Node electricity price determination method and device, storage medium and electronic equipment

Abstract

The invention discloses a node electricity price determination method and device, a storage medium and electronic equipment, and relates to the field of power energy transmission and supply, and the node electricity price determination method comprises the steps: obtaining a target data set, each piece of data in the target data set is used for representing electricity price description information of each power grid node in a power grid system of the target region in a historical period; the average electricity price of the target region is obtained by inputting the target data set into the autoregressive cycle network; the target data set is input into a graph convolutional neural network, the blocking electricity price of each power grid node is obtained, and the blocking electricity price represents extra cost caused by power grid transmission capacity limitation; and determining the node electricity price of each power grid node based on the average electricity price and the predicted blocking electricity price of each power grid node. According to the invention, the technical problem of low electricity price prediction accuracy of the power grid node in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the technical field of electric energy transmission and supply, and in particular to a method and device for determining a node electricity price, a storage medium and an electronic device. Background Art

[0002] Electricity is an indispensable factor in promoting national economic development and improving people's living standards and quality. However, unlike general commodities, electricity cannot be stored in large quantities. Therefore, it is usually necessary to set an electricity price that can ensure a balance between income and expenditure under the lowest cost operating goal in order to meet electricity demand at different times.

[0003] In related technologies, due to the differences in living standards in different regions, it is usually more reasonable for a region to adopt a unified electricity price. Usually, it is done by collecting privacy information such as market supply and demand relations, market structure and market rules, and calculating safety constraint unit rules, safety constraint economic dispatch formulas and other related formulas based on market supply and demand boundary data to analyze the next day's power grid operating conditions and system constraints, and solving the market equilibrium point, that is, the average price point of system energy, based on the predicted power generation quotation curve and the predicted value of the system.

[0004] By adopting the above method, the voltage of each room, the current and power of each branch and other parameters can be determined based on the report of each grid node, so as to evaluate the operation status and safety of the grid system and calculate the price caused by the blockage of each node.

[0005] However, the above method is only applicable to the node electricity price forecast of the power grid company. Since the power distribution company only has the historical electricity price data unique to each company, there is not enough data to support the simulation of the entire network. Therefore, it can only generate its own trading experience based on the changes in the market environment and the corresponding historical price trends, resulting in the technical problem of low accuracy of the predicted node electricity price.

[0006] With regard to the problem of low accuracy in predicting node electricity prices in related technologies, no effective solution has been proposed so far.

[0007] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0008] The embodiments of the present invention provide a method and device for determining a node electricity price, a storage medium, and an electronic device, so as to at least solve the problem of low accuracy of the predicted node electricity price.

[0009] According to one aspect of an embodiment of the present invention, there is provided a method for determining a node electricity price, comprising: acquiring a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the power grid system of a target area within a historical period; obtaining the average electricity price of the target area by inputting the target data set into an autoregressive cyclic network; obtaining the blocking electricity price of each grid node by inputting the target data set into a graph convolutional neural network, wherein the blocking electricity price represents the additional cost caused by the limitation of the transmission capacity of the power grid; and determining the node electricity price of each grid node based on the average electricity price and the predicted blocking electricity price of each grid node.

[0010] In an exemplary embodiment, the above-mentioned method of obtaining the average electricity price in the target area by inputting the target data set into the autoregressive recurrent network includes: converting each data in the target data set into time series data to obtain the time series data set, wherein one time series data in the time series data set is used to describe the electricity price characteristic information of one grid node among each grid node at T time points, and T is a positive integer greater than or equal to 2; based on the time series data set, training a pre-created autoregressive recurrent network to obtain a trained target autoregressive recurrent network, wherein the input of a current hidden layer in the autoregressive recurrent network includes the output result of the autoregressive recurrent network at the previous time point and the electricity price characteristics of each grid node at the current time point; obtaining the target probability distribution by adding the likelihood function in the target autoregressive network, wherein the likelihood function is used to describe the value range of the average electricity price in the target area when the input of the current hidden layer is given; determining the average electricity price in the target area based on the target probability distribution and the value range.

[0011] In an exemplary embodiment, the above-mentioned determination of the average electricity price in the target area based on the target probability distribution and the value range includes: based on the gradient descent method, determining the best fitting data line from the target probability distribution, wherein the best fitting data line has the highest probability in the time series data set; based on the best fitting data line, determining the average electricity price in the target area from the value range, wherein the average electricity price in the target area includes T sub-electricity prices at T time points.

[0012] In an exemplary embodiment, the above-mentioned method of obtaining the congestion electricity price of each grid node by inputting the target data set into a graph convolutional neural network includes: constructing a graph convolutional neural network based on the target data set and the pre-acquired electricity price description information of the forecast day, wherein the graph convolutional neural network includes M graph nodes, and one graph node among the M graph nodes is used to represent a grid node in each grid node, and M is a positive integer greater than or equal to 2; by performing a convolution operation on the M graph nodes, the node features of the M graph nodes are updated to obtain updated node features, wherein the node features are used to represent the description information of the additional electricity price of a grid node caused by the limitation of the grid transmission capacity; by inputting the updated node features into a multi-layer perceptron, the congestion electricity price of each grid node is obtained.

[0013] In an exemplary embodiment, the above-mentioned convolution operation is performed on M graph nodes to update the node features of the M graph nodes to obtain the updated node features, including: constructing an adjacency matrix based on the weights of the edges between any two connected graph nodes among the M graph nodes; using the weighted average method to aggregate the node features of the M graph nodes to obtain the aggregated node features; and determining the aggregated node features as the updated node features.

[0014] In an exemplary embodiment, the above-mentioned weighted average method is used to aggregate the node features of M graph nodes to obtain the aggregated node features, including: obtaining one graph node from the M graph nodes in turn as the current graph node; obtaining the neighbor features of Y graph nodes adjacent to and connected to the current graph node, where Y is a positive integer greater than or equal to 1 and less than M; based on the adjacency matrix and the weighted average method, the neighbor features and the current node features of the current graph node are aggregated to obtain the aggregated current node features.

[0015] In an exemplary embodiment, the node electricity price of each grid node is determined based on the average electricity price and the predicted congestion electricity price of each grid node, including: obtaining a grid node from each grid node in turn as the current grid node; obtaining the current predicted congestion electricity price of the current grid node from the predicted congestion electricity prices of each grid node; and summing the average electricity price, the current predicted congestion electricity price and the network loss electricity price to obtain the node electricity price of the current grid node, wherein the network loss electricity price represents the electricity cost generated due to line loss during power transmission.

[0016] According to another aspect of an embodiment of the present invention, there is also provided a device for determining a node electricity price, comprising: a first acquisition unit, for acquiring a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the power grid system of the target area within a historical period; a first processing unit, for obtaining the average electricity price of the target area by inputting the target data set into an autoregressive cyclic network; a second processing unit, for obtaining the blocking electricity price of each grid node by inputting the target data set into a graph convolutional neural network, wherein the blocking electricity price represents the additional cost caused by the limitation of the transmission capacity of the power grid; and a third processing unit, for determining the node electricity price of each grid node based on the average electricity price and the predicted blocking electricity price of each grid node.

[0017] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for determining the node electricity price when running.

[0018] According to another aspect of an embodiment of the present invention, there is provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method for determining the node electricity price through the computer program.

[0019] Through the above-mentioned embodiments provided by the present application, the average electricity price of the target area is predicted using an autoregressive recurrent network, and the congestion electricity price of each node in the power grid system of the target area is predicted using a graph convolutional neural network, and finally the node electricity price of each power grid node is determined based on the predicted average electricity price and the congestion electricity price of the node. This method of predicting node electricity prices based on artificial intelligence can not only analyze spatial information, but also effectively capture the relationship and dynamic changes between power grid nodes, thereby avoiding dependence on large amounts of data, reducing computing resources, and achieving the technical effect of improving the accuracy of the prediction results of node electricity prices. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0022] Figure 1It is a hardware structure block diagram of a mobile terminal of an optional method for determining node electricity price according to an embodiment of the present invention;

[0023] Figure 2 is a flow chart of an optional method for determining a node electricity price according to an embodiment of the present invention;

[0024] Figure 3 is an overall flow chart of an optional method for determining node electricity price according to an embodiment of the present invention;

[0025] Figure 4 is a schematic diagram of an optional data processing process for node electricity price prediction according to an embodiment of the present invention;

[0026] Figure 5 is a schematic diagram of a module in an optional DeepAR model according to an embodiment of the present invention;

[0027] Figure 6 is a schematic diagram of an optional process of predicting the congestion price of each grid node by using the GCN model according to an embodiment of the present invention;

[0028] Figure 7 is a structural block diagram of an optional device for determining node electricity price according to an embodiment of the present invention;

[0029] Figure 8 It is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for determining a node electricity price according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor (Central Processing Unit, MCU) or a programmable logic device (Field Programmable Gate Array, FPGA)) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0033] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the node electricity price in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0034] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0035] In this embodiment, a method for determining a node electricity price is provided. Figure 2 is a flow chart of a method for determining a node electricity price according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps S202-S208:

[0036] Step S202: Acquire a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the grid system of the target area within a historical period;

[0037] The target data set includes, but is not limited to, obtaining the weather of the target area from the published weather forecast data, obtaining the historical average electricity price of the target area in the historical period from the power grid disclosure, coal price, historical west-to-east power transmission data, historical load, etc. In addition, holiday information can also be added.

[0038] Step S204: obtaining the average electricity price of the target area by inputting the target data set into the autoregressive loop network;

[0039] Step S206: Obtaining the blocking price of each grid node by inputting the target data set into the graph convolutional neural network, wherein the blocking price represents the additional cost caused by the transmission capacity limitation of the grid;

[0040] Step S208: Determine the node electricity price of each grid node based on the average electricity price and the predicted congestion electricity price of each grid node.

[0041] In the embodiment of the present application, the autoregressive recurrent network may be, but is not limited to, an autoregressive recurrent neural network (DeepAR model) based on deep learning, which is used for probabilistic time series prediction. The model can generate probabilistic predictions for time series data, that is, predict the value and its uncertainty at a certain point in the future. The core of the DeepAR model is that it does not predict a certain value, but predicts a probability distribution, which enables the model to provide information about the uncertainty of the prediction.

[0042] Among them, the main features of the DeepAR model include:

[0043] (1) Autoregressive property: The DeepAR model uses an autoregressive mechanism, which means that the model takes into account past values ​​when predicting future values. This autoregressive property enables the model to capture long-term dependencies in time series data;

[0044] (2) Recurrent Neural Network: The model uses a recurrent neural network (RNN), especially a long short-term memory network (LSTM), to process sequence data, enabling it to handle complex patterns and seasonal changes in time series data;

[0045] (3) Probabilistic prediction: DeepAR models not only output a predicted value, but also a probability distribution, usually including mean and standard deviation, which allows users to understand the uncertainty of the prediction;

[0046] (4) Monte Carlo sampling: During the prediction process, the DeepAR model uses Monte Carlo sampling to generate multiple possible future paths, thus providing more information about the distribution of future values;

[0047] (5) Processing multiple time series: The DeepAR model can process multiple related time series, which enables the model to learn the correlation characteristics between different time series and improve the accuracy of the overall prediction.

[0048] Graph Convolutional Networks (GCN) is a deep learning model specifically designed to process graph structured data. It extracts the features of nodes in the graph through convolution operations and uses the structural information of the graph for learning (which can also be understood as spatial information).

[0049] The data processed by GCN is a graph, which consists of nodes and edges. Nodes represent entities or objects, and edges represent the relationship between nodes. Each node has a feature vector representing the attributes of the node, which can be numerical, categorical, etc.

[0050] The entities or objects represented by the graph nodes may be, but are not limited to, grid nodes in the power grid system of the target area. According to the categories, grid nodes include load nodes, power generation nodes, and transmission nodes. The grid nodes may be, but are not limited to, used for transmitting electric energy and controlling the flow of electric energy. Grid nodes may be engines, transformers, links of transmission lines, or any other electrical equipment in the power system.

[0051] It should be noted that the nodes of the GCN model in the embodiments of the present application can also be understood as graph nodes, or nodes for short.

[0052] By using the above DeepAR model, we can conduct a more comprehensive study of the electricity price description information of the target area in the historical period, predict a probability distribution of the average electricity price of the target area, and determine the average electricity price of the target area in the future from the probability distribution.

[0053] Using the above GCN model, we can effectively capture the relationship and dynamic changes between nodes, so as to determine the node characteristics of each node at a future time point, that is, the electricity price description characteristics of each grid node, so as to accurately calculate the congestion price of each grid node.

[0054] The average electricity price of the above-mentioned target area may be, but is not limited to, the predicted average electricity price of the target area on the next day. The average electricity price refers to the average price of electricity transactions in the electricity market within a certain period of time. It may be calculated based on factors such as supply and demand, power generation costs and market rules; congestion price refers to the additional cost caused by power grid transmission capacity limitations or bottlenecks. When a part of the power grid reaches its maximum transmission capacity, the flow of additional electricity may be blocked, resulting in congestion price.

[0055] In the above way, the average electricity price of the target area is predicted by using an autoregressive recurrent network, and the congestion electricity price of each node in the power grid system of the target area is predicted by using a graph convolutional neural network. Finally, the node electricity price of each power grid node is determined based on the predicted average electricity price and the node congestion electricity price. This method of predicting node electricity prices based on artificial intelligence can not only analyze spatial information, but also effectively capture the relationship and dynamic changes between power grid nodes, thereby avoiding dependence on large amounts of data, reducing computing resources, and achieving the technical effect of improving the accuracy of the prediction results of node electricity prices.

[0056] In an optional example, the above method of obtaining the average electricity price of the target area by inputting the target data set into the autoregressive loop network includes:

[0057] Convert each data in the target data set into time series data to obtain a time series data set, wherein one time series data in the time series data set is used to describe the electricity price characteristic information of one grid node in each grid node at T time points, where T is a positive integer greater than or equal to 2;

[0058] Based on the time series data set, a pre-created autoregressive recurrent network is trained to obtain a trained target autoregressive recurrent network, wherein the input of a current hidden layer in the autoregressive recurrent network includes the output result of the autoregressive recurrent network at the previous time point and the electricity price characteristics of each grid node at the current time point;

[0059] By adding a likelihood function to the target autoregressive network, the target probability distribution is obtained, where the likelihood function is used to describe the range of the average electricity price in the target area when the input of the current hidden layer is given;

[0060] Based on the target probability distribution and value range, the average electricity price in the target area is determined.

[0061] The method of converting each data in the target data set into time series data may be, but is not limited to, adjusting all data into a format of 96 points per day to facilitate model learning. The so-called 96-point format per day is to sample or aggregate data into a sequence of 96 time points per day. This format is usually used to represent data recorded every 15 minutes in a day, because there are 24 hours in a day and 4 time points per hour (24×4=96). This conversion method can help users observe and analyze trends and patterns of data within a day.

[0062] The specific process of data processing can be found in Figure 4 As shown in the figure, after obtaining the initial data, the interpolation method is used to fill in the missing data, divide the data, and expand the data according to the similar day method. After the data processing is completed, the processed data is formatted, that is, according to the daily trading time, the data is converted into a format of 96 points a day, and the converted data is used as the input of the DeepAR model and the GCN model.

[0063] For example, if you continuously monitor the real-time data of electricity prices, you can obtain data in a 96-point format for one day by following the steps below:

[0064] (1) Sampling: If the data is recorded continuously, it may be necessary to sample the raw data every 15 minutes to ensure that there are exactly 96 data points in a day;

[0065] (2) Aggregation: If the data is recorded at intervals other than every 15 minutes, it may be necessary to aggregate the data in shorter time intervals (such as finding the average, maximum, or minimum value) to generate data points every 15 minutes;

[0066] (3) Time alignment: Ensure that the data for all days starts and ends at the same time point, for example, starting from midnight of each day and ending at midnight of the next day;

[0067] (4) Missing value processing: If there is no data at certain time points, these missing values ​​may need to be filled in. The value of the previous time point, the value of the next time point, or the interpolation method can be used.

[0068] The missing parts are filled in using interpolation, and the data are further divided according to the dry season and the flood season. Finally, the amount of data information is increased by the similar day method, which is to find a few days with high time correlation (the previous 1-3 days and the previous 7 days) and high feature correlation (the most similar weather and load), and incorporate their features and corresponding electricity prices into the data set.

[0069] After obtaining the time series data, construct an autoregressive recurrent network structure, that is, Figure 5 The hidden layer shown can be understood as the calculation output part of the autoregressive recurrent network h i，t0-1 , and then pass this result as part of the input to the next hidden layer, and also input the result z of the entire model output at the previous moment i，t-1 (or the actual result of the previous moment), the independent variable x of the next moment i，t The update process of the hidden layer is specifically expressed by the following formula (1):

[0070]

[0071] in, It is a given parameter representing the LSTM cell state used in the autoregressive recurrent network, h() represents an autoregressive recurrent network, and h is the output result of the hidden layer of the autoregressive recurrent network.

[0072] During the training process, since all data are known, z i，t-1 The actual result of the previous moment is used, while in the prediction part after the model training, the result of the previous moment needs to be used.

[0073] It can be seen that in the autoregressive recurrent network, the input of each hidden layer includes the result of the DeepAR model output at the previous time point, the input data at the next time point (referred to as the next moment), and the intermediate result of the output of the previous hidden layer, which realizes the layer-by-layer transmission of the electricity price feature information of each node in the network, and uses the low-level features learned in the previous layer as the input of the next hidden layer, so as to extract more advanced features.

[0074] After the training is completed, the target autoregressive recurrent network is obtained, and then the likelihood function is added to the target autoregressive network Make the output prediction value a probability distribution. Generally, you can choose Gaussian distribution (suitable for real continuous distribution data) or negative binomial distribution (suitable for integer data). θ() is the neural network model being trained.

[0075] Among them, the likelihood function can be but is not limited to using Gaussian distribution, because the electricity price prediction result is a price, so it is usually a continuously changing number, or at least approximated to have 3 decimal places. The formula of Gaussian distribution is as follows:

[0076]

[0077]

[0078] Among them, μ is the mathematical expectation, σ 2 is the variance, w is the weight obtained in the likelihood function training, and b is the error term.

[0079] Since there is no negative variance, in order to ensure that the variance is positive, a softplus activation function is added here, which can be regarded as a smoothed version of the forced relu non-negative correction function, making the variance change continuous and non-negative.

[0080] Through the above method, the overall formula of the DeepAR model processing process is obtained, as shown in the following formula (3):

[0081]

[0082] The first equal sign is to write the original joint probability distribution as the probability product form of autoregression. The second equal sign is to express the autoregression result in the form of probability using the likelihood function.

[0083] and It has also been explained in the above embodiment, where h() is the autoregressive recurrent network; the likelihood function This has also been explained in the above-mentioned embodiments.

[0084] It should be noted that the above method of using the DeepAR model to calculate the average electricity price in the target area is only an example and is not limited thereto. It can also be replaced by other methods according to actual conditions.

[0085] By using the time series data in the above embodiment to train the DeepAR model, the prediction result output by the model is the day-ahead average electricity price obtained by predicting the electricity supply and demand in the target area, referred to as the average electricity price in the target area.

[0086] In an exemplary embodiment, the above-mentioned determination of the average electricity price of the target area based on the target probability distribution and the value range includes:

[0087] Based on the gradient descent method, the best fitting data line is determined from the target probability distribution, where the best fitting data line has the highest probability on the time series data set;

[0088] Based on the best fitting data line, the average electricity price of the target area is determined from the value range, wherein the average electricity price of the target area includes T sub-electricity prices at T time points.

[0089] It should be noted that the result output by the target autoregressive recurrent network is the value range or interval range of the average electricity price of the power grid system in the target area at 96 points on the next day, including multiple data including upper and lower limits. Therefore, it is necessary to use the likelihood function to determine the average electricity price of the power grid system in the target area on the next day from the value range.

[0090] Among them, the likelihood function is used to describe the "possibility" of observing a certain data sample under given parameters. In the embodiment of the present application, the maximum likelihood estimation method is used to maximize the likelihood function to find the most "appropriate" parameter value to determine the best fitting data line from multiple data of the target probability distribution.

[0091] Then, based on the best fitting data line, the average electricity price of the target area is determined from the value range.

[0092] In combination with the description in the above embodiments, it can be seen that in the process of constructing and training the autoregressive recurrent network, the data acquisition preprocessing process includes an interpolation method for filling in data missings, thereby ensuring the integrity of the data and the accuracy of the prediction results, and utilizing the average electricity price of the power grid system in the target area output by the self-regressive recurrent network to improve the accuracy of the average electricity price.

[0093] In an exemplary embodiment, the above method of obtaining the congestion price of each grid node by inputting the target data set into the graph convolutional neural network includes:

[0094] Based on the target data set and the pre-acquired electricity price description information of the forecast day, a graph convolutional neural network is constructed, wherein the graph convolutional neural network includes M graph nodes, one graph node among the M graph nodes is used to represent one grid node among each grid node, and M is a positive integer greater than or equal to 2;

[0095] By performing a convolution operation on the M graph nodes, the node features of the M graph nodes are updated to obtain updated node features, wherein the node features are used to represent the description information of the additional electricity price of a grid node due to the limitation of the grid transmission capacity;

[0096] By inputting the updated node features into the multi-layer perceptron, the congestion price of each grid node is obtained.

[0097] First, refer to Figure 6 The construction process of the graph convolutional neural network (GCN) is briefly introduced. Specifically, in the graph structure of GCN, each grid node (including power generation and power consumption) is a node of the graph structure, and the circuit connection between the nodes is the edge of the graph structure.

[0098] In the process of building the GCN model, the transmission capacity is used as the feature of the edge; the node electricity price and the congestion electricity price are used as the features of the node; the coal price, bidding controls, disclosed load, west-to-east power transmission, the overall electricity price of the target area and holidays are used as the features of the entire graph structure.

[0099] Because in the past time period, the day-ahead transaction was once every 15 minutes, therefore, the output result of the previous 15-minute time point is obtained every 15 minutes every day, and it is used as the input of the current 15-minute time point, and so on. The GCN model is built and used to predict the congestion electricity price of each grid node.

[0100] Secondly, in the process of building the GCN model, it is necessary to build a node feature matrix, an edge connection relationship matrix, an adjacency matrix, and full graph information, etc. Then, the node features of the graph nodes in the GCN model are updated in the form of convolution, and finally, weight changes are added after each convolution layer to complete the construction of the GCN model. The above processes will be explained in conjunction with specific embodiments.

[0101] The following are explanations of several matrices used in the process of building the GCN model:

[0102] (1) Node feature matrix, the shape is [number of nodes, number of node features], X i Represents the features of node i; edge feature matrix, the shape is [number of edges, number of edge features];

[0103] (2) A matrix representing edge connection relationships, with a shape of [2, number of edges] (the first row represents the start node index of the edge, and the second row represents the end node index of the edge);

[0104] (3) Node position matrix, the shape is [number of nodes, 2] (only the two-dimensional position information of the geographic location is used, and the height is not considered).

[0105] In order to aggregate node feature information, geographic location information can be directly attached to the node features.

[0106] By performing convolution operations on the graph nodes in the GCN model, the node features of M graph nodes are updated, and then the updated node features are input into the MLP (multi-layer perceptron) to obtain the congestion price of each grid node.

[0107] As an optional implementation method, the above method updates the node features of the M graph nodes by performing a convolution operation on the M graph nodes to obtain updated node features, including:

[0108] Based on the weights of the edges between any two connected graph nodes among the M graph nodes, an adjacency matrix is ​​constructed;

[0109] Using the weighted average method, the node features of M graph nodes are aggregated to obtain the node features after aggregation;

[0110] The node features after the aggregation process are determined as updated node features.

[0111] First, the following symbols are defined for the graph nodes in the GCN model: N = 1, 2, 3, ... n, representing the number of all graph nodes (referred to as nodes); X∈R d Represents the characteristics of all nodes, X i Represents the characteristics of the i-th node; A∈R N×N represents the adjacency matrix, A ij Represents the weight of the edge between node i and node j.

[0112] If there is no edge between nodes i and j, then A ij =0, if yes, the weight of the edge between node i and node j is determined by the following formula (4):

[0113]

[0114] It should be noted that, in the absence of a self-loop, A ii =0.

[0115] Among them, the adjacency matrix is ​​an important data structure used to represent the relationship between nodes in the graph structure. It is a square matrix whose elements represent the connection between nodes. For an undirected graph, the adjacency matrix is ​​symmetrical, that is, if node i is connected to node j, then the elements in the i-th row and j-th column and the j-th row and i-th column in the adjacency matrix are both 1 (or weight value); for a directed graph, the adjacency matrix is ​​not necessarily symmetrical.

[0116] In GCN, the adjacency matrix A is usually used in conjunction with the node feature matrix X to update the feature representation of the node through a graph convolution operation.

[0117] As an optional example, the above method updates the node features of the M graph nodes by performing a convolution operation on the M graph nodes to obtain updated node features, including:

[0118] Based on the weights of the edges between any two connected graph nodes among the M graph nodes, an adjacency matrix is ​​constructed;

[0119] Using the weighted average method, the node features of M graph nodes are aggregated to obtain the node features after aggregation;

[0120] The node features after the aggregation process are determined as updated node features.

[0121] Among them, the method of constructing the adjacency matrix can refer to the description in the above embodiment. After constructing the adjacency matrix, the node features of M graph nodes are aggregated by weighted averaging method, and the aggregated node features are determined as updated node features.

[0122] It should be noted that in the graph convolutional neural network, by aggregating the graph nodes, each graph node can aggregate its own features and the features of its neighboring nodes, so that the graph convolutional neural network can effectively process graph structured data and achieve excellent performance in various graph-related tasks.

[0123] Aggregation processing enables the model to learn embedded representations of graph nodes, which can be used for various tasks such as node classification, graph classification, and prediction.

[0124] As an optional implementation method, the weighted average method is used to aggregate the node features of the M graph nodes to obtain the aggregated node features, including:

[0125] Obtain one graph node from the M graph nodes in turn as the current graph node;

[0126] Get the neighbor features of Y graph nodes that are adjacent to and connected to the current graph node, where Y is a positive integer greater than or equal to 1 and less than M;

[0127] Based on the adjacency matrix and the weighted average method, the neighbor features and the current node features of the current graph node are aggregated to obtain the aggregated current node features.

[0128] Specifically, after creating the adjacency matrix, the node features of the i-th node after aggregating the neighbor node information are obtained through the weighted average method: As shown in the following formula (5):

[0129]

[0130] From this, we can get the following formula (6):

[0131] X * =AX (6)

[0132] While aggregating neighbor information, it is also necessary to add a self-loop to combine the own features with the neighbor features to update the node. Otherwise, the features of the own node will be ignored, that is, formula (6) becomes the following formula (7):

[0133] X * =(A+I)X (7)

[0134] Where I is the identity matrix, and the new adjacency matrix is ​​set to the following formula (8):

[0135]

[0136] In order to unify the problem that the feature values ​​of nodes with many edges after aggregation are much larger than those of nodes with few edges due to the different numbers and weights of connected edges, the node features of the nodes are normalized before updating themselves.

[0137] In the embodiment of the present application, a diagonal matrix D is also set. The information contained in the diagonal matrix is ​​the degree of each node. The node is a matrix composed of multiple weights, wherein one weight among the multiple weights represents the influence degree of a neighbor node on the current node. The diagonal matrix can be represented by, but not limited to, the following formula (9):

[0138]

[0139] right After symmetric normalization, the following formula (10) is obtained:

[0140]

[0141] After completing the above steps, add the node linear transformation matrix W and activation function σ to the node after the neighbor node information is aggregated (which can also be understood as the aggregation node) to construct a complete trainable convolution layer. W is used as a weight matrix to transform the node features after the geographic information is aggregated to a specified dimension, and σ can be a function similar to ReLU (set to ReLU in the embodiment of the present application). The convolution transformation process of the node in each layer is expressed as the following formula (11):

[0142]

[0143] Among them, l represents the lth convolutional layer.

[0144] Compared with the traditional node electricity price calculation which relies on complex grid structure and power flow analysis methods, which not only require a large amount of historical data support but also involve cumbersome calculation processes, the technical solution in the embodiment of the present application can effectively capture the relationship and dynamic changes between power grid nodes by introducing a graph convolutional neural network, avoiding dependence on a large amount of data, thereby analyzing the spatial information of each power grid node in the power grid system of the target area and obtaining the congestion situation of each power grid node.

[0145] Secondly, compared with the complex simulation calculations and high-performance computing resource requirements of power flow calculations, the embodiments of the present application complete the prediction of the day-ahead node electricity price in a shorter time cost under relatively low computing resource requirements.

[0146] In an exemplary embodiment, the node electricity price of each grid node is determined based on the average electricity price and the predicted congestion electricity price of each grid node, including:

[0147] Obtain one grid node from each grid node in turn as the current grid node;

[0148] Obtaining a current predicted congestion price of a current grid node from predicted congestion prices of each grid node;

[0149] The average electricity price, the current predicted congestion electricity price and the network loss electricity price are summed to obtain the node electricity price of the current power grid node, where the network loss electricity price represents the electricity cost caused by line loss during power transmission.

[0150] It is easy to understand that the calculation of node electricity prices in the power grid system can be achieved by, but not limited to, the following formula (12):

[0151] Node marginal price (LMP) = system energy price + congestion price + network loss price (12)

[0152] Among them, the node marginal price can be understood as the node price of the power grid node. The node electricity price reflects the actual cost of electricity at a specific node, including the production cost, transmission cost and potential congestion cost of electricity; the system energy price can be understood as the average electricity price in the above-mentioned target area, which refers to the average price of electricity transactions in the power market within a certain period of time. It may be calculated based on factors such as supply and demand, power generation costs and duration rules; the congestion price refers to the fact that when a part of the power grid reaches its maximum transmission capacity, the flow of additional electricity may be blocked, resulting in a congestion price.

[0153] The network loss price refers to the cost incurred due to line loss during the transmission of electricity. There will be a certain proportion of energy loss during the long-distance transmission of electricity, and this part of the loss needs to be included in the electricity price.

[0154] In actual operations, the calculation of node electricity prices may involve complex market rules and algorithms. To ensure the fairness and efficiency of the electricity market, grid operators and market regulators usually formulate corresponding rules to manage and adjust node electricity prices.

[0155] For ease of calculation, in this embodiment, the network loss electricity price is usually regarded as a fixed value, that is, the network loss price remains basically unchanged. In the process of predicting the node electricity price, the change of this feature is also ignored, and only the average electricity price in the target area and the congestion price of the node are considered.

[0156] Specifically, taking a grid node in the power grid system as an example, the congestion electricity price of the grid node is predicted by the above-mentioned graph convolutional neural network, and then the sum of the congestion electricity price of the grid node, the average electricity price of the target area and the network loss electricity price is determined as the node electricity price of the grid node on the next day.

[0157] It should be noted that the target area in the embodiment of the present application can be but is not limited to a province or a city. The predicted average electricity price, congestion price and node electricity price are all data sequences including 96 points a day, that is, the congestion price of a power grid node includes 96 values, and these 96 values ​​correspond to the dynamic congestion price every 15 minutes within 24 hours of the next day (the next day). Similarly, the average electricity price is also a data sequence of 96 points a day, which is used to indicate the dynamic change of the average electricity price of the target area every 15 minutes within 24 hours of the next day. The data form of the node electricity price of a power grid node predicted is the same as the data form of the average electricity price and the congestion price, which will not be repeated here.

[0158] For ease of understanding, the following Figure 3 The overall flow chart shown further explains the method for determining the node electricity price mentioned above.

[0159] S302, acquiring and processing data for average electricity prices in a target area;

[0160] Specifically, it includes data in multiple dimensions. For example, the weather in the target area is obtained from weather forecast data, and the historical average electricity price, coal price and other information in the target area are obtained from power grid disclosures. The target area can be but is not limited to a province or a city, or several adjacent provinces or cities.

[0161] After obtaining the above initial data, all data need to be adjusted to a format of 96 points a day to facilitate model learning. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0162] S304, constructing a DeepAR model to predict the average day-ahead electricity price in the target area;

[0163] For details, please refer to the description of building and training the DeepAR model in the above embodiment, which will not be repeated here.

[0164] S306, acquiring and processing node electricity price data for each grid node;

[0165] The acquisition method and processing process of this part of data are the same as the content of the above step S302.

[0166] S308, constructing a GCN model to predict node congestion electricity prices;

[0167] For details, please refer to the description of building and training the GCN model in the above embodiment, which will not be repeated here.

[0168] S310, summing up the predicted average electricity price of the target area, the congestion electricity price of each grid node, and the network loss electricity price to obtain the predicted node electricity price of each node on the next day.

[0169] The technical solutions in the embodiments of the present application can be adapted to different power network structures and operating conditions, have strong universality, and can be extended to other regions.

[0170] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present invention.

[0171] In this embodiment, a device for determining a node electricity price is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0172] Figure 7 1 is a structural block diagram of a device for determining a node electricity price according to an embodiment of the present invention, the device comprising:

[0173] The first acquisition unit 702 is used to acquire a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the power grid system of the target area within the historical period; the first processing unit 704 is used to obtain the average electricity price of the target area by inputting the target data set into the autoregressive cyclic network; the second processing unit 706 is used to obtain the blocking electricity price of each grid node by inputting the target data set into the graph convolutional neural network, wherein the blocking electricity price represents the additional cost caused by the limitation of the power grid transmission capacity; the third processing unit 708 is used to determine the node electricity price of each grid node based on the average electricity price and the predicted blocking electricity price of each grid node.

[0174] In an exemplary embodiment, the first processing unit 704 includes: a first processing module, which is used to convert each data in the target data set into time series data to obtain a time series data set, wherein one time series data in the time series data set is used to describe the electricity price characteristic information of one of the grid nodes at T time points, and T is a positive integer greater than or equal to 2; a first training module, which is used to train a pre-created autoregressive recurrent network based on the time series data set to obtain a trained target autoregressive recurrent network, wherein the input of a current hidden layer in the autoregressive recurrent network includes the output result of the autoregressive recurrent network at the previous time point and the electricity price characteristics of each grid node at the current time point; a second processing module, which is used to obtain a target probability distribution by adding a likelihood function to the target autoregressive network, wherein the likelihood function is used to describe the value range of the average electricity price in the target area when the input of the current hidden layer is given; a third processing module, which is used to determine the average electricity price in the target area based on the target probability distribution and the value range.

[0175] In an exemplary embodiment, the above-mentioned third processing module includes: a first processing sub-module, used to determine the best fitting data line from the target probability distribution based on the gradient descent method, wherein the best fitting data line has the highest probability in the time series data set; a second processing sub-module, used to determine the average electricity price of the target area from the value range based on the best fitting data line, wherein the average electricity price of the target area includes T sub-electricity prices at T time points.

[0176] In an exemplary embodiment, the second processing unit 706 includes: a construction module, which is used to construct a graph convolutional neural network based on the target data set and the pre-acquired electricity price description information of the forecast day, wherein the graph convolutional neural network includes M graph nodes, and one graph node among the M graph nodes is used to represent one grid node among each grid node, and M is a positive integer greater than or equal to 2; an update module, which is used to update the node features of the M graph nodes by performing a convolution operation on the M graph nodes to obtain updated node features, wherein the node features are used to represent the description information of the additional electricity price of a grid node caused by the limitation of the grid transmission capacity; a fourth processing module, which is used to obtain the blocking electricity price of each grid node by inputting the updated node features into a multi-layer perceptron.

[0177] In an exemplary embodiment, the above-mentioned update module includes: a construction submodule, which is used to construct an adjacency matrix based on the weights of the edges between any two connected graph nodes among the M graph nodes; a fifth processing module, which is used to aggregate the node features of the M graph nodes using a weighted average method to obtain the aggregated node features; and a sixth processing module, which is used to determine the aggregated node features as the updated node features.

[0178] In an exemplary embodiment, the fifth processing module includes: a first acquisition submodule, used to sequentially acquire a graph node from M graph nodes as the current graph node; a second acquisition submodule, used to acquire neighbor features of Y graph nodes adjacent to and connected to the current graph node, where Y is a positive integer greater than or equal to 1 and less than M; a third processing submodule, which aggregates the neighbor features and the current node features of the current graph node based on the adjacency matrix and weighted average method to obtain the current node features after aggregation.

[0179] In an exemplary embodiment, the third unit 708 includes: a first acquisition module, used for a second acquisition module, used to sequentially acquire a grid node from each grid node as the current grid node; to acquire the current predicted blocking electricity price of the current grid node from the predicted blocking electricity prices of each grid node; and a seventh processing module, used to sum the average electricity price, the current predicted blocking electricity price and the network loss electricity price to obtain the node electricity price of the current grid node, wherein the network loss electricity price represents the electricity cost generated due to line loss during power transmission.

[0180] Through the above device, the average electricity price of the target area is predicted by using an autoregressive recurrent network, and the congestion electricity price of each node in the power grid system of the target area is predicted by using a graph convolutional neural network. Finally, the node electricity price of each power grid node is determined based on the predicted average electricity price and the node congestion electricity price. This method of predicting node electricity prices based on artificial intelligence can not only analyze spatial information, but also effectively capture the relationship and dynamic changes between power grid nodes, thereby avoiding dependence on large amounts of data, reducing computing resources, and achieving the technical effect of improving the accuracy of the prediction results of node electricity prices.

[0181] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when run.

[0182] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0183] S1, obtaining a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the grid system of the target area within a historical period;

[0184] S2, by inputting the target data set into the autoregressive loop network, the average electricity price of the target area is obtained;

[0185] S3, by inputting the target data set into the graph convolutional neural network, the congestion price of each grid node is obtained, where the congestion price represents the additional cost caused by the transmission capacity limitation of the grid;

[0186] S4, determining the node electricity price of each grid node based on the average electricity price and the predicted congestion electricity price of each grid node.

[0187] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0188] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0189] An embodiment of the present invention further provides an electronic device, such as Figure 8 As shown, it includes a memory 802 and a processor 804. The memory 802 may include, but is not limited to, the first acquisition unit 702, the first processing unit 704, the second processing unit 706, and the third processing unit 708 in the above-mentioned node electricity price determination device. In addition, other module units in the above-mentioned node electricity price determination device may also be included, which will not be repeated in this example.

[0190] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0191] S1, obtaining a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the grid system of the target area within a historical period;

[0192] S2, by inputting the target data set into the autoregressive loop network, the average electricity price of the target area is obtained;

[0193] S3, by inputting the target data set into the graph convolutional neural network, the congestion price of each grid node is obtained, where the congestion price represents the additional cost caused by the transmission capacity limitation of the grid;

[0194] S4, determining the node electricity price of each grid node based on the average electricity price and the predicted congestion electricity price of each grid node.

[0195] In an exemplary embodiment, a person of ordinary skill in the art can understand that Figure 8 The structure shown is for illustration only. Figure 8 The structure of the electronic device is not limited. For example, the electronic device may also include Figure 8 More or fewer components (such as network interfaces, etc.) as shown in, or with Figure 8 Different configurations are shown.

[0196] Among them, the memory 802 can be used to store software programs and modules, such as the program instructions / modules corresponding to the node electricity price determination method and the node electricity price determination device in the embodiment of the present application. The processor 804 executes various functional applications and data processing by running the software programs and modules stored in the memory 802, that is, the above-mentioned node electricity price determination method is realized. The memory 802 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 802 may further include a memory remotely arranged relative to the processor 804, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. Among them, the memory 802 can be specifically used, but not limited to, for storing multiple standard commands and multiple decision nodes.

[0197] Optionally, the transmission device 806 is used to receive or send data via a network. Specific examples of the network may include wired networks and wireless networks. In one example, the transmission device 806 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers via a network cable so as to communicate with the Internet or a local area network. In one example, the transmission device 806 is a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0198] In addition, the electronic device mentioned above further includes: a display 808; and a connection bus 810, which is used to connect various module components in the electronic device mentioned above.

[0199] In other embodiments, the electronic device may be a node in a distributed system, wherein the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting the multiple nodes in the form of network communication. Among them, the nodes may form a peer-to-peer (P2P) network, and any form of computing device, such as a server, terminal, or other electronic device, may become a node in the blockchain system by joining the peer-to-peer network.

[0200] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0201] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0202] The embodiments of the present application also provide a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any one of the above method embodiments.

[0203] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0204] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a node electricity price, characterized in that: include: Acquire a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the grid system of the target area within a historical period; By inputting the target data set into an autoregressive cycle network, an average electricity price in the target area is obtained; By inputting the target data set into a graph convolutional neural network, a congestion price of each grid node is obtained, wherein the congestion price represents the additional cost caused by the transmission capacity limitation of the grid; Based on the average electricity price and the predicted congestion electricity price of each grid node, a node electricity price of each grid node is determined.

2. The method according to claim 1, characterized in that The method of obtaining the average electricity price of the target area by inputting the target data set into an autoregressive cycle network comprises: Convert each data in the target data set into time series data to obtain a time series data set, wherein one time series data in the time series data set is used to describe the electricity price characteristic information of one grid node in each grid node at T time points, where T is a positive integer greater than or equal to 2; Based on the time series data set, a pre-created autoregressive recurrent network is trained to obtain a trained target autoregressive recurrent network, wherein the input of a current hidden layer in the autoregressive recurrent network includes the output result of the autoregressive recurrent network at a previous time point and the electricity price characteristics of each grid node at a current time point; A target probability distribution is obtained by adding a likelihood function to the target autoregressive network, wherein the likelihood function is used to describe the range of values ​​of the average electricity price in the target area when a current hidden layer input is given; Based on the target probability distribution and the value range, an average electricity price in the target area is determined.

3. The method according to claim 2, characterized in that The determining the average electricity price of the target area based on the target probability distribution and the value range includes: Determine a best-fit data line from the target probability distribution based on a gradient descent method, wherein the best-fit data line has the highest probability on the time series data set; Based on the best fitting data line, the average electricity price of the target area is determined from the value range, wherein the average electricity price of the target area includes T sub-electricity prices at the T time points.

4. The method according to claim 1, characterized in that The method of obtaining the congestion price of each grid node by inputting the target data set into a graph convolutional neural network comprises: Based on the target data set and the pre-acquired electricity price description information of the forecast day, constructing the graph convolutional neural network, wherein the graph convolutional neural network includes M graph nodes, one graph node among the M graph nodes is used to represent one grid node among the respective grid nodes, and M is a positive integer greater than or equal to 2; By performing a convolution operation on the M graph nodes, node features of the M graph nodes are updated to obtain updated node features, wherein the node features are used to represent description information of an additional electricity price of the one grid node due to transmission capacity limitation of the grid; The congestion electricity price of each grid node is obtained by inputting the updated node features into a multi-layer perceptron.

5. The method according to claim 4, characterized in that The updating the node features of the M graph nodes by performing a convolution operation on the M graph nodes to obtain updated node features includes: Constructing an adjacency matrix based on the weights of the edges between any two connected graph nodes among the M graph nodes; Using a weighted average method, the node features of the M graph nodes are aggregated to obtain aggregated node features; The node features after the aggregation process are determined as the updated node features.

6. The method according to claim 5, characterized in that The node features of the M graph nodes are aggregated by using a weighted average method to obtain the node features after the aggregation process, including: Obtain one graph node from the M graph nodes in sequence as the current graph node; Obtain neighbor features of Y graph nodes adjacent to and connected to the current graph node, where Y is a positive integer greater than or equal to 1 and less than M; Based on the adjacency matrix and the weighted average method, the neighbor features and the current node features of the current graph node are aggregated to obtain the aggregated current node features.

7. The method according to any one of claims 1 to 6, characterized in that The determining the node electricity price of each grid node based on the average electricity price and the predicted congestion electricity price of each grid node comprises: Sequentially acquiring one grid node from each of the grid nodes as the current grid node; Acquire the current predicted congestion power price of the current grid node from the predicted congestion power prices of the respective grid nodes; The average electricity price, the current predicted congestion electricity price and the network loss electricity price are summed to obtain the node electricity price of the current power grid node, wherein the network loss electricity price represents the electricity cost generated due to line loss during power transmission.

8. A device for determining node electricity price, characterized in that: include: A first acquisition unit is used to acquire a target data set, wherein each data in the target data set is used to represent the electricity price description information of each grid node in the grid system of the target area within a historical period; A first processing unit, configured to obtain an average electricity price in the target area by inputting the target data set into an autoregressive cycle network; A second processing unit is configured to obtain a congestion price of each grid node by inputting the target data set into a graph convolutional neural network, wherein the congestion price represents an additional cost caused by a transmission capacity limitation of the grid; The third processing unit is used to determine the node electricity price of each grid node based on the average electricity price and the predicted congestion electricity price of each grid node.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program can be executed by a terminal device or a computer to execute the method described in any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

11. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images