Data prediction method and system, computer equipment and storage medium

By obtaining and integrating data from multiple prediction models in the power spot market transaction decision system, the problems of new energy 'quantity, price' risks and market complexity in the power spot market are solved, and the accuracy of market information prediction and new energy are improved. The rationality of the returns.

CN119006039APending Publication Date: 2024-11-22CHINA THREE GORGES CORPORATION
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411257425.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-09
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In the spot power market, the 'quantity and price' risks of new energy increase, the market design is complex, the price fluctuates greatly, and operation safety and strategic behavior of participants bring decision-making problems, which requires accurate market information prediction and Auxiliary services.

Method used

By obtaining the actual data of the day and the first prediction results of multiple pre-deployed original models, calculate the prediction error and perform multi-dimensional evaluation, determine the target weights of each model, fuse the model and select the target model with the highest comprehensive evaluation, and update the data to obtain accuracy The prediction results are applied to each node of the power spot market trading decision system.

Benefits of technology

It improves the prediction accuracy and the accuracy of declaration volume in the trading decision process of the electricity spot market, and ensures reasonable returns of new energy under market-oriented conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006039A_ABST
    Figure CN119006039A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power grids, in particular to a data prediction method and system, computer equipment and a storage medium, and the method comprises the steps: obtaining the actual data of the day and a first prediction result, and determining the prediction error of each original model according to the actual data of the day and the first prediction result, performing multi-dimensional evaluation on each original model to obtain an evaluation result corresponding to each original model; determining a target weight of each original model according to the prediction error, the evaluation result and a preset initial weight, fusing the original models to obtain a fusion model, and selecting a target model from the fusion model and each original model; updating the actual data of the day to the day-ahead data and inputting the data into the target model to obtain a target prediction result; the target model obtained through operations such as model evaluation, fusion and good selection is the model with the highest comprehensive evaluation, so that the target prediction result obtained based on the target model has high accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of power grid technology, and in particular to a data prediction method, system, computer device and storage medium. Background Art

[0002] The power spot market aims to reflect the real-time price of power supply and demand through a real-time bidding mechanism, improve system efficiency and economy, and promote the integration of renewable energy. Under the conditions of the spot market, the "quantity and price" risks of new energy will increase significantly. Faced with the complexity of market design, price fluctuations, operational safety challenges, and strategic behavior of participants, many professional power market auxiliary services and decision-making manufacturers have emerged.

[0003] Then, how to ensure reasonable returns of new energy under market conditions and provide excellent auxiliary services and accurate market information forecasts for power generation entities and other market participants has become the current problem to be solved. Summary of the invention

[0004] In order to solve the above technical problems, the present disclosure provides a data prediction method, system, computer device and storage medium.

[0005] In a first aspect, the present invention provides a data prediction method, which is applied to predict each node of the power spot market quotation strategy, and includes:

[0006] The first prediction results obtained by respectively predicting the actual data of the day and multiple pre-deployed original models based on the day-ahead data, wherein the actual data of the day and the day-ahead data both correspond to the current node; the prediction error of each original model is determined according to the actual data of the day and the multiple first prediction results, and the prediction function of each original model is evaluated in multiple dimensions to obtain the evaluation result of each original model; the target weight of each original model is calculated according to the prediction error, the evaluation result and the preset initial weight, and each original model is fused according to the corresponding target weight to obtain a fusion model; the evaluation result of each original model and the second prediction result obtained by the fusion model based on the day-ahead data are evaluated in combination with the actual data of the day, and the target model of the current node is selected from the fusion model and the multiple original models; the actual data of the day is updated to the day-ahead data, and the updated day-ahead data is input into the target model to obtain the target prediction result corresponding to the current node.

[0007] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0008] The data prediction method provided in this embodiment, after obtaining the actual data of the day and the first prediction result, first determines the prediction error of each original model according to the actual data of the day and multiple prediction results, and obtains the evaluation result corresponding to each original model through multi-dimensional evaluation of each original model; then, determines the target weight of each original model according to the prediction error, the evaluation result and the preset initial weight, fuses each original model according to the corresponding target weight to obtain a fusion model, and selects the target model of the current node from the fusion model and each original model; finally, updates the actual data of the day to the advance data, and inputs the updated day-ahead data into the target model to obtain the target output result corresponding to the current node; in the above process, the target model obtained through the evaluation, fusion and optimization of the model is the model with the highest comprehensive evaluation, so that the target prediction result obtained based on the target model has higher accuracy; and because each node in the power spot market transaction decision process is predicted according to the above process, the final predicted declared quantity is more accurate.

[0009] In a second aspect, the present invention provides a data prediction device, comprising:

[0010] An acquisition module is used to obtain the actual data of the day and the first prediction results obtained by multiple pre-deployed original models based on the day-ahead data prediction, wherein the actual data of the day and the day-ahead data both correspond to the current node; an evaluation module is used to determine the prediction error of each original model based on the actual data of the day and the multiple first prediction results, and to perform multi-dimensional evaluation on the prediction function of each original model to obtain the evaluation result of each original model; a fusion module is used to calculate the target weight of each original model based on the prediction error, the evaluation result and the preset initial weight, and to fuse each original model according to the corresponding target weight to obtain a fusion model; a selection module is used to evaluate the evaluation result of each original model and the second prediction result obtained by the fusion model based on the day-ahead data prediction, combined with the actual data of the day, and select the target model of the current node from the fusion model and the multiple original models; a prediction module is used to update the actual data of the day to the day-ahead data, and input the updated day-ahead data into the target model to obtain the target prediction result corresponding to the current node.

[0011] In a third aspect, the present invention provides a computer device, comprising:

[0012] The memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the data prediction method of the first aspect and any one of its embodiments by executing the computer instructions.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the data prediction method of the first aspect and any of its embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0015] In order to more clearly illustrate the embodiments of the present invention 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.

[0016] Figure 1 A schematic diagram of a power spot market transaction decision system provided by an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a flow chart of a data prediction method provided by an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a flow chart of a method for determining target weights provided in an embodiment of the present invention;

[0019] Figure 4 A structural connection diagram of a data prediction device provided by an embodiment of the present invention;

[0020] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0023] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0024] According to an embodiment of the present invention, a data prediction method embodiment 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.

[0025] In this embodiment, a data prediction method is provided, which is applied to the power spot market transaction decision system, and is specifically applied to each node of the power spot market transaction decision process. The process of power spot market transaction decision includes 6 nodes, namely, meteorological prediction node, load prediction node, power generation prediction node, power prediction node, electricity price prediction node and transaction strategy prediction node, and each prediction node is responsible for the prediction operation of a type of data. There is an association between the prediction nodes, and the so-called association means that the prediction output of one / multiple nodes is the prediction input of another node, and the prediction operation of each node is completed with the support of the power spot market transaction decision system according to the data prediction method provided by the embodiment of the present invention. Therefore, before describing the data prediction method in detail, the power spot market transaction decision system is briefly described.

[0026] like Figure 1 The electricity spot market transaction decision system shown includes: a data base layer, a model deployment layer, a target range training layer and an advanced application layer.

[0027] The data base layer is used to obtain the day-ahead data and the actual data of the day required for the current node to perform forecasting operations from platforms such as trading centers, meteorological stations, and new energy stations through external interfaces. It is also used to store the forecast results generated by each model in the model deployment layer for subsequent calls.

[0028] The model deployment layer is used to provide two types of services. The first type of service is the environment support for model deployment, and the second type of service is model management / scheduling. Among them, the first type of service is mainly used to deploy the environment for the original model provided by each manufacturer and the fusion model generated by the fusion of the original model by the advanced application layer; the second type of service is used to call the day-ahead prediction result and the actual data of the day corresponding to the current node from the data base layer (the day-ahead prediction result is the result obtained by predicting the day based on the day-ahead data), and send the called data to the range training layer; the second type of service is also used to use the original model and the fusion model to complete the prediction operation of the current node respectively, and store the prediction result of the target model as the input information of the subsequent associated node of the current node in the data base layer, or send the prediction result of the target model to the trading center. The target model is the preferred model of the current node selected by the advanced application layer from the original model and the fusion model. It should be noted that the model deployment layer is also used to store the prediction results output by other models in the data base layer.

[0029] The target range training layer includes the meteorological forecast target range, load forecast target range, power generation forecast target range, power forecast target range, electricity price forecast target range and trading strategy forecast target range. Each target range conducts a multi-dimensional evaluation of the prediction function of each model in the model deployment layer based on the day-ahead prediction results and current data of the corresponding node transmitted by the model deployment layer, and uploads the evaluation results corresponding to each model to the advanced application layer. The evaluation results include the evaluation values ​​corresponding to each dimension.

[0030] The advanced application layer is used to determine the second weight corresponding to each model according to the evaluation results corresponding to each model; it is also used to fuse the original models according to the weights, and send the fused models to the model deployment layer for environmental deployment; it is also used to select the target model corresponding to the current node from the original models and the fused models according to the evaluation results of each original model and the evaluation results of the fused model.

[0031] In a transaction decision-making process of the electricity spot market, each node needs to execute the data prediction method provided by the embodiment of the present invention, and the prediction of the electricity spot market quotation is completed only after all nodes have completed the execution. Figure 2 is a flow chart of a data prediction method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0032] S201, obtaining actual data of the day and first prediction results obtained by multiple pre-deployed original models based on the day-ahead data predictions.

[0033] Among them, the actual data of the day (also known as D-day data) can be real data obtained from the data monitoring platform through an external interface and stored in the data base layer. The actual data of the day is not limited to the day, but can also include the current cycle. The specific duration of the current cycle is not limited in this embodiment, and technicians in this field can choose according to actual conditions. The day-ahead data (also known as D-1 day data) can be understood as historical data, which can be data before the day, or data from the previous collection cycle of the current collection cycle.

[0034] Among them, the actual data of the day and the day-ahead data both correspond to the current node. It can be understood that the actual data of the day and the day-ahead data are the same type of data at different times. The day-ahead data is the input data of the prediction operation of the current node; the first prediction result is the output data of the prediction operation of the current node, and the first prediction result is the prediction of the D-day data based on the D-1 day data; the actual data of the day is the real data corresponding to the current node. Each node corresponds to a set of day-ahead data and the actual data of the day. It should be noted that the prediction operation of the current node is implemented through the pre-deployed original model, and the prediction operation of the D-day data has been completed on D-1 day. Therefore, the first prediction result is stored in the data base layer after the prediction operation is completed on D-1 day.

[0035] The pre-deployed original model is the prediction model provided by various manufacturers and deployed in the power spot market transaction decision system. The original model is a trained model with multiple data prediction functions, and the prediction function corresponds to the node and is used to perform the prediction operation of the corresponding node.

[0036] Specifically, the actual data of the day and the first prediction results obtained by each original model based on the day-ahead data prediction are obtained from the data base layer.

[0037] Exemplarily, the data collection cycle in this embodiment is one day, so the actual data of the day is the data of the day, and the day-ahead data is the data before the day.

[0038] If the current node is a weather forecast node, the original model will perform the weather forecast function. The day-ahead data is the weather data of the day before, and the actual data of the day is the weather data of the day. The weather data includes but is not limited to wind speed, irradiance, temperature and humidity, precipitation and other data.

[0039] If the current node is a load forecasting node, then the original model will perform the load forecasting function. The day-ahead data includes the day-ahead meteorological data and the day-ahead power load data, and the day-ahead actual data is the day-ahead power load data.

[0040] If the current node is a power generation prediction node, then the original model will perform the output prediction function. The day-ahead data includes the day-ahead meteorological data and the day-ahead new energy output data, and the day's actual data is the day's new energy output data.

[0041] If the current node is a power prediction node, the original model will perform the power prediction function. The day-ahead data includes the day-ahead meteorological data, the day-ahead station production data and the day-ahead power data. The actual data for the day is the day-ahead power data.

[0042] If the current node is an electricity price prediction node, then the original model will perform the electricity price prediction function. The day-ahead data includes the day-ahead electricity load data, the day-ahead new energy output data, the day-ahead power data, the day-ahead market boundary condition data, the inter-provincial transaction data and the day-ahead clearing electricity price data. The actual data for the day is the day-ahead node electricity price data.

[0043] If the current node is a trading strategy prediction node, then the original model will execute the trading strategy prediction function. The day-ahead data includes day-ahead power data, node electricity price data and day-ahead transaction data. The actual data of the day is the trading strategy of the day.

[0044] S202, determining the prediction error of each original model based on the actual data of the day and the multiple first prediction results, and performing a multi-dimensional evaluation on the prediction function of each original model to obtain an evaluation result of each original model.

[0045] The actual data of the day and the first prediction result are both group data corresponding to the time series rather than a single data. The prediction error can be understood as the difference between the first prediction result and the actual data of the day. If the first prediction result output by the i-th original model is recorded as M i (t), and the actual data of the day is recorded as C(t), then the prediction error σ of the i-th original model is i (t) is expressed as: i (t) = M i (t)-C(t).

[0046] Specifically, before step S201, a comprehensive evaluation index is constructed, and the index is a dimension, including but not limited to kurtosis, accuracy, maximum positive error, etc. According to the above dimensions, based on the actual data of the day, the prediction error of the first prediction result is compared, the prediction function of the original model is evaluated, and the evaluation value corresponding to each dimension of each original model is obtained, that is, the evaluation result.

[0047] Step S203, calculating the target weight of each original model according to the prediction error, the evaluation result and the preset initial weight, and fusing each original model according to the corresponding target weight to obtain a fusion model.

[0048] Specifically, first, a fusion prediction error is constructed according to the prediction errors corresponding to each original model and the preset initial weights, and the fusion prediction error is solved with the goal of minimizing the fusion prediction error to obtain the first weights corresponding to each original model. Secondly, the prediction function of the original model is analyzed subjectively and objectively according to the evaluation results corresponding to each original model to obtain the second weights corresponding to each original model. Thirdly, the first weight and the second weight are weighted together to obtain the target weight. Finally, each original model is fused according to the target weight corresponding to the original model to obtain a fusion model.

[0049] It should be noted that the fusion operation exists in every prediction operation of each node. For example, in the complete electricity spot market transaction decision process, when the current node is a meteorological forecast node, it is necessary to perform a model fusion operation according to steps S201 to S203, and the obtained fusion model is the fusion model corresponding to the meteorological forecast node. When the node switches to the load forecast node, it is still necessary to perform the model fusion operation according to steps S201 to S203, and the obtained fusion model is the fusion model corresponding to the load forecast node. For another example, the current node is a meteorological forecast node. Before predicting the meteorological data of D-day based on the meteorological data of D-1, it is necessary to perform a model fusion operation, and the obtained fusion model is the fusion model corresponding to the meteorological forecast node on D-1; before predicting the meteorological data of D+1 based on the meteorological data of D-day, it is also necessary to perform a model fusion operation, and the obtained fusion model is the fusion model corresponding to the meteorological forecast node on D-day. It can be seen that the fusion operation needs to be performed in every prediction operation of each node.

[0050] Step S204, based on the evaluation result of each original model and the second prediction result obtained by the fusion model based on the day-ahead data prediction, evaluation is performed in combination with the actual data of the day, and a target model for the current node is selected from the fusion model and multiple original models.

[0051] Specifically, after obtaining the fusion model, first input the day-ahead data into the fusion model for prediction to obtain the second prediction result. Secondly, based on the actual data of the day, the prediction error of the second prediction result is obtained by comparison, and the prediction function of the fusion model is evaluated in multiple dimensions to obtain the evaluation result. Finally, according to the evaluation results of each original data and the evaluation results of the fusion model, the target model of the current node is determined from the original model and the fusion model. The target model can be understood as the model with the highest prediction accuracy among the original model and the fusion model.

[0052] Exemplarily, in an optional embodiment, the target model is selected in the following manner:

[0053] First, the total evaluation value of the original model is determined according to the evaluation result of the original model.

[0054] The evaluation results of the original model include the evaluation values ​​corresponding to each dimension. Therefore, the total evaluation value of the original model can be the sum of the evaluation values ​​corresponding to each dimension. For example, the evaluation values ​​of the original evaluation model are P1, P2, ..., P n .

[0055] Then, based on the second prediction result and the actual data of the day, the prediction function of the fusion model is evaluated in multiple dimensions to obtain the overall evaluation value of the fusion model.

[0056] The total evaluation value P of the fusion model R It can be the sum of the evaluation values ​​of each dimension of the fusion model.

[0057] Finally, the total evaluation value of each original model is compared with the total evaluation value of the fusion model, and the model corresponding to the highest total evaluation value is determined as the target model of the current node.

[0058] Set P1, P2, ..., P n With P R Compare and select the maximum value (i.e. the highest total evaluation value) P max , P max The corresponding model is determined as the target model of the current node.

[0059] Step S205, update the actual data of the day into the day-ahead data, and input the updated day-ahead data into the target model to obtain the target prediction result corresponding to the current node.

[0060] Specifically, the D-day data is updated to the day-ahead data, and the updated day-ahead data is input into the target model as the day-ahead data of D+1 day, and the data of D+1 day is predicted to obtain the target prediction result corresponding to the current node. The method for updating the day-ahead data specifically can be to add the D-day data to the day-ahead data and update the day-ahead data using a sliding window method.

[0061] The data prediction method provided in this embodiment, after obtaining the actual data of the day and the first prediction result, first determines the prediction error of each original model according to the actual data of the day and multiple prediction results, and obtains the evaluation result corresponding to each original model through multi-dimensional evaluation of each original model; then, determines the target weight of each original model according to the prediction error, the evaluation result and the preset initial weight, fuses each original model according to the corresponding target weight to obtain a fusion model, and selects the target model of the current node from the fusion model and each original model; finally, updates the actual data of the day to the advance data, and inputs the updated day-ahead data into the target model to obtain the target output result corresponding to the current node; in the above process, the target model obtained through the evaluation, fusion and optimization of the model is the model with the highest comprehensive evaluation, so that the target prediction result obtained based on the target model has higher accuracy; and because each node in the power spot market transaction decision process is predicted according to the above process, the final predicted declared quantity is more accurate.

[0062] In an optional embodiment, a method for determining target weights is provided, such as Figure 3 As shown, the evaluation results include the evaluation values ​​corresponding to each dimension. The target weights of each original model are calculated based on the prediction error, the evaluation results and the preset initial weights, including:

[0063] Step S301 , performing a weighted sum operation on the prediction errors according to the preset initial weights corresponding to the original models, to obtain a fused prediction error.

[0064] Specifically, the preset initial weight is the initial weight value configured by the system for each original model during the fusion operation, and the preset initial weight of each original model may be the same.

[0065] Exemplarily, the preset initial weights corresponding to the original models are respectively recorded as θ1, θ2, ..., θ m , the prediction errors corresponding to each original model are denoted as σ1(t), σ2(t), …, σ m (t), then the fusion prediction error σ(t) is expressed as:

[0066]

[0067] Step S302, with the goal of minimizing the fusion prediction error and with the preset initial weights and constants as constraints, determine the first weights corresponding to the original models respectively.

[0068] Specifically, the objective function is solved under the constraints of the constraints to obtain a set of optimal weights that can minimize the fusion prediction error, and the set of optimal weights includes the first weights corresponding to each original model.

[0069] Exemplarily, the objective function is:

[0070]

[0071] The constraints are:

[0072]

[0073] Step S303: determining the second weights corresponding to the original models according to the evaluation values ​​corresponding to the dimensions of each original model.

[0074] Specifically, the above step S303 includes:

[0075] Step S3031, using the entropy weight method, calculate the third weights corresponding to each original model according to the evaluation values ​​corresponding to each original model in each dimension.

[0076] Exemplarily, the relative importance of each original model is objectively evaluated using the entropy weight method to obtain the third weight, and the third weight is determined as follows:

[0077] (1) Construct the following multi-dimensional evaluation value matrix:

[0078]

[0079] Among them, x mn Represents the evaluation value of the mth original model in the nth dimension.

[0080] (2) Standardize the data and eliminate the dimensional differences of each dimension through positive dimension standardization:

[0081]

[0082] Get the new data matrix Y 1 :

[0083]

[0084] (3) Calculate the numerical weight P of each dimension to each original model ij :

[0085]

[0086] (4) Calculate the information entropy e of each dimension j ;

[0087] If P ij =0, then define

[0088]

[0089] (5) Determine the entropy weight ω of each dimension j :

[0090]

[0091] (6) Determine the third weight Z of the i-th original model i :

[0092]

[0093] Step S3032, using the hierarchical analysis method, the fourth weights corresponding to the original models are calculated according to the evaluation values ​​corresponding to each dimension of each original model.

[0094] Exemplarily, the relative importance of each original model is subjectively evaluated using the hierarchical analysis method to obtain the fourth weight, and the fourth weight is determined as follows:

[0095] (1) Construct the judgment matrix A through pairwise comparison n×n ,X ij is the comparison result of the relative importance of the i-th evaluation index and the j-th evaluation index:

[0096]

[0097] (2) After the construction of the judgment matrix A is completed, since its construction is affected to a certain extent by the subjective judgment of the decision maker, a consistency test is required. The definition of the consistency ratio CR and the calculation method of the consistency index CI are used to calculate the eigenvector μ corresponding to the maximum eigenvalue max of the judgment matrix A. Where n is the number of evaluation indicators. CR = CI / RI, RI is a random consistency index, and its value is obtained by looking up the table according to the order of the matrix. It is generally believed that when CR < 0.1, the judgment matrix is ​​consistent:

[0098]

[0099] (3) Calculate the indicator weight:

[0100] Calculate the row geometric mean of the judgment matrix A row by row:

[0101]

[0102] Normalize the average values ​​of the rows of matrix A to determine the objective evaluation weight coefficient a of each original model i :

[0103]

[0104] Determine the fourth weight R of the i-th original model i :

[0105]

[0106] Step S3033, normalizing the third weight and the fourth weight to obtain the second weights corresponding to the original models.

[0107] Exemplarily, the second weight is determined as follows:

[0108]

[0109] Among them, k is the swing coefficient between the third weight and the fourth weight, and its initial value is 1, which means that the third weight and the fourth weight are equally important.

[0110] Step S304: determine a target weight according to the first weight and the second weight.

[0111] For example, the first weight θ i and the second weight γ i Combine the weights and calculate the target weight λ of the original model i , determined as follows:

[0112]

[0113] The method for determining the target weights provided in this embodiment not only takes into account the subjective and objective evaluation of the importance of each original model, but also takes into account the prediction error of the model. The target weights determined in this way not only ensure the timeliness of the prediction model, but also improve the accuracy of the prediction results.

[0114] Exemplarily, after step S205, the target prediction result is added to the day-ahead data corresponding to the successor node of the current node.

[0115] For example, since there is an association relationship between the nodes in the electricity spot market transaction decision process, when a node completes its corresponding prediction operation, it is necessary to add its prediction result to the day-ahead data of its subsequent associated node.

[0116] The transaction decision-making process of the power spot market includes weather forecast nodes, load forecast nodes, power forecast nodes, electricity price forecast nodes and trading strategy forecast nodes. Among them, the successor nodes of the weather forecast node include load forecast nodes, power forecast nodes and power forecast nodes; the successor nodes of the load forecast node and power forecast node are both electricity price forecast nodes, and the successor nodes of the electricity price forecast node and power forecast node are both trading strategy forecast nodes. Furthermore, when the current node is a meteorological forecast node, the predicted meteorological data (i.e., the target forecast result) output by the target model corresponding to the meteorological forecast node is added to the day-ahead data corresponding to the load forecast node, the power generation forecast node, and the power forecast node respectively; when the current node is a load forecast node, the predicted compliance data output by the target model corresponding to the load forecast node is added to the day-ahead data corresponding to the electricity price forecast node; when the current node is a power generation forecast node, the predicted power generation data output by the target model corresponding to the power generation forecast node is added to the day-ahead data corresponding to the electricity price forecast node; when the current node is an electricity price forecast node, the predicted electricity price data output by the target model corresponding to the electricity price forecast node is added to the day-ahead data corresponding to the trading strategy forecast node; when the current node is a power forecast node, the predicted power generation data output by the target model corresponding to the power forecast node is added to the day-ahead data corresponding to the trading strategy forecast node.

[0117] In an optional embodiment, when the current node is a trading strategy prediction node, after step S205, the target prediction result is uploaded to the trading center.

[0118] If the current node is a trading strategy prediction node, then when step S205 is completed, it means that the power spot market trading decision process is completed. At this time, the target prediction result can be directly uploaded to the trading center. The target prediction result is the predicted trading strategy, which can be understood as the predicted power declaration amount.

[0119] In an optional embodiment, when the current node is a trading strategy prediction node, after step S205, the target prediction result can also be displayed on the client, and the staff can manually adjust the target prediction result and upload the adjusted data to the trading center.

[0120] In the present embodiment, a data prediction device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of 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 conceived.

[0121] This embodiment provides a data prediction device, such as Figure 4As shown, including:

[0122] The acquisition module 401 is used to acquire the actual data of the day and the first prediction results obtained by multiple pre-deployed original models based on the prediction of the day-ahead data, wherein the actual data of the day and the day-ahead data both correspond to the current node.

[0123] The evaluation module 402 is used to determine the prediction error of each original model based on the actual data of the day and multiple first prediction results, and to perform multi-dimensional evaluation on the prediction function of each original model to obtain the evaluation result of each original model.

[0124] The fusion module 403 is used to calculate the target weight of each original model according to the prediction error, the evaluation result and the preset initial weight, and fuse the original models according to the corresponding target weight to obtain a fusion model.

[0125] The selection module 404 is used to evaluate the evaluation results of each original model and the second prediction result obtained by the fusion model based on the day-ahead data prediction, combined with the actual data of the day, and select the target model of the current node from the fusion model and multiple original models.

[0126] The prediction module 405 is used to update the actual data of the day into the day-ahead data, and input the updated day-ahead data into the target model to obtain the target prediction result corresponding to the current node.

[0127] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0128] The data prediction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0129] The embodiment of the present invention also provides a computer device having the above Figure 4 The data prediction device shown.

[0130] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0131] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0132] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0133] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device 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.

[0134] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0135] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0136] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0137] The above are only specific embodiments of the present disclosure, so that those skilled in the art can understand or implement the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to these embodiments herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data prediction method, characterized in that: The data prediction method is applied to the power spot market transaction decision system, and is specifically applied to each node of the power spot market transaction decision process, including: Acquire actual data of the day and first prediction results obtained by multiple pre-deployed original models based on the prediction of the day-ahead data, wherein the actual data of the day and the day-ahead data both correspond to the current node; Determine the prediction error of each of the original models according to the actual data of the day and the first prediction results, and perform a multi-dimensional evaluation on the prediction function of each of the original models to obtain an evaluation result of each of the original models; Calculate the target weight of each of the original models according to the prediction error, the evaluation result and the preset initial weight, and fuse the original models according to the corresponding target weight to obtain a fusion model; According to the evaluation result of each of the original models and the second prediction result obtained by the fusion model based on the day-ahead data prediction, evaluation is performed in combination with the actual data of the day, and a target model of the current node is selected from the fusion model and the plurality of original models; The actual data of the day is updated to the data of the day before, and the updated data of the day before is input into the target model to obtain the target prediction result corresponding to the current node.

2. The data prediction method according to claim 1, characterized in that: The evaluation result includes an evaluation value corresponding to each dimension, and the target weight of each original model is calculated according to the prediction error, the evaluation result and the preset initial weight, including: According to the preset initial weights corresponding to the original models, the prediction errors are weighted summed to obtain a fused prediction error; Taking the minimum fusion prediction error as the goal and the preset initial weight and constant as the constraints, determining the first weights corresponding to the original models respectively; Determine the second weights corresponding to the original models according to the evaluation values ​​corresponding to the dimensions of the original models; The target weight is determined according to the first weight and the second weight.

3. The data prediction method according to claim 2, characterized in that: The determining the second weights corresponding to the original models according to the evaluation values ​​corresponding to the dimensions of the original models respectively includes: Calculate the third weights corresponding to the original models according to the evaluation values ​​corresponding to the dimensions of the original models by using the entropy weight method; Calculate the fourth weights corresponding to the original models according to the evaluation values ​​corresponding to the dimensions of the original models by using the hierarchical analysis method; The third weight and the fourth weight are normalized to obtain second weights corresponding to the original models.

4. The data prediction method according to claim 1, characterized in that: The nodes include: a meteorological forecast node, a load forecast node, a power generation forecast node, a power forecast node, an electricity price forecast node and a trading strategy forecast node, wherein the successor nodes of the meteorological forecast node include the load forecast node, the power generation forecast node and the power forecast node; the successor nodes of the load forecast node and the power generation forecast node are the electricity price forecast node, and the successor nodes of the electricity price forecast node and the power forecast node are the trading strategy prediction nodes.

5. The data prediction method according to claim 2, characterized in that: The step of evaluating the current day actual data based on the evaluation result of each of the original models and the second prediction result obtained by the fusion model based on the previous day data, and selecting the target model of the current node from the fusion model and the multiple original models includes: Determining a total evaluation value of the original model according to the evaluation result of the original model; According to the second prediction result and the actual data of the day, a multi-dimensional evaluation is performed on the prediction function of the fusion model to obtain an overall evaluation value of the fusion model; The total evaluation value of each of the original models is compared with the total evaluation value of the fusion model, and the model corresponding to the highest total evaluation value is determined as the target model of the current node.

6. The data prediction method according to claim 1, characterized in that: When the current node is not the last node of the power spot market transaction decision process, after updating the actual data of the day into the day-ahead data and inputting the updated day-ahead data into the target model to obtain the target prediction result corresponding to the current node, the method further includes: The target prediction result is uploaded to the trading center.

7. The data prediction method according to claim 1, characterized in that: When the current node is not the last node of the power spot market transaction decision process, after updating the actual data of the day into the day-ahead data and inputting the updated day-ahead data into the target model to obtain the target prediction result corresponding to the current node, the method further includes: The target prediction result is added to the day-ahead data corresponding to the successor node of the current node.

8. A data prediction device, characterized in that: include: An acquisition module, used to acquire actual data of the day and first prediction results obtained by a plurality of pre-deployed original models based on predictions of the day-ahead data, wherein both the actual data of the day and the day-ahead data correspond to the current node; An evaluation module, used to determine the prediction error of each of the original models according to the actual data of the day and the first prediction results, and to perform a multi-dimensional evaluation on the prediction function of each of the original models to obtain an evaluation result of each of the original models; A fusion module, used for calculating the target weight of each of the original models according to the prediction error, the evaluation result and the preset initial weight, and fusing each of the original models according to the corresponding target weight to obtain a fusion model; A selection module, configured to evaluate the actual data of the day according to the evaluation result of each of the original models and the second prediction result obtained by the fusion model based on the day-ahead data prediction, and select a target model of the current node from the fusion model and the plurality of original models; The prediction module is used to update the actual data of the day into the data of the day before, and input the updated data of the day before into the target model to obtain the target prediction result corresponding to the current node.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the data prediction method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the data prediction method according to any one of claims 1 to 7.