Method and device for determining power prediction result, storage medium and electronic device

Abstract

The invention discloses a power prediction result determination method and device, a storage medium and an electronic device, and relates to the field of computers, and the method comprises the steps: obtaining the meteorological prediction data of m meteorological sources, carrying out the power prediction of the obtained m meteorological prediction data through m groups of power prediction models, and obtaining m * n power prediction results, each group of power prediction models comprises n power prediction models, the m meteorological sources are in one-to-one correspondence with the m groups of power prediction models, and m and n are positive integers; determining model weights of n first power prediction models corresponding to the first meteorological source, and integrating n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result; and determining meteorological source weights of the m meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

Description

Technical Field

[0001] The present application relates to the field of new energy, and specifically, to a method and device for determining power prediction results, a storage medium and an electronic device. Background Art

[0002] Affected by factors such as weather and topography, the power generation of wind farms is random and volatile. The access of large-scale wind farms to the power grid has brought important challenges to the dispatching and stable operation of the power grid. The short-term power forecasting technology of wind farms uses weather forecast data to predict the power generation of wind farms in the next 1-3 days, which can provide data support for power grid dispatching and station operation and maintenance management.

[0003] With the development and application of artificial intelligence and big data technology, the current short-term power forecasting method for wind farms is mainly based on data-driven statistical methods. Machine learning models including random forests, lightweight hoists, support vector machines, etc. have been applied to a certain extent in the field of wind power forecasting. For the short-term power forecasting problem of the same wind farm, the forecast results given by different models are not the same, and the forecast accuracy has advantages and disadvantages under different conditions. In addition, the same short-term wind power forecasting model has different forecast results depending on the different meteorological forecast data sources used. How to select the possible optimal result from multiple forecast results for power grid dispatching and station operation and maintenance decisions is an important issue facing wind power forecasting. At present, the usual practice is to backtest the historical accuracy of the forecast results of different models over a period of time, and select the model with the highest accuracy as the model for outputting wind power forecast results in the future. This model evaluation and optimization method only uses a single accuracy evaluation indicator, which cannot comprehensively evaluate the forecasting effect of the model under different scenarios, and the forecast results for a period of time in the future are all generated by the selected single model, which cannot integrate the advantages of each model. When facing changes in meteorological characteristics or station operation conditions during operation, the adaptability of the forecast results is poor.

[0004] Regarding the related technologies, currently a single prediction method cannot comprehensively evaluate the prediction effect of the model under different scenarios, resulting in poor adaptability of the prediction results. No effective solution has been proposed yet.

[0005] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0006] The embodiments of the present application provide a method and device for determining power prediction results, a storage medium and an electronic device, so as to at least solve the problem in the related art that a single prediction method cannot comprehensively evaluate the prediction effect of the model in different scenarios, resulting in poor adaptability of the prediction results.

[0007] According to one aspect of an embodiment of the present application, a method for determining a power prediction result is provided, comprising: obtaining meteorological forecast data of m types of meteorological sources, and performing power prediction on the obtained m meteorological forecast data respectively through m groups of power prediction models to obtain m*n power prediction results, wherein each group of power prediction models includes n power prediction models, the m types of meteorological sources correspond one-to-one to the m groups of power prediction models, and m and n are both positive integers; determining model weights of n first power prediction models corresponding to a first meteorological source, and integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results; determining meteorological source weights of the m types of meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0008] In an exemplary embodiment, meteorological forecast data of m kinds of meteorological sources are obtained, and power prediction is performed on the m meteorological forecast data respectively through m groups of power prediction models to obtain m*n power prediction results. The method also includes: obtaining historical meteorological forecast data of m kinds of meteorological sources, and obtaining actual power data corresponding to the m historical meteorological forecast data; constructing m groups of meteorological features according to the m historical meteorological forecast data; using the m groups of meteorological features as input and the m actual power data as input, respectively training power prediction models through n types of model training methods to obtain the m groups of power prediction models.

[0009] In an exemplary embodiment, the model weights of n first power prediction models corresponding to the first meteorological source are determined, including: evaluating each first power prediction model according to t evaluation indicators to obtain n groups of evaluation indicator values, wherein each group of evaluation indicator values ​​includes t evaluation indicator values ​​corresponding to the t evaluation indicators, and t is a positive integer; determining the model weights of the n first power prediction models according to the n groups of evaluation indicator values.

[0010] In an exemplary embodiment, the model weights of the n first power prediction models are determined according to the n groups of evaluation index values, including: performing forward processing on the n groups of evaluation index values ​​to obtain n groups of forward-normalized evaluation index values, and constructing an index evaluation matrix according to the n groups of forward-normalized evaluation index values; performing normalization processing on the index evaluation matrix to obtain a normalized index evaluation matrix; determining the initial model weights of the n first power prediction models according to the normalized index evaluation matrix; and correcting the n initial model weights according to the time period accuracy of the n first power prediction models to obtain n model weights.

[0011] In an exemplary embodiment, the n initial model weights are corrected according to the time period accuracy of the n first power prediction models to obtain n model weights, including: calculating the time period accuracy of the n first power prediction models in s time periods to obtain n*s time period accuracy rates; determining the average time period accuracy rate of the n first power prediction models in each time period to obtain s average time period accuracy rates; when it is determined that the difference between the time period accuracy rate of the second power prediction model in the i-th time period and the time period accuracy rate average value is greater than a preset threshold, correcting the model weights of the n first power prediction models in the i-th time period according to the difference, wherein the n first power prediction models include the second power prediction model.

[0012] In an exemplary embodiment, determining the meteorological source weights of the m meteorological sources includes: determining the wind speed prediction accuracy and the wind direction prediction accuracy of the m meteorological sources; determining the meteorological prediction accuracy of the m meteorological sources according to the wind speed prediction accuracy and the wind direction prediction accuracy; and calculating the meteorological prediction accuracy of the m meteorological sources by formula Determine the meteorological source weights of the m meteorological sources, where w 气象，i is the meteorological source weight of the ith meteorological source, acc 气象，i is the weather forecast accuracy of the i-th weather source.

[0013] According to another aspect of an embodiment of the present application, a device for determining a power prediction result is also provided, including: an acquisition module, used to acquire meteorological forecast data of m types of meteorological sources, and perform power prediction on the obtained m meteorological forecast data through m groups of power prediction models respectively, to obtain m*n power prediction results, wherein each group of power prediction models includes n power prediction models, the m types of meteorological sources correspond one-to-one to the m groups of power prediction models, and m and n are both positive integers; a determination module, used to determine the model weights of n first power prediction models corresponding to a first meteorological source, and integrate the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results; an integration module, used to determine the meteorological source weights of the m types of meteorological sources, and integrate the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0014] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for determining the power prediction result when running.

[0015] According to another aspect of an embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method for determining the power prediction result through the computer program.

[0016] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps of the method described in each embodiment of the present application are implemented.

[0017] The present application provides a method for determining a power prediction result, which comprises first obtaining meteorological prediction data of m kinds of meteorological sources, and performing power prediction on the obtained m meteorological prediction data respectively through m groups of power prediction models to obtain m*n power prediction results, wherein each group of power prediction models includes n power prediction models, and the m kinds of meteorological sources correspond to the m groups of power prediction models one-to-one, and m and n are both positive integers; determining the model weights of the n first power prediction models corresponding to the first meteorological source, and then integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m kinds of meteorological sources include the first meteorological source, and the m groups of power prediction models include the n first power prediction models, m*n power prediction results including the n first power prediction results; determining the meteorological source weights corresponding to the m meteorological sources, and finally integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result; adopting the above scheme, a model integration method for short-term power prediction of wind farms is provided, which integrates the prediction results of multiple models, explores and combines the advantages of multiple algorithms, overcomes the instability and system deviation of a single prediction method, and makes the prediction performance more stable and accurate; thereby solving the problem in related technologies that the prediction effect of the model under different scenarios cannot be comprehensively evaluated by a single prediction method, resulting in poor adaptability of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0019] In order to more clearly illustrate the embodiments of the present application 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, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0020] Figure 1 It is a hardware structure block diagram of a computer terminal of a method for determining a power prediction result according to an embodiment of the present application;

[0021] Figure 2 is a flow chart of a method for determining a power prediction result according to an embodiment of the present application;

[0022] Figure 3 is a flow chart of a process for integrating multiple power prediction results according to an embodiment of the present application;

[0023] Figure 4 It is a structural block diagram of a device for determining a power prediction result according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. 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 creative work should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] The method embodiments provided in the embodiments of the present application can be executed in a computer terminal or a similar computing device. Taking running on a computer terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a computer terminal for determining a power prediction result according to an embodiment of the present application. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor (Central Processing Unit, CPU) or a programmable logic device (Field Programmable Gate Array, FPGA)) and a memory 104 for storing data, wherein the above-mentioned computer terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0027] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the power prediction result in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0028] A wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0029] In this embodiment, a method for determining a power prediction result is provided. Figure 2 is a flow chart of a method for determining a power prediction result according to an embodiment of the present application, such as Figure 2 As shown, the method comprises the following steps:

[0030] Step S202, obtaining meteorological forecast data of m meteorological sources, and performing power forecasting on the obtained m meteorological forecast data respectively through m groups of power forecasting models, to obtain m*n power forecasting results, wherein each group of power forecasting models includes n power forecasting models, the m meteorological sources correspond to the m groups of power forecasting models one by one, and m and n are both positive integers;

[0031] It should be noted that the various meteorological sources mentioned in the embodiments of the present application generally refer to weather forecast information from different institutions, using different models or algorithms, and based on different data sets. In meteorology and related application fields, weather forecast data is one of the key inputs for wind farm power prediction, because meteorological factors such as wind speed, wind direction, temperature, and humidity directly affect the efficiency and capacity of wind power generation.

[0032] Step S204, determining the model weights of the n first power prediction models corresponding to the first meteorological source, and integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results;

[0033] Step S206, determining the meteorological source weights of the m meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0034] Through the present application, meteorological forecast data of m kinds of meteorological sources are first obtained, and power prediction is performed on the obtained m meteorological forecast data through m groups of power prediction models respectively, to obtain m*n power prediction results, wherein each group of power prediction models contains n power prediction models, and the m kinds of meteorological sources correspond to the m groups of power prediction models one by one, and m and n are both positive integers; the model weights of the n first power prediction models corresponding to the first meteorological source are determined, and then the n first power prediction results corresponding to the first meteorological source are integrated according to the n model weights to obtain a second power prediction result, wherein the m kinds of meteorological sources include the first meteorological source, and the m groups of power prediction models include the n first power prediction models. The m*n power prediction results include the n first power prediction results; the meteorological source weights corresponding to the m meteorological sources are determined, and finally the m second power prediction results are integrated according to the m meteorological source weights to obtain a third power prediction result; the above scheme is adopted to provide a model integration method for short-term power prediction of wind farms, which integrates the prediction results of multiple models, explores and combines the advantages of multiple algorithms, overcomes the instability and system deviation of a single prediction method, and makes the prediction performance more stable and accurate; thereby solving the problem in the related technology that the prediction effect of the model under different scenarios cannot be comprehensively evaluated by a single prediction method, resulting in poor adaptability of the prediction results.

[0035] In an exemplary embodiment, meteorological forecast data of m kinds of meteorological sources are obtained, and power prediction is performed on the m meteorological forecast data respectively through m groups of power prediction models to obtain m*n power prediction results. The method also includes: obtaining historical meteorological forecast data of m kinds of meteorological sources, and obtaining actual power data corresponding to the m historical meteorological forecast data; constructing m groups of meteorological features according to the m historical meteorological forecast data; using the m groups of meteorological features as input and the m actual power data as input, respectively training power prediction models through n types of model training methods to obtain the m groups of power prediction models.

[0036] First, it is necessary to obtain multiple basic models for integration (i.e., the power prediction model mentioned above) through data collection, feature selection, and model training. The main steps include:

[0037] Step 1: Data collection. Collect historical meteorological forecast data from multiple meteorological sources. The data elements should at least include wind speed, wind direction, temperature, humidity, air pressure, precipitation, etc. The hub height of existing wind turbines is usually above 70 meters. The wind speed and wind direction in the meteorological forecast data should at least include different height layers such as 10 meters, 50 meters, 100 meters, and 150 meters, covering the height that the wind turbine blades can reach. Elements such as temperature, humidity, air pressure, and precipitation are values ​​at ground height. Collect actual power data of wind farms and align them in time with meteorological forecast data from multiple meteorological sources, with a time resolution of 15 minutes.

[0038] Step 2: Feature construction. For each meteorological source, the meteorological elements with a greater impact on the actual power (such as a linear correlation coefficient greater than 0.3) are screened out through correlation analysis, principal component analysis and other methods as feature inputs for basic model training. The feature elements screened out by different meteorological sources may be different. In addition, the time features of the meteorological element time series, such as month, day, and hour, are extracted as feature inputs for basic model training.

[0039] Step 3: Model training. For each meteorological source, the features obtained in step 2 are used as input and the actual power is used as output. A variety of methods are used to train the short-term prediction model for wind farm power generation to obtain multiple power prediction results. The various methods should have different technical routes and mathematical principles, so as to reflect the correlation between meteorological forecast data and actual power from different dimensions as much as possible. Specifically, the prediction methods may include linear regression, support vector machine, multi-layer perceptron, random forest, long short-term memory network, etc.

[0040] Through the above steps, multiple short-term prediction models of wind farm power generation based on different meteorological sources (ie, m*n power prediction models) are obtained as the basic model for model integration.

[0041] Optionally, determining the model weights of n first power prediction models corresponding to the first meteorological source includes: evaluating each first power prediction model according to t evaluation indicators to obtain n groups of evaluation indicator values, wherein each group of evaluation indicator values ​​includes t evaluation indicator values ​​corresponding to the t evaluation indicators, and t is a positive integer; determining the model weights of the n first power prediction models according to the n groups of evaluation indicator values.

[0042] The process of determining the model weight of each first power prediction model includes: firstly establishing a multi-dimensional index evaluation system, which may specifically include the following indexes:

[0043] (1) Accuracy. The most important indicator of wind farm power forecasting is accuracy, which is usually calculated on a daily basis. The accuracy calculation formulas are defined differently in different provinces where different stations are located. The most commonly used method is to calculate based on the root mean square error, which is:

[0044] Where acc_day is the daily accuracy, n is the number of data points in one day. In the case of 15-minute resolution, there are 96 data points in one day. pi is the predicted power, p ri is the actual power. According to the daily accuracy, the average accuracy of each basic model acc_day_mean is further calculated as the accuracy parameter of the basic model.

[0045] (2) Accuracy variance. In terms of prediction stability, the prediction accuracy variance of each basic model is calculated based on the daily accuracy data to reflect the fluctuation of the daily accuracy. The variance calculation formula is:

[0046] Where σ is the daily accuracy variance, and k is the number of days of backtest data.

[0047] (3) Correlation coefficient. The correlation coefficient can reflect the consistency of the change trend between the predicted power and the actual power, and measure the prediction effect of the model on the power change trend. The daily correlation coefficient calculation formula is:

[0048] Where corr_day is the correlation coefficient between the predicted power and the actual power for a single day. is the daily average power forecast, is the daily average value of actual power. According to the daily correlation coefficient, the average correlation coefficient of each basic model is further calculated corr_day_mean.

[0049] (4) Mean absolute error. The accuracy is calculated based on the relative error calculated based on the actual power. It is a relative indicator. The absolute value of the power of the predicted deviation is also very important for the operation of the wind farm. The calculation formula of the mean absolute error is:

[0050] Where MAE is the mean absolute error and l is the total number of points on the backtest date.

[0051] After obtaining n groups of evaluation index values ​​according to the statistical method of the above different index values, the model weight of each first power prediction model is determined according to the n groups of evaluation index values.

[0052] Optionally, the model weights of the n first power prediction models are determined according to the n groups of evaluation index values, including: performing forward processing on the n groups of evaluation index values ​​to obtain n groups of forward-normalized evaluation index values, and constructing an index evaluation matrix according to the n groups of forward-normalized evaluation index values; performing normalization processing on the index evaluation matrix to obtain a normalized index evaluation matrix; determining the initial model weights of the n first power prediction models according to the normalized index evaluation matrix; and correcting the n initial model weights according to the time period accuracy of the n first power prediction models to obtain n model weights.

[0053] The process of determining the model weight according to the evaluation index value includes: in an optional embodiment, assuming that there are m models based on a certain meteorological source, the average accuracy, accuracy variance, correlation coefficient, and mean absolute error of each model are calculated respectively to form an index evaluation matrix A, where element a ij Represents the j-th evaluation index value of the ith model.

[0054] Among the four indicators involved in the indicator evaluation matrix A, the higher the average accuracy and correlation coefficient, the better the model prediction effect, and the lower the accuracy variance and mean absolute error, the better the model prediction effect. In order to further determine the model weight, it is necessary to perform a positive transformation on the accuracy variance and mean absolute error so that the higher their values ​​are, the better the model prediction effect is. For the accuracy variance indicator, 1-σ is defined as a new variance type indicator. For the mean absolute error indicator, Cap-MAE is defined as a new absolute error type indicator, where Cap is the installed capacity of the wind farm. The matrix composed of the values ​​of each indicator after positive transformation is defined as B (that is, the above-mentioned indicator evaluation matrix), where the element b ij Represents the j-th positive evaluation index value of the ith model.

[0055] For the indicator evaluation matrix B, the matrix elements are normalized by column to avoid the influence of different magnitudes on the weight determination. That is, for each element b ij , the normalized value is calculated by the following method to form the normalized index evaluation matrix C.

[0056] For the matrix evaluation index C, calculate the sum of the values ​​of each index in each row, that is, Press the formula to d i Normalize to get the final weight coefficient w of each model i .

[0057] Finally, since the overall power generation of wind farms is high at night and low during the day, the peak power load usually occurs in the morning and evening. The power prediction accuracy during the peak wind power generation period and the peak power consumption period is more important for grid dispatching and station operation benefits. Therefore, it is necessary to consider the prediction accuracy by time period. Therefore, the embodiment of the present application also introduces an evaluation index for the accuracy of the time period. The resolution of the wind power prediction time series is 15 minutes, 4 points per hour, and the prediction accuracy calculation formula for each hour is: For each basic model, calculate the prediction accuracy of each hour on all backtest dates, and average the prediction accuracy of the same hour on different dates to get the time period accuracy acc_hour_mean. Then modify the model weight according to the time period accuracy to get the final model weight (i.e. the n model weights mentioned above).

[0058] Optionally, the n initial model weights are corrected according to the time period accuracies of the n first power prediction models to obtain n model weights, including: calculating the time period accuracies of the n first power prediction models in s time periods to obtain n*s time period accuracies; determining the average time period accuracies of the n first power prediction models in each time period to obtain s average time period accuracies; when it is determined that the difference between the time period accuracy of the second power prediction model in the i-th time period and the average time period accuracy is greater than a preset threshold, correcting the model weights of the n first power prediction models in the i-th time period according to the difference, wherein the n first power prediction models include the second power prediction model.

[0059] In order to consider the difference in prediction performance of each model in different time periods, based on the above model weight determination method, the weight is modified in combination with the prediction accuracy of each model in different time periods. The time period accuracy of each model in each time period of 24 hours a day is calculated, represented by acc_hour_mean_i_k, i=1~m represents the model number, and k=1~24 represents the time period number.

[0060] If the difference between the accuracy of a model in a certain time period and the average accuracy of the remaining models in the time period (i.e. the above-mentioned time period accuracy average) is above 5% (i.e. the above-mentioned preset threshold), it is considered that the accuracy of the model (i.e. the above-mentioned second power prediction model) in the time period is significantly higher, and this difference is recorded as α, and the corrected weight of the model in the time period is w′ i =w i +α, the modified weights of the remaining models in this period are

[0061] Through the above implementation, the model weights of the various power prediction models are obtained, and the power prediction results of the various models are weighted according to the model weights to obtain a model integration result based on a single meteorological source.

[0062] Specifically, for the time period where the weights are not corrected, the model ensemble prediction result =

[0063]

[0064] For the time period of weight correction,

[0065] Finally, the above two parts of prediction results are sorted in chronological order to obtain the model integration prediction results based on a single meteorological source.

[0066] Optionally, determining the meteorological source weights of the m meteorological sources includes: determining the wind speed prediction accuracy and wind direction prediction accuracy of the m meteorological sources; determining the meteorological prediction accuracy of the m meteorological sources according to the wind speed prediction accuracy and the wind direction prediction accuracy; and calculating the meteorological prediction accuracy of the m meteorological sources by formula Determine the meteorological source weights of the m meteorological sources, where w 气象，i is the meteorological source weight of the ith meteorological source, acc 气象，i is the weather forecast accuracy of the i-th weather source.

[0067] Different meteorological sources may have systematic deviations in the prediction of meteorological elements due to differences in data sources, technical routes, theoretical assumptions, etc. The power prediction model based on a single meteorological source is affected by the systematic deviation of the meteorological source, and the prediction results may be poor under some conditions. It is impossible to obtain accurate prediction results by integrating the results of the power prediction model based on a single meteorological source. Therefore, it is necessary to integrate the characteristics of multiple meteorological sources and further integrate the results of the power prediction models based on different meteorological sources.

[0068] The process of determining the meteorological source weights for each meteorological source includes:

[0069] Step 1: Calculate the prediction accuracy of each meteorological source.

[0070] For each meteorological source, the prediction accuracy of wind speed and wind direction, the main meteorological elements affecting wind power, is calculated, that is, where acc 风速，i represents the wind speed prediction accuracy of the i-th meteorological source, ws represents the wind direction (wind speed), acc 风向，i represents the wind direction prediction accuracy of the i-th meteorological source, and wd represents the wind direction.

[0071] With acc 风速，i and acc 风向，iThe average value of is taken as the prediction accuracy of the i-th meteorological source acc 气象，i .

[0072] Step 2: Calculate the meteorological source weight of each meteorological source.

[0073] w 气象，i represents the meteorological source weight of the ith meteorological source, and the calculation formula is:

[0074] After obtaining the meteorological source weights of each meteorological source, the power prediction results based on each meteorological source (i.e., the m second power prediction results) are weighted according to the meteorological source weights to obtain the integrated result of the power prediction model based on multiple meteorological sources (i.e., the third power prediction result), that is: This completes the integration of multi-model prediction results.

[0075] Through the above embodiments, the present application establishes a multi-dimensional evaluation index system for wind power prediction results of multiple meteorological sources and multiple models, conducts a comprehensive evaluation of different prediction results, and uses the multi-dimensional index evaluation results to obtain the integrated weight of the basic model. The weight correction of the time period accuracy and the prediction system deviation of multiple meteorological sources are considered to achieve the integration of wind power prediction results and realize a more stable and accurate short-term power forecast of wind farms.

[0076] In order to better understand the method for determining the power prediction result described in the above embodiment, the present application embodiment is combined with Figure 3 The integration process of the power prediction results of multiple models is further explained, such as Figure 3 As shown, firstly, the basic model is trained, data is collected, and feature engineering is constructed from the collected data. Then, a variety of model training methods are used for model training, such as linear regression, random forest, long short-term memory network and other model training methods, to obtain a variety of power prediction models.

[0077] After power forecasting is performed using the multiple power forecasting models obtained and multiple power forecasting results are obtained, the power forecasting results are integrated based on the power forecasting model of a single meteorological source. A multi-dimensional indicator system is first constructed, including accuracy, accuracy variance, correlation coefficient, and mean absolute error. Model weights are then calculated based on these evaluation indicators, and the time period accuracy indicator is used to correct the model weights to obtain the model weights of multiple power forecasting models under a single meteorological source. The model weights are used to integrate the power forecasting results under a single meteorological source to obtain the above-mentioned m second power forecasting results.

[0078] Finally, the power prediction model based on multiple meteorological sources is integrated. First, the prediction accuracy of different meteorological sources is used to determine the meteorological source weight of each meteorological source. The meteorological source weight is used to integrate the m second power prediction results to obtain an accurate and stable wind power model integration result.

[0079] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0080] Optionally, in the present embodiment, a device for determining a power prediction result is also provided, and the device is applied to the above-mentioned system embodiment and preferred implementation mode, and the description thereof will not be repeated. As used below, the term "module" may implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0081] Figure 4 : is a structural block diagram of a device for determining a power prediction result according to an embodiment of the present application, the device comprising:

[0082] An acquisition module 42 is used to acquire meteorological forecast data of m meteorological sources, and perform power forecasting on the m meteorological forecast data obtained by using m groups of power forecasting models to obtain m*n power forecasting results, wherein each group of power forecasting models includes n power forecasting models, and the m meteorological sources correspond to the m groups of power forecasting models one by one, and m and n are both positive integers;

[0083] A determination module 44 is used to determine the model weights of the n first power prediction models corresponding to the first meteorological source, and integrate the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results;

[0084] The integration module 46 is used to determine the meteorological source weights of the m meteorological sources, and integrate the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0085] Through the above scheme, the meteorological forecast data of m kinds of meteorological sources are first obtained, and the power prediction is performed on the m meteorological forecast data respectively through m groups of power prediction models to obtain m*n power prediction results, wherein each group of power prediction models contains n power prediction models, the m kinds of meteorological sources correspond to the m groups of power prediction models one by one, and m and n are both positive integers; the model weights of the n first power prediction models corresponding to the first meteorological source are determined, and then the n first power prediction results corresponding to the first meteorological source are integrated according to the n model weights to obtain a second power prediction result, wherein the m kinds of meteorological sources include the first meteorological source, and the m groups of power prediction models include the n first power prediction models. The invention discloses a method for integrating the m*n power prediction results of a wind farm into a model, wherein the m*n power prediction results include the n first power prediction results; the meteorological source weights corresponding to the m meteorological sources are determined; and finally the m second power prediction results are integrated according to the m meteorological source weights to obtain a third power prediction result; the above scheme is adopted to provide a model integration method for short-term power prediction of a wind farm, which integrates the prediction results of multiple models, explores and combines the advantages of multiple algorithms, overcomes the instability and system deviation of a single prediction method, and makes the prediction performance more stable and accurate; thereby solving the problem in the related technology that the prediction effect of the model under different scenarios cannot be comprehensively evaluated by a single prediction method, resulting in poor adaptability of the prediction results.

[0086] In an exemplary embodiment, the acquisition module 42 is also used to obtain historical meteorological forecast data of m meteorological sources, and to obtain actual power data corresponding to the m historical meteorological forecast data; to construct m groups of meteorological features based on the m historical meteorological forecast data; to use the m groups of meteorological features as input and the m actual power data as input, to perform power prediction model training using n model training methods respectively, to obtain the m groups of power prediction models.

[0087] In an exemplary embodiment, the above-mentioned determination module 44 is also used to evaluate each first power prediction model according to t evaluation indicators to obtain n groups of evaluation indicator values, wherein each group of evaluation indicator values ​​includes t evaluation indicator values ​​corresponding to the t evaluation indicators, and t is a positive integer; and determine the model weights of the n first power prediction models according to the n groups of evaluation indicator values.

[0088] In an exemplary embodiment, the above-mentioned determination module 44 is also used to perform positive processing on the n groups of evaluation index values ​​to obtain n groups of positively evaluated index values, and construct an index evaluation matrix based on the n groups of positively evaluated index values; perform normalization processing on the index evaluation matrix to obtain a normalized index evaluation matrix; determine the initial model weights of the n first power prediction models based on the normalized index evaluation matrix; and correct the n initial model weights according to the time period accuracy of the n first power prediction models to obtain n model weights.

[0089] In an exemplary embodiment, the above-mentioned determination module 44 is also used to calculate the time period accuracy of the n first power prediction models in s time periods to obtain n*s time period accuracy rates; determine the time period accuracy average value of the n first power prediction models in each time period to obtain s time period accuracy average values; when it is determined that the difference between the time period accuracy of the second power prediction model in the i-th time period and the time period accuracy average value is greater than a preset threshold, the model weights of the n first power prediction models in the i-th time period are corrected according to the difference, wherein the n first power prediction models include the second power prediction model.

[0090] In an exemplary embodiment, the integrated module 46 is further used to determine the wind speed prediction accuracy and wind direction prediction accuracy of the m types of meteorological sources; determine the meteorological prediction accuracy of the m types of meteorological sources according to the wind speed prediction accuracy and the wind direction prediction accuracy; and calculate the meteorological prediction accuracy of the m types of meteorological sources by formula Determine the meteorological source weights of the m meteorological sources, where w 气象，i is the meteorological source weight of the ith meteorological source, acc 气象，i is the weather forecast accuracy of the i-th weather source.

[0091] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0092] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0093] S1, obtaining meteorological forecast data of m meteorological sources, and performing power forecasting on the obtained m meteorological forecast data respectively through m groups of power forecasting models, to obtain m*n power forecasting results, wherein each group of power forecasting models includes n power forecasting models, the m meteorological sources correspond to the m groups of power forecasting models one by one, and m and n are both positive integers;

[0094] S2, determining the model weights of n first power prediction models corresponding to the first meteorological source, and integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results;

[0095] S3, determining the meteorological source weights of the m meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0096] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0097] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0098] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0099] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0100] S1, obtaining meteorological forecast data of m meteorological sources, and performing power forecasting on the obtained m meteorological forecast data respectively through m groups of power forecasting models, to obtain m*n power forecasting results, wherein each group of power forecasting models includes n power forecasting models, the m meteorological sources correspond to the m groups of power forecasting models one by one, and m and n are both positive integers;

[0101] S2, determining the model weights of n first power prediction models corresponding to the first meteorological source, and integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results;

[0102] S3, determining the meteorological source weights of the m meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0103] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0104] An embodiment of the present application further provides a computer program product, comprising a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program product, and when the computer program is executed by a processor, the steps of the method described in each embodiment of the present application are implemented.

[0105] Optionally, in this embodiment, the above computer program may be configured to implement the following steps when executed by a processor:

[0106] S1, obtaining meteorological forecast data of m meteorological sources, and performing power forecasting on the obtained m meteorological forecast data respectively through m groups of power forecasting models, to obtain m*n power forecasting results, wherein each group of power forecasting models includes n power forecasting models, the m meteorological sources correspond to the m groups of power forecasting models one by one, and m and n are both positive integers;

[0107] S2, determining the model weights of n first power prediction models corresponding to the first meteorological source, and integrating the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results;

[0108] S3, determining the meteorological source weights of the m meteorological sources, and integrating the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

[0109] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0110] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0111] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a power prediction result, characterized in that: include: Obtain meteorological forecast data of m meteorological sources, and perform power forecast on the obtained m meteorological forecast data through m groups of power forecast models respectively, to obtain m*n power forecast results, wherein each group of power forecast models includes n power forecast models, the m meteorological sources correspond to the m groups of power forecast models one by one, and m and n are both positive integers; Determine the model weights of n first power prediction models corresponding to a first meteorological source, and integrate the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results; The meteorological source weights of the m meteorological sources are determined, and the m second power prediction results are integrated according to the m meteorological source weights to obtain a third power prediction result.

2. The method according to claim 1, characterized in that: Before obtaining meteorological forecast data of m meteorological sources and performing power forecast on the m meteorological forecast data respectively through m groups of power forecast models to obtain m*n power forecast results, the method further includes: Obtaining historical meteorological forecast data of m meteorological sources, and obtaining actual power data corresponding to the m historical meteorological forecast data; constructing m groups of meteorological characteristics according to the m historical meteorological forecast data; Taking the m groups of meteorological features as input and the m actual power data as input, power prediction model training is performed using n types of model training methods respectively to obtain the m groups of power prediction models.

3. The method according to claim 1, characterized in that Determining model weights of n first power prediction models corresponding to the first meteorological source includes: Each first power prediction model is evaluated according to t evaluation indicators to obtain n groups of evaluation indicator values, wherein each group of evaluation indicator values ​​includes t evaluation indicator values ​​corresponding to the t evaluation indicators, t is a positive integer; The model weights of the n first power prediction models are determined according to the n groups of evaluation index values.

4. The method according to claim 3, characterized in that: Determining model weights of the n first power prediction models according to the n groups of evaluation index values ​​includes: Performing positive processing on the n groups of evaluation index values ​​to obtain n groups of positively evaluated index values, and constructing an index evaluation matrix according to the n groups of positively evaluated index values; Normalizing the indicator evaluation matrix to obtain a normalized indicator evaluation matrix; Determining the initial model weights of the n first power prediction models according to the normalized index evaluation matrix; The n initial model weights are modified according to the time period accuracy of the n first power prediction models to obtain n model weights.

5. The method according to claim 4, characterized in that The n initial model weights are modified according to the time period accuracy of the n first power prediction models to obtain the n model weights, including: Calculate the time period accuracy of the n first power prediction models in s time periods to obtain n*s time period accuracy; Determine the average time period accuracy of the n first power prediction models in each time period, and obtain s average time period accuracy values; When it is determined that the difference between the time period accuracy of the second power prediction model in the i-th time period and the average time period accuracy is greater than a preset threshold, the model weights of the n first power prediction models in the i-th time period are corrected according to the difference, wherein the n first power prediction models include the second power prediction model.

6. The method according to claim 1, characterized in that Determining the meteorological source weights of the m meteorological sources includes: Determining the wind speed prediction accuracy and wind direction prediction accuracy of the m types of meteorological sources; Determine the meteorological forecast accuracy of the m meteorological sources according to the wind speed forecast accuracy and the wind direction forecast accuracy; By formula Determine the meteorological source weights of the m meteorological sources, where w 气象，i is the meteorological source weight of the ith meteorological source, acc 气象，i is the weather forecast accuracy of the i-th weather source.

7. A device for determining a power prediction result, characterized in that: include: An acquisition module is used to acquire meteorological forecast data of m meteorological sources, and perform power forecast on the m meteorological forecast data obtained by using m groups of power forecast models to obtain m*n power forecast results, wherein each group of power forecast models includes n power forecast models, and the m meteorological sources correspond to the m groups of power forecast models one by one, and m and n are both positive integers; A determination module, used to determine the model weights of n first power prediction models corresponding to a first meteorological source, and integrate the n first power prediction results corresponding to the first meteorological source according to the n model weights to obtain a second power prediction result, wherein the m types of meteorological sources include the first meteorological source, the m groups of power prediction models include the n first power prediction models, and the m*n power prediction results include the n first power prediction results; An integration module is used to determine the meteorological source weights of the m meteorological sources, and integrate the m second power prediction results according to the m meteorological source weights to obtain a third power prediction result.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6 when executed.

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 6 through the computer program.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images