Multi-source composite node electricity price prediction method and system

Abstract

The invention discloses a multi-source composite node electricity price prediction method and system, and the method employs the big data analysis and machine learning technology, comprehensively considers a plurality of influence factors, carries out the strategy-level dynamic analysis, clustering, integration and fusion of multi-source electricity price prediction models (including multi-type, multi-manufacturer and multi-algorithm team prediction models), and achieves the prediction of the electricity price. And a composite electricity price prediction model is constructed to effectively improve the node electricity price prediction precision and provide reliable support for stable operation of the electricity market.

Description

Technical Field

[0001] The present invention relates to an electricity price prediction model and system, and in particular to a multi-source composite node electricity price prediction method and system. Background Art

[0002] The power system has gradually changed from a single centralized power supply mode to a multi-source distributed power supply mode. Due to its singleness and limitations, the traditional electricity price forecasting method has been difficult to adapt to the current complex power market needs. The complexity of the power market is reflected in the diversity of power sources, the volatility of load demand, and the influence of external environmental factors (such as weather, policies, etc.). As the core element of the power market, the prediction accuracy of electricity prices is directly related to the operating efficiency and economic benefits of the power system. There are many existing technologies for predicting electricity prices in the power market. For example, CN202011555954 "Method, system, device and storage medium for predicting marginal electricity prices in the power spot market" discloses a method for predicting node marginal electricity prices by processing power load forecast curves and simulated quotation curves of power generation enterprises; CN202011200533 "A method, system, computer device and storage medium for predicting electricity prices probabilistically" discloses a method for predicting electricity prices using feature clustering technology. However, since the electricity market is affected by a variety of factors, including but not limited to electricity supply and demand, fuel prices, policies and regulations, climate change, and electricity trading behavior, the above traditional single-source models (such as a single manufacturer, algorithm team) and methods usually rely on historical data and single-factor analysis, and are unable to deal with these complex factors and are unable to fully reflect the dynamic changes in the electricity market. In the actual electricity market, the electricity price of each node depends not only on the local power generation and load conditions, but also on the power flow of neighboring nodes and the external environment. Therefore, a single prediction model is difficult to provide accurate electricity price prediction results, which may lead to inefficient operation of the power system and may even cause systemic risks in the power market. Summary of the invention

[0003] Purpose of the invention: In view of the above problems, the present invention proposes a multi-source composite node electricity price prediction method and system, which can effectively improve the accuracy of node electricity price prediction and provide reliability support for the stable operation of the power market.

[0004] Technical solution: The technical solution adopted by the present invention is a multi-source composite node electricity price prediction method, comprising the following steps:

[0005] Step S1: predict that sources A-X are nodes F1-F in the specified area. N Training electricity price prediction models;

[0006] Step S2: Get all nodes F1 to F in the specified area. NThe actual electricity price data within the specified time period T constitutes the validation data set;

[0007] Step S3, obtaining the electricity price prediction model of each prediction source in the system within the specified time period T, and performing N The electricity price prediction results are used to form a prediction data set;

[0008] Step S4, using the validation data set and the prediction data set output by each prediction source model, calculate the scene adaptability weight coefficient c associated with the prediction source s ;

[0009] Step S5, using the validation data set and the prediction data set output by each prediction source model, calculate the environmental adaptability weight coefficient c associated with the prediction source w ;

[0010] Step S6: construct a composite model for a specific node in the region and calculate the accuracy weight coefficient ∈ associated with the prediction source model i ; Using the prediction source scene adaptability weight coefficient c s , Environmental adaptability weight coefficient c w and the accuracy weight coefficient of the electricity price prediction model ∈ i Calculate the multi-source fusion weight coefficient taking into account the scenario and environmental adaptability, and weight the output of each sub-model according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m ;

[0011] Step S7: on the forecast day, use the multi-source composite electricity price forecast model f that matches the node m Predict the electricity price for the node and evaluate the prediction results;

[0012] Step S8, screening and adjusting the prediction source sub-models participating in the composite model according to the composite model evaluation result;

[0013] Step S9, repeating steps S3 to S8 to obtain a new composite model.

[0014] Calculate the scenario adaptability weight coefficient c associated with the prediction source s A preferred solution comprises the following steps:

[0015] Step S1: Divide all nodes F1~FN in the region into K categories according to the scenario, marked as S1~SK respectively; where S1={F1,F2,...F k1},S2={F k1+1 ,F k1+2 ,...F k2}, ..., Sk = {F k2+1 ,F k2+2 ,...F N};

[0016] Step S2: Obtain the nodes F1 to FN in the verification data set Dv in the specified time period T = [t1, t2, ..., t nT ,] the actual electricity price P F1 ~P FN , where the historical electricity price P of any node Fi It is expressed as:

[0017] P Fi =[P Fi,t1 ,P Fi,t2 ,...,P Fi,tnT ]

[0018] In the formula, i is the node number, nT is the total number of prediction points in time period T;

[0019] Step S3: Obtain the prediction source X for nodes F1 to F in the short-term prediction data set Dp. N The predicted electricity price P within the specified time period T X1 ~P XN , where the historical electricity price P of any node Xi It is expressed as:

[0020] P Xi =[P Xi,t1 ,P Xi,t2 ,...,P Xi,tnT ]

[0021] Step S4: Calculate the prediction source X model for node F i The accuracy of electricity price prediction at time tj is R xi,tj , the calculation formula is:

[0022]

[0023] Step S5: Calculate the prediction source X model for node F i The average electricity price prediction accuracy within time period T is calculated as follows:

[0024]

[0025] In the formula, R xi,T For the prediction source X model, node F i The average electricity price prediction accuracy in time period T;

[0026] Step S6: Calculate the prediction accuracy R of the prediction source X model for the scenario Sk in the time period T k,T , the calculation formula is:

[0027]

[0028] In the formula, k1 is the number of nodes contained in the scene Sk;

[0029] Step S7: k,T After normalization, the scene adaptability weight coefficient cs of the prediction source X for a certain type of scene is obtained.

[0030] Calculate the environmental adaptability weight coefficient c associated with the prediction source w A preferred solution comprises the following steps:

[0031] Step S1: All nodes F1 to F N According to the regional average temperature in the specified time period T, it is divided into K categories, marked as W1~WK; where W1={F1,F2,...F k1}, W2={F k1+1 ,F k1+2 ,...F k2}, ..., Wk = {F k2+1 ,F k2+2 ,...F N};

[0032] Step S2: Obtain nodes F1 to F in the verification data set Dv respectively N In the specified time period T=[t1,t2,...,t nT ,] the actual electricity price P F1 ~P FN ;

[0033] Step S3: Obtain the prediction source X for nodes F1 to F in the short-term prediction data set Dp. N The predicted electricity price P within the specified time period T X1 ~P XN ;

[0034] Step S4: Calculate the prediction source X model for node F i The electricity price prediction accuracy W at time tj xi,tj , the calculation formula is:

[0035]

[0036] Step S5: Calculate the prediction accuracy W of the average electricity price of the node Fi in the time period T by the prediction source X model xi,T , the calculation formula is:

[0037]

[0038] Where W xi,T For the prediction source X model, node F i The average electricity price prediction accuracy in time period T;

[0039] Step S6: Calculate the prediction accuracy W of the prediction source X model for the climate category Wk in the time period T k,T , the calculation formula is:

[0040]

[0041] In the formula, k2 is the number of nodes contained in the environment class Wk;

[0042] Step S7: k,T Normalize and get the adaptability weight coefficient c of the prediction source X to a certain type of environment w .

[0043] Calculate the accuracy weight coefficient ∈ associated with the prediction source model i A preferred solution comprises the following steps:

[0044] Step S1: Selecting an accuracy threshold δ;

[0045] Step S2: Calculate the prediction accuracy H of the average electricity price of the station F in the specified time period T by the model Mi (i=1,2...n) i ;

[0046] Step S3: If Hi>δ, then include the model Mi into the list of models to be composited;

[0047] Step S4: repeat steps S2 to S3 to obtain the best electricity price prediction model list Ms = [M1, M2..., Mn];

[0048] Step S5: construct the basic composite model M of node F, and assume that the electricity price forecast value of the composite model M at time t in the historical time series T is M(t), and its mathematical expression is:

[0049]

[0050] In the formula, ∈ i is the weight coefficient, satisfying

[0051] Step S6: Considering the actual electricity price P(t) of node F at time t, construct the following least squares problem:

[0052]

[0053] Where e is the composite model error, e(t) is the prediction error of a single model, and n is the number of models;

[0054] Solve the least squares problem and determine the optimal weight coefficient combination ∈ i ~∈ k .

[0055] Furthermore, the multi-source fusion weight coefficient considering the scenario and environmental adaptability is calculated, and the output of each sub-model is weighted according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m , including the following steps:

[0056] Step S1: Obtain the scene category of the node F;

[0057] Step S2: Obtain the current environment category of node F;

[0058] Step S3: record any sub-model in the optimal electricity price prediction model list Ms as Mi, and calculate its scenario adaptability weight coefficient c according to the prediction source to which the model Mi belongs. si ;

[0059] Step S4: Calculate the environmental adaptability weight coefficient c according to the prediction source to which the model Mi belongs wi ;

[0060] Step S5: Calculate the accuracy weight coefficient ∈ of model Mi i ;

[0061] Step S6: Calculate the fusion weight coefficient, the formula is:

[0062]

[0063] In the formula, λ i is the fusion weight coefficient, n is the number of models;

[0064] Step S7: Iterate the best electricity price prediction model list Ms, repeat steps S4 to S6, and obtain the fusion weight coefficient vector λ of all models to be combined;

[0065] Step S8: Weight the output of each sub-model according to the fusion weight coefficient vector λ of all models to be combined to obtain the multi-source composite electricity price prediction model f m .

[0066] The present invention proposes a multi-source composite node electricity price prediction system, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the multi-source composite node electricity price prediction method is implemented when the processor executes the computer program.

[0067] The present invention proposes a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the multi-source composite node electricity price prediction method when executing the computer program.

[0068] The present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the multi-source composite node electricity price prediction method is implemented.

[0069] The present invention provides a computer program product, including a computer program and / or instructions, which implement the multi-source composite node electricity price prediction method when executed by a processor.

[0070] Beneficial effects: Compared with the prior art, the present invention has the following advantages: This method can identify the technical adaptability of the prediction source by comprehensively considering multiple factors such as scenarios and environments, thereby constructing a specific multi-source composite electricity price prediction model for each power system node in the region on the basis of comprehensive optimization, and generating optimized prediction results. In addition, the method can dynamically select and adjust the composite model according to the operation of the composite prediction model and multiple types of triggering means to improve the accuracy of electricity price prediction for all time periods. Compared with the prior art, the present invention has the ability to integrate node electricity price prediction models of multiple prediction sources (including but not limited to multiple manufacturers, algorithm teams, researchers, etc.) in one stop. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a schematic diagram of the process flow of the multi-source composite node electricity price prediction method of the present invention;

[0072] Figure 2 It is a working flow chart of the multi-source composite node electricity price prediction system described in the present invention. DETAILED DESCRIPTION

[0073] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0074] Example 1

[0075] The multi-source composite node electricity price prediction method described in the present invention utilizes big data analysis and machine learning technology, comprehensively considers multiple influencing factors, and constructs a composite electricity price prediction model by performing strategy-level dynamic analysis, clustering, integration and fusion on multi-source electricity price prediction models (including multi-category, multi-factory, and multi-algorithm team prediction models) to effectively improve the accuracy of node electricity price prediction and provide reliability support for the stable operation of the power market. The execution process of this method is as follows: Figure 1 As shown, the specific steps are as follows:

[0076] Step S1: The prediction sources A~X train an electricity price prediction model for nodes F1~FN in a specified area, and upload and deploy it in the disclosed system;

[0077] Step S2: Obtain the actual electricity price data of all nodes F1 to FN in the specified area within the specified time period T to form a verification data set Dv;

[0078] Step S3: Obtain the electricity price prediction results of nodes F1 to FN from the electricity price prediction models of each prediction source in the system within the specified time period T to form a prediction data set Dp;

[0079] Step S4: using the validation data set and the prediction data set of each prediction source model, calculating the scene adaptability weight coefficient cs associated with the prediction source;

[0080] Step S5: using the validation data set and the prediction data set of each prediction source model, calculating the environmental adaptability weight coefficient cw associated with the prediction source;

[0081] Step S6: For a specific node in the region, use the validation data set and the prediction data set of each prediction source model to calculate the accuracy weight coefficient ∈ associated with the prediction source model i ;

[0082] Step S7: Construct a multi-source fusion electricity price prediction mathematical model for specific nodes in the region. Use the prediction source scene adaptability weight coefficient cs, environmental adaptability weight coefficient cw and electricity price prediction model accuracy weight coefficient ∈ i Determine the fusion weight coefficient λ of each sub-model in the composite model i ;

[0083] Step S8: On the forecast day, use the multi-source composite electricity price forecasting model fm that matches the node to forecast the electricity price of the node, and evaluate the forecast results.

[0084] Step S9: According to the composite model evaluation result, screen and adjust the prediction source sub-models participating in the composite model, and repeat steps S3 to S9 to obtain a new composite model.

[0085] The core technologies in the multi-source composite node electricity price prediction method described in the present invention include four parts: prediction source scene adaptability evaluation and weight coefficient calculation, prediction source environment adaptability evaluation and weight coefficient calculation, electricity price prediction model accuracy evaluation and basic composite model construction, and multi-source fusion weight coefficient calculation taking into account scene and environmental adaptability. They are introduced in detail below.

[0086] (1) Prediction source scenario adaptability assessment and weight coefficient calculation

[0087] Different prediction sources adopt different technical routes, and their prediction models or products also have different skills and characteristics. For example, the electricity price prediction model of a certain prediction source may have good prediction accuracy in offshore scene nodes, but perform poorly in onshore scene nodes. In order to accurately predict node electricity prices, it is necessary to perform a quantitative evaluation of the scene adaptability of the electricity price prediction model of each prediction source. The present invention discloses a scene adaptability evaluation method, and the specific process is as follows:

[0088] Step S1: Divide all nodes F1~FN in the region into K categories according to the scenarios, including but not limited to onshore mountain nodes, offshore nodes, offshore nodes, desert nodes, etc., marked as S1~SK respectively; where S1={F1,F2,...F k1},S2={F k1+1 ,F k1+2 ,...F k2}, ..., Sk = {F k2+1 ,F k2+2 ,...F N}.

[0089] Step S2: Obtain the nodes F1 to FN in the verification data set Dv in the specified time period T = [t1, t2, ..., t nT ,] the actual electricity price P F1 ~P FN , where the historical electricity price PFi of any node can be expressed as:

[0090] P Fi =[P Fi,t1 ,P Fi,t2 ,...,P Fi,tnT ]

[0091] Where i is the node number, nT is the total number of prediction points in time period T, usually every 15 minutes is a prediction time point, that is, 1 day contains 96 prediction points.

[0092] Step S3: Obtain the prediction source X for nodes F1 to F in the short-term prediction data set Dp. N The predicted electricity price P within the specified time period T X1 ~P XN , where the historical electricity price PXi of any node can be expressed as:

[0093] P Xi =[P Xi,t1 ,P Xi,t2 ,...,P Xi,tnT ]

[0094] Step S4: Calculate the prediction source X model for node F i The accuracy of electricity price prediction at time tj is R xi,tj , the calculation formula is:

[0095]

[0096] Step S5: Calculate the prediction source X model for node F i The average electricity price prediction accuracy R in time period T xi,T , the calculation formula is:

[0097]

[0098] Step S6: Calculate the prediction accuracy R of the prediction source X model for the scenario Sk in the time period T k,T , the calculation formula is:

[0099]

[0100] Where k1 is the number of nodes contained in the scene Sk.

[0101] Step S7: R 1,T ~R k,T Quantifies the adaptability of the prediction source X to scenarios S1~SK, R k,T The larger the value, the better the prediction skill of the prediction source X for this type of scene. In order to reflect the scene adaptability of the prediction source in the composite model, it is normalized to obtain the scene adaptability weight coefficient cs of the prediction source X for a certain type of scene. The calculation formula is as follows:

[0102]

[0103] In the formula, R T It refers to the adaptability of the prediction source X to a certain type of scene in S1~SK, where k is the number of categories into which all nodes F1~FN in the region are divided according to the scene.

[0104] (2) Evaluation of environmental adaptability of prediction sources and calculation of weight coefficients

[0105] Similarly, the present invention discloses an environmental adaptability assessment method suitable for evaluating electricity price prediction models. The method uses the average temperature of a node in a specified time period as a feature to analyze the prediction skill of the prediction source environment. The specific process is as follows:

[0106] Step S1: All nodes F1 to F N According to the regional average temperature in the specified time period T, it is divided into K categories, marked as W1~WK; where W1={F1,F2,...F k1}, W2={F k1+1 ,F k1+2 ,...F k2}, ..., Wk = {F k2+1 ,F k2+2 ,...F N}.

[0107] Step S2: Obtain nodes F1 to F in the verification data set Dv respectively N In the specified time period T=[t1,t2,...,t nT ,] the actual electricity price P F1 ~PFN 。

[0108] Step S3: Obtain the predicted electricity price P of the prediction source X for nodes F1 to F in the short-term prediction dataset Dp N within the specified time period T. X1 ~P XN 。

[0109] Step S4: Calculate the electricity price prediction accuracy W of the prediction source X model for node F i at time tj, and the calculation formula is the same as before. xi,tj

[0110] Step S5: Calculate the average electricity price prediction accuracy W of the prediction source X model for node Fi within the time period T xi,T and the calculation formula is the same as before.

[0111] Step S6: Calculate the environmental correlation electricity price prediction accuracy W of the prediction source X model for climate type Wk within the time period T k,T and the calculation formula is:

[0112]

[0113] In the formula, k2 is the number of nodes included in environmental type Wk.

[0114] Step S7: W 1,T ~W k,T quantifies the adaptability of the prediction source X to environmental types W1 to WK. The larger the value of W k,T , the better the prediction skills of the prediction source X for this type of environment (temperature range). Similarly, perform normalization processing on it to obtain the adaptability weight coefficient cw of the prediction source X to a certain type of environment, and the calculation formula is as follows:

[0115]

[0116] In the formula, W T represents the fitness of the prediction source X to a certain type of environment.

[0117] (3) Evaluation of the accuracy of the electricity price prediction model and construction of the basic composite model

[0118] In the fusion prediction stage, the system's focus object changes to specific nodes F within the region and the electricity price prediction models M1 to MN deployed by each prediction source for them. The system uses the following steps to screen the best electricity price prediction models M1 to Mn (where n < N) for fusion and construct the basic composite model associated with node F:

[0119] Step S1: Select the accuracy threshold δ;

[0120] Step S2: According to the above steps, the prediction accuracy Hi of the average electricity price of the station F in the specified time period T by the calculation model Mi (i=1,2...n) is calculated, and the calculation formula is the same as before.

[0121] Step S3: If Hi>δ, the model Mi is included in the list of models to be combined; otherwise, the model is discarded.

[0122] Step S4: Repeat steps S2 to S3 to obtain the best electricity price prediction model list Ms = [M1, M2..., Mn].

[0123] Step S5: Construct the basic composite model M of node F. Assume that the electricity price forecast value of the composite model M at time t in the historical time series T is M(t), and its mathematical expression can be written as:

[0124]

[0125] In the formula, ∈ i is the weight coefficient, satisfying

[0126] Step S6: Optimizing the weight coefficient of the composite model. Considering the actual electricity price P(t) of node F at time t, the prediction error of the composite model is recorded as e(t), that is, In order to obtain the optimal weight coefficient ∈ of the combined model on the entire validation time series, the following least squares problem is constructed:

[0127]

[0128] Solving the above optimization problem, we can determine the optimal weight coefficient combination ∈ i ~∈ k , substituted into the model expression, the basic composite model can be built.

[0129] (4) Calculation of multi-source fusion weight coefficients taking into account scene and environmental adaptability

[0130] Part (3) introduces the determination of the optimal weight of the composite prediction model of a specific node in the least squares sense through accuracy analysis within a specified period. Conventional methods usually only use this weight sequence as the combined prediction weight for the next prediction period. The present invention also proposes a method for determining a fusion weight coefficient, which can comprehensively consider the scene and environmental adaptability in the composite model, thereby maximizing the dominant role of the advantage model in the composite model. Let the specific node in the region be denoted as F, and the steps for determining its fusion weight coefficient are as follows:

[0131] Step S1: Obtain the scene category of the node F;

[0132] Step S2: Obtain the current environment (temperature range) category of node F;

[0133] Step S3: any sub-model in the optimal electricity price prediction model list Ms is recorded as Mi. According to the prediction source to which the model Mi belongs, the scenario adaptability weight coefficient csi is calculated using method (1);

[0134] Step S4: Calculate the environmental adaptability weight coefficient cwi of the model Mi according to the prediction source to which it belongs using method (2);

[0135] Step S5: Calculate the accuracy weight coefficient ∈ of model Mi using method (3) i ;

[0136] Step S6: Calculate the fusion weight coefficient, the formula is:

[0137]

[0138] Step S7: Iterate the best electricity price prediction model list Ms, repeat steps S4 to S6, and obtain the fusion weight coefficient vector λ of all models to be combined;

[0139] Step S8: Substitute λ into the mathematical expression of the composite model M to obtain a multi-source composite model that considers multiple factors such as scene and environmental adaptability, and perform fusion prediction from the next prediction moment.

[0140] Example 2

[0141] The multi-source composite node electricity price prediction system of the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-source composite node electricity price prediction method is implemented.

[0142] The working flow diagram of the multi-source composite node electricity price prediction system of the present invention is as follows: Figure 2 shown.

[0143] Calculate the multi-source fusion weight coefficient taking into account the scenario and environmental adaptability, and weight the output of each sub-model according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m , including the following steps:

[0144] Step S1: Obtain the scene category of the node F;

[0145] Step S2: Obtain the current environment category of node F;

[0146] Step S3: record any sub-model in the optimal electricity price prediction model list Ms as Mi, and calculate its scenario adaptability weight coefficient c according to the prediction source to which the model Mi belongs. si ;

[0147] Step S4: Calculate the environmental adaptability weight coefficient c according to the prediction source to which the model Mi belongs wi ;

[0148] Step S5: Calculate the accuracy weight coefficient ∈ of model Mi i ;

[0149] Step S6: Calculate the fusion weight coefficient, the formula is:

[0150]

[0151] In the formula, λ i is the fusion weight coefficient, n is the number of models;

[0152] Step S7: Iterate the best electricity price prediction model list Ms, repeat steps S4 to S6, and obtain the fusion weight coefficient vector λ of all models to be combined;

[0153] Step S8: Weight the output of each sub-model according to the fusion weight coefficient vector λ of all models to be combined to obtain the multi-source composite electricity price prediction model f m .

[0154] Example 3

[0155] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the multi-source composite node electricity price prediction method when executing the computer program.

[0156] Example 4

[0157] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the multi-source composite node electricity price prediction method is implemented.

[0158] Example 5

[0159] In one embodiment, a computer program product is provided, including a computer program and / or instructions, which implement the multi-source composite node electricity price prediction method when executed by a processor.

[0160] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0161] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A multi-source composite node electricity price prediction method, characterized in that: The following steps are involved: Step S1: predict that sources A-X are nodes F1-F in the specified area. N Training electricity price prediction models; Step S2: Get all nodes F1 to F in the specified area. N The actual electricity price data within the specified time period T constitutes the validation data set; Step S3, obtaining the electricity price prediction model of each prediction source in the system within the specified time period T, and performing N The electricity price prediction results are used to form a prediction data set; Step S4, using the validation data set and the prediction data set output by each prediction source model, calculate the scene adaptability weight coefficient c associated with the prediction source s ; Step S5, using the validation data set and the prediction data set output by each prediction source model, calculate the environmental adaptability weight coefficient c associated with the prediction source w ; Step S6: construct a composite model for a specific node in the region and calculate the accuracy weight coefficient ∈ associated with the prediction source model i ; Using the prediction source scene adaptability weight coefficient c s , Environmental adaptability weight coefficient c w and the accuracy weight coefficient of the electricity price prediction model ∈ i Calculate the multi-source fusion weight coefficient taking into account the scenario and environmental adaptability, and weight the output of each sub-model according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m ; Step S7: on the forecast day, use the multi-source composite electricity price forecast model f that matches the node m Predict the electricity price for the node and evaluate the prediction results; Step S8, screening and adjusting the prediction source sub-models participating in the composite model according to the composite model evaluation result; Step S9, repeating steps S3 to S8 to obtain a new composite model.

2. The multi-source composite node electricity price prediction method according to claim 1 is characterized by: Calculate the scenario adaptability weight coefficient c associated with the prediction source s , including the following steps: Step S1: Divide all nodes F1~FN in the region into K categories according to the scenario, marked as S1~SK respectively; where S1={F1,F2,...F k1 },S2={F k1+1 ,F k1+2 ,...F k2 }, ..., Sk = {F k2+1 ,F k2+2 ,...F N }; Step S2: Obtain the nodes F1 to FN in the verification data set Dv in the specified time period T = [t1, t2, ..., t nT ,] the actual electricity price P F1 ~P FN , where the historical electricity price P of any node Fi It is expressed as: P Fi =[P Fi,t1 ,P Fi,t2 ,...,P Fi,tnT ] In the formula, i is the node number, nT is the total number of prediction points in time period T; Step S3: Obtain the prediction source X for nodes F1 to F in the short-term prediction data set Dp. N The predicted electricity price P within the specified time period T X1 ~P XN , where the historical electricity price P of any node Xi It is expressed as: P Xi =[P Xi,t1 ,P Xi,t2 ,...,P Xi,tnT ] Step S4: Calculate the prediction source X model for node F i The accuracy of electricity price prediction at time tj is R xi,tj , the calculation formula is: Step S5: Calculate the prediction source X model for node F i The average electricity price prediction accuracy within the time period T is calculated as follows: In the formula, R xi,T For the prediction source X model, node F i The average electricity price prediction accuracy in time period T; Step S6: Calculate the prediction accuracy R of the prediction source X model for the scenario Sk in the time period T k,T , the calculation formula is: In the formula, k1 is the number of nodes contained in the scene Sk; Step S7: k,T After normalization, the scene adaptability weight coefficient cs of the prediction source X for a certain type of scene is obtained.

3. The multi-source composite node electricity price prediction method according to claim 1, characterized in that: Calculate the environmental adaptability weight coefficient c associated with the prediction source w , including the following steps: Step S1: All nodes F1 to F N According to the regional average temperature in the specified time period T, it is divided into K categories, marked as W1~WK; where W1={F1,F2,...F k1 }, W2={F k1+1 ,F k1+2 ,...F k2 }, ..., Wk = {F k2+1 ,F k2+2 ,...F N }; Step S2: Obtain nodes F1 to F in the verification data set Dv respectively N In the specified time period T=[t1,t2,...,t nT ,] the actual electricity price P F1 ~P FN ; Step S3: Obtain the prediction source X for nodes F1 to F in the short-term prediction data set Dp. N The predicted electricity price P within the specified time period T X1 ~P XN ; Step S4: Calculate the prediction source X model for node F i The electricity price prediction accuracy W at time tj xi,tj , the calculation formula is: Step S5: Calculate the prediction accuracy W of the average electricity price of the node Fi in the time period T by the prediction source X model xi,T , the calculation formula is: Where W xi,T For the prediction source X model, node F i The average electricity price prediction accuracy in time period T; Step S6: Calculate the prediction accuracy W of the prediction source X model for the climate category Wk in the time period T k,T , the calculation formula is: In the formula, k2 is the number of nodes contained in the environment class Wk; Step S7: k,T Normalize and get the adaptability weight coefficient c of the prediction source X to a certain type of environment w .

4. The multi-source composite node electricity price prediction method according to claim 1, characterized in that: Construct a composite model and calculate the accuracy weight coefficient ∈ associated with the prediction source model i , including the following steps: Step S1: Selecting an accuracy threshold δ; Step S2: Calculate the prediction accuracy H of the average electricity price of the station F in the specified time period T by the model Mi (i=1,2...n) i ; Step S3: If Hi>δ, then include the model Mi into the list of models to be composited; Step S4: repeat steps S2 to S3 to obtain the best electricity price prediction model list Ms = [M1, M2..., Mn]; Step S5: construct the basic composite model M of node F, and assume that the electricity price forecast value of the composite model M at time t in the historical time series T is M(t), and its mathematical expression is: In the formula, ∈ i is the weight coefficient, satisfying Step S6: Considering the actual electricity price P(t) of node F at time t, construct the following least squares problem: Where e is the composite model error, e(t) is the prediction error of a single model, and n is the number of models; Solve the least squares problem and determine the optimal weight coefficient combination ∈ i ~∈ k .

5. The multi-source composite node electricity price prediction method according to claim 1, characterized in that: Calculate the multi-source fusion weight coefficient taking into account the scenario and environmental adaptability, and weight the output of each sub-model according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m , including the following steps: Step S1: Obtain the scene category of the node F; Step S2: Obtain the current environment category of node F; Step S3: record any sub-model in the optimal electricity price prediction model list Ms as Mi, and calculate its scenario adaptability weight coefficient c according to the prediction source to which the model Mi belongs. si ; Step S4: Calculate the environmental adaptability weight coefficient c according to the prediction source to which the model Mi belongs wi ; Step S5: Calculate the accuracy weight coefficient ∈ of model Mi i ; Step S6: Calculate the fusion weight coefficient, the formula is: In the formula, λ i is the fusion weight coefficient, n is the number of models; Step S7: Iterate the best electricity price prediction model list Ms, repeat steps S4 to S6, and obtain the fusion weight coefficient vector λ of all models to be combined; Step S8: Weight the output of each sub-model according to the fusion weight coefficient vector λ of all models to be combined to obtain the multi-source composite electricity price prediction model f m .

6. A multi-source composite node electricity price prediction system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the multi-source composite node electricity price prediction method according to any one of claims 1 to 5 is implemented.

7. The multi-source composite node electricity price prediction system according to claim 6, characterized in that: Calculate the multi-source fusion weight coefficient taking into account the scenario and environmental adaptability, and weight the output of each sub-model according to the multi-source fusion weight coefficient to obtain the multi-source composite electricity price prediction model f m , including the following steps: Step S1: Obtain the scene category of the node F; Step S2: Obtain the current environment category of node F; Step S3: record any sub-model in the optimal electricity price prediction model list Ms as Mi, and calculate its scenario adaptability weight coefficient c according to the prediction source to which the model Mi belongs. si ; Step S4: Calculate the environmental adaptability weight coefficient c according to the prediction source to which the model Mi belongs wi ; Step S5: Calculate the accuracy weight coefficient ∈ of model Mi i ; Step S6: Calculate the fusion weight coefficient, the formula is: In the formula, λ i is the fusion weight coefficient, n is the number of models; Step S7: Iterate the best electricity price prediction model list Ms, repeat steps S4 to S6, and obtain the fusion weight coefficient vector λ of all models to be combined; Step S8: Weight the output of each sub-model according to the fusion weight coefficient vector λ of all models to be combined to obtain the multi-source composite electricity price prediction model f m .

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the multi-source composite node electricity price prediction method according to any one of claims 1 to 5 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multi-source composite node electricity price prediction method according to any one of claims 1 to 5 is implemented.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or instruction is executed by a processor, the multi-source composite node electricity price prediction method according to any one of claims 1 to 5 is implemented.

Images