A charging pile power prediction method and device, electronic equipment and storage medium

Abstract

The application provides a charging pile power prediction method and device, equipment and storage medium, and relates to the electric vehicle charging technology field. The method comprises the following steps: acquiring real-time charging power data of a to-be-detected charging pile; inputting the real-time charging power data into a pre-trained charging power prediction model to obtain a charging power prediction result of the to-be-detected charging pile; wherein the charging power prediction model is obtained by training a causal perception dynamic graph neural network based on historical charging power data samples of a charging network in which the charging pile is located. The charging power prediction model of the application is obtained based on the causal perception dynamic graph neural network, can clearly simulate the causal relationship and dynamics between key nodes in the charging network load prediction, and compared with the traditional charging power prediction method, the method provided in the application can generate accurate charging power prediction results, and provides a reliable basis for efficient management and resource optimization of the power grid.

Description

Technical Field

[0001] This application relates to the field of electric vehicle charging technology, specifically to a charging pile power prediction method, device, electronic equipment, and storage medium. Background Technology

[0002] The widespread adoption of electric vehicles (EVs) is rapidly transforming transportation systems; however, the increasing demand for EV charging poses significant challenges to grid stability and efficient power management. EV charging networks (hereinafter referred to as charging networks) play a crucial role in facilitating the transition to electric transportation and require accurate and robust load forecasting technologies to predict future electricity demand and ensure reliable and sustainable charging infrastructure.

[0003] Accurate charging network load forecasting is beneficial for grid operators in alleviating peak demand pressures and optimizing generation and distribution resources to effectively meet the charging needs of electric vehicles. In planning and investment decisions, reliable forecasting is a key input for infrastructure expansion and upgrades, ensuring sufficient capacity and resilience to support the growing electric vehicle fleet. Furthermore, charging network load forecasting forms the basis for providing consumers with personalized charging cost insights and encouraging behavioral adjustments to flatten the load curve, thereby improving grid efficiency and minimizing charging costs.

[0004] Most existing methods uniformly treat all nodes and their connections in a charging network during representation learning, neglecting two key aspects that limit their prediction accuracy. First, existing methods fail to distinguish between correlation and true causation, which can lead to less robust and interpretable predictions, especially as the network dynamics evolve. Second, existing methods fail to recognize that some charging network nodes may have a disproportionately larger impact on the overall network load behavior. Because the unique role of these influential "key" nodes is not explicitly identified and utilized during the prediction process, critical information may not be fully utilized, leading to performance degradation and insufficient information, making it difficult for traditional prediction methods to achieve accurate predictions. Summary of the Invention

[0005] The purpose of this application is to provide a charging pile power prediction method, device, electronic device, and storage medium, which can improve the accuracy and reliability of charging power prediction results.

[0006] In a first aspect, embodiments of this application provide a method for predicting the power of a charging pile, including:

[0007] Obtain real-time charging power data of the charging pile to be tested;

[0008] The real-time charging power data is input into a pre-trained charging power prediction model to obtain the charging power prediction result of the charging pile to be detected; wherein, the charging power prediction model is obtained by training a causal perception dynamic graph neural network with historical charging power data samples of the charging network where the charging pile is located.

[0009] In some embodiments of this application, the charging power prediction model is pre-trained in the following manner:

[0010] Historical charging power data of each charging pile in the charging network is obtained, and the historical charging power data is preprocessed to obtain a historical charging power data sample.

[0011] The historical charging power data samples are input into the causal perception dynamic graph neural network for training, and the charging power prediction model is obtained after the preset training cutoff condition is reached.

[0012] In some embodiments of this application, the causal-aware dynamic graph neural network includes:

[0013] The causal relationship graph learning module is used to construct a causal relationship graph between charging piles based on the historical charging power data samples, and to perform dynamic graph convolution operation on the causal relationship graph to obtain a causal relationship representation;

[0014] The key relationship graph learning module is used to identify influential key nodes in the charging network, construct a key relationship graph with key nodes as components, and perform dynamic graph convolution operation on the key relationship graph to obtain a key relationship representation; wherein, the node represents the charging pile in the charging network.

[0015] The time representation fusion module is used to fuse the causal relationship representation and the key relationship representation in time representation to obtain a refined time representation;

[0016] The output layer is used to generate prediction results based on the refining time representation.

[0017] In some embodiments of this application, the causal graph learning module is specifically used for:

[0018] The input historical charging power data samples are normalized to obtain normalized data samples.

[0019] Based on the greedy equivalence search framework for time segments of normalized data samples, causal relationships in the normalized data samples are discovered, and a causal relationship graph between charging piles is constructed based on the causal relationships.

[0020] The normalized data sample time segment greedy equivalence search framework is constructed based on the generalized scoring function and the traditional greedy equivalence search algorithm.

[0021] In some embodiments of this application, the key relationship graph learning module is specifically used for:

[0022] A centrality metric based on historical load dependencies is used to evaluate the centrality score of each node, and nodes whose centrality scores fall within a preset proportion of the highest ranking are considered key nodes.

[0023] In some embodiments of this application, the time representation fusion module is specifically used for:

[0024] The causal relationship representation and the key relationship representation are combined in a channel-wise manner to obtain a combined feature. Two temporal 1D convolutional layers are then applied to the combined feature to obtain a refined temporal representation.

[0025] In some embodiments of this application, the output layer includes a multilayer perceptron.

[0026] Secondly, embodiments of this application provide a charging pile power prediction device, comprising:

[0027] The acquisition module is used to acquire real-time charging power data of the charging pile to be tested;

[0028] The prediction module is used to input the real-time charging power data into a pre-trained charging power prediction model to obtain the charging power prediction result of the charging pile to be detected; wherein, the charging power prediction model is obtained by training a causal perception dynamic graph neural network with historical charging power data samples of the charging network where the charging pile is located.

[0029] Thirdly, this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in the first aspect.

[0030] Fourthly, this application provides a computer-readable storage medium having computer-readable instructions stored thereon, which can be executed by a processor to implement the method described in the first aspect.

[0031] Compared to existing technologies, the charging pile power prediction method provided in this application obtains real-time charging power data of the charging pile to be tested; inputs the real-time charging power data into a pre-trained charging power prediction model to obtain the charging power prediction result of the charging pile to be tested; wherein, the charging power prediction model is obtained by training a causal-aware dynamic graph neural network using historical charging power data samples of the charging network in which the charging pile is located. The charging power prediction model of this application, based on a causal-aware dynamic graph neural network, can clearly simulate the causal relationships and dynamics between key nodes in charging network load prediction. Compared with traditional charging power prediction methods, the method proposed in this application can generate accurate charging power prediction results, providing a reliable basis for efficient power grid management and resource optimization. Attached Figure Description

[0032] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0033] Figure 1 A flowchart of a charging pile power prediction method provided in an embodiment of this application is shown;

[0034] Figure 2 A schematic diagram illustrating the prediction process of a charging power prediction model provided in an embodiment of this application is shown.

[0035] Figure 3 A schematic diagram illustrating the prediction process of another charging power prediction model provided in an embodiment of this application is shown.

[0036] Figure 4 A schematic diagram of a charging pile power prediction device provided in an embodiment of this application is shown;

[0037] Figure 5 A schematic diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0038] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0039] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0040] Furthermore, the terms "first" and "second," etc., are used to distinguish different objects, not to describe a specific order. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to those processes, methods, products, or apparatuses.

[0041] Please refer to Figure 1 , Figure 1 This is a flowchart of a charging pile power prediction method provided in an embodiment of this application, including the following steps S101 to S102:

[0042] S101. Obtain the real-time charging power data of the charging pile to be tested.

[0043] S102. Input the real-time charging power data into a pre-trained charging power prediction model to obtain the charging power prediction result of the charging pile to be detected; wherein, the charging power prediction model is obtained by training a causal perception dynamic graph neural network with historical charging power data samples of the charging network where the charging pile is located.

[0044] The following describes how to pre-train a charging power prediction model. Specifically, the pre-training process includes the following steps:

[0045] S1. Obtain historical charging power data of each charging pile in the charging network, and after preprocessing the historical charging power data, obtain historical charging power data samples.

[0046] Historical charging power data refers to the historical charging power data of each charging pile in the charging network. The preprocessing refers to cleaning the historical charging power data to remove incomplete, non-standard, and inaccurate data.

[0047] S2. Input the historical charging power data samples into the causal perception dynamic graph neural network for training, and obtain the charging power prediction model after reaching the preset training cutoff condition.

[0048] The preset training cutoff condition can be that the number of training cycles reaches a set number, or the prediction error is less than a preset threshold.

[0049] Specifically, the aforementioned causal perception dynamic graph neural network includes: a causal relationship graph learning module, a key relationship graph learning module, a time representation fusion module, and an output layer.

[0050] The causal relationship graph learning module is used to construct a causal relationship graph between charging piles based on the historical charging power data samples, and to perform dynamic graph convolution operation on the causal relationship graph to obtain a causal relationship representation;

[0051] The causal relationship graph learning module constructs a causal relationship graph between charging piles based on the historical charging power data samples. Specifically, this includes: normalizing the input historical charging power data samples to obtain normalized data samples; discovering causal relationships in the normalized data samples based on a greedy equivalence search framework for time segments of the normalized data samples; and constructing a causal relationship graph between charging piles based on the causal relationships. The greedy equivalence search framework for time segments of the normalized data samples is constructed based on a generalized scoring function and a traditional greedy equivalence search algorithm.

[0052] The key relationship graph learning module is used to identify influential key nodes in the charging network, construct a key relationship graph with key nodes as components, and perform dynamic graph convolution operation on the key relationship graph to obtain a key relationship representation; wherein, the node represents the charging pile in the charging network.

[0053] The key relationship graph learning module identifies influential key nodes in the charging network, specifically by: using a centrality metric based on historical load dependencies to evaluate the centrality score of each node, and considering nodes whose centrality scores fall within a preset proportion of higher rankings as key nodes.

[0054] The time representation fusion module is used to fuse the causal relationship representation and the key relationship representation in time representation to obtain a refined time representation;

[0055] The time representation fusion module is specifically used to: combine the causal relationship representation and the key relationship representation through channel-wise to obtain a combined feature, and apply two time 1D convolutional layers to the combined feature to obtain a refined time representation.

[0056] The output layer is used to generate prediction results based on the refining time representation.

[0057] The charging pile power prediction method provided in this application introduces a Causal Aware Dynamic Graph Neural Network (CADGN), which can explicitly simulate the causal relationships and dynamics between key nodes in charging network load prediction. CADGN utilizes two key modules: a Causal Aware Graph Learning Module (CAGLM) and a Key Relationship Graph Learning Module (CRGLM). CAGLM infers a causal relationship graph from historical load data, representing the causal relationships between charging piles, and performs dynamic graph convolution on this causal graph to learn causal dependency representations. CRGLM focuses on modeling the complex dynamics between the most influential charging piles in the network, representing their interconnectivity by identifying "key" nodes and constructing a dynamic key relationship graph. By dynamically capturing and representing these relationships, CRGLM extracts key information that traditional graph-based models often overlook. The representations learned in these modules are then fused and processed through a Temporal Representation Fusion Module (TRFM) and the output layer, thereby improving prediction accuracy, especially in short- and medium-term predictions.

[0058] Please refer to Figure 2 The following section provides a detailed introduction to the charging power prediction model.

[0059] a) Constructing a causal graph using the Causal Graph Learning Module (CAGLM): By constructing a causal graph reflecting causal relationships, and then performing causal dynamic graph learning, key causal representations between charging stations are extracted. First, the input historical charging power data samples of the Electric Vehicle Charging Network (EVCS) are normalized. A Time Segment Greedy Equivalence Search (TFGES) framework is designed using the Generalized Score Function (GSF) and the traditional greedy equivalence search algorithm, focusing on the discovery of causal relationships within shorter periods, thus enabling the handling of larger datasets and more complex causal structures.

[0060] Step a) above specifically includes the following 7 steps:

[0061] a1. Construct a causal graph reflecting these causal relationships, and perform causal dynamic graph learning based on this causal structure to extract meaningful representations. CAGLM first normalizes the input EVCS historical charging power data samples to ensure that all features contribute equally to the model training process and improve convergence during optimization. The mathematical expression for the normalization operation is:

[0062]

[0063] in, Represents a normalized data matrix. N represents the average value of X over time. T X represents the number of historical time steps in the review window. (t)This represents the load vector at time step t, where t represents the review window N. T The t-th time step in the time series.

[0064] a2. Subsequently, based on the Generalized Score Function (GSF) and the traditional Greedy Equivalence Search algorithm, a Temporal Segment Greedy Equivalence Search (TFGES) framework was designed for causal discovery. The TFGES framework focuses on normalized data segments. Where t m ∈[1,t e ), t b ,t e ∈[1,N T ] represent the start and end times of the time segment, respectively, and tx <t e By analyzing shorter periods, such as N... T Within a historical time step range of [12,72] hours, the TFGES framework can identify causal relationships that may be hidden in longer and more complex time series, enabling it to handle larger datasets and more complex causal structures.

[0065] a3. To balance model complexity, candidate maps are evaluated using GSF based on the fit between the candidate map and the observed data. This method utilizes RKHS (Reproducing Kernel Hilbert Space) to capture dependencies in a nonparametric manner, thus enabling it to handle nonlinear causal relationships and diverse data distributions. Cross-validation is used within the TFGES framework, by... Divided into N Q Training and test sets are used to mitigate overfitting. For a given vertex V... ca 、E ca causal directed acyclic graph Cross-validation is defined as follows:

[0066]

[0067] in, This represents the entire dataset, partitioned for cross-validation. This represents the regression function learned from the q-th training set. This represents the number obtained on the q-th test set. This is the q-th test set in the cross-validation process, where F represents the charging power prediction mapping function, and N... Q Denotes the number of cross-validation folds, q∈[1,N] Q ] represents the q-th cross-validation fold, N S This indicates the number of electric vehicle charging stations. Indicates the target graph and input data The cross-validation scoring function.

[0068] a4. For each node v in the charging network (i) The regression function in RKHS is expressed as:

[0069]

[0070] In the formula, v (i) Let Z represent the i-th individual variable of the charging network. (i) For the image Chinese v (i) The set of parent nodes. Function F tf (·) indicates that RKHS H v Parent node Z within (i) Mapping to v (i) The true underlying function, U (i) This represents the noise or error term. The maximum log-likelihood of this regression is given by the following equation:

[0071]

[0072] Among them, K X and K Z It is the kernel matrix representing the pairwise similarity between time steps in the feature space, λ is the regularization parameter, and t m Indicates the time step. Let I denote the maximum log-likelihood scoring function. Let I denote the identity matrix. The optimization problem is formulated as finding the DAG that minimizes the combined objective. :

[0073]

[0074] Where, N s yes The number of variables (or nodes) In a given data segment The model fit of node i is evaluated under the condition of parameter Φ. For calculation It provides the foundation because It aggregates the log-likelihood scores from multiple cross-validation folds. Φ contains the specific parameter configurations and optimization criteria of the TFGES framework, and Ω is the regularization graph. Complexity penalty.

[0075] a5. In establishing the cause-effect graph structure Subsequently, CAGLM uses Stacked Graph Convolutional Networks (SGCN), whose architecture is similar to... The encoded inferences about causality remain consistent. Traditional GCN layers employ propagation rules:

[0076] Guided by the adjacency structure of the graph, information is aggregated from adjacent nodes:

[0077] H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) W (l) );

[0078] in, Let d be the node feature matrix of all EVCS nodes in layer 1. l Let W be the number of features in the l-th layer of the GCN; D is the corresponding angle matrix, ensuring proper normalization during message passing; W is the number of features in the l-th layer of the GCN. (l) Let represent the trainable weights associated with the l-th layer, and σ(·) represent the hyperbolic tangent (tanh) activation function.

[0079] To enhance the model's generalization ability, CAGLM constructs a temporarily masked data representation X. m , as the initial input feature H (0) To start. This representation is achieved by applying an application with a shielding ratio R. d Random masking operator M ca (R d This is created to help models learn robust representations that are less sensitive to noise and missing information in the data.

[0080] a6. Implementing GCN in CAGLM includes batch normalization (BN) and dropout layers distributed between convolutional layers. Below is a detailed description of SGCN using a two-layer GCN architecture in CAGLM:

[0081]

[0082] in, d represents the intermediate feature representation generated at the output of the i-th SGCN block in CAGLM. g The feature dimension of the last layer GCN output. and The trainable parameters of the first and second GCN layers are encapsulated in the i-th SGCN block, respectively, emphasizing that each layer and each block learns a different set of convolutional filters. Furthermore, the operator *G ca This indicates the graph convolution operation performed, using A ca As a guiding adjacency structure. Finally, D(η) represents the application of a probability η to prevent overfitting and encourage the sparsity of the learned representations.

[0083] To further enrich the feature representation, CAGLM includes an additional component focused on capturing short-term local time patterns from the raw charging power data. A multilayer perceptron (MLP) directly and linearly processes the masked data segments. Generate output representation This representation captures the shallow and local temporal dynamics that occur throughout the entire period. Intermediate representation obtained from GCN. Carrying global causal information and local temporal representation Adding these two elements produces This is a more comprehensive concatenated feature representation. This fusion process enriches the expressiveness of the learned representation and helps Causal Aware Dynamic Graph Neural Networks (CADGN) to model complex interactions in charging networks.

[0084] a7. Finally, to refine these feature maps, two additional standard one-dimensional (temporal) convolutional layers are applied consecutively. These layers process... Generate H (o) A concise representation:

[0085]

[0086] in, and This represents two temporal convolutional operations, each parameterized with learnable kernel weights and combined with non-linear activations to model complex patterns within the temporal dimension of the input features. By stacking these convolutional layers, CAGLM performs hierarchical feature extraction, where higher-level features are extracted from... This captures more abstract and complex patterns from the fused causal and temporal information. From a CAGLM perspective, this multi-level causal representation... As a refined summary of load dynamics, it captures the causal structure of the network and local time trends within the input data segments.

[0087] b) The Critical Relationship Graph Learning Module (CRGLM) performs dynamic learning of key node relationships: identifying and recognizing key nodes with significant influence in the charging network. A centrality score for each node is evaluated using a centrality metric based on historical load dependencies. Then, a dynamic critical relationship graph is used to represent the relationships between these key nodes, further enhancing prediction accuracy by dynamically capturing and representing these relationships.

[0088] Step b) above specifically includes the following 5 steps:

[0089] b1. Identify a subset of key nodes (influential charging stations) and then dynamically learn from a graph representing the relationships between them. To identify key nodes, a centrality metric based on historical load dependencies is used. First, pairwise relationship scores are calculated. (i,j) Indicating v in the charging network (i) With v (j) The degree of influence between them is calculated using the following formula:

[0090]

[0091] In the formula, X (i,k) Let t be the input feature vector of charging pile i at time step k; s This represents a time shift to explain the underlying lag relationship; N T w represents the historical time window considered in calculating the score. (i,j) As learnable weights, the terms allow the model to dynamically emphasize or de-emphasize specific pairwise charging pile relationships during training.

[0092] b2. The CRGLM module calculates the centrality score s(i) for each charging station i. This score aggregates the relationship scores of the charging station with all other charging stations in the network, aiming to identify charging stations with significant overall influence.

[0093]

[0094] b3, CRGLM uses a threshold parameter R k A Top-K selection mechanism, guided by centrality scores, identifies the most influential charging stations. This process formally defines the set of key nodes C. cr as follows:

[0095] C cr =Top-K(S,R) k )={i|rank(s (i) S)≤R k};

[0096] In the formula, rank(s) i S) represents the centrality score of charging pile i across all centrality scores s. i Rank in a set This model only considers centrality scores that fall on the highest R-squared. k Charging stations within the specified proportion are considered key nodes, thus focusing attention on the most important interactions.

[0097] b4. These identified key nodes become the constituent elements of the key relationship graph. (The graph...) It inherits the same set of nodes as the complete charging network, namely V cr =V. However, the adjacency matrix A cr Unlike a complete network diagram, A cr Specifically designed to represent the interactions between key charging stations. The time-dependent adjacency tensor A at time step t. cr Each entry within The definition is as follows:

[0098]

[0099] Here, i and j represent different charging stations. Essentially, A cr A dynamic graph structure is encoded, in which a network of connections is established between key nodes and other nodes. The time-varying nature of this adjacency structure enables CRGLM to capture the evolving influence between these key nodes throughout the charging power prediction process.

[0100] b5. Identify key nodes and form an initial graph. First, CRGLM utilizes dynamic graphs to learn structure and cross-N cr The SGCN layer is used to process the masked data X. m Each layer generates an intermediate representation. Specifically, it captures patterns around key nodes. The adjacency matrix A of the key graph is used in the convolution operation. cr Similar to CAGLM, The representations are aggregated and then concatenated with individual, locally concentrated temporal representations. The fused feature representations of CRGLM are further refined through a two-layer convolution process similar to CAGLM, resulting in the key relation representations of the final output of CRGLM.

[0101] c) The Temporal Representation Fusion Module (TRFM) performs temporal feature fusion: The TRFM module acts as a bridge, fusing the temporal patterns extracted from the Causal Relationship Graph Learning Module (CAGLM) and the Key Relationship Graph Learning Module (CRGLM) to generate the final prediction output. This TRFM module can stitch together causal relationships and the dynamic relationship information of key nodes, thereby effectively improving the accuracy of prediction.

[0102] Step c) above specifically includes the following steps:

[0103] The TRFM module acts as a bridge, receiving outputs from both the CAGLM and CRGLM modules. and First, through channel-wise and Combined. Then, two temporal 1D convolutional layers are applied to the combined features to produce a temporally rich output representation:

[0104]

[0105] in, This represents the refining time representation of the final output of the TRFM module; and The TRFM module represents consecutive temporal convolution operations within a module, each associated with its own set of learnable parameters. ψ(·) represents the rectified linear unit (ReLU) activation function, adding complexity and nonlinearity. By handling cascaded representations in this way, the TRFM module aims to discover... and The complex interactions between the temporal patterns present in the data ultimately provide richer information and a stronger foundation for the downstream output layer to generate prediction results.

[0106] d) Output layer generates prediction results: The refined time representation from the TRFM module is then used. The results are fed into the output layer to generate a prediction of the charging power of the charging station. The input is obtained by... With time step t b up to time step t e The raw, pre-processed (unshielded) charging power data of charging piles is concatenated to form the output. Finally, the input to this concatenation is processed by an MLP, which consists of a series of fully connected layers that learn to map the fused spatiotemporal representation into a target load prediction, effectively capturing the various relationships learned in the modules of CADGN. Mathematically, this output generation process can be represented as:

[0107]

[0108] in, Indicates from t e +1 to t e +N P The time step is used to predict the final charging pile load output by the causal perception dynamic graph neural network.

[0109] The performance of the causal-aware dynamic graphical neural network during training was evaluated using the mean squared error (MSE) loss function, which compares the generated predictions. With the corresponding fundamental truth value To calculate:

[0110]

[0111] in, This represents the loss of the model parameters θ, which includes parameters from all component modules, such as SGCN layers, traditional convolutional layers, MLPs, and other learnable components. It is minimized through an optimization algorithm. The driving model generates charging pile load predictions.

[0112] During model training, the model parameters are adjusted based on the comparison results of the loss function, and the optimal charging power prediction model is obtained after the preset training cutoff condition is met.

[0113] After obtaining the charging power prediction model, the real-time charging power data of the charging pile to be tested can be input into the charging power prediction model to predict the charging power of the charging pile in the future within a set time period. Finally, the prediction results are sent to the grid operator.

[0114] like Figure 3 As shown, the charging pile load data (i.e., charging power data) is acquired in real time, the data is preprocessed, and then model training and real-time prediction are performed. The prediction results are sent to the grid operator, who can adjust the power supply of the charging network. This provides a reliable basis for the efficient management and resource optimization of the grid, and effectively improves the stability and management efficiency of the grid.

[0115] The charging pile power prediction method provided in this application has the following beneficial effects:

[0116] a) Improve the robustness and explanatory power of forecasts.

[0117] The CADGN framework infers and represents causal relationships between charging stations, rather than just correlations, through the Causal Graph Learning Module (CAGLM). This allows the model to capture true causal dependencies. The Key Relationship Graph Learning Module (CRGLM) focuses on modeling the dynamic relationships between the most influential key nodes in the network. These nodes have a significant impact on the overall load behavior of the network. By dynamically capturing and representing these relationships, the model can utilize important information that traditional graph models overlook, thereby improving the robustness and interpretability of predictions.

[0118] b) Better prediction performance

[0119] CADGN uses the Temporal Representation Fusion Module (TRFM) to fuse temporal patterns extracted from the Causal Relationship Graph Learning Module (CAGLM) and the Key Relationship Graph Learning Module (CRGLM). This module helps integrate information on causal relationships and key node relationships, further improving prediction accuracy.

[0120] In the above embodiments, a charging pile power prediction method is provided. Correspondingly, this application also provides a charging pile power prediction device. The charging pile power prediction device provided in this application can implement the above-described charging pile power prediction method. This charging pile power prediction device can be implemented through software, hardware, or a combination of both. For example, the charging pile power prediction device may include integrated or separate functional modules or units to perform the corresponding steps in the above methods. Please refer to... Figure 4 As shown, the charging pile power prediction device 10 includes:

[0121] The acquisition module 101 is used to acquire the real-time charging power data of the charging pile to be tested;

[0122] The prediction module 102 is used to input the real-time charging power data into a pre-trained charging power prediction model to obtain the charging power prediction result of the charging pile to be detected; wherein, the charging power prediction model is obtained by training a causal perception dynamic graph neural network with historical charging power data samples of the charging network where the charging pile is located.

[0123] In some embodiments of this application, the apparatus further includes a training module for pre-training the charging power prediction model in the following manner:

[0124] Historical charging power data of each charging pile in the charging network is obtained, and the historical charging power data is preprocessed to obtain a historical charging power data sample.

[0125] The historical charging power data samples are input into the causal perception dynamic graph neural network for training, and the charging power prediction model is obtained after the preset training cutoff condition is reached.

[0126] In some embodiments of this application, the causal-aware dynamic graph neural network includes:

[0127] The causal relationship graph learning module is used to construct a causal relationship graph between charging piles based on the historical charging power data samples, and to perform dynamic graph convolution operation on the causal relationship graph to obtain a causal relationship representation;

[0128] The key relationship graph learning module is used to identify influential key nodes in the charging network, construct a key relationship graph with key nodes as components, and perform dynamic graph convolution operation on the key relationship graph to obtain a key relationship representation; wherein, the node represents the charging pile in the charging network.

[0129] The time representation fusion module is used to fuse the causal relationship representation and the key relationship representation in time representation to obtain a refined time representation;

[0130] The output layer is used to generate prediction results based on the refining time representation.

[0131] In some embodiments of this application, the causal graph learning module is specifically used for:

[0132] The input historical charging power data samples are normalized to obtain normalized data samples.

[0133] Based on the greedy equivalence search framework for time segments of normalized data samples, causal relationships in the normalized data samples are discovered, and a causal relationship graph between charging piles is constructed based on the causal relationships.

[0134] The normalized data sample time segment greedy equivalence search framework is constructed based on the generalized scoring function and the traditional greedy equivalence search algorithm.

[0135] In some embodiments of this application, the key relationship graph learning module is specifically used for:

[0136] A centrality metric based on historical load dependencies is used to evaluate the centrality score of each node, and nodes whose centrality scores fall within a preset proportion of the highest ranking are considered key nodes.

[0137] In some embodiments of this application, the time representation fusion module is specifically used for:

[0138] The causal relationship representation and the key relationship representation are combined in a channel-wise manner to obtain a combined feature. Two temporal 1D convolutional layers are then applied to the combined feature to obtain a refined temporal representation.

[0139] In some embodiments of this application, the output layer includes a multilayer perceptron.

[0140] This application also provides an electronic device corresponding to the method provided in the foregoing embodiments. The electronic device may be an electronic device for a client, such as a mobile phone, laptop, tablet computer, desktop computer, etc., to execute the above-described charging pile power prediction method.

[0141] Please refer to Figure 5 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 5 As shown, the electronic device 20 includes: a processor 200, a memory 201, a bus 202, and a communication interface 203. The processor 200, the communication interface 203, and the memory 201 are connected via the bus 202. The memory 201 stores a computer program that can run on the processor 200. When the processor 200 runs the computer program, it executes the charging pile power prediction method provided in any of the foregoing embodiments of this application.

[0142] The memory 201 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 203 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0143] Bus 202 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 201 is used to store programs. After receiving an execution instruction, the processor 200 executes the program. The charging pile power prediction method disclosed in any of the foregoing embodiments of this application can be applied to the processor 200, or implemented by the processor 200.

[0144] The processor 200 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 200 or by instructions in software form. The processor 200 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 201. The processor 200 reads the information in memory 201 and, in conjunction with its hardware, completes the steps of the above method.

[0145] The electronic device provided in this application embodiment and the charging pile power prediction method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0146] This application also provides a computer-readable storage medium corresponding to the charging pile power prediction method provided in the foregoing embodiments, which stores a computer program (i.e., a program product) thereon. When the computer program is run by a processor, it executes the charging pile power prediction method provided in any of the foregoing embodiments.

[0147] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0148] The computer-readable storage medium provided in the above embodiments of this application and the charging pile power prediction method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0149] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application.

Claims

1. A charging pile power prediction method, characterized in that, The method comprises the following steps: obtaining real-time charging power data of a charging pile to be detected; inputting the real-time charging power data into a pre-trained charging power prediction model to obtain a charging power prediction result of the charging pile to be detected; wherein the charging power prediction model is obtained by training a causal-aware dynamic graph neural network based on historical charging power data samples of a charging network in which the charging pile is located; the causal-aware dynamic graph neural network comprises: a causal relationship graph learning module, configured to construct a causal relationship graph between charging piles according to the historical charging power data samples, and perform a dynamic graph convolution operation on the causal relationship graph to obtain a causal relationship representation; a key relationship graph learning module, configured to identify influential key nodes in the charging network, construct a key relationship graph with the key nodes as constituent elements, and perform a dynamic graph convolution operation on the key relationship graph to obtain a key relationship representation; wherein the nodes represent charging piles in the charging network; a time representation fusion module, configured to fuse the causal relationship representation and the key relationship representation to obtain a refined time representation; an output layer, configured to generate a prediction result based on the refined time representation; the causal relationship graph learning module is specifically configured to: normalize the input historical charging power data samples to obtain normalized data samples; discover causal relationships in the normalized data samples based on a normalized data sample time segment greedy equivalence search framework, and construct a causal relationship graph between charging piles according to the causal relationships; wherein the normalized data sample time segment greedy equivalence search framework is constructed based on a generalized score function and a traditional greedy equivalence search algorithm; the key relationship graph learning module is specifically configured to: use a centrality measurement based on historical load dependency relationships to evaluate the centrality scores of each node, and regard nodes with centrality scores falling within a preset proportion of a high ranking as key nodes; the time representation fusion module is specifically configured to: combine the causal relationship representation and the key relationship representation through channel-wise to obtain a combined feature, and apply two time 1-dimensional convolution layers to the combined feature to obtain a refined time representation.

2. The method of claim 1, wherein, The charging power prediction model is pre-trained in the following manner: obtain historical charging power data of each charging pile in the charging network, and obtain historical charging power data samples after preprocessing the historical charging power data; input the historical charging power data samples into the causal-aware dynamic graph neural network for training, and obtain the charging power prediction model after a preset training cutoff condition is reached.

3. The method of claim 1, wherein, The output layer comprises a multilayer perceptron.

4. A charging pile power prediction device, characterized in that, The method comprises the following steps: an obtaining module, configured to obtain real-time charging power data of a charging pile to be detected; a prediction module, configured to input the real-time charging power data into a pre-trained charging power prediction model to obtain a charging power prediction result of the charging pile to be detected; wherein the charging power prediction model is obtained by training a causal-aware dynamic graph neural network based on historical charging power data samples of a charging network in which the charging pile is located; The causal-aware dynamic graph neural network comprises: a causal relationship graph learning module configured to construct a causal relationship graph between charging piles according to the historical charging power data samples, and perform a dynamic graph convolution operation on the causal relationship graph to obtain a causal relationship representation; a key relationship graph learning module configured to identify influential key nodes in a charging network, construct a key relationship graph with the key nodes as constituent elements, and perform a dynamic graph convolution operation on the key relationship graph to obtain a key relationship representation; wherein the nodes represent charging piles in the charging network; a time representation fusion module configured to fuse the causal relationship representation and the key relationship representation to obtain a refined time representation; an output layer configured to generate a prediction result according to the refined time representation; the causal relationship graph learning module is specifically configured to: normalize the input historical charging power data samples to obtain normalized data samples; discover causal relationships in the normalized data samples based on a normalized data sample time segment greedy equivalence search framework, and construct a causal relationship graph between charging piles according to the causal relationships; wherein the normalized data sample time segment greedy equivalence search framework is constructed based on a generalized score function and a traditional greedy equivalence search algorithm; the key relationship graph learning module is specifically configured to: use a centrality measure based on historical load dependency relationships to evaluate the centrality scores of each node, and regard nodes with centrality scores falling within a preset proportion of a higher ranking as key nodes; the time representation fusion module is specifically configured to: combine the causal relationship representation and the key relationship representation through channel-wise to obtain combined features, and apply two time 1-dimensional convolution layers to the combined features to obtain a refined time representation.

5. An electronic device comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 3 when executing the computer program.

6. A computer-readable storage medium, characterized in that, A computer readable instruction is stored thereon, and the computer readable instruction can be executed by a processor to implement the method of any one of claims 1 to 3.

Images