Electricity price probability prediction model construction method and device based on big data, storage medium and terminal

Abstract

The invention discloses a big data-based electricity price probability prediction model construction method and device, a storage medium and a terminal, and the method comprises the steps: obtaining an electricity price feature data set of a target region within a preset time period, the electricity price feature data set comprising a plurality of electricity price feature data sets; carrying out image processing on the electricity price characteristic data group set to obtain an electricity price characteristic graph, and dividing the electricity price characteristic graph through a preset division mode to obtain a plurality of parameter characteristic blocks and electricity price labels corresponding to the parameter characteristic blocks, forming a training data set based on all the parameter feature blocks and the electricity price tag corresponding to each parameter feature block; and training a preset deep learning network model based on the training data set to obtain an electricity price probability prediction model. According to the method, market change characteristics can be fully reflected, the reliability of an electricity price prediction result is improved, and the obtained model can better embody the advantages of the model in terms of space and time, so that the model can learn and adapt to a complex electricity price change mode.

Description

Technical Field

[0001] The present invention relates to the technical field of power market, and in particular to a method and device for constructing an electricity price probability prediction model based on big data, a storage medium and a terminal. Background Art

[0002] As the degree of market competition in the power industry gradually deepens, the electricity wholesale market settlement based on the node electricity price theory will change with different market conditions, and the volatility of electricity prices will become increasingly obvious. It is very important to conduct scientific and accurate electricity price forecasts as a reference coefficient for the actual electricity price.

[0003] The probability of electricity price is a reference factor for electricity price pricing that is often used in the power industry. That is, the bidding strategy of power generators participating in the wholesale market often refers to the probability of electricity price to estimate future electricity prices, and hopes to achieve future electricity supply layout through its better prediction of electricity prices. Therefore, the introduction of scientific and feasible probability prediction of electricity price is of great significance to the main body of power generation.

[0004] The significance of electricity price probability mining is to discover and understand price changes and trends in the electricity market, and to help decision makers formulate effective energy policies and market strategies. The fluctuation of price probability in the electricity market has important risk management significance for energy companies, electricity consumers and investors. Through the mining and analysis of electricity prices, it is possible to predict price probability trends, identify risks and opportunities, and formulate corresponding risk management strategies, thereby improving energy utilization efficiency and reliability, and reducing the costs and risks caused by supply and demand imbalances. Promoting the development of renewable energy and formulating reasonable energy policies will promote the healthy operation and sustainable development of the energy market.

[0005] Artificial intelligence algorithm is an analytical method in the current electricity price probability prediction method. It is mainly based on machine learning and deep learning. It extracts features of key factors affecting electricity prices to achieve prediction. It can better predict the trend of electricity price probability fluctuations, and can also better respond to peak and valley electricity prices. It has better prediction accuracy, so it has gradually become the main research direction in recent years. However, the current artificial intelligence algorithm has the problem of a single type of characteristic value, that is, the extracted features only contain relevant information disclosed by the market or the results obtained after simple addition, subtraction and ratio operations between market disclosed information. There is no obvious distinction between the weights of each characteristic value, and the correlation between each feature is not strong, making it difficult to improve the prediction accuracy to a higher level. There is also a problem that some characteristic values ​​are missing, resulting in the failure of the algorithm; secondly, the traditional BP neural network used also has the problem that the gradient descent process is prone to gradient vanishing or overfitting, which makes the model unable to converge to the optimal point, so there is a problem of low prediction accuracy.

[0006] At the same time, the existing electricity price probability forecasting methods also have the problem of a single algorithm type, which limits the diversity and flexibility of electricity price probability forecasting, fails to highlight the inherent characteristics of each feature and the relationship between each feature, resulting in poor forecasting results in certain specific situations. Therefore, relying solely on a single algorithm may not be able to fully capture these complex relationships, thus affecting the accuracy and reliability of the forecast. Summary of the invention

[0007] In view of the shortcomings of the prior art mentioned above, the purpose of the present application is to provide a method and device for constructing an electricity price probability prediction model based on big data, a storage medium and a terminal, so as to solve the problems that the existing electricity price probability prediction method has a single type of characteristic values, the correlation between the characteristics is not strong, resulting in low accuracy of electricity price probability prediction, and the single type of algorithm used cannot fully capture the complex relationship between the characteristic values, which in turn affects the accuracy and reliability of electricity price probability prediction.

[0008] In a first aspect, the present application provides a method for constructing an electricity price probability prediction model based on big data, comprising:

[0009] Acquire a set of electricity price characteristic data groups in a target area within a preset time period, wherein the set of electricity price characteristic data groups includes a plurality of electricity price characteristic data groups, and the electricity price characteristic data groups include a momentary electricity price at a collection moment and a plurality of relevant parameter characteristic subgroups;

[0010] The electricity price characteristic data set is processed into an image to obtain an electricity price characteristic graph, the electricity price characteristic graph is divided into multiple parameter characteristic blocks and an electricity price label corresponding to each parameter characteristic block by a preset division method, and a training data set is formed based on all the parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block;

[0011] Based on the training data set, a preset deep learning network model is trained to obtain an electricity price probability prediction model;

[0012] Among them, the preset deep learning network model is a CNN_LSTM combined model, and multiple related parameter feature subgroups include an empirical reference feature subgroup, and the empirical reference feature subgroup includes a maximum startup ratio feature value, a reference price ratio feature value, a moment startup capacity feature value, a moment reference price ratio feature value and a negative electricity price state feature value.

[0013] In one embodiment of the present application, the electricity price characteristic data set is processed into an image to obtain an electricity price characteristic graph, including:

[0014] The momentary electricity prices and characteristic values ​​of relevant parameters in each of the electricity price characteristic data groups are horizontally sorted in a preset order, and each data in the sorted electricity price characteristic data group is used as a single pixel, and the data value of each data is used as the pixel value of the corresponding pixel. All the sorted electricity price characteristic data groups are arranged vertically in sequence to form an electricity price characteristic graph.

[0015] In one embodiment of the present application, the electricity price characteristic diagram is divided by a preset division method to obtain a plurality of parameter characteristic blocks and the electricity price label corresponding to each of the parameter characteristic blocks includes:

[0016] The electricity price characteristic diagram is divided into a plurality of divided characteristic blocks in a sliding window division manner, a module formed by all relevant parameter characteristic values ​​in each of the divided characteristic blocks is used as a parameter characteristic block corresponding to the divided characteristic block, and an electricity price label corresponding to the divided characteristic block is obtained according to a preset electricity price acquisition method based on the moment electricity price in each of the divided characteristic blocks;

[0017] The width of the image formed by all the data pixels in the electricity price characteristic data group is used as the sliding window width.

[0018] In one embodiment of the present application, obtaining the electricity price tag of the partitioned feature block according to a preset electricity price acquisition method based on the moment electricity price in the partitioned feature block includes:

[0019] The electricity price at a preset position in each of the divided feature blocks is used as the electricity price label of the corresponding parameter feature block, or the average of all the electricity prices at the moments in each of the divided feature blocks is used as the electricity price label of the corresponding parameter feature block.

[0020] In one embodiment of the present application, all relevant parameter characteristic values ​​in the electricity price characteristic data group need to be subjected to Min-Max normalization processing.

[0021] In one embodiment of the present application, the multiple related parameter feature subgroups also include a load proportion feature subgroup, a market energy coupling feature subgroup, a coal information feature subgroup, and a meteorological element feature subgroup.

[0022] In one embodiment of the present application, the preset deep learning network model includes an input layer, a convolution layer, an activation function layer, a pooling layer, a dropout layer, multiple time series layers, a transition output layer, an activation function layer and an output layer.

[0023] In a second aspect, the present application provides a device for constructing an electricity price probability prediction model based on big data, including a data acquisition module, a training data set acquisition module and a model training module;

[0024] The data acquisition module is used to acquire a set of electricity price characteristic data groups of a target area within a preset time period, wherein the set of electricity price characteristic data groups includes a plurality of electricity price characteristic data groups, and the electricity price characteristic data groups include a momentary electricity price at a collection moment and a plurality of relevant parameter characteristic subgroups;

[0025] The training data set acquisition module is used to image the electricity price feature data set to obtain an electricity price feature graph, divide the electricity price feature graph in a preset division method to obtain multiple parameter feature blocks and an electricity price label corresponding to each parameter feature block, and form a training data set based on all the parameter feature blocks and the electricity price label corresponding to each parameter feature block;

[0026] The model training module is used to train a preset deep learning network model based on the training data set to obtain an electricity price probability prediction model;

[0027] Among them, the preset deep learning network model is a CNN_LSTM combined model, and multiple related parameter feature subgroups include an empirical reference feature subgroup, and the empirical reference feature subgroup includes a maximum startup ratio feature value, a reference price ratio feature value, a moment startup capacity feature value, a moment reference price ratio feature value and a negative electricity price state feature value.

[0028] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for constructing an electricity price probability prediction model based on big data is implemented.

[0029] In a fourth aspect, the present application provides a terminal, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method for constructing an electricity price probability prediction model based on big data.

[0030] Compared with the prior art, one or more embodiments of the above scheme may have the following advantages or beneficial effects:

[0031] The method for constructing an electricity price probability prediction model based on big data provided by an embodiment of the present invention is applied, and the maximum startup ratio characteristic value, the reference price ratio characteristic value, the moment startup capacity characteristic value, the moment reference price ratio characteristic value and the negative electricity price state characteristic value are combined with other relevant parameter characteristic values ​​as an electricity price characteristic data group. The empirical reference characteristic value has a large correlation with the market energy coupling characteristic value. Therefore, the introduction of the empirical reference characteristic value can fully reflect the market change characteristics and increase the reliability of the electricity price prediction result; the relevant parameter characteristic values ​​in the electricity price characteristic data group are embodied in the form of pixels to form an electricity price characteristic map, and the relevant parameter characteristic values ​​are normalized to strengthen the correlation between the features, and the electricity price characteristic map is divided to realize the conversion of electricity price features into image features. The CNN_LSTM model is trained based on a training data set with image features, which can better reflect the spatial and temporal advantages between the features, so that the model can learn and adapt to complex electricity price change patterns. The electricity price probability prediction model using this method can make the obtained electricity price probability results more accurate, solving the problem of low accuracy and reliability of current electricity price probability prediction.

[0032] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0034] Figure 1 Shown is a flow chart of the method for constructing an electricity price probability prediction model based on big data as described in an embodiment of the present application.

[0035] Figure 2 Shown is a schematic diagram of the division of an electricity price characteristic graph in the method for constructing an electricity price probability prediction model based on big data described in an embodiment of the present application.

[0036] Figure 3 The figure shows a structural schematic diagram of a preset deep learning network model in the method for constructing an electricity price probability prediction model based on big data described in an embodiment of the present application.

[0037] Figure 4 Shown is a schematic diagram comparing the electricity price probability and the actual electricity price in the Shanxi market obtained by using the electricity price probability prediction model in the method for constructing an electricity price probability prediction model based on big data described in an embodiment of the present application.

[0038] Figure 5It is a schematic diagram of model performance parameters of the electricity price probability prediction model in the method for constructing an electricity price probability prediction model based on big data described in an embodiment of the present application.

[0039] Figure 6 Shown is a schematic diagram of the structure of the device for constructing an electricity price probability prediction model based on big data as described in an embodiment of the present application.

[0040] Figure 7 Shown is a schematic diagram of the structure of the terminal described in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following will describe the implementation methods of the present invention in detail with reference to the accompanying drawings and embodiments, so that the implementation process of how the present invention applies technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that as long as there is no conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other, and the technical solutions formed are all within the protection scope of the present invention.

[0042] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application, and thus the drawings only show components related to the present application rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.

[0043] The following embodiments of the present application provide a method and device for constructing an electricity price probability prediction model based on big data, a storage medium and a terminal, which solve the problem that the existing electricity price prediction method has a single type of characteristic value, resulting in low electricity price prediction accuracy, and the single type of algorithm used cannot fully capture the complex relationship between the characteristic values, thereby affecting the accuracy and reliability of electricity price prediction.

[0044] The following will describe in detail the principles and implementation methods of a method and device for constructing an electricity price probability prediction model based on big data, a storage medium and a terminal of this embodiment in conjunction with the accompanying drawings, so that those skilled in the art can understand the method and device for constructing an electricity price probability prediction model based on big data, a storage medium and a terminal of this embodiment without creative work.

[0045] like Figure 1 As shown, this embodiment provides a method for constructing an electricity price probability prediction model based on big data, including the following steps.

[0046] Step S101, obtaining a set of electricity price characteristic data groups of a target area within a preset time period.

[0047] The electricity price characteristic data set includes multiple electricity price characteristic data sets, each of which includes the electricity price at the time of collection and multiple related parameter characteristic subgroups. Furthermore, the multiple related parameter characteristic subgroups include an experience reference characteristic subgroup, wherein the experience reference characteristic subgroup includes a maximum startup ratio characteristic value, a reference price ratio characteristic value, a startup capacity characteristic value at the time, a reference price ratio characteristic value at the time, and a negative electricity price state characteristic value.

[0048] Step S102, the electricity price characteristic data set is imaged to obtain an electricity price characteristic graph, the electricity price characteristic graph is divided by a preset division method to obtain multiple parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block, and a training data set is formed based on all the parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block.

[0049] Step S103: training a preset deep learning network model based on the training data set to obtain an electricity price probability prediction model. The preset deep learning network model is a CNN_LSTM combined model.

[0050] The method for constructing an electricity price probability prediction model based on big data provided by an embodiment of the present invention combines the maximum startup ratio characteristic value, the reference price ratio characteristic value, the moment startup capacity characteristic value, the moment reference price ratio characteristic value and the negative electricity price state characteristic value with other relevant parameter characteristic values ​​as an electricity price characteristic data group. The empirical reference characteristic value has a large correlation with the market energy coupling characteristic value. Therefore, the introduction of the empirical reference characteristic value can fully reflect the market change characteristics and increase the reliability of the electricity price prediction result. The relevant parameter characteristic values ​​in the electricity price characteristic data group are embodied in the form of pixels to form an electricity price characteristic graph, and the relevant parameter characteristic values ​​are normalized to strengthen the correlation between the features, and the electricity price characteristic graph is divided to realize the conversion of the electricity price features into image features. The CNN_LSTM model is trained based on a training data set with image features, which can better reflect the spatial and temporal advantages between the features, so that the model can learn and adapt to complex electricity price change patterns. The electricity price probability prediction model using this method can make the obtained electricity price probability results more accurate, solving the problem of low accuracy and reliability of the current electricity price probability prediction.

[0051] The following is a detailed description of steps S101 to S103 of the method for constructing an electricity price probability prediction model based on big data in this embodiment.

[0052] Step S101, obtaining a set of electricity price characteristic data groups of a target area within a preset time period.

[0053] The area to be predicted for electricity prices is taken as the target area, and a set of electricity price characteristic data groups in the target area within a preset time period can be obtained by big data data capture and other methods. The set of electricity price characteristic data groups includes multiple electricity price characteristic data groups. Specifically, multiple collection moments can be set every day, and characteristic data collection is performed for every day within a period of time. The characteristic data collected at each moment is a set of electricity price characteristic data groups. For example, characteristic data collection can be performed every fifteen minutes, and the collection continues for half a year, and then all the collected electricity price characteristic data groups are combined into an electricity price characteristic data group set.

[0054] The electricity price characteristic data group specifically includes the instantaneous electricity price at the time of collection and multiple related parameter characteristic subgroups. Further, multiple related parameter characteristic subgroups include an empirical reference characteristic subgroup. Among them, the empirical reference characteristic subgroup includes the maximum startup ratio characteristic value, the reference price ratio characteristic value, the instant startup capacity characteristic value, the instant reference price ratio characteristic value and the negative electricity price state characteristic value. Among them, the value-taking logic of the maximum startup ratio characteristic value is "When the bidding unit space (ie conventional unit)" / "maximum bidding space" <0.5, the maximum startup ratio characteristic value is equal to 0.5; "When the bidding unit space (ie conventional unit)" / "maximum bidding space">0.5, the maximum startup ratio characteristic value is equal to "bidding unit space" / "maximum bidding space"". The logic for taking the reference price ratio feature value is "For the data collected at the first time of each day, the reference price = the "Day-ahead clearing price" at the latest time of the previous day; for data collected at other times, if the "maximum power-on ratio" > 50%, then the reference price = the "reference price" of the first data / (the "maximum power-on ratio" of the first data - 50%)*(the "maximum power-on ratio" at the same time - 50%), if the "maximum power-on ratio" <= 50%, then the reference price = 0; further, if "reference price = 0", then the reference price ratio feature value = -1, otherwise the reference price ratio feature value = "reference price" / 1500.". The logic for taking the feature value of power-on capacity at a given moment is "maximum power-on capacity × maximum power-on ratio". The logic for taking the feature value of the reference price ratio at a given moment is "if "reference price = 0", then the reference ratio = -1; otherwise the reference price ratio at a given moment = "reference price" / "the maximum value of the reference price of the day"". The logic for determining the negative electricity price state characteristic value is "if the reference price < 0, the negative electricity price state is 1, otherwise it is 0".

[0055] Conventional units are conventional generating units, the maximum startup ratio is the maximum startup ratio of conventional units, and the maximum startup capacity is the maximum startup power capacity of conventional units.

[0056] At the same time, multiple relevant parameter feature subgroups also include load proportion feature subgroup, market energy coupling feature subgroup, coal information feature subgroup and meteorological element feature subgroup. The load proportion feature subgroup includes the unified dispatching load characteristic value ("unified dispatching load" / "total installed capacity"), wind power load ("total wind power" / "wind power capacity"), photovoltaic load ("total photovoltaic" / "photovoltaic capacity"), conventional unit load (total conventional unit = unified dispatching load-wind power-photovoltaic-non-market unit; conventional unit load proportion = total conventional unit / historical conventional unit total maximum value*1.2), non-market unit output ("non-market unit output" / "total installed capacity (i.e., total installed capacity of State Grid New Energy Cloud)") and other related line load proportions. Non-market units are non-market power generating units.

[0057] The market energy coupling feature subgroups include bidding unit space (“bidding unit space (i.e. conventional units)” / “thermal power capacity”), wind power flexibility (the first data of each day, wind power flexibility = the second data “wind power flexibility”*2-the third data “wind power flexibility”; the last data of each day, wind power flexibility = the second to last data “wind power flexibility”*2-the third to last data “wind power flexibility”; other data, wind power flexibility = “wind power load (proportion)” at the corresponding moment-“wind power load (proportion)” at the previous moment), photovoltaic flexibility (the first data of each day, photovoltaic flexibility = the second data “photovoltaic flexibility”*2-the third data “photovoltaic flexibility”; the last data of each day, photovoltaic flexibility = the second to last data “photovoltaic flexibility”*2-the third to last data “photovoltaic flexibility”; other data, photovoltaic flexibility = the corresponding moment =The current moment's "PV load (percentage)" - the previous moment's "PV load (percentage)")), bidding capacity ratio ("bidding unit space (i.e. conventional units)" / ("unified dispatching load" + "|interconnection line load|")), new energy output ratio (("total wind power" + "total PV power") / ("unified dispatching load" + "|interconnection line load|")), unit switching (divided into three time periods every day: 0:15-8:00 in the morning, value 3; 8:15-16:00 in the daytime, value 2; 16:15-0:00 in the evening), maximum wind power output ratio (maximum wind power load on the day / total wind power installed capacity), wind power ramp-down difference ((wind power load - maximum wind power load on the day) / maximum wind power load), minimum wind power output ratio (minimum wind power load on the day / total wind power installed capacity), wind power ramp-down difference ((wind power load - minimum wind power load on the day) / minimum wind power load), etc.

[0058] The coal information feature subgroup includes coal price (coal price data), etc. The meteorological element feature subgroup includes temperature (temperature data), precipitation (precipitation data), and snowfall (snowfall), etc.

[0059] After obtaining the electricity price characteristic data set, there is a problem that the characteristic values ​​of the relevant parameters in the electricity price characteristic data set are of different dimensions due to the wide variety of characteristic values. In order to solve this problem, it is necessary to normalize all the characteristic values ​​of the relevant parameters in the electricity price characteristic data set. Specifically, the Min-Max normalization method can be used to normalize the characteristic values ​​of the relevant parameters in the electricity price characteristic data set, and the expression is:

[0060]

[0061] Among them, the characteristic value of each relevant parameter in the electricity price characteristic data group is x 1 , x 2 , LL x n ,y i For x i The normalized value of Min-Max. The normalization process is performed on the characteristic values ​​of relevant parameters in all electricity price characteristic data groups.

[0062] Step S102, the electricity price characteristic data set is imaged to obtain an electricity price characteristic graph, the electricity price characteristic graph is divided by a preset division method to obtain multiple parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block, and a training data set is formed based on all the parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block.

[0063] In order to enable the acquired one-dimensional data to enter the convolutional neural network for training, the acquired electricity price feature data set must first be processed into images. Specifically, the moment electricity prices and the characteristic values ​​of each related parameter in each electricity price feature data set are first sorted horizontally in a preset order to obtain a sorted electricity price feature data set; then all the sorted electricity price feature data sets are arranged vertically in sequence, and each data in the sorted electricity price feature data set is taken as a single pixel, and the data value of each data is taken as the pixel value of the corresponding pixel, so as to form an electricity price feature map of the grayscale image type. A partial example form of an electricity price feature map is as follows: Figure 2 shown.

[0064] Then, the formed electricity price characteristic graph is divided by a preset division method to obtain a plurality of parameter characteristic blocks and an electricity price label corresponding to each parameter characteristic block. Figure 2As shown, the preset division method can be a sliding window division method, specifically, the image width formed by all data pixels in the electricity price feature data group is used as the sliding window width, the sliding window length is set, and the sliding window is slid vertically on the electricity price feature map with a fixed step length to obtain multiple division feature blocks. Then, for each division feature block, the module formed by all relevant parameter feature values ​​in the division feature block is used as the parameter feature block of the division feature block, and the electricity price label of the division feature block is obtained according to the preset electricity price acquisition method based on the moment electricity price in the division feature block. The electricity price label of the division feature block is the electricity price label corresponding to the parameter feature block of the division feature block.

[0065] Furthermore, the electricity price label of the corresponding partitioned feature block is obtained according to the preset electricity price acquisition method based on the moment electricity price in the partitioned feature block. Specifically, the moment electricity price at the preset position in the partitioned feature block can be set as the electricity price label of the corresponding parameter feature block, for example, the moment electricity price at the bottom of the partitioned feature block is used as the electricity price label of the corresponding reference feature block. In addition, the average of all moment electricity prices in the partitioned feature block can be used as the electricity price label of the corresponding parameter feature block. It should be noted that the electricity price label of the corresponding partitioned feature block can also be obtained by other suitable methods, and there is no fixed restriction on it here.

[0066] It should be noted that the moment electricity prices and the characteristic values ​​of the relevant parameters in each electricity price characteristic data group can also be sorted vertically in a preset order, and all the sorted electricity price characteristic data groups are arranged horizontally in sequence to obtain the corresponding electricity price characteristic graph. At this time, the sliding window needs to be slid horizontally on the electricity price characteristic graph to obtain the divided characteristic blocks. The implementation process is similar to the above-mentioned multiple parameter characteristic blocks and the electricity price label acquisition process corresponding to each parameter characteristic block, and it will not be described in detail here.

[0067] All parameter feature blocks and the electricity price label corresponding to each parameter feature block form the training data set.

[0068] Step 103: Train a preset deep learning network model based on the training data set to obtain an electricity price probability prediction model.

[0069] Since the various load characteristic values ​​and various meteorological element characteristic values ​​in the electricity price characteristic data group are all time series data, and the LSTM network can learn and utilize the temporal dependencies in these data, it can more accurately predict the probability of electricity prices; and the design of the LSTM network can enable the preset deep learning network model to memorize long-term dependencies, so that the probability of electricity prices is affected by the data of the previous few days or even weeks. Therefore, the preset deep learning network model designed in this embodiment is designed with an LSTM network part. Since CNN can help extract features related to electricity price changes from complex multidimensional data, these features can then be used by LSTM for more accurate predictions. Therefore, the preset deep learning network model used in this embodiment is a CNN_LSTM combined model. The model combining CNN and LSTM combines the feature extraction capability of CNN and the sequence modeling capability of LSTM, so that the model can learn and adapt to complex electricity price change patterns; and this hybrid model is flexible in design and can be adjusted and optimized according to specific data characteristics and prediction requirements.

[0070] Compared with traditional machine learning algorithms, the CNN_LSTM combined model is more complex and has more network layers. Therefore, it can learn more in-depth and complex information in the eigenvalues. At the same time, it avoids the problem of overfitting in traditional machine learning algorithms and improves the overall accuracy and stability of the prediction algorithm model.

[0071] The specific preset deep learning network model includes input layer, convolution layer, activation function layer, pooling layer, dropout layer, multiple time series layers, transition output layer, activation function layer and output layer, etc. Figure 3 FIG. 2 is a schematic diagram of a structure of a preset deep learning network model in an embodiment of the present application. Figure 3As shown in the figure, the following is the specific process of each layer of the preset deep learning network model processing the training data set: first, the training data set is input into the input layer in batches; then the convolution layer performs deep feature extraction on the training data, where the convolution kernel size can be set to 3×3 and the step size can be set to 1; the ReLU activation function is used to improve the model's ability to handle nonlinear features and enhance the accuracy of the fitting results; then the pooling layer uses maximum pooling to aggregate multiple features, selects the most influential features or average features, thereby reducing the spatial volume of the data, improving the model's computing performance, and effectively reducing the probability of overfitting. The pooling size can be set to 2×2 and the step size can be set to 1; the dropout layer temporarily randomly discards some neural network units from the network with a certain probability, thereby reducing the number of neurons. The mutual dependence between them ensures that the model can extract important features and prevent overfitting; then use three time series layers to receive the pooled features, capture the long-term dependencies and time series relationships in the features, so as to accurately analyze the information with long-term time dependence characteristics such as the supply and demand relationship that affects the electricity price, establish time series relationships, and greatly improve the accuracy of the model. The above three time series layers constitute a deep time series hierarchy, in which the hidden layer size in each layer can be set to 256, 128, and 64, and the output of each layer is the input of the next layer; the transition output layer is the transition before the output result, that is, it is used to summarize the results calculated by the above layers; then the ReLU activation function is used again to convert the input signal of the neural network into an output signal; the output model outputs the result through the output layer.

[0072] By adopting the above data processing process, after training the preset deep learning network model based on the training data set, the electricity price probability prediction model can be obtained. After obtaining the electricity price probability prediction model, the electricity price probability at the corresponding time can be obtained by collecting the electricity price feature data group of the target area at a certain time and inputting the electricity price feature data group into the electricity price probability prediction model.

[0073] In order to verify the accuracy of the electricity price probability prediction model constructed by the electricity price probability prediction model construction method based on big data in this embodiment, this embodiment uses the electricity price probability prediction model for the Shanxi market obtained by the electricity price probability prediction model construction method based on big data in this embodiment to predict the electricity price probability of the Shanxi market on a certain day. Figure 4 This is a schematic diagram comparing the probability of electricity prices in Shanxi market on a certain day with the actual electricity prices. Figure 5 Figure 1 is a schematic diagram of the performance parameters of the probability prediction model for electricity prices in Shanxi market. Figure 4 and Figure 5As shown, the accuracy of the electricity price probability prediction model constructed by the electricity price probability prediction model construction method based on big data in this embodiment can reach 97%, and the difference between the electricity price probability and the actual electricity price is not large. Therefore, it can be seen that the electricity price probability prediction model obtained by the electricity price probability prediction model construction method based on big data in this embodiment can provide more accurate electricity price probability reference data for power generation companies and wholesale markets and other power generation sides. Through the trend of electricity price probability, identification of risks and opportunities, it can help decision makers formulate effective energy policies and market strategies, thereby improving energy utilization efficiency and reliability, reducing the costs and risks caused by imbalance between supply and demand, promoting the development of renewable energy and formulating reasonable energy policies, thereby promoting the healthy operation and sustainable development of the energy market.

[0074] The method for constructing an electricity price probability prediction model based on big data provided by an embodiment of the present invention combines the maximum startup ratio characteristic value, the reference price ratio characteristic value, the moment startup capacity characteristic value, the moment reference price ratio characteristic value and the negative electricity price state characteristic value with other relevant parameter characteristic values ​​as an electricity price characteristic data group. The empirical reference characteristic value has a large correlation with the market energy coupling characteristic value. Therefore, the introduction of the empirical reference characteristic value can fully reflect the market change characteristics and increase the reliability of the electricity price prediction result. The relevant parameter characteristic values ​​in the electricity price characteristic data group are embodied in the form of pixels to form an electricity price characteristic graph, and the relevant parameter characteristic values ​​are normalized to strengthen the correlation between the features, and the electricity price characteristic graph is divided to realize the conversion of the electricity price features into image features. The CNN_LSTM model is trained based on a training data set with image features, which can better reflect the spatial and temporal advantages between the features, so that the model can learn and adapt to complex electricity price change patterns. The electricity price probability prediction model using this method can make the obtained electricity price probability results more accurate, solving the problem of low accuracy and reliability of the current electricity price probability prediction.

[0075] like Figure 6 As shown, this embodiment provides a device for constructing an electricity price probability prediction model based on big data, including a data acquisition module, a training data set acquisition module and a model training module.

[0076] The data acquisition module is used to obtain a set of electricity price characteristic data groups in a target area within a preset time period. The set of electricity price characteristic data groups includes multiple electricity price characteristic data groups, and the electricity price characteristic data group includes the electricity price at the time of collection and multiple related parameter characteristic subgroups.

[0077] The training data set acquisition module is used to image the electricity price characteristic data set to obtain an electricity price characteristic graph, divide the electricity price characteristic graph into multiple parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block by a preset division method, and form a training data set based on all parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block.

[0078] The model training module is used to train the preset deep learning network model based on the training data set to obtain the electricity price probability prediction model.

[0079] Among them, the preset deep learning network model is a CNN_LSTM combined model, and multiple related parameter feature subgroups include an empirical reference feature subgroup, and the empirical reference feature subgroup includes a maximum startup ratio feature value, a reference price ratio feature value, a momentary startup capacity feature value, a momentary reference price ratio feature value and a negative electricity price state feature value.

[0080] The apparatus for constructing a probability prediction model of electricity price based on big data provided by an embodiment of the present invention combines the characteristic value of the maximum startup ratio, the characteristic value of the reference price ratio, the characteristic value of the startup capacity at the moment, the characteristic value of the reference price ratio at the moment and the characteristic value of the negative electricity price state with the characteristic values ​​of other relevant parameters as an electricity price characteristic data group. The empirical reference characteristic value has a large correlation with the characteristic value of the market energy coupling. Therefore, the introduction of the empirical reference characteristic value can fully reflect the characteristics of market changes and increase the reliability of the electricity price prediction result. The characteristic values ​​of the relevant parameters in the electricity price characteristic data group are embodied in the form of pixels to form an electricity price characteristic graph, and the normalization of the characteristic values ​​of the relevant parameters can strengthen the correlation between the features, and the electricity price characteristic graph is divided to realize the conversion of the electricity price features into image features. The CNN_LSTM model is trained based on a training data set with image features, which can better reflect the advantages of space and time between the features, so that the model can learn and adapt to complex electricity price change patterns. The electricity price probability prediction model using this method can make the obtained electricity price probability results more accurate, solving the problem of low accuracy and reliability of the current electricity price probability prediction.

[0081] The present application embodiment also provides a computer-readable storage medium. A person of ordinary skill in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the processor through a program, and the program can be stored in a computer-readable storage medium, and the storage medium is a non-transitory medium, such as a random access memory, a read-only memory, a flash memory, a hard disk, a solid-state hard disk, a magnetic tape, a floppy disk, an optical disc, and any combination thereof. The above-mentioned storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a digital video disc (DVD)), or a semiconductor medium (for example, a solid-state disk (SSD)), etc.

[0082] like Figure 7 As shown, an embodiment of the present application provides a terminal.

[0083] The terminal of this embodiment includes a processor and a memory connected to each other; the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory, so that all or part of the steps in the method of the above embodiment can be implemented when the terminal executes.

[0084] The beneficial effects of all or part of the steps of the method in the above embodiment are the same as the beneficial effects obtained by applying the terminal provided by the embodiment of the present invention, and will not be described in detail here.

[0085] It should be noted that the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage. Similarly, the processor may also be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0086] Although the embodiments disclosed in the present invention are as above, the contents described are only embodiments adopted for facilitating the understanding of the present invention and are not intended to limit the present invention. Any technician in the technical field to which the present invention belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present invention, but the protection scope of the present invention shall still be subject to the scope defined in the attached claims.

Claims

1. A method for constructing an electricity price probability prediction model based on big data, comprising: Acquire a set of electricity price characteristic data groups in a target area within a preset time period, wherein the set of electricity price characteristic data groups includes a plurality of electricity price characteristic data groups, and the electricity price characteristic data groups include a momentary electricity price at a collection moment and a plurality of relevant parameter characteristic subgroups; The electricity price characteristic data set is processed into an image to obtain an electricity price characteristic graph, the electricity price characteristic graph is divided into multiple parameter characteristic blocks and an electricity price label corresponding to each parameter characteristic block by a preset division method, and a training data set is formed based on all the parameter characteristic blocks and the electricity price label corresponding to each parameter characteristic block; Based on the training data set, a preset deep learning network model is trained to obtain an electricity price probability prediction model; Among them, the preset deep learning network model is a CNN_LSTM combined model, and multiple related parameter feature subgroups include an empirical reference feature subgroup, and the empirical reference feature subgroup includes a maximum startup ratio feature value, a reference price ratio feature value, a moment startup capacity feature value, a moment reference price ratio feature value and a negative electricity price state feature value.

2. The model building method according to claim 1, characterized in that: The electricity price characteristic graph obtained by processing the electricity price characteristic data set into an image includes: The momentary electricity prices and characteristic values ​​of relevant parameters in each of the electricity price characteristic data groups are horizontally sorted in a preset order, and each data in the sorted electricity price characteristic data group is used as a single pixel, and the data value of each data is used as the pixel value of the corresponding pixel. All the sorted electricity price characteristic data groups are arranged vertically in sequence to form an electricity price characteristic graph.

3. The model building method according to claim 2, characterized in that: Dividing the electricity price characteristic diagram by a preset division method to obtain a plurality of parameter characteristic blocks and an electricity price label corresponding to each of the parameter characteristic blocks includes: The electricity price characteristic diagram is divided into a plurality of divided characteristic blocks in a sliding window division manner, a module formed by all relevant parameter characteristic values ​​in each of the divided characteristic blocks is used as a parameter characteristic block corresponding to the divided characteristic block, and an electricity price label corresponding to the divided characteristic block is obtained according to a preset electricity price acquisition method based on the moment electricity price in each of the divided characteristic blocks; The width of the image formed by all the data pixels in the electricity price characteristic data group is used as the sliding window width.

4. The model building method according to claim 3, characterized in that: Acquiring the electricity price tag of the partitioned feature block according to a preset electricity price acquisition method based on the moment electricity price in the partitioned feature block includes: The electricity price at a preset position in each of the divided feature blocks is used as the electricity price label of the corresponding parameter feature block, or the average of all the electricity prices at the moments in each of the divided feature blocks is used as the electricity price label of the corresponding parameter feature block.

5. The model building method according to claim 1, characterized in that: All relevant parameter characteristic values ​​in the electricity price characteristic data group need to be subjected to Min-Max normalization processing.

6. The model building method according to claim 1, characterized in that: The multiple related parameter characteristic subgroups also include a load proportion characteristic subgroup, a market energy coupling characteristic subgroup, a coal information characteristic subgroup and a meteorological element characteristic subgroup.

7. The model building method according to claim 1, characterized in that: The preset deep learning network model includes an input layer, a convolution layer, an activation function layer, a pooling layer, a dropout layer, multiple time series layers, a transition output layer, an activation function layer and an output layer.

8. A device for constructing an electricity price probability prediction model based on big data, characterized in that: It includes a data acquisition module, a training data set acquisition module and a model training module; The data acquisition module is used to acquire a set of electricity price characteristic data groups of a target area within a preset time period, wherein the set of electricity price characteristic data groups includes a plurality of electricity price characteristic data groups, and the electricity price characteristic data groups include a momentary electricity price at a collection moment and a plurality of relevant parameter characteristic subgroups; The training data set acquisition module is used to image the electricity price feature data set to obtain an electricity price feature graph, divide the electricity price feature graph in a preset division method to obtain multiple parameter feature blocks and an electricity price label corresponding to each parameter feature block, and form a training data set based on all the parameter feature blocks and the electricity price label corresponding to each parameter feature block; The model training module is used to train a preset deep learning network model based on the training data set to obtain an electricity price probability prediction model; Among them, the preset deep learning network model is a CNN_LSTM combined model, and multiple related parameter feature subgroups include an empirical reference feature subgroup, and the empirical reference feature subgroup includes a maximum startup ratio feature value, a reference price ratio feature value, a moment startup capacity feature value, a moment reference price ratio feature value and a negative electricity price state feature value.

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 method for constructing an electricity price probability prediction model based on big data as described in any one of claims 1 to 7 is implemented.

10. A terminal comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the method for constructing an electricity price probability prediction model based on big data according to any one of claims 1 to 7.

Images