Electricity price prediction method and device based on combined time sequence model, equipment and medium
By constructing a combined timing model, including MLR, ARIMA, and ESN models, and returning processing and weighted summary, it solves the challenge of electricity prices in the power market, improves the accuracy and robustness of the prediction, reduces the calculation costs Essence
Patent Information
- Application Number
- CN202411423034.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-12
- Publication Date
- 2025-01-28
AI Technical Summary
The prior art is difficult to effectively predict electricity prices in the power market, especially in the context of renewable energy volatility and intermittentity, resulting in increased uncertainty in power supply.
Adopting the electric price prediction method based on the combined timing model, by constructing a multi -linear regression (MLR) model, a self -regulating point moving average (ARIMA) model, and echo status network (ESN) model, combined with naturalized processing and weighted methods, Perform electricity prices.
Improve the accuracy and robustness of electricity prediction, reduce the amount of calculation and calculation costs, and enhance the adaptability and predictive accuracy of the model.
Smart Images
Figure CN119379327A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field related to electricity price prediction, and specifically to an electricity price prediction method, device, equipment and medium based on a combined time series model. Background Art
[0002] With the transformation of the global energy structure and the advancement of the "dual carbon" goals, the power market is undergoing a transition from traditional energy such as coal to renewable energy such as wind and solar energy. This transition has brought complexity to the operation of the power system, because renewable energy has significant volatility and intermittency, resulting in increased uncertainty in power supply. This uncertainty poses new challenges to price forecasting in the power market.
[0003] In the power market, electricity price forecasting is a key link, which helps market participants make more informed investment decisions, conduct effective risk management, and optimize energy consumption plans. Accurate electricity price forecasting is crucial for power companies and market managers because it can guide them in planning future investments, reduce operational risks, and improve market efficiency. Currently, there is a lack of an effective electricity price forecasting method. Summary of the invention
[0004] In view of this, the embodiments of the present application are directed to providing an electricity price prediction method, apparatus, device and medium based on a combined time series model to perform electricity price prediction.
[0005] The present application provides an electricity price prediction method based on a combined time series model, including: constructing an MLR model, an ARIMA model and an ESN model respectively, for predicting electricity prices based on basic data after normalization;
[0006] Collect basic data;
[0007] Normalizing the basic data;
[0008] The MLR model, ARIMA model and ESN model are used to predict the electricity price based on the normalized processing, and three predicted values are obtained;
[0009] The predicted electricity price is obtained by weighted summing of the prediction results.
[0010] In some embodiments, the formula for normalization is:
[0011]
[0012] Among them, x i ∈X represents the i-th real data in the input data set, x′ i ∈X′ is the standardized value, i and n are the index number and total number of input data respectively.
[0013] In some embodiments, the prediction results are weighted and summed, and the formula is:
[0014]
[0015] In the formula, w1, w2, and w3 are weight factors; is the predicted value of the MLR model, is the predicted value of the ARIMA model, is the predicted value of the ESN model.
[0016] In some embodiments, the basic data includes: historical average transaction price.
[0017] The present application provides an electricity price prediction device based on a combined time series model, comprising:
[0018] A construction module is used to construct an MLR model, an ARIMA model and an ESN model respectively, and is used to predict electricity prices based on the basic data after normalization;
[0019] Collection module, used to collect basic data;
[0020] A normalization module, used for normalizing the basic data;
[0021] A prediction module is used to predict the electricity price through the MLR model, the ARIMA model and the ESN model respectively, based on the normalization process, to obtain three prediction values;
[0022] The summation module is used to perform weighted summation on the prediction results to obtain the predicted electricity price.
[0023] The present application provides an electronic device, including:
[0024] A processor, and a memory for storing a program executable by the processor;
[0025] The processor is used to implement the above-mentioned electricity price prediction method based on the combined time series model by running the program in the memory.
[0026] The present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor is enabled to execute the electricity price prediction method based on the combined time series model as described above.
[0027] The present application provides a computer program product, including a computer program, which implements the above-mentioned electricity price prediction method based on the combined time series model when executed by a processor.
[0028] The present application provides a method for predicting electricity prices based on a combined time series model. First, an MLR model, an ARIMA model and an ESN model are constructed respectively, which are used to predict electricity prices based on basic data after normalization processing; basic data are collected; the basic data are normalized; electricity prices are predicted based on the normalized data through the MLR model, the ARIMA model and the ESN model respectively, and three predicted values are obtained; and the predicted results are weighted summed to obtain the predicted electricity price. With such a setting, the solution provided by the present application performs data preprocessing, accelerates the data normalization process, reduces the amount of calculation, and reduces the computing cost. The MLR model, the ARIMA model and the ESN model are trained, and the three models are used to predict the same data set, and the prediction results of the three models are combined by the weighted average method to increase the model robustness and improve the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0030] Figure 1 It is a flowchart of an electricity price prediction method based on a combined time series model provided in one embodiment of the present application.
[0031] Figure 2 This is the basic principle of the ARIMA method in the method provided in one embodiment of the present application.
[0032] Figure 3 This is an ESN structure diagram in a method provided in an embodiment of the present application.
[0033] Figure 4 It is a structural schematic diagram of an electricity price prediction device based on a combined time series model provided in one embodiment of the present application.
[0034] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION
[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0036] Application Overview
[0037] With the realization of the dual carbon goals, the international electricity market is shifting from coal-based power resources to an increasing proportion of renewable energy. Renewable energy is different from traditional power sources. It has the characteristics of large volatility and strong intermittency. The actual output power fluctuates greatly, and the traditional method has low prediction accuracy.
[0038] Traditional methods mainly include multiple linear regression model (MLRM), autoregressive integrated moving average model (AIMAM), etc. However, traditional data-driven methods face the difficulties of large amount of calculation, frequent parameter setting, and the model is more sensitive to parameters, resulting in a long prediction time.
[0039] Traditional prediction methods have higher requirements on data quality and often assume that the relationship between variables is linear. However, in actual situations, many relationships are nonlinear and the calculation cost is high.
[0040] The present invention proposes a combined time series prediction method based on the multivariate linear regression model, the autoregressive integrated moving average model and the echo state network model. First, the data is normalized and the input data value is limited to the interval of 0 to 1. On this basis, a new combined model is designed based on the above prediction method. Compared with other conventional methods, the proposed model is more robust and has higher prediction accuracy, which can effectively reduce the amount of calculation and reduce the calculation cost.
[0041] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.
[0042] Exemplary Methods
[0043] Figure 1 FIG. 1 is a flow chart of an electricity price prediction method based on a combined time series model provided by an embodiment of the present application. Figure 1 As shown, the method includes the following contents.
[0044] Step S110, constructing an MLR model, an ARIMA model and an ESN model respectively, for predicting electricity prices based on the basic data after normalization;
[0045] Step S120, collecting basic data;
[0046] Among them, the basic data include: historical average transaction price.
[0047] Step S130, normalizing the basic data;
[0048] Step S140, predicting the electricity price based on the normalized data by using the MLR model, the ARIMA model and the ESN model to obtain three predicted values;
[0049] Step S150, performing weighted summation on the prediction results to obtain the predicted electricity price.
[0050] With such settings, the solution provided by this application performs data preprocessing, accelerates the data normalization process, reduces the amount of calculation, and reduces the computing cost. The MLR model, ARIMA model, and ESN model are trained, and the three models are used to predict the same data set, and the prediction results of the three models are combined using the weighted average method to increase the model robustness and improve the prediction accuracy.
[0051] In some embodiments, the data is analyzed for medium- and long-term trends, typical monthly trends, and holiday and weekday trends, and the data is normalized. In this way, the adverse effects of dimensional differences can be eliminated to ensure more accurate prediction modeling. The normalization formula is as follows:
[0052]
[0053] In the formula, x i ∈X represents
[0054] The i-th real data in the input data set, x i ′∈X′ is the normalized value, i and n are the index number and total number of input data respectively.
[0055] In the solution provided in this application, three different algorithms, namely, multivariate linear regression (MLR), autoregressive differential moving average model (ARIMA) and echo state network (ESN), are selected to form a prediction model library, and the models are trained as follows:
[0056] 1) Multiple linear regression modeling and training
[0057] The data is modeled using a linear prediction function, and the unknown parameters are estimated by the least squares approximation problem. The MLR model is:
[0058]
[0059] In the formula, is the predicted value of the target variable y; x′1, x′2, x′3, x′4 are input variables; β′1, β′2, β′3, β′4 are regression coefficients, indicating that the input variable x′ j The weight of is the intercept value, which is equal to the expected value when all input variables are 0
[0060] Based on the trend analysis results of the object to be predicted, the multivariate linear regression model is modeled by month, and the corresponding annual, monthly and intra-month type day feature data are organized for training to obtain parameter estimates for different months. The parameter training results are shown in Table 1 below:
[0061] Table 1 Significant parameters of the MLR model
[0062]
[0063]
[0064] As shown in the table, is a constant term, They are the weights of the annual average, monthly average, average of typical days and average of the previous day. When making a specific prediction, it is necessary to call the parameters of the corresponding month to complete the prediction based on the month in which the object to be predicted is located.
[0065] 2) Autoregressive integrated moving average modeling and training
[0066] By connecting the autoregressive (AR) part, the difference (I) part, and the moving average (MA) part, the ARIMA formula can be expressed as:
[0067] φ(L)(1-L) d y t =θ(t)ε t (3)
[0068]
[0069] Where p, d, q are the AR order, MA difference level and order respectively; L is the lag operator; L i It means to lag the time series by i periods.
[0070] The parameters of the ARIMA model need to be selected and estimated according to the actual time series data. Due to the complexity of the optional parameter combinations, AIC (Akaike Index) is selected as the preferred target in the modeling process, and the parameters are estimated using the optimization method based on the spatial network. The values of p, d, and q should be integers greater than zero, and d = 1 is set to keep the data stable.
[0071] Use the data of the previous 364 days as input variables to predict the price under a single-day step. Split the 96-point data of the whole year into 96 groups of data from 00:15 to 24:00, and train and model them separately to obtain 96 groups of model parameters. In actual prediction, the 96-point forecast data for the next day can be obtained at the same time. The process is as follows: Figure 2 As shown:
[0072] The important parameters of the ARIMA model are shown in Table 2:
[0073] Table 2 Important parameters of ARIMA model
[0074] time P D Q 0:15 1 1 3 0:30 1 1 1 0:45 0 1 2 1:00 0 1 3 1:15 1 1 1 1:30 2 1 2 1:45 1 1 1 2:00 2 1 2 2:15 2 1 1 2:30 1 1 2 2:45 1 1 2 3:00 1 1 1
[0075] 3) Echo State Network Modeling and Training
[0076] Figure 3 It is the standard structure of ESN, which mainly consists of three parts: input layer, storage layer and output layer. The values of L, N and K represent the number of units in the corresponding part. in , W out Respectively represent the connection weight matrix from the input layer to the storage layer, and from the storage layer to the output layer, W back , W represent the feedback connection weight matrix from the output layer to the storage layer and the internal connection weight matrix of the storage layer, respectively. Wherein, u(n), x(n), y(n) represent the network input, the network library state vector, and the network output at time slot n, respectively.
[0077] Input the input data into the network model to obtain the partial generation state of the reservoir layer and form a state matrix. Based on this, the network reservoir state vector x(n+1) can be calculated, and the expression is:
[0078] x(n+1)=f(W in u(n+1)+Wx(n)+W back y(n)) (5)
[0079] In the formula, f(·) indicates that the output function of the echo state network is an S-type hyperbolic tangent function. Based on this, the state matrix X of the network can be derived as equation (2), which consists of u(n), x(n) and the feedback between the output layer and the storage layer.
[0080]
[0081] Where n is the total number of network input data, (f out ) -1 is the inverse function of the feedback function.
[0082] The hyperparameters for training the echo state network model are:
[0083] n_inputs=1, the number of neurons in the input layer is 1, which means that under the current model, only one input parameter, historical volume and price, is considered.
[0084] n_outputs=1, the number of neurons in the output layer is 1, which means that under the current model, only a single quantity and price prediction result is needed.
[0085] n_reservoir=500, which indicates the number of neurons in the hidden layer. In principle, the larger the hidden layer, the more accurate the prediction result. However, there is a certain possibility that a larger number of neurons will result in a smaller improvement in accuracy.
[0086] sparsity=0.2 indicates the sparsity of hidden layer connections. Sparse connections help reduce the amount of computation while maintaining the dynamic characteristics of the network.
[0087] random_state=23, random number seed, used to initialize the neural network.
[0088] spectral_radius = 1.2, which indicates the maximum eigenvalue (or spectral radius) of the hidden layer weight matrix, and is used to maintain long-term dependencies.
[0089] Noise=0.005, represents the noise added to the hidden layer state.
[0090] These parameters together define the structure and characteristics of an ESN, allowing the network to be customized for specific time series prediction tasks. By adjusting these parameters, the performance of the network can be optimized to adapt to different application scenarios and data characteristics.
[0091] Execute the above three models respectively to predict the same prediction target. First, organize the corresponding annual and monthly feature data, organize the corresponding monthly feature number according to the date type, and the average transaction price of the previous day, apply the corresponding monthly parameters, and obtain the corresponding prediction formula (2); then, organize the time-sharing data of 364 days before the prediction date as input, apply the ARIMA parameters of the corresponding time points, and predict the price of the next day with a step value of 1; finally, from 00:15 to 24:00 of the next day, organize the first 35040 data as input in sequence, and predict the 96 points of data of the next day in sequence with a step value of 1.
[0092] After completing the predictions of the three different models, you can set different weights for the three different prediction values according to business needs or historical deviations. The formula is:
[0093]
[0094] In the formula, w1, w2, and w3 are weight factors; They are the predicted values of MLR model, ARIMA model and ESN model respectively.
[0095] Exemplary Devices
[0096] The device embodiments of the present application can be used to execute the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.
[0097] Figure 4 FIG. 1 is a block diagram of an electricity price prediction device based on a combined time series model provided by an embodiment of the present application. Figure 4 As shown, the device comprises:
[0098] A construction module 41 is used to construct an MLR model, an ARIMA model and an ESN model respectively, for predicting electricity prices based on the basic data after normalization;
[0099] A collection module 42, used for collecting basic data;
[0100] A normalization module 43, used for normalizing the basic data;
[0101] The prediction module 44 is used to predict the electricity price through the MLR model, the ARIMA model and the ESN model respectively based on the normalized processing to obtain three prediction values;
[0102] The summing module 45 is used to perform weighted summing of the prediction results to obtain the predicted electricity price.
[0103] Exemplary Electronic Devices
[0104] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application. Figure 5 A block diagram of an electronic device according to an embodiment of the present application is illustrated.
[0105] like Figure 5 As shown, electronic device 500 includes one or more processors 510 and memory 520 .
[0106] The processor 510 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 500 to perform desired functions.
[0107] The memory 520 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 510 may run the program instructions to implement the electricity price prediction method based on the combined time series model of each embodiment of the present application described above and / or other desired functions. Various contents such as category correspondences may also be stored in the computer-readable storage medium.
[0108] In one example, the electronic device 500 may further include: an input device 530 and an output device 540 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).
[0109] In addition, the input device 530 may also include, for example, a keyboard, a mouse, an interface, etc. The output device 540 may output various information to the outside, including analysis results, etc. The output device 540 may include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, etc.
[0110] Of course, to simplify, Figure 5 Only some of the components in the electronic device related to the present application are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application conditions.
[0111] Exemplary computer program products and computer-readable storage media
[0112] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the electricity price prediction method based on a combined time series model according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.
[0113] The computer program product may be written in any combination of one or more programming languages to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0114] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the steps of the electricity price prediction method based on a combined time series model according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.
[0115] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
[0116] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.
Claims
1. A method for predicting electricity prices based on a combined time series model, characterized in that: include: The MLR model, ARIMA model and ESN model are constructed respectively to predict electricity prices based on the normalized basic data; Collect basic data; Normalizing the basic data; The MLR model, ARIMA model and ESN model are used to predict the electricity price based on the normalized processing, and three predicted values are obtained; The predicted electricity price is obtained by weighted summing of the prediction results.
2. The electricity price prediction method based on the combined time series model according to claim 1 is characterized in that: The formula for normalization is: Among them, x i ∈X represents the i-th real data in the input data set, x′ i ∈X′ is the standardized value, i and n are the index number and total number of input data respectively.
3. The electricity price prediction method based on the combined time series model according to claim 1 is characterized in that: The prediction results are weighted and summed, and the formula is: In the formula, w1, w2, and w3 are weight factors; is the predicted value of the MLR model, is the predicted value of the ARIMA model, is the predicted value of the ESN model.
4. The electricity price prediction method based on the combined time series model according to claim 1, characterized in that: Basic data include: historical average transaction price.
5. An electricity price prediction device based on a combined time series model, characterized in that: include: A construction module is used to construct an MLR model, an ARIMA model and an ESN model respectively, and is used to predict electricity prices based on the basic data after normalization; Collection module, used to collect basic data; A normalization module, used for normalizing the basic data; A prediction module is used to predict the electricity price through the MLR model, the ARIMA model and the ESN model respectively, based on the normalization process, to obtain three prediction values; The summation module is used to perform weighted summation on the prediction results to obtain the predicted electricity price.
6. An electronic device, characterized in that: include: A processor, and a memory for storing a program executable by the processor; The processor is used to implement the electricity price prediction method based on the combined time series model as described in any one of claims 1 to 4 by running the program in the memory.
7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the electricity price prediction method based on a combined time series model as claimed in any one of claims 1 to 4.
8. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the electricity price prediction method based on a combined time series model according to any one of claims 1 to 4.