Wind power prediction method and device based on quantile regression and medium
Through the method based on quantile regression, the time domain convolution network and elastic network are used to predict wind power, which solves the problem of inaccurate prediction of wind power in the prior art, and achieves more accurate and stable prediction results.
Patent Information
- Application Number
- CN202411761673.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-04-04
AI Technical Summary
The existing technology is difficult to accurately predict wind power, which leads to huge challenges in safe operation and scheduling of power grids.
The wind power power prediction method based on quantile regression is used to predict the wind power power coarse prediction curve of multiple quantiles through time domain convolution network (TCN), and the multiquantile results are combined and integrated through elastic network (Elastic Net), and the fitting with the actual measured power of wind power is fine-tuned to generate the wind power power prediction curve.
The accuracy and robustness of wind power prediction are achieved, and a variety of effective confidence intervals are provided, providing effective support for grid decision-making.
Smart Images
Figure CN119761842A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind power prediction, and in particular to a wind power prediction method, device and medium based on quantile regression. Background Art
[0002] With the continuous increase in installed capacity of new energy, especially wind power and photovoltaic power in recent years, their strong dependence on weather, volatility and intermittency have brought huge uncertainty to the prediction of new energy power, thus bringing huge challenges to the safe operation and scheduling of the power grid.
[0003] Due to this uncertainty, using a single model for prediction may have large errors in some scenarios, and it is often difficult to learn the complex relationship between multiple variables. Even the most advanced prediction models cannot completely eliminate the uncertainty in the prediction. Therefore, there is currently a lack of an accurate wind power prediction solution. Summary of the invention
[0004] The present application provides a wind power prediction method, device and medium based on quantile regression to solve the above problems.
[0005] On the one hand, the present application provides a wind power prediction method based on quantile regression, the method comprising: predicting multiple quantiles based on a time domain convolutional network (TCN) to obtain a rough prediction curve of wind power under different probability distributions; performing a multi-quantile result combination and integration operation based on an elastic network (Elastic Net), fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
[0006] In one implementation of the present application, the process of generating a rough wind power prediction curve is specifically as follows: based on the sequence-to-point prediction operation of the time domain convolutional network TCN, the weather forecast for several hours before and after the time point to be predicted is used as input; based on the convolutional layer of the time domain convolutional network TCN, the relationship between meteorological data and wind power is learned, and several quantiles (y T1 ,y T2 , …, y Tn ), where 0<T1<T2<…<T n <1.
[0007] In one implementation of the present application, the method further includes: constructing an association between meteorological data within a time window and corresponding output through a time domain convolutional network (TCN).
[0008] In one implementation of the present application, the method further includes: in the part of the loss function, adding a penalty for the interval width so that the confidence interval is as narrow as possible while covering the true curve; wherein the process of penalizing the interval width is specifically as follows: for each quantile T i Calculate the quantile loss function. The final loss function is the mean of each quantile loss function plus the interval width penalty term, as shown in the following formula:
[0009]
[0010] in, α is the regularization coefficient, which is used to control the interval width.
[0011] In one implementation of the present application, the combined prediction based on the elastic network Elastic Net is specifically as follows: construct wind power output related features; wherein the wind power output related features specifically include: wind speed, wind direction, temperature, air pressure and humidity; mark important time periods, and convert different types of weather conditions into numerical forms; construct an elastic network and regression task, and dynamically optimize the regression results through the objective function; splice the multi-quantile results of the rough wind power prediction curve under the different probability distributions and the wind power output related features according to the time index, and input the spliced data into the elastic network for Bayesian optimization; after updating the iterative optimization, output the wind power prediction curve.
[0012] In one implementation of the present application, the objective function is as follows:
[0013] Loss=MSE+α·(λ1·L1+λ2·L2)
[0014] Among them, MSE (Mean Squared Error) is the mean square error, which is used to measure the difference between the predicted value and the true value. α controls the strength of the overall regularization. λ1 and λ2 control the influence of L1 and L2 regularization respectively, usually satisfying λ1+λ2=1.
[0015] In one implementation of the present application, the Bayesian optimization process is specifically as follows: a Gaussian process probability model is established to model the objective function, and the acquisition function is improved using the expected value, and its mathematical expression is as follows:
[0016] EI(x)=E[max(f(x)-f(x + )-ξ,0)]
[0017] Among them, f(x) is the unknown objective function that we want to maximize; x + represents the best position observed so far, representing (λ1,λ2); ξ>0 is a small positive number defined by the user to balance the relationship between exploration and exploitation.
[0018] In one implementation of the present application, the method further includes: each time a new point is selected for evaluation according to the acquisition function, the probability model is updated with the obtained new data until the objective function is minimized.
[0019] The present application also provides a wind power prediction device based on quantile regression, the device comprising: at least one processor; and 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 complete: predicting multiple quantiles based on a time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing a combination and integration operation of multiple quantile results based on an elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
[0020] The present application also provides a non-volatile computer storage medium for wind power prediction based on quantile regression, storing computer executable instructions, characterized in that the computer executable instructions are executed by a processor to achieve: predicting multiple quantiles based on a time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing a multi-quantile result combination and integration operation based on an elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
[0021] The present application provides a wind power prediction method, device and medium based on quantile regression, which has the following beneficial effects:
[0022] (1) A flexible new energy power probability prediction algorithm based on quantile regression is constructed. By regularizing the difference between quantiles, the confidence interval is made as narrow as possible while covering the actual output curve. This can provide multiple valid confidence intervals and provide effective support for decision-making.
[0023] (2) A flexible non-parametric combined prediction method is constructed, which can be applied to different quantile prediction curves or to the fusion of different model output results, thereby improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0025] Figure 1 A flow chart of a wind power prediction method based on quantile regression provided in an embodiment of the present application;
[0026] Figure 2 A simplified diagram of the TCN model provided in the embodiment of the present application;
[0027] Figure 3 A probability prediction effect diagram of wind power prediction and 95% confidence interval provided in an embodiment of the present application;
[0028] Figure 4 A schematic diagram of a wind power prediction device based on quantile regression provided in an embodiment of the present application. DETAILED DESCRIPTION
[0029] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0030] With the continuous increase in installed capacity of new energy, especially wind power and photovoltaic power in recent years, its strong dependence on weather, volatility, and intermittency have brought huge uncertainty to the prediction of new energy power, thus bringing great challenges to the safe operation and dispatch of power grids. Due to this uncertainty, the use of a single model for prediction may have large errors in some scenarios, and it is often difficult to learn such complex relationships between multiple variables. Even the most advanced prediction models cannot completely eliminate the uncertainty in the prediction. Therefore, it is very important to introduce probabilistic prediction based on quantile regression. It can help estimate the value of new energy output at different probability levels, not only focusing on a single trend, but also considering the distribution form of the data. On the other hand, how to make effective output predictions based on multiple quantile predictions is also crucial. This application combines probabilistic prediction and combined prediction algorithms, and uses a TCN-based algorithm to perform flexible multiple quantile regression predictions on the day-ahead wind power output. After that, this application adopts a combined prediction strategy to further effectively quantify and manage these uncertainties and improve the robustness and stability of the overall prediction.
[0031] Specifically, the embodiment of the present application provides a wind power prediction method, device and medium based on quantile regression. The technical solution proposed in the embodiment of the present application is described in detail below with the help of the accompanying drawings.
[0032] Figure 1 The flowchart of the wind power prediction method based on quantile regression provided in the embodiment of the present application is as follows. Figure 1 As shown, the method mainly includes the following steps:
[0033] Step 101: Predict multiple quantiles based on the time domain convolutional network (TCN) to obtain a rough prediction curve of wind power under different probability distributions.
[0034] In the embodiment of the present application, since there may be a time offset between the weather forecast and the actual weather at the station, the present application uses a sequence-to-point prediction, by inputting the weather forecast for several hours before and after the time point to be predicted, and learning the relationship between the weather data and the output through the TCN convolution layer, directly predicting several quantiles of the time point to be predicted. Where 0<τ1<τ2<...<τ n <1.
[0035] Specifically, assuming that this application wants to predict the wind power output quantiles at time t Then the constructed input set is [X t-w , X t-w+1 , ..., X t , X t+1 , ..., X t+w ]∈R D×(2w+1) , where X t ∈R D×1 is D meteorological elements at time t, including wind speed, wind direction, temperature, humidity, pressure, etc. This application uses TCN to build the relationship between the meteorological data in this time window and the corresponding output. Figure 2 As shown, b is the batch size, D is the dimension of the input meteorological data, T is the length of the input time window, and n is the number of quantiles required for prediction.
[0036] Furthermore, in the loss function, a penalty is added to the interval width so that the confidence interval is as narrow as possible while covering the true curve. Specifically, for each quantile T i Calculate the quantile loss function, and the final loss function is the mean of each quantile loss function plus the interval width penalty term:
[0037]
[0038] in, Where a is the regularization coefficient, which is used to control the width of the interval.
[0039] In the embodiment of the present application, the probability prediction effect diagram of wind power prediction and 95% confidence interval is as follows: Figure 3As shown, the blue curve is the actual measurement, the orange curve is the 50% quantile wind power prediction curve, the upper limit of the gray part is the 97.5% quantile prediction curve, the lower limit is the 5% quantile prediction curve, and the gray area is the 95% confidence interval.
[0040] Step 102: performing a multi-quantile result combination integration operation based on the elastic network Elastic Net, fine-tuning and fitting the generated rough wind power prediction curve under the different probability distributions with the wind power measured power, and generating a wind power prediction curve.
[0041] In the embodiment of the present application, feature engineering is first performed to construct wind power output-related features. In addition to basic meteorological features such as wind speed, wind direction, temperature, air pressure, humidity, etc., a series of complex features that may affect wind power output need to be constructed, including the effective wind speed perpendicular to the wind rotor surface and the air density correction factor based on parameters such as temperature and air pressure.
[0042] Furthermore, in order to have sufficient feature support in different weather conditions and different time periods, it is necessary to mark important time periods, including converting different types of weather conditions (sunny, cloudy, rainy, etc.) into numerical form for easy model processing. Categorical variables are converted using binary coding and converted into a set of binary numbers. For example, sunny, cloudy, and rainy days are coded as 00, 01, and 10 respectively; the periodic changes in time are marked by months and quarters; considering the impact of factors such as temperature difference on wind power, a day is divided into three sections, each 8 hours long.
[0043] Furthermore, an elastic network and regression task are constructed, and the regression results are dynamically optimized through the objective function. The objective function can be expressed as:
[0044] Loss=MSE+α·(λ1·L1+λ2·L2)
[0045] Among them, MSE (Mean Squared Error) is the mean square error, which is used to measure the difference between the predicted value and the true value. a controls the strength of the overall regularization. λ1 and λ2 control the influence of L1 and L2 regularization respectively, usually satisfying λ1+λ2=1.
[0046] Furthermore, the generated multi-quantile results of the rough wind power prediction curve and the constructed wind speed, wind direction, temperature, air pressure, humidity, effective wind speed, air density factor, weather type classification, monthly and quarterly periodic characteristics, daily time-sharing characteristics, etc. are spliced together according to the time index, and the spliced data are input into the elastic network.
[0047] Furthermore, the MSE in the loss function is the root mean square error between the measured power and the power prediction, and the Bayesian optimization is used for parameter selection. Bayesian optimization first establishes a Gaussian process probability model to model the objective function mentioned in step c. The acquisition function uses the expected improvement, and its mathematical expression is:
[0048] EI(x)=E[max(f(x)-f(x + )-ξ,0)]
[0049] f(x) is the unknown objective function that this application wants to maximize. + represents the best position observed so far, representing (λ1,λ2) in step c. ξ>0 is a small positive number defined by the user to balance the relationship between exploration and exploitation. The predicted distribution of f(x) is a normal distribution, so the above probability can be converted into a standard normal cumulative distribution function.
[0050] Further, iterative update: each time a new point is selected for evaluation according to the acquisition function, the probability model is updated with the new data obtained, and the above process is repeated until the objective function is minimized. At this point, the output of the elastic network is the wind power prediction result after the combined prediction.
[0051] The above is a wind power prediction method based on quantile regression provided by an embodiment of the present application. Based on the same inventive concept, an embodiment of the present application also provides a wind power prediction device based on quantile regression. Figure 4 A schematic diagram of a wind power prediction device based on quantile regression provided in an embodiment of the present application, such as Figure 4 As shown, the device mainly includes: at least one processor 401; and a memory 402 that is communicatively connected to the at least one processor; wherein the memory 402 stores instructions that can be executed by the at least one processor 401, and the instructions are executed by the at least one processor 401 so that the at least one processor 401 can complete: predicting multiple quantiles based on the time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing a combination and integration operation of multiple quantile results based on the elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under the different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
[0052] In addition, an embodiment of the present application also provides a non-volatile computer storage medium for wind power prediction based on quantile regression, storing computer executable instructions, which are executed by a processor to achieve: predicting multiple quantiles based on a time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing a multi-quantile result combination and integration operation based on an elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
[0053] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0054] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0055] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0056] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0057] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0058] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0059] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A wind power prediction method based on quantile regression, characterized in that: The method comprises: predicting multiple quantiles based on a time domain convolutional network (TCN) to obtain a rough prediction curve of wind power under different probability distributions; performing a combination and integration operation of multiple quantile results based on an elastic network (Elastic Net), fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual wind power measurement to generate a wind power prediction curve.
2. The wind power prediction method based on quantile regression according to claim 1 is characterized in that: The generation process of the rough wind power prediction curve is as follows: The sequence-to-point prediction operation based on the time domain convolutional network (TCN) takes the weather forecast several hours before and after the predicted time point as input; Based on the time domain convolutional network TCN convolutional layer, the relationship between meteorological data and wind power is learned to predict several quantiles (y T1 ,y T2 , …, y Tn ), where 0<T1<T2<…<T n <1.
3. The wind power prediction method based on quantile regression according to claim 2 is characterized in that: The method also includes: constructing a correlation between the meteorological data in the time window and the corresponding output through a time domain convolutional network (TCN).
4. The wind power prediction method based on quantile regression according to claim 2 is characterized in that: The method further includes: in the loss function, adding a penalty for the interval width so that the confidence interval is as narrow as possible while covering the true curve; wherein the process of penalizing the interval width is specifically as follows: for each quantile T i Calculate the quantile loss function. The final loss function is the mean of each quantile loss function plus the interval width penalty term, as shown in the following formula: in,(·) + =max(0,·), α is the regularization coefficient, which is used to control the interval width.
5. The wind power prediction method based on quantile regression according to claim 1 is characterized in that: Based on the Elastic Net combined prediction, specifically: Constructing wind power output related characteristics; wherein the wind power output related characteristics specifically include: wind speed, wind direction, temperature, air pressure and humidity; Mark important periods and convert different types of weather conditions into numerical values; Build elastic networks and regression tasks, and dynamically optimize regression results through objective functions; splicing the multi-quantile results of the rough wind power prediction curve under the different probability distributions and the wind power output-related features according to the time index, and inputting the spliced data into the elastic network for Bayesian optimization; After updating and iterative optimization, the wind power prediction curve is output.
6. The wind power prediction method based on quantile regression according to claim 5 is characterized in that: The objective function is as follows: Loss=MSE+α·(λ1·L1+λ2·L2) Among them, MSE (Mean Squared Error) is the mean square error, which is used to measure the difference between the predicted value and the true value; a is used to control the strength of the overall regularization; λ1 and λ2 control the influence of L1 and L2 regularization respectively, usually satisfying λ1+λ2=1.
7. The wind power prediction method based on quantile regression according to claim 5 is characterized in that: The process of Bayesian optimization is as follows: A Gaussian process probability model is established to model the objective function, and the acquisition function uses expected improvement, and its mathematical expression is as follows: EI(x)=E[max(f(x)-f(x + )-ξ,0)] Among them, f(x) is the unknown objective function that we want to maximize; x + represents the best position observed so far, representing (λ1,λ2); ξ>0 is a small positive number defined by the user to balance the relationship between exploration and exploitation.
8. The method for wind power prediction based on quantile regression according to claim 5, characterized in that: The method further comprises: each time a new point is selected for evaluation according to the acquisition function, the probability model is updated with the obtained new data until the objective function is minimized.
9. A wind power prediction device based on quantile regression, characterized in that: The device comprises: 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 complete: predicting multiple quantiles based on the time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing multi-quantile result combination and integration operations based on the elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.
10. A non-volatile computer storage medium for wind power prediction based on quantile regression, storing computer executable instructions, characterized in that: The computer executable instructions are executed by the processor to achieve: predicting multiple quantiles based on the time domain convolutional network TCN to obtain a rough prediction curve of wind power under different probability distributions; performing a multi-quantile result combination integration operation based on the elastic network Elastic Net, fine-tuning and fitting the generated rough prediction curve of wind power under different probability distributions with the actual measured wind power, and generating a wind power prediction curve.