Electric power spot market day-ahead time-phased average price prediction method, device and equipment

Abstract

The invention provides a day-ahead time-phased average price prediction method, device and equipment for an electric power spot market, and relates to the technical field of electricity price prediction, and the method comprises the steps: obtaining a market boundary data prediction result of the electric power spot market in a first preset time period in the future; according to a market boundary data prediction result, predicting a day-ahead time-division thermal power load rate mean value of the electric power spot market in a first preset time period; inputting the market boundary data prediction result and the day-ahead time-divided thermal power load rate mean value into a pre-trained electric power spot market day-ahead time-divided average price regression prediction model; and obtaining a day-ahead time-segmented average price predicted value of the electric power spot market in the first preset time period, which is output by the day-ahead time-segmented average price regression prediction model of the electric power spot market. According to the method, the day-ahead time-phased average price of the electric power spot market can be accurately predicted to a certain extent.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity price forecasting, and in particular to a method, device and equipment for forecasting the day-ahead time-segment average price in the electricity spot market. Background Art

[0002] With the continuous advancement of the development of new energy, the scale of new energy continues to expand. On the one hand, the difficulty of balancing electricity supply and demand is prominent, especially in the face of long-term scenarios of continuous lack of wind and solar power for new energy or long-term scenarios of continuous large-scale development of new energy, which causes large fluctuations in the system electricity spot market price. How to accurately predict the average price of the electricity spot market by time period the day before becomes the key; on the other hand, in the medium- and long-term transactions in the electricity spot market, the size of the position ratio after considering the average price of the electricity spot market by time period the day before is closely related to the price difference between the medium- and long-term electricity spot price by time period the day before and the average price of the electricity spot price. If the prediction ability of the long-term electricity spot price by time period the day before is insufficient, then participating in medium- and long-term market transactions will face huge market return risks.

[0003] Based on this, how to accurately predict the average price of the electricity spot market by time period in the day before has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of this, a method, device and equipment for predicting the average price of the electricity spot market by time period the day before is provided, so as to accurately predict the average price of the electricity spot market by time period the day before to a certain extent.

[0005] The present invention adopts the following technical solution: In a first aspect, the present invention provides a method for predicting the average price of a power spot market by time period, comprising: Obtaining a prediction result of market boundary data of the electricity spot market within a first preset time period in the future; According to the prediction result of the market boundary data, predict the average value of the thermal power load rate of the power spot market in the first preset time period in the day before; The market boundary data prediction result and the average thermal power load rate in the time period the day before are input into a pre-trained time period average price regression prediction model for the electricity spot market the day before, and the time period average price prediction value of the electricity spot market in the first preset time period is obtained as output by the time period average price regression prediction model for the electricity spot market the day before.

[0006] Optionally, the training process of the day-ahead time-based average price regression prediction model of the electricity spot market includes: Acquire historical market boundary data, clearing price data and thermal power load rate data of the power spot market within a second preset historical time period; the historical market boundary data includes sub-historical market boundary data of multiple dimensions; Based on the historical market boundary data and the clearing price data, a target dimension is selected from all the dimensions; the correlation coefficient between the sub-historical market boundary data corresponding to the target dimension and the clearing price data is greater than a preset coefficient threshold; Model training is performed based on the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data to obtain a regression prediction model for the day-ahead time-based average price of the electricity spot market.

[0007] Optionally, the selecting a target dimension from all the dimensions based on the historical market boundary data and the clearing price data specifically includes: For each of the dimensions, based on Spearman correlation analysis, according to the sub-historical market boundary data of the dimension and the clearing price data, calculate the correlation coefficient corresponding to the dimension; Determining whether the correlation coefficient is greater than the preset coefficient threshold; If the correlation coefficient is greater than the preset coefficient threshold, the dimension is determined to be the target dimension.

[0008] Optionally, the performing of model training according to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data to obtain the day-ahead time-based average price regression prediction model of the power spot market specifically includes: According to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data, respectively train the pre-built random forest algorithm prediction model, XGboost algorithm prediction model and lightGBM algorithm prediction model; After training the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model, integrated training is performed based on the prediction results of the three models to obtain a meta-model; the output result of the meta-model is used as the output result of the day-ahead time-based average price regression prediction model of the electricity spot market.

[0009] Optionally, the metamodel is expressed as follows:

[0010] in, It indicates the forecast result of the average price of the electricity spot market in the day-ahead period within the forecast period. Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The initial forecast result of the average price of the electricity spot market in the day-ahead period during the forecast period obtained by the model, Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The error between the initial prediction result of the day-ahead time-segment average price of the electricity spot market in the previous historical cycle of the time period to be predicted obtained by the model and the corresponding actual value.

[0011] Optionally, predicting the average value of the thermal power load rate of the power spot market in the first preset time period by time in the day before according to the market boundary data prediction result specifically includes: According to the market area electricity load, market area power transmission or reception, new energy output value and unit online capacity in the market boundary data prediction results, the average thermal power load rate in the time period on the day before is calculated.

[0012] In a second aspect, the present invention further provides a device for predicting the average price of the electricity spot market by time period, comprising: A first acquisition module is used to obtain a prediction result of market boundary data of the electricity spot market within a first preset time period in the future; A first prediction module is used to predict the average value of the thermal power load rate of the power spot market in the first preset time period according to the prediction result of the market boundary data; The second prediction module is used to input the market boundary data prediction result and the average thermal power load rate in the time period the day before into a pre-trained electricity spot market time period average price regression prediction model, and obtain the day-ahead time period average price prediction value of the electricity spot market within the first preset time period output by the electricity spot market time period average price regression prediction model.

[0013] In a third aspect, the present invention further provides a device for predicting the average price of the electricity spot market by time period, comprising: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can implement the method for predicting the day-ahead time-based average price in the electricity spot market as described above.

[0014] The present invention adopts the above technical scheme, a method for predicting the day-ahead time-period average price of the electricity spot market, including: obtaining the market boundary data prediction results of the electricity spot market in the first preset time period in the future; predicting the day-ahead time-period average thermal power load rate of the electricity spot market in the first preset time period according to the market boundary data prediction results; inputting the market boundary data prediction results and the day-ahead time-period average thermal power load rate into a pre-trained day-ahead time-period average price regression prediction model for the electricity spot market, and obtaining the day-ahead time-period average price prediction value of the electricity spot market in the first preset time period output by the day-ahead time-period average price regression prediction model for the electricity spot market.

[0015] Based on this, since the time-sharing electricity volume of new energy and the average time-sharing thermal power load rate of the day before that affect the average time-sharing price of the electricity spot market, it is possible to carry out supply and demand evaluation through long-term quantitative analysis means, and with the coupling relationship characteristics of the historical load rate of thermal power-quotation distribution, a scientific and reasonable long-term electricity spot market time-sharing price can be formed. Therefore, considering the time-sharing distribution characteristics of the thermal power load rate can improve the accuracy of the prediction results of the day-ahead time-sharing price. In the present invention, by inputting the prediction results of the market boundary data and the average time-sharing thermal power load rate of the day before into a pre-trained prediction model, the prediction value of the day-ahead time-sharing price output by the prediction model is obtained. It can be seen that the present invention takes into account the time-sharing distribution characteristics of the thermal power load rate when predicting the day-ahead time-sharing price. Therefore, the present invention can accurately predict the day-ahead time-sharing price of the electricity spot market to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a method for predicting the average price of the electricity spot market by time period on the day before the event provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of a training process of a day-ahead time-based average price regression prediction model for a power spot market provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a device for predicting the average price of the electricity spot market by time period on the day before the event provided by an embodiment of the present invention; Figure 4 The present invention provides a schematic diagram of the structure of a device for predicting the average price of the electricity spot market by time period on the day before. DETAILED DESCRIPTION

[0018] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0019] Figure 1 1 is a flow chart of a method for predicting the average price of the electricity spot market by time period on the day before the event provided by an embodiment of the present invention. Figure 1 As shown in the figure, the method for predicting the average price of the electricity spot market by time period includes: Step 101: Obtain market boundary data prediction results of the electricity spot market within a first preset time period in the future.

[0020] Specifically, the dimensions of market boundary data include wind speed, solar irradiance, cloud cover, rainfall, temperature, online capacity of units, power load in the market area, power delivered or received in the market area, and new energy output value.

[0021] In the embodiment of the present invention, all dimensions are pre-screened, and the day-ahead average price of the electricity spot market is predicted based on the selected target dimensions. Therefore, the market boundary data prediction results include the sub-market boundary data prediction results corresponding to each target dimension.

[0022] It should be noted that the method for obtaining market boundary data prediction results is an existing technology and will not be described in detail here.

[0023] Step 102: According to the prediction result of the market boundary data, the average value of the thermal power load rate in the power spot market in the first preset time period is predicted.

[0024] Specifically, based on the prediction results of the market boundary data, the average value of the thermal power load rate in the power spot market in the first preset time period is predicted, which may specifically include: Based on the market boundary data forecast results of the market area electricity load, the market area outbound or incoming electricity, the new energy output value and the unit online capacity, the average thermal power load rate in each time period on the day before is calculated.

[0025] Here, based on the historical long-term real data and the time series autoregression method, the future long-term market area power load, the market area outbound or inbound power, the new energy output value and the unit online capacity forecast are carried out, and combined with the calculation of the thermal power time-based load rate, the future long-term thermal power time-based load rate (that is, the average of the thermal power load rate in the previous time period) is obtained, and its expression is as follows: ...... (1) in, represents the average thermal power load rate in different time periods the day before, Indicates the electricity load in the market area, Indicates the amount of electricity delivered or received in the market area. Represents the output value of new energy. Indicates the online capacity of the unit.

[0026] Step 103: Input the prediction results of the market boundary data and the average thermal power load rate in different time periods the day before into a pre-trained regression prediction model for the average price in the time period of the day before the electricity spot market, and obtain the predicted value of the average price in the time period of the day before the electricity spot market in the first preset time period output by the regression prediction model for the average price in the time period of the day before the electricity spot market.

[0027] The embodiment of the present invention adopts the above-mentioned technical scheme. Since the time-sharing electricity amount of new energy and the average value of the thermal power load rate in the time period a day that affect the average price of the electricity spot market in the time period a day can be used to carry out supply and demand evaluation through long-term quantitative analysis means, and with the coupling relationship characteristics of the historical load rate of thermal power and the quotation distribution, a scientific and reasonable long-term electricity spot market time period average price a day ago is formed. Therefore, considering the time-sharing distribution characteristics of the thermal power load rate can improve the accuracy of the prediction results of the time period average price a day ago. In the present invention, by inputting the prediction results of the market boundary data and the average value of the thermal power load rate in the time period a day a day into a pre-trained prediction model, the prediction value of the time period average price output by the prediction model is obtained. It can be seen that the present invention takes into account the time-sharing distribution characteristics of the thermal power load rate when predicting the average price in the time period a day. Therefore, the present invention can accurately predict the average price in the time period a day of the electricity spot market to a certain extent.

[0028] Figure 2 Schematic diagram of the training process of a day-ahead time-based average price regression prediction model for the electricity spot market provided by an embodiment of the present invention. Figure 2 As shown, this process includes: Step 201: Obtain historical market boundary data, clearing price data and thermal power load rate data of the electricity spot market within a second preset time period in history; the historical market boundary data includes sub-historical market boundary data of multiple dimensions.

[0029] Specifically, the length of the second preset time period is an integer multiple of the length of the time period to be predicted. For example, when the length of the time period to be predicted is 7 days, the length of the second preset time period is an integer multiple of 7 days. The time periods may specifically refer to each hour of each day (i.e., 1 o'clock, 2 o'clock, 3 o'clock, ..., 24 o'clock). The present invention is described by taking the time periods referring to each hour of each day as an example.

[0030] In a specific example, for the dimension of wind speed, its sub-historical market boundary data may include the wind speed at each hour of each day in the second preset time period. The same is true for other dimensions, which will not be repeated here. The clearing price data may include the clearing price at each hour of each day in the second preset time period. The thermal power load rate data may include the thermal power load rate at each hour of each day in the second preset time period.

[0031] Step 202: Based on the historical market boundary data and the clearing price data, a target dimension is selected from all dimensions; the correlation coefficient between the sub-historical market boundary data and the clearing price data corresponding to the target dimension is greater than a preset coefficient threshold.

[0032] Specifically, based on historical market boundary data and clearing price data, target dimensions are selected from all dimensions, which may include: (1) For each dimension, based on the Spearman correlation analysis, the correlation coefficient corresponding to the dimension is calculated according to the sub-historical market boundary data and clearing price data of the dimension.

[0033] Specifically, the correlation coefficient can be calculated using the following formula (2): : ...... (2) in, The value range is: -1< <1.

[0034] t =1, 2, 3, ..., 24, M is equal to the total number of days in the second preset time period, Indicates that the corresponding dimension (such as wind speed) is in the first j Day's t The value at the time, Indicates the corresponding dimension in the second preset time period t The mean at time instant, for example, t =1, This is equal to adding up the values ​​of the corresponding dimension at 1:00 every day in the second preset time period and then dividing by M The obtained value.

[0035] Indicates the electricity spot market in the second preset time period j Day's t The clearing price at time, Indicates the electricity spot market in the second preset time period t The mean at time instant, for example, t =1, It is equal to the sum of the clearing prices of the electricity spot market at 1:00 every day in the second preset time period, and then divided by M The obtained value.

[0036] In addition, the correlation coefficient can also be calculated using the following calculation formula (3): : ...... (3) in, .

[0037]

[0038] In this example, the collection , The differences of the corresponding elements in form the ranked difference set , and consider the correlation in 5 sections (i.e. - ).

[0039] (2) Determine whether the correlation coefficient is greater than a preset coefficient threshold.

[0040] Specifically, considering the actual difference between the correlation of market boundary data and nonlinear segmented correlation, the correlation is generally 0.5< <1. Therefore, the value of the preset coefficient threshold can be 0.5.

[0041] (3) If the correlation coefficient is greater than the preset coefficient threshold, the dimension is determined to be the target dimension.

[0042] Step 203: Model training is performed based on the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension to obtain a regression prediction model for the day-ahead time-based average price of the electricity spot market.

[0043] Here, according to the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension, the model training is carried out to obtain the regression prediction model of the average price of the electricity spot market by time period on the day before, which can specifically include: (1) According to the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension, the pre-built random forest algorithm prediction model, XGboost algorithm prediction model and lightGBM algorithm prediction model are trained respectively.

[0044] Specifically, the random forest algorithm prediction model is a prediction model built based on the random forest algorithm. Random forest (RF) is a classic machine learning algorithm, which is composed of its weak model decision regression tree (classification and regression tree, CART) combined with the bagging algorithm method and random feature subspace (random subspace method, RSM). CART can be used for both regression and classification, and the corresponding random forest can also be used for classification and regression analysis. Among them, CART is a binary recursive segmentation technology, which divides the current sample set into 2 subsets at each node (except the leaf node). By constructing multiple decision trees and aggregating their prediction results for decision-making, the accuracy and stability of the model are improved.

[0045] According to the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension, the process of training the pre-built random forest algorithm prediction model includes: (a) Based on the Bootstrap Method, samples are extracted from the sub-historical market boundary data and historical FM market clearing prices corresponding to the target dimension.

[0046] (b) Use the extracted samples to train the random forest algorithm prediction model. During the training process, the random forest algorithm prediction model is evaluated based on the out-of-bag samples. The evaluation results are used to correct errors and parameters. The evaluation results are marked as oob_score, and oob score is returned. , which is expressed as follows: ...... (4) in, It represents the residual sum of squares of the day-ahead clearing price in the electricity spot market in the corresponding time period, It represents the total sum of squares of the distances between the market boundary data of the electricity spot market in the corresponding time period, represents the number of samples, Indicates the sample number, Indicates The model regression value of samples, Indicates The true market clearing price of samples is express The average of the true market clearing prices of the samples.

[0047] The XGboost algorithm prediction model is a prediction model built based on the extreme gradient boosting (XGBoost) algorithm. XGBoost is a machine learning algorithm based on a tree model, which has great advantages in processing tabular data. When samples are missing, the algorithm can automatically select the appropriate split direction and can sample in columns, which not only prevents overfitting but also improves the calculation speed. Secondly, XGBoost can rank the importance of each feature variable to the model output, which has the advantages of facilitating the selection of model structure, reducing irrelevant sample types, improving the efficiency of collecting necessary samples, and improving the calculation speed and prediction accuracy of the model. Modeling the XGBoost algorithm based on the tree model is to determine the tree model structure and the output value of its leaf nodes.

[0048] According to the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension, the process of training the pre-built XGboost algorithm prediction model includes: (a) The first decision tree based on the XGboost algorithm prediction model fits the sub-historical market boundary data corresponding to the target dimension and calculates the residual between the actual value and the predicted value.

[0049] (b) Introduce the next decision tree based on the residual and fit the residual until the residual meets the preset residual value requirements, then stop introducing the decision tree.

[0050] (c) Accumulate the fitting values ​​of each decision tree to obtain the final fitting result of the XGboost algorithm, which is expressed as follows: ...... (5) in, represents the objective function, represents the loss function, Ω represents the regularization penalty function, Indicates The true value of the market clearing price corresponding to the samples, Indicates The predicted value of the market clearing price corresponding to the samples, Indicates The output prediction value of a decision tree is represents the number of samples, Represents the number of decision trees.

[0051] The lightGBM algorithm prediction model is a prediction model built based on the Light Graddient Boosting Machine (lightGBM) algorithm. lightGBM uses decision trees as weak learners. The gradient boosting algorithm of LightGBM is consistent with the XGBoost algorithm, both of which are second-order Taylor expansion fitting losses with regularization terms. The core of LightGBM is the definition of the loss function, the iterative process of gradient boosting, and the Leaf-wise growth strategy. Through these steps, large-scale data can be efficiently processed and good prediction results can be achieved. Based on the lightGBM algorithm, a lightGBM algorithm prediction weak model set is constructed, and its expression is as follows: ...... (6) in, represents the loss function; represents the true value of the market clearing price; Before t -The model predicted value of the market clearing price in round 1; Represents the regular term, including L 1 Regularization term and L 2 regularization terms, C is a constant term; Gain is a gain function, is the sum of the first-order derivatives of the left child node; is the sum of the first-order derivatives of the right child node; is the sum of the second-order derivatives of the left child node; is the sum of the second-order derivatives of the right child node; Refers to the complexity cost incurred each time a new leaf node is generated; S is the number of iterative splits of the decision tree.

[0052] It should be noted that the above three models can use cross-validation and grid search methods during the training process.

[0053] (2) After the random forest algorithm prediction model, the XGboost algorithm prediction model, and the lightGBM algorithm prediction model are trained, an integrated training is performed based on the prediction results of the three models to obtain a meta-model; the output result of the meta-model is used as the output result of the day-ahead time-based average price regression prediction model of the electricity spot market.

[0054] Specifically, the prediction results of different single models have different accuracy and deviation. The use of an integrated learning model to fuse multiple prediction models can significantly improve the prediction accuracy. At the same time, the Stacking model has strong heterogeneous data processing and generalization capabilities, which can reduce the error of a single model at the same time. Ensemble learning uses the prediction results of multiple basic models as input features and then trains a "meta-model" to make the final prediction. The goal of this meta-model is to find the best combination of the prediction results of the basic models to give the final prediction.

[0055] The metamodel is expressed as follows: ...... (7) in, It indicates the forecast result of the average price of the electricity spot market in the day-ahead period within the forecast period. Indicates the use of the random forest algorithm prediction model, XGboost algorithm prediction model, and lightGBM algorithm prediction model. The initial forecast results of the average price of the electricity spot market in the forecast period obtained by the model are as follows: Indicates the use of the random forest algorithm prediction model, XGboost algorithm prediction model, and lightGBM algorithm prediction model. The error between the initial prediction result of the day-ahead time-segment average price of the electricity spot market in the previous historical period of the forecast time period obtained by the model and the corresponding actual value.

[0056] In summary, the process of training the prediction model and optimizing parameters based on the sub-historical market boundary data and thermal power load rate data corresponding to the target dimension includes: (1) Model training: Use the selected features and related data to train the model. This process usually includes steps such as parameter estimation and model optimization. (2) Model verification: Verify the accuracy and reliability of the model. (3) Feature engineering: Filter, transform or combine features to optimize and improve the prediction ability of the model. (4) Model evaluation: Use evaluation indicators such as accuracy, recall rate, F1 value, etc. to evaluate the performance of the model and judge the stability and accuracy of the model. (5) Model optimization: According to the evaluation results, adjust and optimize the model to improve the prediction accuracy, including steps such as adjusting model parameters and adding or deleting features.

[0057] It should be noted that the training method of the above-mentioned electricity spot market day-ahead time-based average price regression prediction model is an existing technology. Therefore, the above only briefly introduces the training process of the model. Those skilled in the art can easily learn the specific implementation details of the model training process based on the above content.

[0058] Based on a general inventive concept, the present invention also provides a device for predicting the average price of the electricity spot market by time period on the day before. Figure 3 Schematic diagram of the structure of a device for predicting the average price of the electricity spot market by time period according to the embodiment of the present invention. Figure 3 As shown, the device comprises: The first acquisition module 31 is used to obtain the market boundary data prediction result of the power spot market within a first preset time period in the future.

[0059] The first prediction module 32 is used to predict the average value of the thermal power load rate in the power spot market in the first preset time period according to the prediction result of the market boundary data.

[0060] The second prediction module 33 is used to input the prediction results of market boundary data and the average value of thermal power load rate in different time periods the day before into a pre-trained regression prediction model for the average price in the time period of the day before the electricity spot market, and obtain the predicted value of the average price in the time period of the day before the electricity spot market within the first preset time period output by the regression prediction model for the average price in the time period of the day before the electricity spot market.

[0061] Optionally, the device for predicting the average price of the electricity spot market by time period on the day before the present invention further includes: The second acquisition module is used to obtain historical market boundary data, clearing price data and thermal power load rate data of the electricity spot market within a second preset historical time period; the historical market boundary data includes sub-historical market boundary data of multiple dimensions.

[0062] A selection module is used to select a target dimension from all the dimensions based on the historical market boundary data and the clearing price data; the correlation coefficient between the sub-historical market boundary data corresponding to the target dimension and the clearing price data is greater than a preset coefficient threshold.

[0063] The training module is used to perform model training according to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data to obtain the day-ahead time-based average price regression prediction model of the electricity spot market.

[0064] Optionally, select a module that can be used for: (1) For each of the dimensions, based on Spearman correlation analysis, the correlation coefficient corresponding to the dimension is calculated according to the sub-historical market boundary data of the dimension and the clearing price data.

[0065] (2) Determine whether the correlation coefficient is greater than the preset coefficient threshold.

[0066] (3) If the correlation coefficient is greater than the preset coefficient threshold, the dimension is determined to be the target dimension.

[0067] Optional training modules can be used to: (1) According to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data, the pre-built random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model are trained respectively.

[0068] (2) After the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model are trained, integrated training is performed based on the prediction results of the three models to obtain a meta-model; the output result of the meta-model is used as the output result of the day-ahead time-based average price regression prediction model of the electricity spot market.

[0069] Optionally, the metamodel expression is as follows:

[0070] in, It indicates the forecast result of the average price of the electricity spot market in the day-ahead period within the forecast period. Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The initial forecast result of the average price of the electricity spot market in the day-ahead period during the forecast period obtained by the model, Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The error between the initial prediction result of the day-ahead time-segment average price of the electricity spot market in the previous historical cycle of the time period to be predicted obtained by the model and the corresponding actual value.

[0071] Optionally, the first prediction module 32 may be specifically used for: According to the market area electricity load, market area power transmission or reception, new energy output value and unit online capacity in the market boundary data prediction results, the average thermal power load rate in the time period on the day before is calculated.

[0072] Based on a general inventive concept, the present invention also provides a device for predicting the average price of the electricity spot market by time period on the day before. Figure 4 Schematic diagram of the structure of a device for predicting the average price of the electricity spot market by time period on the day before the event provided by the embodiment of the present invention. Figure 4 As shown, the day-ahead time-period average price forecasting device 400 for the electricity spot market includes: at least one processor 410; and, A memory 430 is communicatively connected to at least one processor 410; wherein, The memory 430 stores instructions 420 executable by at least one processor 410 , and the instructions 420 are executed by at least one processor 410 to enable the at least one processor 410 to implement the method for predicting the day-ahead time-based average price in the electricity spot market as described above.

[0073] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0074] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0075] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0076] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0077] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0078] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0079] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0080] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0081] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for predicting the average price of electricity spot market by time period on the day before, characterized in that: include: Obtaining a prediction result of market boundary data of the electricity spot market within a first preset time period in the future; According to the prediction result of the market boundary data, predict the average value of the thermal power load rate of the power spot market in the first preset time period in the day before; The market boundary data prediction result and the average thermal power load rate in the time period the day before are input into a pre-trained time period average price regression prediction model for the electricity spot market the day before, and the time period average price prediction value of the electricity spot market in the first preset time period is obtained as output by the time period average price regression prediction model for the electricity spot market the day before.

2. The method for predicting the average price of the electricity spot market by time period according to claim 1 is characterized in that: The training process of the day-ahead time-based average price regression prediction model of the electricity spot market includes: Acquire historical market boundary data, clearing price data and thermal power load rate data of the power spot market within a second preset historical time period; the historical market boundary data includes sub-historical market boundary data of multiple dimensions; Based on the historical market boundary data and the clearing price data, a target dimension is selected from all the dimensions; the correlation coefficient between the sub-historical market boundary data corresponding to the target dimension and the clearing price data is greater than a preset coefficient threshold; Model training is performed based on the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data to obtain a regression prediction model for the day-ahead time-based average price of the electricity spot market.

3. The method for predicting the average price of the electricity spot market by time period according to claim 2 is characterized in that: The selecting a target dimension from all the dimensions based on the historical market boundary data and the clearing price data specifically includes: For each of the dimensions, based on Spearman correlation analysis, according to the sub-historical market boundary data of the dimension and the clearing price data, calculate the correlation coefficient corresponding to the dimension; Determining whether the correlation coefficient is greater than the preset coefficient threshold; If the correlation coefficient is greater than the preset coefficient threshold, the dimension is determined to be the target dimension.

4. The method for predicting the average price of the electricity spot market by time period according to claim 2 is characterized in that: The model training is performed according to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data to obtain the day-ahead time-based average price regression prediction model of the power spot market, specifically including: According to the sub-historical market boundary data corresponding to the target dimension and the thermal power load rate data, respectively train the pre-built random forest algorithm prediction model, XGboost algorithm prediction model and lightGBM algorithm prediction model; After training the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model, integrated training is performed based on the prediction results of the three models to obtain a meta-model; the output result of the meta-model is used as the output result of the day-ahead time-based average price regression prediction model of the electricity spot market.

5. The method for predicting the average price of the electricity spot market by time period according to claim 4 is characterized in that: The meta-model is expressed as follows: in, It indicates the forecast result of the average price of the electricity spot market in the day-ahead period within the forecast period. Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The initial forecast result of the average price of the electricity spot market in the day-ahead period during the forecast period obtained by the model, Indicates the use of the random forest algorithm prediction model, the XGboost algorithm prediction model and the lightGBM algorithm prediction model. The error between the initial prediction result of the day-ahead time-segment average price of the electricity spot market in the previous historical cycle of the time period to be predicted obtained by the model and the corresponding actual value.

6. The method for predicting the average price of the electricity spot market by time period according to claim 1 is characterized in that: The step of predicting the average thermal power load rate of the power spot market in the first preset time period according to the market boundary data prediction result specifically includes: According to the market area electricity load, market area power transmission or reception, new energy output value and unit online capacity in the market boundary data prediction results, the average thermal power load rate in the time period on the day before is calculated.

7. A device for predicting the average price of electricity spot market by time period, characterized in that: include: A first acquisition module is used to obtain a prediction result of market boundary data of the electricity spot market within a first preset time period in the future; A first prediction module is used to predict the average value of the thermal power load rate of the power spot market in the first preset time period according to the prediction result of the market boundary data; The second prediction module is used to input the market boundary data prediction result and the average thermal power load rate in the time period the day before into a pre-trained electricity spot market time period average price regression prediction model, and obtain the day-ahead time period average price prediction value of the electricity spot market within the first preset time period output by the electricity spot market time period average price regression prediction model.

8. A device for predicting the average price of electricity spot market by time period, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can implement the method for predicting the day-ahead time-based average price in the electricity spot market as described in any one of claims 1 to 6.

Images