Node electricity price prediction method and device

Abstract

The invention discloses a node electricity price prediction method and device. The method comprises the following steps: acquiring electricity price data and a topological relation of a plurality of target nodes in a first time period, including a to-be-predicted first node and an associated second node; the first time period is divided into a plurality of sub time periods with preset duration, the electricity price data of each node in each sub time period form a one-dimensional array, and all the one-dimensional arrays form a two-dimensional matrix; forming a three-dimensional tensor by the two-dimensional matrixes of the plurality of nodes according to the topological relation; analyzing the three-dimensional tensor by using a pre-trained node electricity price prediction model containing a three-layer convolutional neural network, and predicting the electricity price of the first node in a future time period; and the model analyzes the electricity price relation of each node in each sub-time period, the time sequence electricity price relation of each node and the electricity price relation between the nodes. According to the invention, the technical problem that a correlation node electricity price prediction scheme lacks consideration of a relation between nodes and the prediction result is not high in accuracy is solved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method and device for predicting node electricity prices. Background Art

[0002] With the continuous development of my country's electricity spot market, day-ahead price forecasts are becoming increasingly important for power generation companies to optimize power generation scheduling and reduce operating costs. Accurate price forecasts can help power generation companies rationally schedule the start-up and shutdown sequences and power generation of their generator sets, ensuring increased power output during peak price periods to maximize profits while reducing power generation during low price periods to avoid unnecessary financial losses. Furthermore, price forecasts can support cross-time and cross-regional power trading decisions. By accurately predicting prices, power generation companies can better manage market risks and mitigate the uncertainty caused by price fluctuations.

[0003] Existing node electricity price prediction methods include time series analysis, regression models, machine learning, and deep learning. However, existing methods often find it difficult to capture the topological relationship between nodes, thereby ignoring the most important blocking factor in the formation of node electricity prices. At the same time, they also do not take into account the electricity price relationship between nodes and are unable to handle the complex spatial influencing factors on the formation of node electricity prices.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The embodiments of the present application provide a node electricity price prediction method and device to at least solve the technical problem that related node electricity price prediction schemes lack consideration of the relationship between nodes and the prediction results are not accurate.

[0006] According to one aspect of an embodiment of the present application, a node electricity price prediction method is provided, including: obtaining electricity price data of multiple target nodes within a first time period, and obtaining a topological relationship between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node; evenly dividing the first time period into multiple first sub-time periods of preset duration, forming all electricity price data of each target node in each first sub-time period into a first one-dimensional array, forming all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and forming multiple first two-dimensional matrices corresponding to multiple target nodes into a first three-dimensional tensor based on the topological relationship; using a pre-trained node electricity price prediction model to analyze the first three-dimensional tensor to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between the multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of the multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0007] Optionally, obtaining electricity price data of multiple target nodes in the first time period includes: for each target node, collecting multiple electricity price data of the target node in the first time period according to a preset sampling frequency.

[0008] Optionally, after obtaining electricity price data of multiple target nodes in the first time period, the method further includes: performing data preprocessing on all obtained electricity price data, wherein the data preprocessing method includes at least one of the following: outlier deletion processing, null value interpolation processing, and data standardization processing.

[0009] Optionally, multiple first two-dimensional matrices corresponding to multiple target nodes are combined into a first three-dimensional tensor based on the topological relationship, including: taking the first node as the starting point, performing a breadth-first search traversal on the multiple target nodes based on the topological relationship to obtain a one-dimensional topological sequence corresponding to the multiple target nodes; stacking the first two-dimensional matrices corresponding to each target node based on the order relationship of each target node in the one-dimensional topological sequence to obtain a first three-dimensional tensor.

[0010] Optionally, the training process of the node electricity price prediction model includes: constructing an initial prediction model, wherein the initial prediction model includes at least three layers of convolutional neural networks; obtaining electricity price data of multiple target nodes in multiple first historical time periods and a second historical time period after each first historical time period, wherein the first historical time period is the same length as the first time period, and the second historical time period is the same length as the second time period; performing data preprocessing on all obtained electricity price data; for each first historical time period, evenly dividing the first historical time period into multiple second sub-time periods of preset length, forming all electricity price data of each target node in each second sub-time period into a second one-dimensional array, and dividing each All second one-dimensional arrays corresponding to the target nodes form a second two-dimensional matrix, and based on the topological relationship, multiple second two-dimensional matrices corresponding to multiple target nodes are combined into a second three-dimensional tensor, the second three-dimensional tensor is used as a training sample, and the electricity price data of the second historical time period after the first historical time period is used as the sample label corresponding to the training sample; the determined multiple training samples are divided into a training set and a validation set; the initial prediction model is iteratively trained using the training set, and the performance of the trained model is verified using the validation set; if the performance verification fails, the trained model is iteratively trained using the training set again; if the performance verification passes, the trained model is used as the node electricity price prediction model.

[0011] Optionally, the initial prediction model includes: a width convolutional neural network, a depth convolutional neural network, a point convolutional neural network, a double-layer feedforward neural network and a Gaussian error linear unit connected in sequence, a first residual block connected to the output of the width convolutional neural network and the input of the point convolutional neural network respectively, and a second residual block connected to the output of the width convolutional neural network and the input of the double-layer feedforward neural network respectively; wherein, the width convolutional neural network is used to analyze the relationship between the electricity price data of multiple target nodes, the depth convolutional neural network is used to analyze the relationship between the electricity price data of multiple sub-time periods corresponding to each target node, and the point convolutional neural network is used to analyze the relationship between the multiple electricity price data of each target node in each sub-time period.

[0012] Optionally, after obtaining the target electricity price data of the first node in the second time period in the future, the method also includes: adjusting the power generation strategy of the second time period based on the target electricity price data, wherein, when the target electricity price data is greater than a preset electricity price threshold, it is determined to increase the power generation; when the target electricity price data is equal to the electricity price threshold, it is determined to keep the current power generation unchanged; when the target electricity price data is less than the electricity price threshold, it is determined to reduce the power generation.

[0013] According to another aspect of the embodiment of the present application, a node electricity price prediction device is also provided, including: an acquisition module for acquiring electricity price data of multiple target nodes in a first time period, and acquiring the topological relationship between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node; a data processing module for evenly dividing the first time period into multiple first sub-time periods of preset duration, forming all electricity price data of each target node in each first sub-time period into a first one-dimensional array, forming all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and forming multiple first two-dimensional matrices corresponding to multiple target nodes into a first three-dimensional tensor based on the topological relationship; an analysis module for analyzing the first three-dimensional tensor using a pre-trained node electricity price prediction model to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between the multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of the multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes. According to another aspect of the embodiment of the present application, a node electricity price prediction device is also provided, including:

[0014] According to another aspect of an embodiment of the present application, a computer program product is further provided, the computer program product comprising: a computer program, wherein when the computer program is executed by a processor, the above-mentioned node electricity price prediction method is implemented.

[0015] According to another aspect of an embodiment of the present application, an electronic device is provided, which includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the above-mentioned node electricity price prediction method through the computer program.

[0016] In an embodiment of the present application, by obtaining electricity price data of multiple target nodes in a first time period and combining the topological relationship between the nodes, the original unstructured data is converted into a structured three-dimensional tensor. This not only helps the model to process and understand the data more effectively, but also can integrate the time series characteristics of electricity prices and the spatial dependencies between nodes in a unified mathematical representation. A three-layer convolutional neural network is used to analyze the three-dimensional tensor data. Each layer of the convolutional network is responsible for capturing different levels of features of the electricity price data. The first layer of the breadth convolutional neural network focuses on the spatial relationship between multiple target nodes, the second layer of the depth convolutional network deeply analyzes the electricity price fluctuation characteristics of each node in the time series, and the third layer of the point convolutional network further integrates channel information and learns the complex time series relationship between electricity price data. This multi-level feature learning mechanism can understand the electricity price data more comprehensively and deeply, thereby improving the accuracy of the prediction model, and thus solving the technical problem that the relevant node electricity price prediction scheme lacks consideration of the relationship between nodes and the prediction results are not accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 This is a flow chart of an optional node electricity price prediction method according to an embodiment of the present application;

[0019] Figure 2 is a schematic structural diagram of an optional node electricity price prediction device according to an embodiment of the present application;

[0020] Figure 3 It is a schematic structural diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0022] It should be noted that the terms "first", "second", etc. in the specification, claims, and drawings of the present application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. 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 variations thereof 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.

[0023] In order to better understand the embodiments of the present application, some nouns or terms that appear in the description of the embodiments of the present application are first translated and explained as follows:

[0024] The locational marginal price (LMP) refers to the marginal cost of increasing unit electricity demand at a specific geographic location or supply area (i.e., a node) in the electricity market. It reflects the incremental cost of consuming an additional kilowatt-hour of electricity at a specific time and location. LMP not only takes into account the cost of generation, but also factors such as transmission line losses and congestion costs. It is a key indicator of the supply-demand balance in the electricity market. LMPs vary across time and space, resulting in different prices at different times and locations. LMPs are influenced by local electricity supply and demand. When supply exceeds demand, prices may fall or even become negative. When demand exceeds supply, prices rise. LMPs provide important economic signals for market participants, influencing the production decisions of power generation companies and the electricity consumption behavior of consumers. Calculating LMPs requires considering multiple factors, such as the marginal cost of generators, system capacity, network losses, and line congestion.

[0025] Example 1

[0026] According to an embodiment of the present application, a node electricity price prediction method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] Figure 1 is a flow chart of a node electricity price prediction method provided according to an embodiment of the present application, such as Figure 1 As shown, the method includes the following steps:

[0028] Step S102: obtaining electricity price data of a plurality of target nodes in a first time period, and obtaining a topological relationship between the plurality of target nodes, wherein the plurality of target nodes include: a first node to be predicted and a plurality of second nodes associated with the first node;

[0029] Step S104: evenly divide the first time period into a plurality of first sub-time periods of preset lengths, combine all electricity price data of each target node in each first sub-time period into a first one-dimensional array, combine all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and combine the plurality of first two-dimensional matrices corresponding to the plurality of target nodes into a first three-dimensional tensor based on a topological relationship;

[0030] Step S106: Use the pre-trained node electricity price prediction model to analyze the first three-dimensional tensor to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0031] The following describes the various steps of the node electricity price prediction method in combination with the specific implementation process.

[0032] First, electricity price data of multiple target nodes within a first time period is obtained, and the topological relationship between the multiple target nodes is obtained, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node, the second node belongs to the same area as the first node, or is related to the power supply and demand of the first node, wherein the multiple second nodes associated with the first node may be nodes belonging to the same area as the first node or related to the power demand of the first node.

[0033] As an optional implementation, when acquiring electricity price data, for each target node, multiple electricity price data of the target node within a first time period may be collected according to a preset sampling frequency.

[0034] For example, a first time period is first determined, which can be any length of past time, such as the past 30 days, 90 days, or longer, depending on the amount of training data required by the prediction model and the accuracy of the prediction. Then, multiple target nodes are selected from the power network, including a first node (N0) for which the node electricity price needs to be predicted and multiple second nodes (N1, N2, ..., NM) directly or indirectly connected to the node on the power network. For each target node, the node's electricity price data within the first time period is collected according to a preset sampling frequency (for example, once every 15 minutes). The selection of the sampling frequency depends on the trading rules of the electricity spot market and the availability of data. Usually, the electricity market electricity price is released in hours or shorter periods (such as 15 minutes).

[0035] Optionally, after obtaining electricity price data of multiple target nodes in the first time period, data preprocessing may be performed on all obtained electricity price data, wherein the data preprocessing method includes at least one of the following: outlier deletion processing, null value interpolation processing, and data standardization processing.

[0036] Furthermore, the first time period is evenly divided into multiple first sub-time periods of preset length, and all electricity price data of each target node in each first sub-time period are combined into a first one-dimensional array. Each one-dimensional array contains the electricity price sequence of the node in a single sub-time period, which is convenient for capturing the short-term fluctuation characteristics of electricity prices. All first one-dimensional arrays corresponding to each target node are combined into a first two-dimensional matrix. The rows of the matrix represent different sub-time periods, and the columns represent the electricity price data points of the node in the sub-time period. This structure is conducive to the model learning time series characteristics and periodic changes in electricity prices at the same time, and based on the topological relationship, multiple first two-dimensional matrices corresponding to multiple target nodes are combined into a first three-dimensional tensor.

[0037] As an optional implementation, the process of forming a first three-dimensional tensor from multiple first two-dimensional matrices corresponding to multiple target nodes based on topological relationships can be carried out in the following steps: starting from the first node, a breadth-first search is performed on the multiple target nodes based on the topological relationship (i.e., the connection method between the nodes) to obtain a one-dimensional topological sequence corresponding to the multiple target nodes, which represents the relative position and connectivity of the nodes; the first two-dimensional matrices corresponding to each target node are stacked according to the order relationship of each target node in the one-dimensional topological sequence to obtain a first three-dimensional tensor, the dimension of which is the length of the sub-time period × the number of data points × the number of nodes. It can simultaneously reflect the electricity price data and topological relationships of all target nodes in each sub-time period, and provides structured input for the model to capture the spatial correlation between nodes and the complex influencing factors of electricity prices.

[0038] For example, suppose there are five nodes in the power network, namely N0 (the first node), N1, N2, N3, and N4 (the second node). Electricity price data for these nodes is collected for the past 30 days. Each day's electricity price data is subdivided into 96 15-minute time points as the data for the first time period. The ultimate goal is to predict the electricity price for node N0 for the next three days. First, these 30 days are evenly divided into 30 first sub-time periods, each of which is one day. For each node, such as N0, all 96 electricity price data points in each first sub-time period (each day) are collected and organized to form 30 first one-dimensional arrays 1*D. For example, the electricity price data for node N0 on the first day can be represented as a one-dimensional array containing 96 elements [Price1, Price2, ..., Price96]. All the first one-dimensional arrays corresponding to each target node are stacked to form a first two-dimensional matrix 1*D*N. For node N0, this will be a 30-row × 96-column matrix, where each row represents the electricity price sequence for one day. Similarly, repeat the above steps for nodes N1, N2, N3, and N4 to obtain their respective two-dimensional matrices. Based on the topological relationship of the power network, the connection order between nodes is determined. Here, starting from N0, a breadth-first search algorithm is used to traverse all relevant nodes to obtain a one-dimensional topological sequence G = [N0, N 1, …, N4], for example, the order is N0→N1→N2→N3→N4. According to this one-dimensional topological sequence, the two-dimensional matrices of N0, N1, N2, N3, and N4 are stacked in order to form a three-dimensional tensor M*D*N. The dimension of this tensor will be the length of the sub-time period (30) × the number of data points (96) × the number of nodes (5).

[0039] After obtaining the above-mentioned first three-dimensional tensor, the first three-dimensional tensor is analyzed using a pre-trained node electricity price prediction model to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0040] As an optional implementation, the training process of the node electricity price prediction model can adopt the following steps:

[0041] S1, constructing an initial prediction model, wherein the initial prediction model includes at least three layers of convolutional neural networks;

[0042] As an optional implementation, the initial prediction model may include: a width convolutional neural network, a depth convolutional neural network, a point convolutional neural network, a double-layer feedforward neural network and a Gaussian error linear unit connected in sequence, a first residual block connected to the output of the width convolutional neural network and the input of the point convolutional neural network respectively, and a second residual block connected to the output of the width convolutional neural network and the input of the double-layer feedforward neural network respectively. The introduction of the residual block is to alleviate the gradient vanishing problem during the model training process and make the training of the deep network more stable and efficient; wherein, the width convolutional neural network is used to analyze the relationship between the electricity price data of multiple target nodes, the depth convolutional neural network is used to analyze the relationship between the electricity price data of multiple sub-time periods corresponding to each target node, and the point convolutional neural network is used to analyze the relationship between the multiple electricity price data of each target node in each sub-time period.

[0043] Among them, the breadth convolutional neural network uses a 1*M convolution kernel to aggregate information on the three-dimensional tensor M*D*N along the node dimension M. The size of the convolutional data is D*N*1. The breadth convolutional neural network is used to capture the mutual influence factors of electricity prices caused by the spatial relationship between the focus node and the remaining nodes. The formula is:

[0044] x (l) =BN(σ(Conv(x (l-1) ,k=M)))

[0045] Where x (l-1) Represents the output of the previous layer, x (l) is the output of the current layer, BN represents batch normalization, Conv represents the convolution operation, k = M represents the convolution kernel size, σ represents the activation function, and the Gaussian error linear unit is used. The Gaussian error linear unit formula is:

[0046]

[0047] Where x represents the input of a neuron in the neural network, that is, the activation value or feature passed from the previous layer.

[0048] Then, for the D*N*1 data block, it passes through the linear head and the nonlinear head respectively, where the linear head is a residual connection, and the nonlinear head includes two layers of convolutional neural networks: the first layer is a deep convolutional neural network, in order to capture the temporal relationship within the block, and the second layer is a point convolutional neural network, in order to capture the temporal relationship between blocks. The first residual block and the second residual block are respectively connected between the output of the wide convolutional neural network and the input of the point convolutional neural network, and between the output of the wide convolutional neural network and the input of the double-layer feedforward neural network.

[0049] Depthwise convolution applies convolution to each input channel independently to capture the temporal feature relationship within the block. The formula is:

[0050] x (l) =BN(σ(Conv(x (l-1) ,k=K)))

[0051] In the formula, the convolution kernel size k=K can be adjusted according to actual conditions.

[0052] Point convolution performs 1x1 convolution on each position, which is mainly used to integrate information between channels and capture the temporal relationship between blocks. The formula is:

[0053] x (l) =BN(σ(Conv(x (l-1) ,k=1)))

[0054] Then, the output of the multi-layer separable convolution part is input into a two-layer feedforward neural network structure, where the activation function is a Gaussian error linear unit, and finally the predicted output is obtained.

[0055] S2, obtaining electricity price data of multiple target nodes in multiple first historical time periods and a second historical time period following each first historical time period, wherein the first historical time period is the same length as the first time period, and the second historical time period is the same length as the second time period;

[0056] S3, preprocessing all acquired electricity price data;

[0057] S4, for each first historical time period, evenly divide the first historical time period into multiple second sub-time periods of preset lengths, combine all electricity price data of each target node in each second sub-time period into a second one-dimensional array, combine all second one-dimensional arrays corresponding to each target node into a second two-dimensional matrix, and combine multiple second two-dimensional matrices corresponding to multiple target nodes into a second three-dimensional tensor based on the topological relationship, use the second three-dimensional tensor as a training sample, and use the electricity price data of the second historical time period after the first historical time period as the sample label corresponding to the training sample;

[0058] S5, dividing the determined multiple training samples into a training set and a validation set;

[0059] S6, iteratively train the initial prediction model using the training set, and use the validation set to verify the performance of the trained model;

[0060] S7: If the performance verification fails, the trained model is iteratively trained using the training set again. If the performance verification passes, the trained model is used as the node electricity price prediction model.

[0061] Once the model passes performance verification, that is, it has good predictive capabilities, it will be saved and used as the final "node electricity price prediction model". The model will then be used to analyze the first three-dimensional tensor to obtain the target electricity price data of the first node in the second time period in the future.

[0062] As an optional implementation, after obtaining the target electricity price data of the first node in the second future time period, the power generation strategy for the second time period can also be adjusted according to the target electricity price data, including: "When the target electricity price data is greater than the preset electricity price threshold, it indicates that the electricity price in the future time period may be high, and the power generation enterprise should consider increasing the power generation to fully utilize the high electricity price to make profits and increase electricity sales revenue, so it is determined to increase the power generation; when the target electricity price data is equal to the electricity price threshold, it is determined to keep the current power generation unchanged to avoid unnecessary cost increases; when the target electricity price data is less than the electricity price threshold, the power generation enterprise should consider reducing the power generation and reducing the power output during the low electricity price period to avoid losses or unnecessary operating costs, so it is determined to reduce the power generation.

[0063] In an embodiment of the present application, by obtaining electricity price data of multiple target nodes in a first time period and combining the topological relationship between the nodes, the original unstructured data is converted into a structured three-dimensional tensor. This not only helps the model to process and understand the data more effectively, but also can integrate the time series characteristics of electricity prices and the spatial dependencies between nodes in a unified mathematical representation. A three-layer convolutional neural network is used to analyze the three-dimensional tensor data. Each layer of the convolutional network is responsible for capturing different levels of features of the electricity price data. The first layer of the breadth convolutional neural network focuses on the spatial relationship between multiple target nodes, the second layer of the depth convolutional network deeply analyzes the electricity price fluctuation characteristics of each node in the time series, and the third layer of the point convolutional network further integrates channel information and learns the complex time series relationship between electricity price data. This multi-level feature learning mechanism can understand the electricity price data more comprehensively and deeply, thereby improving the accuracy of the prediction model, and thus solving the technical problem that the relevant node electricity price prediction scheme lacks consideration of the relationship between nodes and the prediction results are not accurate.

[0064] Example 2

[0065] According to an embodiment of the present application, a node electricity price prediction device for implementing the node electricity price prediction method in embodiment 1 is also provided. Figure 2 As shown, the node electricity price prediction device at least includes: an acquisition module 21, a data processing module 22 and an analysis module 23, wherein:

[0066] An acquisition module 21 is configured to acquire electricity price data of a plurality of target nodes within a first time period and to acquire a topological relationship between the plurality of target nodes, wherein the plurality of target nodes include: a first node to be predicted and a plurality of second nodes associated with the first node;

[0067] a data processing module 22 configured to evenly divide the first time period into a plurality of first sub-time periods of preset duration, combine all electricity price data of each target node in each first sub-time period into a first one-dimensional array, combine all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and combine the plurality of first two-dimensional matrices corresponding to the plurality of target nodes into a first three-dimensional tensor based on a topological relationship;

[0068] The analysis module 23 is used to analyze the first three-dimensional tensor using a pre-trained node electricity price prediction model to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0069] The following describes the functions of each module of the node electricity price prediction device in conjunction with the specific implementation process.

[0070] The acquisition module acquires electricity price data of multiple target nodes in a first time period, which may include, and acquires topological relationships between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node.

[0071] As an optional implementation, when acquiring electricity price data, for each target node, multiple electricity price data of the target node within a first time period may be collected according to a preset sampling frequency.

[0072] Optionally, after obtaining electricity price data of multiple target nodes in the first time period, data preprocessing may be performed on all obtained electricity price data, wherein the data preprocessing method includes at least one of the following: outlier deletion processing, null value interpolation processing, and data standardization processing.

[0073] The data processing module evenly divides the first time period into multiple first sub-time periods of preset length, organizes all electricity price data of each target node in each first sub-time period into a first one-dimensional array, organizes all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and organizes multiple first two-dimensional matrices corresponding to multiple target nodes into a first three-dimensional tensor based on the topological relationship.

[0074] As an optional implementation, the process of forming a first three-dimensional tensor from multiple first two-dimensional matrices corresponding to multiple target nodes based on topological relationships can be carried out in the following steps: taking the first node as the starting point, performing a breadth-first search traversal on multiple target nodes based on the topological relationship to obtain a one-dimensional topological sequence corresponding to the multiple target nodes; stacking the first two-dimensional matrices corresponding to each target node based on the order relationship of each target node in the one-dimensional topological sequence to obtain a first three-dimensional tensor.

[0075] After obtaining the above-mentioned first three-dimensional tensor, the analysis module uses the pre-trained node electricity price prediction model to analyze the first three-dimensional tensor to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0076] As an optional implementation, the training process of the node electricity price prediction model can adopt the following steps:

[0077] S1, constructing an initial prediction model, wherein the initial prediction model includes at least three layers of convolutional neural networks;

[0078] As an optional implementation, the initial prediction model includes: a width convolutional neural network, a depth convolutional neural network, a point convolutional neural network, a double-layer feedforward neural network and a Gaussian error linear unit connected in sequence, a first residual block connected to the output of the width convolutional neural network and the input of the point convolutional neural network respectively, and a second residual block connected to the output of the width convolutional neural network and the input of the double-layer feedforward neural network respectively; wherein, the width convolutional neural network is used to analyze the relationship between the electricity price data of multiple target nodes, the depth convolutional neural network is used to analyze the relationship between the electricity price data of multiple sub-time periods corresponding to each target node, and the point convolutional neural network is used to analyze the relationship between the multiple electricity price data of each target node in each sub-time period.

[0079] S2, obtaining electricity price data of multiple target nodes in multiple first historical time periods and a second historical time period following each first historical time period, wherein the first historical time period is the same length as the first time period, and the second historical time period is the same length as the second time period;

[0080] S3, preprocessing all acquired electricity price data;

[0081] S4, for each first historical time period, evenly divide the first historical time period into multiple second sub-time periods of preset lengths, combine all electricity price data of each target node in each second sub-time period into a second one-dimensional array, combine all second one-dimensional arrays corresponding to each target node into a second two-dimensional matrix, and combine multiple second two-dimensional matrices corresponding to multiple target nodes into a second three-dimensional tensor based on the topological relationship, use the second three-dimensional tensor as a training sample, and use the electricity price data of the second historical time period after the first historical time period as the sample label corresponding to the training sample;

[0082] S5, dividing the determined multiple training samples into a training set and a validation set;

[0083] S6, iteratively train the initial prediction model using the training set, and use the validation set to verify the performance of the trained model;

[0084] S7: If the performance verification fails, the trained model is iteratively trained using the training set again. If the performance verification passes, the trained model is used as the node electricity price prediction model.

[0085] After the node electricity price prediction model is obtained in the above manner, the analysis module uses the first three-dimensional tensor to analyze and obtain the target electricity price data of the first node in the future second time period.

[0086] As an optional implementation, after obtaining the target electricity price data of the first node in the second time period in the future, the power generation strategy of the second time period can also be adjusted based on the target electricity price data. When the target electricity price data is greater than the preset electricity price threshold, it is determined to increase the power generation; when the target electricity price data is equal to the electricity price threshold, it is determined to keep the current power generation unchanged; when the target electricity price data is less than the electricity price threshold, it is determined to reduce the power generation.

[0087] It should be noted that each module in the node electricity price prediction device in the embodiment of the present application corresponds one-to-one to each implementation step of the node electricity price prediction method in Example 1. Since a detailed description has been given in Example 1, some details not reflected in this embodiment can be referred to Example 1 and will not be elaborated here.

[0088] Example 3

[0089] According to an embodiment of the present application, a computer program product is further provided, which includes a computer program, wherein when the computer program is executed by a processor, the node electricity price prediction method in Example 1 is implemented.

[0090] According to an embodiment of the present application, a non-volatile storage medium is also provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the node electricity price prediction method in Example 1 by running the computer program.

[0091] According to an embodiment of the present application, a processor is further provided, which is used to run a computer program, wherein the node electricity price prediction method in Example 1 is executed when the computer program is running.

[0092] According to an embodiment of the present application, an electronic device is also provided, which includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the node electricity price prediction method in Example 1 through the computer program.

[0093] Specifically, when the computer program is running, the following steps are executed: obtaining electricity price data of multiple target nodes within a first time period, and obtaining a topological relationship between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node; evenly dividing the first time period into multiple first sub-time periods of preset duration, forming all electricity price data of each target node in each first sub-time period into a first one-dimensional array, forming all first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and forming multiple first two-dimensional matrices corresponding to multiple target nodes into a first three-dimensional tensor based on the topological relationship; using a pre-trained node electricity price prediction model to analyze the first three-dimensional tensor to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between the multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of the multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

[0094] As an optional implementation, the electronic device may be in the form of a mobile terminal, a computer terminal or a similar computing device. Figure 3 The hardware structure block diagram of an electronic device for implementing a node electricity price prediction method is shown in FIG. Figure 3As shown, the electronic device 30 may include one or more (illustrated as 302a, 302b, ..., 302n in the figure) processors 302 (the processor 302 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 304 for storing data, and a transmission device 306 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 3 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 3 More or fewer components than shown, or with Figure 3 Different configurations shown.

[0095] It should be noted that the one or more processors 302 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the electronic device 30. As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0096] The memory 304 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the node electricity price prediction method in the embodiment of the present application. The processor 302 executes various functional applications and data processing by running the software programs and modules stored in the memory 304, that is, implementing the vulnerability detection method of the above-mentioned application. The memory 304 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 304 may further include a memory remotely located relative to the processor 302, and these remote memories may be connected to the electronic device 30 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.

[0097] The transmission device 306 is used to receive or send data via a network. Specific examples of the aforementioned network may include a wireless network provided by the communications provider of the electronic device 30. In one embodiment, the transmission device 306 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In one embodiment, the transmission device 306 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0098] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the electronic device 30 .

[0099] The serial numbers of the embodiments are for description only and do not represent the advantages or disadvantages of the embodiments.

[0100] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0102] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected to achieve the purpose of the present embodiment according to actual needs.

[0103] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program code.

[0105] The above is only a preferred embodiment 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 node electricity price prediction method, characterized in that: include: Obtaining electricity price data of multiple target nodes in a first time period, and obtaining a topological relationship between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node; The first time period is evenly divided into a plurality of first sub-time periods of preset duration, all electricity price data of each target node in each of the first sub-time periods are combined into a first one-dimensional array, all the first one-dimensional arrays corresponding to each target node are combined into a first two-dimensional matrix, and the plurality of the first two-dimensional matrices corresponding to the plurality of target nodes are combined into a first three-dimensional tensor according to the topological relationship; The first three-dimensional tensor is analyzed using a pre-trained node electricity price prediction model to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

2. The method according to claim 1, characterized in that Obtain electricity price data of multiple target nodes in a first time period, including: For each of the target nodes, a plurality of electricity price data of the target node in the first time period are collected according to a preset sampling frequency.

3. The method according to claim 1, characterized in that After acquiring electricity price data of multiple target nodes in the first time period, the method further includes: Data preprocessing is performed on all acquired electricity price data, wherein the data preprocessing method includes at least one of the following: abnormal value deletion processing, null value interpolation processing, and data standardization processing.

4. The method according to claim 1, characterized in that: According to the topological relationship, the first two-dimensional matrices corresponding to the target nodes are combined into a first three-dimensional tensor, including: Taking the first node as a starting point, performing a breadth-first search traversal on the plurality of target nodes according to the topological relationship, and obtaining a one-dimensional topological sequence corresponding to the plurality of target nodes; The first two-dimensional matrices corresponding to the target nodes are stacked according to the order relationship of the target nodes in the one-dimensional topological sequence to obtain the first three-dimensional tensor.

5. The method according to claim 1, characterized in that The training process of the node electricity price prediction model includes: Constructing an initial prediction model, wherein the initial prediction model at least includes the three-layer convolutional neural network; Obtaining electricity price data of a plurality of target nodes in a plurality of first historical time periods and a second historical time period after each of the first historical time periods, wherein the first historical time period is the same length as the first time period, and the second historical time period is the same length as the second time period; Perform data preprocessing on all acquired electricity price data; For each first historical time period, the first historical time period is evenly divided into a plurality of second sub-time periods of the preset duration, all electricity price data of each target node in each second sub-time period are formed into a second one-dimensional array, all the second one-dimensional arrays corresponding to each target node are formed into a second two-dimensional matrix, and multiple second two-dimensional matrices corresponding to multiple target nodes are formed into a second three-dimensional tensor according to the topological relationship, the second three-dimensional tensor is used as a training sample, and the electricity price data of the second historical time period after the first historical time period is used as a sample label corresponding to the training sample; Dividing the determined plurality of training samples into a training set and a validation set; The initial prediction model is iteratively trained using the training set, and the performance of the trained model is verified using the verification set; if the performance verification fails, the trained model is iteratively trained again using the training set; if the performance verification passes, the trained model is used as the node electricity price prediction model.

6. The method according to claim 5, characterized in that The initial prediction model includes: a sequentially connected wide convolutional neural network, a deep convolutional neural network, a point convolutional neural network, a double-layer feedforward neural network and a Gaussian error linear unit, a first residual block connected to the output of the wide convolutional neural network and the input of the point convolutional neural network respectively, and a second residual block connected to the output of the wide convolutional neural network and the input of the double-layer feedforward neural network respectively; Among them, the breadth convolutional neural network is used to analyze the relationship between the electricity price data of multiple target nodes, the deep convolutional neural network is used to analyze the relationship between the electricity price data of multiple sub-time periods corresponding to each target node, and the point convolutional neural network is used to analyze the relationship between the multiple electricity price data of each target node in each sub-time period.

7. The method according to claim 1, characterized in that After obtaining the target electricity price data of the first node in the future second time period, the method further includes: The power generation strategy for the second time period is adjusted according to the target electricity price data, wherein, when the target electricity price data is greater than a preset electricity price threshold, the power generation is determined to be increased; when the target electricity price data is equal to the electricity price threshold, the current power generation is determined to be maintained unchanged; and when the target electricity price data is less than the electricity price threshold, the power generation is determined to be reduced.

8. A node electricity price prediction device, characterized in that: include: An acquisition module, used to acquire electricity price data of multiple target nodes in a first time period, and acquire a topological relationship between the multiple target nodes, wherein the multiple target nodes include: a first node to be predicted and multiple second nodes associated with the first node; a data processing module, configured to evenly divide the first time period into a plurality of first sub-time periods of preset duration, form all electricity price data of each target node in each of the first sub-time periods into a first one-dimensional array, form all the first one-dimensional arrays corresponding to each target node into a first two-dimensional matrix, and form a plurality of the first two-dimensional matrices corresponding to a plurality of the target nodes into a first three-dimensional tensor according to the topological relationship; An analysis module is used to analyze the first three-dimensional tensor using a pre-trained node electricity price prediction model to obtain the target electricity price data of the first node in the future second time period, wherein the node electricity price prediction model includes a three-layer convolutional neural network, which is used to analyze the relationship between multiple electricity price data of each target node in each first sub-time period, the relationship between the electricity price data of multiple first sub-time periods corresponding to each target node, and the relationship between the electricity price data of multiple target nodes.

9. A computer program product, characterized in that include: A computer program, wherein when the computer program is executed by a processor, the node electricity price prediction method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the node electricity price prediction method according to any one of claims 1 to 7 through the computer program.

Images