Machine learning and adaptive mechanism-based spot electricity price prediction method and computing device

Abstract

The invention provides a method for predicting spot electricity price based on machine learning and an adaptive mechanism and computing equipment, and is applied to the technical field of electricity price prediction. The method comprises the following steps: acquiring a dynamic data set; calculating basic characteristics by using the dynamic data set, wherein the basic characteristics comprise a thermal power generating unit bidding space; constructing composite features based on the basic features to form a factor library; selecting features used for modeling in the factor library, calculating segmentation indexes of the electricity price in the first time, and finding out the bidding space of the thermal power generating unit corresponding to the demarcation electricity price; and independently establishing a prediction model for each segmented index, splicing data results of the segmented index prediction models, and performing spot electricity price prediction. By using the method of the invention, accuracy and reliability of electricity price prediction can be effectively increased.

Description

Technical Field

[0001] The present invention relates to the field of electricity price forecasting, and particularly to a method and a computing device for forecasting spot electricity prices based on machine learning and an adaptive mechanism. Background Art

[0002] The adoption of a dual-track electricity price mechanism usually leads to a gap between the electricity costs of power generation enterprises and the electricity consumption expenditures of power users, causing losses to stakeholders. At the same time, due to the instability of the output of new energy units, the electricity price fluctuates greatly during some periods. Electricity price forecasting can help power generation side units formulate more reasonable power generation plans, and at the same time help power load aggregators better plan products, promote the efficient utilization of power resources, achieve environmental friendliness, and reduce waste.

[0004] Therefore, a method for forecasting spot electricity prices based on machine learning and an adaptive mechanism is needed to cope with the multi-periodicity of various changes in the day-ahead market electricity price. Summary of the Invention

[0005] The present invention aims to provide a method and a computing device for forecasting spot electricity prices based on machine learning and an adaptive mechanism, and solve the problem of constructing a dynamic data set for forecasting spot electricity prices when the power generation unit changes its bidding strategy due to various factors such as the power supply and demand relationship.

[0006] According to one aspect of the present invention, a method for forecasting spot electricity prices based on machine learning and an adaptive mechanism is provided, including:

[0007] Obtaining a dynamic data set, and obtaining the dynamic data set on a device with a CPU and / or a GPU, wherein the obtained data set is transmitted through a communication network based on a communication protocol;

[0008] Calculating basic features by using the dynamic data set, where the basic features include the bidding space of thermal power units;

[0009] Constructing composite features based on the basic features to form a factor library;

[0010] Selecting features for modeling in the factor library, calculating the segmented index of the electricity price within the first time, and finding out the bidding space of the thermal power units corresponding to the demarcation electricity price;

[0011] Establishing a prediction model for each segmented index separately, splicing the data results of the segmented index prediction models, and performing spot electricity price forecasting.

[0012] According to some embodiments, the obtaining of the dynamic data set includes:

[0013] Obtaining a dynamic data set of electricity price, power supply and demand, and new energy power generation data;

[0014] Preprocess the dynamic data set and divide the dynamic data set into historical data and prediction data.

[0015] According to some embodiments, the factor library includes basic features of the bidding space of thermal power units and composite features constructed based on the basic features.

[0016] According to some embodiments, selecting features for modeling in the factor library includes:

[0017] Select historical data from the factor library, intercept the data of the first n days, and calculate the correlation between all features and the electricity price;

[0018] Sort according to the correlation, and select the n features with the highest correlation for modeling.

[0019] According to some embodiments, calculating the segmented index of the electricity price in the first time period includes:

[0020] Select the data of the first n days from the historical data, use the statistical index of the electricity price to divide the electricity price into the first price stage, the second price stage and the third price stage, and find out the thermal power bidding space corresponding to the boundary electricity price as the segmented index for dividing the data.

[0021] According to some embodiments, separately establishing a prediction model for each segmented index includes:

[0022] Construct prediction models for the first price stage, the second price stage and the third price stage respectively for segmented electricity price prediction, and retain the time stamps corresponding to the prediction data.

[0023] According to some embodiments, splicing the data results of the segmented index prediction models for spot electricity price prediction includes:

[0024] Sort according to the time stamp, merge the data results of the segmented electricity price prediction, and obtain the predicted electricity price for the whole stage;

[0025] Use polynomial regression to smooth the predicted electricity price for the whole stage to obtain the final predicted electricity price.

[0026] According to some embodiments, the obtaining of the dynamic data set further includes: filling the missing values in the dynamic data set with adjacent data.

[0027] According to another aspect of the present invention, there is provided a computer program product, including:

[0028] A computer program, which when executed by a processor implements any of the foregoing methods.

[0029] According to another aspect of the present invention, there is provided a computing device, including:

[0030] a processor; and

[0031] a memory storing a computer program which, when executed by the processor, causes the processor to execute any of the foregoing methods.

[0032] According to the exemplary embodiments of the present invention, by systematically collecting and processing data, it is possible to ensure that the dynamic data set used is as complete as possible, reduce the impact of missing values on subsequent analysis, the basic features can effectively capture the key factors affecting electricity price fluctuations, thereby providing richer information support for the model, and the composite features provide more dimensional information for the model, thereby improving the prediction accuracy; by separately modeling each segment, specific patterns within each interval can be captured more accurately, and then by splicing the prediction results of each segment model, a full-stage prediction sequence can be formed, making full use of the advantages of each segment model and enhancing the overall prediction effect.

[0033] According to the exemplary embodiments, a method for predicting spot electricity price based on machine learning and an adaptive mechanism, for multiple machine learning algorithms, selects the most suitable model for each segment characteristic according to different scenarios, thereby improving the prediction accuracy. This method can better understand and respond to market dynamics, help power generation companies optimize their bidding strategies, and also contribute to effective load management by grid operators, promoting the efficient operation of the entire power system.

[0034] It should be understood that the above general description and the following detailed description are only exemplary and do not limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments.

[0036] Figure 1 A flowchart showing a method for predicting spot electricity price based on machine learning and an adaptive mechanism according to an exemplary embodiment.

[0037] Figure 2 A schematic flowchart showing the process of establishing a prediction model according to an exemplary embodiment.

[0038] Figure 3 A block diagram showing a computing device according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. Like reference numerals in the figures denote like or similar parts, and thus their repetitive description will be omitted.

[0040] In addition, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present invention. However, those skilled in the art will realize that the technical solutions of the present invention may be practiced without one or more of the specific details, or may be implemented using other methods, components, devices, steps, etc. In other cases, well-known methods, devices, implementations, or operations are not shown or described in detail to avoid obscuring aspects of the present invention.

[0041] The block diagrams shown in the drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0042] The flowcharts shown in the drawings are merely illustrative and do not necessarily include all the content and operations / steps, nor are they necessarily executed in the order described. For example, some operations / steps may be decomposed, while some operations / steps may be combined or partially combined, so the actual execution order may change according to the actual situation.

[0043] It should be understood that although terms such as first, second, and third may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Thus, the first component discussed below may be referred to as the second component without departing from the teachings of the concept of the present invention. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0044] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of the relevant data need to comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entrances are provided for the user to select to authorize or reject.

[0045] Those skilled in the art can understand that the accompanying drawings are only schematic diagrams of exemplary embodiments, and the modules or processes in the accompanying drawings are not necessarily essential for implementing the present invention. Therefore, they cannot be used to limit the protection scope of the present invention.

[0046] In academia and industry, the methods for predicting spot electricity prices can be roughly divided into statistical methods, traditional machine learning, and deep learning methods. Statistical methods are mainly based on probability models, statistically analyzing historical data to establish a prediction model to fit the short-term trend of future electricity prices. However, statistical methods usually rely on assumptions such as data stationarity and autocorrelation. The supply and demand relationship in the electricity market is affected by various factors such as weather and policies, making it difficult to ensure the stability and accuracy of electricity price prediction. Traditional machine learning methods mainly focus on the relationship between independent variables and dependent variables in historical data and are used to predict future electricity prices, including methods such as linear regression, exponential regression, support vector machines, classification and regression trees, bagging and random forest regression tree models, etc. However, traditional machine learning algorithms usually require a large amount of historical data and effective features. Facing the non-linear, time-varying, and uncertain factors in the electricity market, the robustness of the algorithms is difficult to ensure. Deep learning methods mainly use historical data and algorithms such as backpropagation to learn a set of parameters to fit the historical environment and are used to predict future electricity prices.

[0047] The bidding strategies of power generation units are affected by various factors. Existing research and methods do not consider the impact of the bidding strategies on the data relationship of the power generation side, resulting in large fluctuations and losses during low and high electricity price periods. At the same time, affected by various factors, the bidding strategies of thermal power units may have large fluctuations in the same bidding space, which are not reflected in the existing models.

[0048] The various changes in the day-ahead market electricity price have multiple periodicities. Among them, the weekly, monthly, annual changes, etc. overlap and influence each other, forming a complex time pattern, making the mapping relationship between electricity price and power supply and demand change faster.

[0049] The present invention provides a method for predicting spot electricity prices based on machine learning and an adaptive mechanism, which can comprehensively consider factors such as the change gradient of price trends, the impact of the bidding strategies of power generation units, and the output of new energy units. According to the prediction day, training data is dynamically constructed to model and predict the electricity price, and the accuracy of electricity price prediction is effectively improved by combining data processing.

[0050] Before describing the embodiments of the present application, some terms or concepts related to the embodiments of the present application are explained.

[0051] Thermal power bidding space: refers to the price range allowed for thermal power plants or thermal power units to participate in bidding in the electricity market.

[0052] The exemplary embodiments of the present invention are described below with reference to the accompanying drawings.

[0053] Figure 1 The flowchart shows a method for predicting spot electricity prices based on machine learning and an adaptive mechanism according to an exemplary embodiment.

[0054] According to an exemplary embodiment, refer to Figure 1 , which shows the method steps for predicting spot electricity prices based on machine learning and an adaptive mechanism. Through this method, the accuracy and reliability of electricity price prediction can be improved.

[0055] In S101, obtain a dynamic data set.

[0056] According to an exemplary embodiment, obtain the dynamic data set on a device with a CPU / GPU. The data set is obtained through data transmission based on a communication protocol over a communication network, and a suitable communication protocol is selected for data transmission; collect historical data of the power market, including but not limited to electricity prices, power supply and demand relationships, and new energy power generation. Preprocess the dynamic data set to separate the historical data for training and the prediction data for testing or actual prediction. Handle missing values with neighboring data to ensure data integrity. By systematically collecting and organizing data, it can be ensured that the dynamic data set used is as complete as possible, reducing the impact of missing values or outliers on subsequent analysis.

[0057] In S103, calculate basic features using the dynamic data set, where the basic features include the bidding space of thermal power units.

[0058] Based on the collected data, calculate basic features using the dynamic data set. Calculating the basic features makes the meaning of each input variable clearer, facilitating understanding of the working principle of the model and the impact of each feature on the final prediction result. The basic features include the bidding space of thermal power units, which also reflects the actual operating conditions of thermal power units in the power market, enhancing the connection between the model and the actual business scenario. Calculating the basic feature of the bidding space of thermal power units for the above dynamic data set can effectively capture the key factors affecting electricity price fluctuations, thereby providing richer information support for the model. Among them, the thermal power bidding space refers to the power demand in the province after considering new energy output, non-marketized output, and inter-provincial power transmission. The calculation formula is:

[0059] Thermal power bidding space = Provincial load demand + Tie line power - New energy unit output - Non-marketized output

[0060] Among them, the competitive bidding space of thermal power reflects the power share that thermal power units still need to provide to the market after meeting new energy and non-marketized power sources; the provincial load demand refers to the actual power demand in a certain province at a certain moment; the tie line refers to the line connecting the power grids of two provinces, and its main function is to transmit power across provinces. The tie line power refers to the power transmitted between two inter-provincial power grids; the output of new energy units refers to the actual power generation of renewable energy power generation facilities such as wind energy and solar energy for a period of time; the non-marketized output refers to the power sources that do not participate in the market competition for power supply, such as nuclear power plants, hydropower plants or cogeneration units.

[0061] According to the calculated competitive bidding space of thermal power, analyze the competitiveness of thermal power units under the current market conditions, and accordingly adjust the power generation plan or the strategy of participating in the market. It is possible to decide whether to increase or decrease the power generation according to the competitive bidding space of thermal power, optimize the cost structure, which helps to conduct power dispatching more precisely and ensure the safe and stable power supply.

[0062] In S105, composite features are constructed based on the basic features to form a factor library.

[0063] According to the exemplary embodiment, composite features are constructed based on the basic features, new composite features are created using two or more basic features, and all the calculated basic features and composite features are organized into a comprehensive factor library, providing rich feature options for subsequent model selection.

[0064] The factor library consists of the basic features of the competitive bidding space of thermal power units and the composite features constructed based on the basic features. The composite features constructed based on the basic features, for example, are constructed into composite features through the two basic features of "provincial load demand" and "new energy output" using the functions in the function set. The functions in the function set generally include various operators, such as arithmetic operations, relational operations, logical operations, conditional operations, etc.; the function operations include sine, cosine, arcsine, arccosine, logarithm, exponent, power, maximum, minimum, square root, etc. A large number of composite features can be derived from the basic factors through operations, and then a factor library is formed.

[0065] In the electricity price prediction task, the algorithm hopes to retain the features with the greatest correlation with the electricity price. The correlation coefficient, also known as the Pearson correlation coefficient, is introduced in the selection process. The Pearson correlation coefficient is used to evaluate the fitness of each factor in the factor library. The Pearson correlation coefficient reflects the mutual relationship and the direction of correlation between two variables. According to the correlation, the feature factors with the greatest contribution are selected. The larger the correlation coefficient value, the higher the correlation with the electricity price and the higher the fitness, and the more suitable for modeling. The calculation formula of the Pearson correlation coefficient is:

[0066]

[0067] Among them, Cov(x, y) represents the covariance of two variables, and Var(x) and Var(y) represent the variances of the two variables respectively.

[0068] By combining different basic features, complex non-linear relationships hidden in the data can be revealed. By screening a large number of composite features, the most predictive feature subset can be selected to further optimize the model performance, enabling the model to better adapt to different market environments and technological advancements.

[0069] In S107, select the features for modeling in the factor library, calculate the segmented index of the electricity price within the first time period, and find the thermal power bidding space corresponding to the demarcation electricity price.

[0070] According to the exemplary embodiment, select the features for modeling in the factor library, select historical data from the factor library, intercept the data of the previous n days, calculate the correlation between all features and the electricity price, sort according to the correlation, and select the n features with the highest correlation for modeling.

[0071] Calculate the segmented index of the electricity price within the first time period. From the historical data, intercept the data of the previous n days. Using the statistical index of the electricity price, divide the electricity price into the first price stage, the second price stage, and the third price stage. Find the thermal power bidding space corresponding to the demarcation electricity price as the segmented index for dividing the data. Based on the data of the recent n days, use the statistical index of the electricity price to divide the electricity price into low electricity price, medium electricity price, and high electricity price, and find the thermal power bidding space corresponding to the demarcation electricity price as the segmented index for dividing the data.

[0072] Using the calculated critical thermal power bidding space, divide the prediction data into three stages: low electricity price, medium electricity price, and high electricity price. Finally, merge the data of the low electricity price, medium electricity price, and high electricity price stages, and use the fractional digit stage to form a dynamic data set for modeling.

[0073] In S109, establish a prediction model separately for each segmented index, splice the data results of the segmented index prediction models, and perform spot electricity price prediction.

[0074] Establish a prediction model separately for each segmented index. According to the three electricity price segments defined by the divided first price stage, second price stage, and third price stage, construct prediction models respectively for segmented electricity price prediction, and retain the time stamps corresponding to the prediction data. After the prediction of the three-tier data is completed, splice the data results of the segmented electricity price prediction, and sort them according to the time stamps to obtain the predicted electricity price for the entire stage.

[0075] Finally, perform smoothing processing on the predicted electricity price for the entire stage by means of polynomial regression, etc., to obtain the final predicted electricity price.

[0076] For the predicted spot electricity price, conduct power generation side strategy analysis. When it is predicted that the electricity spot price is higher than the marginal production cost of the power plant, the power plant can increase the power generation volume to the maximum level of its production capacity, thereby improving the resource utilization efficiency and the power plant's revenue; for the user side, commodity producers can combine electricity price prediction, utilize the fluctuations of the electricity price, adjust their production electricity consumption behavior, increase production activities during low electricity price periods, further reduce the comprehensive cost accounting of products, and help build a competitive advantage in price leadership for products. Not only can both the power generation side and the user side benefit from the spot electricity price prediction, but it also promotes the balance between supply and demand.

[0077] In this exemplary embodiment, the basic features can effectively capture the key factors affecting electricity price fluctuations, and the composite features provide deeper market insights for power generation companies, helping to formulate more scientific and reasonable bidding strategies; by dividing the electricity price into different intervals of low electricity price segments, medium electricity price segments, and high electricity price segments, and performing refined segmented processing, the segmented indicators can be dynamically adjusted according to changes in market conditions, enabling the model to better adapt to specific patterns within each interval and improving the prediction accuracy within the segment.

[0078] Figure 2 A flow diagram showing the establishment of a prediction model according to an exemplary embodiment is shown.

[0079] According to the exemplary embodiment, refer to Figure 2 , the establishment of the prediction model is divided into three segmented indicators, namely low electricity price segment, normal price segment, and high electricity price segment. Establishing a prediction model separately for each segmented indicator helps to improve the prediction accuracy within a specific range.

[0080] According to the above embodiment, obtain historical data and prediction data, fill in the missing values with adjacent data, and calculate the basic features; it is also possible to use the attention mechanism to calculate the weight of each feature in the historical data and combine neural network or machine learning algorithms for prediction.

[0081] For the establishment of the prediction model in each stage, quantile truncation processing is required. Select appropriate quantiles (such as 5% and 95%) to define the truncation points of the lower and upper bounds, sort the electricity prices in the dynamic dataset, and calculate the corresponding electricity price thresholds based on the selected quantiles. Set the electricity prices lower than the lower bound quantile to be equal to the value of the lower bound quantile, and set the electricity prices higher than the upper bound quantile to be equal to the value of the upper bound quantile, which can reduce the impact of outliers on the model; after performing quantile truncation, continue to execute the process of basic feature calculation, composite feature construction, and feature selection to ensure that all features used for modeling have undergone the same quantile truncation processing.

[0082] A regression model can be established using data after quantile truncation, and algorithms suitable for time series prediction such as XGBoost can be selected. According to the constructed factor library, through selection, the features participating in the calculation will be formed. These selected feature factors are used as X, and the target electricity price is used as Y for modeling training. The trained model is used for electricity price prediction to obtain segmented prediction results.

[0083] The final prediction result is obtained by combining the outputs of each segmented model. The data results of the segmented electricity price prediction are spliced and sorted by timestamp to ensure the correct chronological order of the predicted electricity prices. The predicted data is combined for overall prediction to ensure the prediction consistency within the entire price range.

[0084] According to some embodiments, the accuracy of the prediction model is an important step to ensure the model performance. The difference between the predicted value and the actual value can be measured by the average single-point deviation, specifically, the average of the absolute error or squared error at each time point. The calculation formula for evaluating the prediction accuracy of the algorithm by the average single-point deviation is:

[0085]

[0086] where n is the number of predicted electricity prices, and i is the i-th predicted point.

[0087] Truncating the data in time focuses on the supply-demand and electricity price relationship under the recent generator bidding strategies. At the same time, a factor library is constructed using specific algorithms, and the feature factors with the greatest contribution are selected according to the correlation. This way of combining the dynamic construction of the dataset and feature selection effectively improves the prediction accuracy of the spot electricity price.

[0088] Since the electricity market is highly dynamic, the prediction model should have good adaptability and be able to quickly respond to changes in market conditions. The trained model of the present invention can be used to predict the spot electricity price within a certain future time period, and the model can be continuously adjusted to adapt to market changes.

[0089] According to some embodiments, the algorithm backtests the effect of the day-ahead electricity price prediction algorithm from January 1, 2024 to July 31, 2024, and selects the data of the previous 13 months before the prediction day as historical data for rolling training and prediction on each prediction day. For example, if the prediction day is January 1, 2024, the historical data that can be obtained is from December 1, 2022 to December 31, 2023. The comparison algorithm uses fixed features such as power supply-demand and new energy output, and does not perform segmented processing.

[0090] The algorithm effect is as follows:

[0091] Backtest month The algorithm in this paper Non-segmented algorithm 2024-01 73.32％ 71.61％ 2024-02 72.64％ 69.13％ 2024-03 77.76％ 78.30％ 2024-04 86.21％ 83.37％ 2024-05 84.01％ 78.60％ 2024-06 91.02％ 81.98％ 2024-07 96.69％ 96.41％ Average 83.14％ 79.99％

[0092] In this backtest, the algorithm of the present invention was compared with the prediction results of the non-segmented algorithm. The data covered the backtest results from January 2024 to July 2024. The average return rate of the algorithm of the present invention during the backtest period was 83.14%, significantly higher than 79.99% of the non-segmented algorithm, indicating that the algorithm of the present invention has better prediction ability as a whole and reacts faster to changes in the power supply and demand environment.

[0093] In terms of algorithm stability, the algorithm of the present invention performs relatively more stably, and still can maintain good prediction results when the market fluctuates greatly; on the contrary, the performance of the non-segmented algorithm fluctuates relatively more, especially during the holiday period in February 2024 and the high-wind season, and the prediction accuracy of electricity prices is significantly lower than the average level.

[0094] According to the exemplary embodiment, the algorithm of the present invention shows better adaptability and stability in the environment of power supply and demand changes and changes in unit bidding strategies, making it a more robust spot electricity price prediction algorithm.

[0095] A method for predicting spot electricity prices based on machine learning and an adaptive mechanism proposed by the present invention can, by deeply mining the potential relationships in the data, construct a rich feature set, and can more comprehensively capture the factors affecting electricity prices based on machine learning; combined with the adaptive mechanism, the model can dynamically adjust parameters according to the latest market conditions, enhancing the model's understanding ability of key factors and ensuring good prediction performance in a rapidly changing environment. This method not only improves the accuracy and reliability of electricity price prediction, but also provides a powerful decision-making support tool for participants in the power market, can better respond to market dynamics, help power generation manufacturers optimize bidding strategies, and also helps grid operators conduct effective load management, promoting the efficient operation of the entire power system.

[0096] Figure 3 A block diagram of a computing device according to an exemplary embodiment of the present invention is shown.

[0097] As Figure 3 shown, the computing device 30 includes a processor 12 and a memory 14. The computing device 30 may further include a bus 22, a network interface 16, and an I / O interface 18. The processor 12, the memory 14, the network interface 16, and the I / O interface 18 may communicate with each other through the bus 22.

[0098] The processor 12 may include one or more general-purpose CPUs (Central Processing Unit, processors), microprocessors, or application-specific integrated circuits, etc., for executing relevant program instructions.

[0099] The memory 14 may include a machine system-readable medium in the form of volatile memory, such as random access memory (RAM), read-only memory (ROM), and / or cache memory. The memory 14 is used to store one or more programs containing instructions and data. The processor 12 can read the instructions stored in the memory 14 to execute the method according to the embodiments of the present invention as described above.

[0100] The computing device 30 can also communicate with one or more networks through the network interface 16. The network interface 16 can be a wireless network interface.

[0101] The bus 22 can include an address bus, a data bus, a control bus, etc. The bus 22 provides a path for exchanging information between components.

[0102] It should be noted that in the specific implementation process, the computing device 30 may also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above devices may also only include the components necessary to implement the solutions of the embodiments of this specification, and do not necessarily include all the components shown in the figure.

[0103] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps of the above method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, and magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic or optical cards, nano-systems (including molecular memory ICs), network storage devices, cloud storage devices, or any type of medium or device suitable for storing instructions and / or data.

[0104] The embodiments of the present invention also provide a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. The computer program is operable to cause a computer to execute some or all of the steps of any one of the methods described in the above method embodiments.

[0105] Those skilled in the art can clearly understand that the technical solutions of the present invention can be implemented by means of software and / or hardware. The "units" and "modules" in this specification refer to software and / or hardware that can independently complete or cooperate with other components to complete specific functions, where the hardware can be, for example, a field programmable gate array, an integrated circuit, etc.

[0106] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0107] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0108] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some service interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0109] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0110] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0111] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present invention.

[0112] In the above embodiments, the descriptions of the respective embodiments each have their own emphasis. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0113] The exemplary embodiments of the present invention have been specifically shown and described above. It should be understood that the present invention is not limited to the detailed structures, arrangements or implementation methods described herein; on the contrary, the present invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

[0114] Those skilled in the art can clearly understand that the technical solutions of the present invention can be implemented by means of software and / or hardware. The "units" and "modules" in this specification refer to software and / or hardware that can independently complete or cooperate with other components to complete specific functions, where the hardware can be, for example, a field programmable gate array, an integrated circuit, etc.

[0115] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0116] In the above embodiments, the descriptions of the respective embodiments each have their own emphasis. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0117] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some service interfaces, and the indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0118] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0119] In addition, in each embodiment of the present invention, each functional unit can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present invention.

[0121] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0122] The above specifically shows and describes the exemplary embodiments of the present invention. It should be understood that the present invention is not limited to the detailed structures, setting methods, or implementation methods described herein; on the contrary, the present invention is intended to cover various modifications and equivalent settings included within the spirit and scope of the appended claims.

Claims

1. A method for predicting spot electricity prices based on machine learning and an adaptive mechanism, characterized in that, Including: Obtain a dynamic data set, and obtain the dynamic data set on a device with a CPU and / or GPU, wherein the obtained data set performs data transmission based on a communication protocol through a communication network; Calculate basic features by using the dynamic data set, where the basic features include the bidding space of thermal power units; Construct composite features based on the basic features to form a factor library; Select features for modeling in the factor library, calculate the segmented index of the electricity price within the first time period, and find out the bidding space of the thermal power unit corresponding to the demarcation electricity price; Separate prediction models are established for each segmented index, and the data results of the segmented index prediction models are spliced to perform spot electricity price prediction.

2. The method according to claim 1, characterized in that, The obtaining of the dynamic data set includes: Obtain a dynamic data set of electricity price, power supply and demand, and new energy power generation data; Preprocess the dynamic data set and divide the dynamic data set into historical data and prediction data.

3. The method according to claim 1, wherein The factor library includes the basic features of the bidding space of thermal power units and the composite features constructed based on the basic features.

4. The method according to claim 2, wherein Selecting features for modeling in the factor library includes: Select historical data from the factor library, intercept the data of the first n days, and calculate the correlation between all features and the electricity price; Sort according to the correlation and select the n features with the highest correlation for modeling.

5. The method according to claim 2, wherein Calculating the segmented index of the electricity price within the first time period includes: Intercept the data of the first n days from the historical data, use the statistical index of the electricity price to divide the electricity price into the first price stage, the second price stage and the third price stage, and find out the bidding space of the thermal power corresponding to the demarcation electricity price as the segmented index for dividing the data.

6. The method according to claim 2, characterized in that Separate prediction models are established for each segmented index, including: Separate prediction models are constructed for the first price stage, the second price stage and the third price stage to perform segmented electricity price prediction, and the time stamps corresponding to the prediction data are retained.

7. The method according to claim 6, wherein Splicing the data results of the segmented index prediction models to perform spot electricity price prediction includes: Sort according to the time stamp, merge the data results of the segmented electricity price prediction, and obtain the predicted electricity price for the whole stage; Use polynomial regression to smooth the predicted electricity price for the whole stage to obtain the final predicted electricity price.

8. The method according to claim 2, wherein The obtaining of the dynamic data set further includes: filling the missing values in the dynamic data set with adjacent data.

9. A computer program product, characterized in that, Including: A computer program which, when executed by a processor, implements the method according to any one of claims 1-8.

10. A computing device, characterized in that, Including: A processor; And A memory storing a computer program, which when executed by the processor causes the processor to execute the method according to any one of claims 1-8.

Images