mismatch and quality fluctuation problems through time-varying window function synchronization constraints and a dynamic confidence allocation mechanism. This method has significant value for improving the operation of charging infrastructure and enhancing the efficiency of power grid dispatching. Summary of the Invention

[0005] To address the technical problems existing in the prior art, this invention provides a method, medium, and system for multi-source data fusion of people, vehicles, charging stations, and networks that significantly improves the accuracy of multi-source data fusion and effectively supports intelligent decision-making in charging networks.

[0006] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows: A method for fusing multi-source data from people, vehicles, charging stations, and networks includes the following steps: S1. Based on vehicle BMS battery voltage signal Charging pile current Grid voltage fluctuation V abc(t) Constructing a time-varying window function based on three types of high-frequency data. And apply vehicle-pile synchronization constraints; S2. Use the time-varying window function generated in step S1. Input the spectrum estimation model, solve the objective function, and output the aligned vehicle-pile-network spectrum characteristic vector F. sync ; S3. Convert the vector F output from step S2... sync User behavior time-series data, vehicle SOC sequence signals, charging pile power, and grid load rate are all input into a four-branch neural network. Feature weights are calculated through a spatiotemporal attention mechanism. The charging pile-grid branch uses an interaction layer to achieve feature alignment, while the user-vehicle branch assigns semantic associations through attention weights, outputting a fused feature F. fused ; S4. Calculate the confidence scores of the four sources: people, vehicles, piles, and networks; S5. Based on the confidence level of step S4 and online status indication function Dynamically allocate weights F in step S3 fused By weight Weighted fusion, hierarchical output of charging load forecasts and grid dispatch instructions.

[0007] Preferably, in step S1, the vehicle-pile synchronization constraint specifically includes: ; Where T is the sampling period (default 2 seconds); This is the threshold for vehicle-pile data synchronization.

[0008] Preferably, in step S2, the objective function is: ; in: The signal represents the vehicle's BMS battery voltage; K represents the total number of window functions; F{} represents the Fourier transform operator; f represents the frequency variable; and t represents the time variable. For the k-th window function; This represents the true value of the power grid load spectrum.

[0009] Preferably, in step S3, the four branches of the neural network are the human branch, the vehicle branch, the charging pile branch, and the network branch; wherein the input data corresponding to the human branch is the charging time preference Ut and the scheduled charging time Tchg; the input data corresponding to the vehicle branch is the battery SOC sequence SOC and the estimated driving range Dest; the input data corresponding to the charging pile branch is the charging power Ppile(t) and the current harmonic distortion rate THDI(t); and the input data corresponding to the network branch is the distribution transformer three-phase voltage Vabc(t) and the load rate ηload(t).

[0010] Preferably, the pile-network feature interaction layer Specifically:

[0011] in This is the 128-dimensional feature vector output by the stake branch; This is the 128-dimensional feature vector output by the network branch; is the trainable weight matrix; [;] is the vector concatenation operation; ReLU(⋅) is the modified linear unit activation function.

[0012] Preferably, confidence level The calculation formula is:

[0013] in: The confidence score for the i-th type of data source is given, where i=1,2,3,4 correspond to people, vehicles, charging piles, and networks, respectively. This represents the number of valid samples within the time window [t-Δt,t]. The theoretical total number of samples should be; The standard deviation of charging pile power; The timestamp of the latest update of the data source; λ is the time decay coefficient; To prevent the removal of the zero constant.

[0014] Preferably, in step S5, the fusion weights of the i-th type of data source are... The calculation formula is:

[0015] in This is a function indicating the online status.

[0016] Preferably, The specific formula is: .

[0017] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, the computer program performing the steps of the method described above when run by a processor.

[0018] The present invention further discloses a multi-source data fusion system for people, vehicles, charging stations, and networks, including a memory and a processor connected to each other. The memory stores a computer program, which executes the steps of the method described above when run by the processor.

[0019] Compared with the prior art, the advantages of the present invention are as follows: This invention presents a cascaded optimization-based multi-source data fusion method for vehicles, charging piles, and the power grid, aiming to address three major technical bottlenecks in existing multi-source fusion methods: signal asynchronicity, modal heterogeneity, and static weights. The invention constructs a four-level closed-loop optimization mechanism—spectrum synchronization, feature fusion, dynamic decision-making, and anti-interference training—to meticulously consider the real-time dynamic characteristics of the charging load. Furthermore, through the synergy of window function constraints, cross-modal interaction layers, dynamic confidence weighting, and domain migration mechanisms, it achieves adaptive compensation strategies under various operating conditions.

[0020] This invention optimizes the multi-window spectrum estimation model using time-varying weighting factors to reduce high-frequency noise interference. Secondly, it constructs a multi-branch neural network with a spatiotemporal attention mechanism to achieve cross-modal feature alignment of user behavior, vehicle status, charging pile operation, and power grid load data. Furthermore, it dynamically allocates data source weights based on the confidence matrix and uses a three-layer fusion architecture to process heterogeneous data in layers. This method significantly improves the accuracy of multi-source data fusion and effectively supports intelligent decision-making in charging networks.

[0021] This invention addresses three major shortcomings in existing human-vehicle-pile-grid data fusion methods: asynchronous vehicle-pile signals, modal semantic gap, and static weight allocation. It provides a multi-source data fusion method based on an improved multi-window interpolation algorithm and neural networks. By constructing a three-stage framework of time-varying weighted spectrum estimation, cross-modal attention fusion, and confidence-based dynamic decision-making, the method reduces charging load prediction errors by more than 40% and improves the cross-scenario adaptability of power grid dispatch commands. Attached Figure Description

[0022] Figure 1 This is a flowchart of an embodiment of the multi-source data fusion method for people-vehicle-pile station-network of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0024] like Figure 1 As shown, the multi-source data fusion method for people-vehicle-pile station-network provided in this embodiment of the invention includes the following steps: S1. Based on vehicle BMS battery voltage signal Charging pile current Grid voltage fluctuation V abc(t) Constructing a time-varying window function based on three types of high-frequency data. And apply vehicle-pile synchronization constraints: ; Where T is the sampling period (default 2 seconds); The threshold for vehicle-pile data synchronization is set to eliminate phase deviation caused by asynchronous vehicle-pile signals; S2. Use the time-varying window function generated in step S1. Input the spectrum estimation model, solve the objective function, and output the aligned vehicle-pile-network spectrum characteristic vector F. sync ; The objective function is: ; in: For vehicle BMS battery voltage signal; K: total number of window functions (K≥3); F{⋅}: Fourier transform operator; f: frequency variable (unit: Hz); t: time variable (unit: seconds); For the k-th window function; This represents the true value of the power grid load spectrum (calibrated via a SCADA system).

[0025] S3. Convert the vector F output from step S2... sync User behavior time-series data, vehicle SOC sequence signals, charging pile power, and grid load rate are all input into a four-branch neural network. Feature weights are calculated through a spatiotemporal attention mechanism. The charging pile-grid branch uses an interaction layer to achieve feature alignment, while the user-vehicle branch assigns semantic associations through attention weights, outputting a fused feature F. fused ; The input definitions for the four branches of a neural network are as follows:

[0026] The pile-network feature interaction layer specifically consists of:

[0027] in , The high-dimensional characteristics of the pile / network branches; This is the 128-dimensional feature vector output by the stake branch; This is the 128-dimensional feature vector output by the network branch; The training weight matrix (dimension 256×256); [;] represents vector concatenation operation; ReLU(⋅) is the modified linear unit activation function; S4. Dynamic confidence weighting Calculate the confidence scores of the four sources: people, vehicles, piles, and network.

[0028] in: Calculate the confidence score for the i-th type of data source (i=1,2,3,4 correspond to people, vehicles, charging piles, and networks, respectively). The number of valid samples within the time window [t-Δt,t] (Δt=5 minutes); The theoretical total number of samples should be; The standard deviation of charging pile power (1-minute window); λ is the latest update timestamp of the data source; λ is the time decay coefficient (default 0.02s). - ¹); To prevent division by zero constant; S5. Layered Integration Decision Based on the confidence level of step S4 and online status indication function Dynamically allocate weights F in step S3 fused By weight Weighted fusion, hierarchical output of charging load forecasts and grid dispatch instructions.

[0029] Where the fusion weight of the i-th type of data source The calculation formula is:

[0030] The online status indicator function has the following formula: .

[0031] This invention presents a cascaded optimization-based multi-source data fusion method for vehicles, charging piles, and the power grid, aiming to address three major technical bottlenecks in existing multi-source fusion methods: signal asynchronicity, modal heterogeneity, and static weights. The invention constructs a four-level closed-loop optimization mechanism—spectrum synchronization, feature fusion, dynamic decision-making, and anti-interference training—to meticulously consider the real-time dynamic characteristics of the charging load. Furthermore, through the synergy of window function constraints, cross-modal interaction layers, dynamic confidence weighting, and domain migration mechanisms, it achieves adaptive compensation strategies under various operating conditions.

[0032] This invention optimizes the multi-window spectrum estimation model using time-varying weighting factors to reduce high-frequency noise interference. Secondly, it constructs a multi-branch neural network with a spatiotemporal attention mechanism to achieve cross-modal feature alignment of user behavior, vehicle status, charging pile operation, and power grid load data. Furthermore, it dynamically allocates data source weights based on the confidence matrix and uses a three-layer fusion architecture to process heterogeneous data in layers. This method significantly improves the accuracy of multi-source data fusion and effectively supports intelligent decision-making in charging networks.

[0033] This invention addresses three major shortcomings in existing human-vehicle-pile-grid data fusion methods: asynchronous vehicle-pile signals, modal semantic gap, and static weight allocation. It provides a multi-source data fusion method based on an improved multi-window interpolation algorithm and neural networks. By constructing a three-stage framework of time-varying weighted spectrum estimation, cross-modal attention fusion, and confidence-based dynamic decision-making, the method reduces charging load prediction errors by more than 40% and improves the cross-scenario adaptability of power grid dispatch commands.

[0034] This invention also discloses a computer-readable storage medium storing a computer program thereon, which, when run by a processor, executes the steps of the method described above. This invention further discloses a multi-source data fusion system for people, vehicles, charging stations, and networks, including an interconnected memory and a processor, wherein the memory stores a computer program, which, when run by a processor, executes the steps of the method described above. The medium and system of this invention, corresponding to the methods described above, also possess the advantages described above.

[0035] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0036] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for fusing multi-source data from people, vehicles, charging stations, and networks, characterized in that, Including the following steps: S1. Based on vehicle BMS battery voltage signal Charging pile current Grid voltage fluctuation V abc(t) Constructing a time-varying window function based on three types of high-frequency data. And apply vehicle-pile synchronization constraints; S2. Use the time-varying window function generated in step S1. Input the spectrum estimation model, solve the objective function, and output the aligned vehicle-pile-network spectrum characteristic vector F. sync ; S3. Convert the vector F output from step S2... sync User behavior time-series data, vehicle SOC sequence signals, charging pile power, and grid load rate are all input into a four-branch neural network. Feature weights are calculated through a spatiotemporal attention mechanism. The charging pile-grid branch uses an interaction layer to achieve feature alignment, while the user-vehicle branch assigns semantic associations through attention weights, outputting a fused feature F. fused ; S4. Calculate the confidence scores of the four sources: people, vehicles, piles, and networks; S5. Based on the confidence level of step S4 and online status indication function Dynamically allocate weights F in step S3 fused By weight Weighted fusion, hierarchical output of charging load forecasts and grid dispatch instructions.

2. The multi-source data fusion method for people-vehicle-pile station-network according to claim 1, characterized in that, In step S1, the vehicle-pile synchronization constraint is specifically as follows: ； Where T is the sampling period; This is the threshold for vehicle-pile data synchronization.

3. The multi-source data fusion method for people-vehicle-pile station-network according to claim 1 or 2, characterized in that, In step S2, the objective function is: ； in: The signal represents the vehicle's BMS battery voltage; K represents the total number of window functions; F{} represents the Fourier transform operator; f represents the frequency variable; and t represents the time variable. For the k-th window function; This represents the true value of the power grid load spectrum.

4. The multi-source data fusion method for people-vehicle-pile station-network according to claim 1 or 2, characterized in that, In step S3, the four branches of the neural network are the human branch, the vehicle branch, the stake branch, and the network branch. The input data for the human branch are charging time preference Ut and scheduled charging duration Tchg; the input data for the vehicle branch are battery SOC sequence SOC and estimated driving range Dest; the input data for the charging pile branch are charging power Ppile(t) and current harmonic distortion rate THDI(t); and the input data for the grid branch are distribution transformer three-phase voltage Vabc(t) and load rate ηload(t).

5. The multi-source data fusion method for people-vehicle-pile station-network according to claim 4, characterized in that, Pile-Network Feature Interaction Layer Specifically: in This is the 128-dimensional feature vector output by the stake branch; This is the 128-dimensional feature vector output by the network branch; is the trainable weight matrix; [;] is the vector concatenation operation; ReLU(⋅) is the modified linear unit activation function.

6. The multi-source data fusion method for people-vehicle-pile station-network according to claim 1 or 2, characterized in that, Confidence The calculation formula is: in: The confidence score for the i-th type of data source is given, where i=1,2,3,4 correspond to people, vehicles, charging piles, and networks, respectively. This represents the number of valid samples within the time window [t-Δt,t]. The theoretical total number of samples should be; The standard deviation of charging pile power; The timestamp of the latest update of the data source; λ is the time decay coefficient; To prevent the removal of the zero constant.

7. The multi-source data fusion method for people-vehicle-pile station-network according to claim 1 or 2, characterized in that, In step S5, the fusion weights of the i-th type of data source are... The calculation formula is: in This is a function indicating the online status. Score the confidence level of the j-th type of data source.

8. The multi-source data fusion method for people-vehicle-pile station-network according to claim 7, characterized in that, The specific formula is: 。 9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-8.

10. A multi-source data fusion system for people, vehicles, charging stations, and networks, comprising interconnected memory and processor, wherein the memory stores a computer program, characterized in that... The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-8.

Images