New energy power prediction error decoupling analysis method

Abstract

The invention relates to the technical field of power prediction, and particularly provides a new energy power prediction error decoupling analysis method, which comprises the steps of determining a first real prediction error according to target weather forecast data and target historical moment power generation power data; obtaining a first prediction error according to the target meteorological measured data, the target historical time period generated power data, the target irradiance data, the target meteorological forecast data, the target historical moment generated power data and the first real prediction error; obtaining a second prediction error according to the target weather forecast data, the target historical moment power generation power data and the first real prediction error; obtaining a third prediction error according to the target weather forecast data, the target historical moment power generation power data and the first real prediction error; and determining an error decoupling analysis result of power prediction according to the first prediction error, the second prediction error and the third prediction error. According to the invention, key influence factors in the prediction process can be accurately identified.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power prediction, in particular to a new energy power prediction error decoupling analysis method. BACKGROUND

[0002] Solar energy and wind energy, as the most representative renewable energy, have been widely developed and utilized due to their wide distribution, huge reserves, and clean and green advantages. Unlike conventional power sources such as thermal power and nuclear power, which have continuous and adjustable and controllable output, wind power and photovoltaic power are affected by meteorological factors and have strong randomness, volatility and intermittency. The large-scale access of wind power and photovoltaic power to the power grid makes it difficult to balance power supply and demand, and makes it difficult for the power system to operate safely and stably. Therefore, it is necessary to predict the power of new energy in order to optimize the dispatching of the power system according to the prediction results.

[0003] However, new energy power prediction will be affected by many factors and will produce errors. At present, the theoretical research on new energy power prediction mainly focuses on the research of prediction methods and prediction indexes, and the research on the mechanism of power prediction deviation is relatively lacking. The existing decoupling quantitative analysis of power prediction error is only aimed at a single factor or a certain link factor, and lacks full-link decoupling quantitative analysis of error. It is not clear which endogenous factors, such as data input, model construction and correction strategy, cause the external phenomenon of power prediction error, it is difficult to trace the source of power prediction error to guide the optimization of power prediction model, it is difficult to identify the key influencing factors in the prediction process, and the optimization and improvement effect of prediction accuracy is limited. SUMMARY

[0004] The new energy power prediction error decoupling analysis method provided by the embodiments of the present application at least solves the problem that conventional new energy power prediction lacks decoupling quantitative analysis of error and the optimization and improvement effect of prediction accuracy is limited, and can accurately identify the key influencing factors in the prediction process in order to optimize and improve the prediction accuracy of power in a targeted manner.

[0005] The application provides a new energy power prediction error decoupling analysis method, comprising the steps of: determining a first real prediction error according to target meteorological forecast data and target historical time instant power generation data; calculating the error of a power prediction data input link to obtain a first prediction error according to target meteorological measured data, target historical time period power generation data, target irradiance data, the target meteorological forecast data, the target historical time instant power generation data, and the first real prediction error; calculating the error of a power prediction model construction link to obtain a second prediction error according to the target meteorological forecast data, the target historical time instant power generation data, and the first real prediction error; calculating the error of a power prediction correction strategy link to obtain a third prediction error according to the target meteorological forecast data, the target historical time instant power generation data, and the first real prediction error; and determining an error decoupling analysis result of power prediction according to the first prediction error, the second prediction error, and the third prediction error.

[0006] In an embodiment of the application, the first real prediction error is determined according to the target meteorological forecast data and the target historical time instant power generation data, comprising the steps of: inputting the target meteorological forecast data and the target historical time instant power generation data into a meteorological power conversion model to determine a first original prediction power; and subtracting the first original prediction power from the measured power corresponding to the first original prediction power and taking an absolute value to obtain the first real prediction error.

[0007] In an embodiment of the application, the error of the power prediction data input link is calculated to obtain the first prediction error according to the target meteorological measured data, the target historical time period power generation data, the target irradiance data, the target meteorological forecast data, the target historical time instant power generation data, and the first real prediction error, comprising the steps of: determining a first prediction sub-error according to the target meteorological measured data, the target historical time instant power generation data, and the first real prediction error; the first prediction sub-error being a power prediction error caused by a meteorological prediction error in the power prediction data input link; determining a second prediction sub-error according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error; the second prediction sub-error being a power prediction error caused by a data time sequence length error in the power prediction data input link; determining a third prediction sub-error according to the target irradiance data, the target historical time instant power generation data, and the target meteorological forecast data; the third prediction sub-error being a power prediction error caused by a feature vector dimension error in the power prediction data input link; and summing the first prediction sub-error, the second prediction sub-error, and the third prediction sub-error to obtain the first prediction error.

[0008] In one embodiment of the present application, the first prediction sub-error is determined according to the target meteorological observation data, the target historical time instant power generation data, and the first real prediction error, including the steps of: inputting the target meteorological observation data and the target historical time instant power generation data into a meteorological power conversion model to determine a first theoretical prediction power; subtracting the first theoretical prediction power from the measured power corresponding to the first theoretical prediction power and taking an absolute value to obtain a first theoretical prediction error; and subtracting the first theoretical prediction error from the first real prediction error and taking an absolute value to obtain the first prediction sub-error.

[0009] In one embodiment of the present application, the second prediction sub-error is determined according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error, including the steps of: inputting the target meteorological forecast data and the target historical time period power generation data into a meteorological power conversion model to determine a second theoretical prediction power; the second theoretical prediction power is a theoretical prediction power under an optimal time scale; subtracting the second theoretical prediction power from the measured power corresponding to the second theoretical prediction power and taking an absolute value to obtain a second theoretical prediction error; and subtracting the first real prediction error from the second theoretical prediction error and taking an absolute value to obtain the second prediction sub-error.

[0010] In one embodiment of the present application, the third prediction sub-error is determined according to the target irradiance data, the target historical time instant power generation data, and the target meteorological forecast data, including the steps of: inputting the target irradiance data and the target historical time instant power generation data into a meteorological power conversion model to determine a first original prediction power; subtracting the first original prediction power from the measured power corresponding to the first original prediction power and taking an absolute value to obtain a first real prediction error; determining an optimal feature vector combination in the feature combination data according to the feature combination data and the target historical time instant power generation data; the feature combination data is the target meteorological forecast data of different feature vector combinations; inputting the optimal feature vector combination and the target historical time instant power generation data into a meteorological power conversion model to determine a third theoretical prediction power; subtracting the third theoretical prediction power from the measured power corresponding to the third theoretical prediction power and taking an absolute value to obtain a third theoretical prediction error; and subtracting the first real prediction error from the third theoretical prediction error and taking an absolute value to obtain the third prediction sub-error.

[0011] In one embodiment of the present application, the error of the power prediction model construction link is calculated according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, to obtain a second prediction error, including the steps of: determining a fourth prediction sub-error according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error; the fourth prediction sub-error is the power prediction error caused by the optimal hyperparameter error in the power prediction model construction link; determining a fifth prediction sub-error according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error; the fifth prediction sub-error is the power prediction error caused by the optimal modeling method in the power prediction data input link; summing the fourth prediction sub-error and the fifth prediction sub-error to obtain the second prediction error.

[0012] In one embodiment of the present application, the fourth prediction sub-error is determined according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, including the steps of: determining the optimal parameter of the meteorological power conversion model and the fourth theoretical prediction power corresponding to the optimal parameter according to the target meteorological forecast data and the target historical time instant power generation data; subtracting the fourth theoretical prediction power from the measured power corresponding to the fourth theoretical prediction power and taking the absolute value to obtain the fourth theoretical prediction error; subtracting the first real prediction error from the fourth theoretical prediction error and taking the absolute value to obtain the fourth prediction sub-error.

[0013] In one embodiment of the present application, the fifth prediction sub-error is determined according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, including the steps of: determining the fifth theoretical prediction power corresponding to the meteorological power conversion model of different modeling methods according to the target meteorological forecast data and the target historical time instant power generation data; subtracting each fifth theoretical prediction power from the corresponding measured power and taking the absolute value to obtain the corresponding fifth theoretical prediction error; the sixth theoretical prediction error is the minimum value in the fifth theoretical prediction error; subtracting the first real prediction error from the sixth theoretical prediction error and taking the absolute value to obtain the fifth prediction sub-error.

[0014] In one embodiment of the present application, the error of the power prediction correction strategy link is calculated according to the target meteorological forecast data, the target historical time power generation data and the first real prediction error, to obtain a third prediction error, comprising the steps of: correcting the first original predicted power according to different correction strategies to obtain a corresponding sixth theoretical predicted power; the first original predicted power is obtained by inputting the target meteorological forecast data and the target historical time power generation data into a meteorological power conversion model; the sixth theoretical predicted power is subtracted from the measured power corresponding to the sixth theoretical predicted power and the absolute value is taken to obtain a seventh theoretical predicted error; the eighth theoretical predicted error is the minimum value in the seventh theoretical predicted error; the first real prediction error and the eighth theoretical predicted error are subtracted and the absolute value is taken to obtain the third prediction error.

[0015] The above technical solutions of the present application have the following beneficial effects compared with the prior art:

[0016] The new energy power prediction error decoupling analysis method disclosed by the present application compares the first prediction error of the power prediction data input link, the second prediction error of the model construction link and the third prediction error of the correction strategy link, so as to calculate the error decoupling analysis result according to the three errors. In this way, the key influencing factors of the link power prediction error can be determined to guide the improvement of the subsequent power prediction accuracy. On this basis, accurate power prediction can facilitate the optimization of the power system, ensure the safe and stable operation of the power system, significantly improve the accommodation capacity of new energy, reduce the balance pressure of the power grid, and enhance the adaptability and operation resilience of the power system to high-proportion new energy access. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other embodiments according to these drawings without creating any inventive labor. In the drawings:

[0018] Figure 1 is a flowchart of the new energy power prediction error decoupling analysis method in the preferred embodiment of the present application.

[0019] Figure 2 is a power prediction error broken line diagram based on different meteorological forecast data in the preferred embodiment of the present application.

[0020] Figure 3 is a power prediction error broken line diagram based on different input data time series length in the preferred embodiment of the present application.

[0021] Figure 4 is a power prediction error broken line diagram based on different input variable feature dimensions in a preferred embodiment of the present application.

[0022] Figure 5 is a power prediction error broken line diagram based on different hyperparameters in a preferred embodiment of the present application.

[0023] Figure 6 is a power prediction error broken line diagram based on different modeling methods in a preferred embodiment of the present application.

[0024] Figure 7 is a power prediction error broken line diagram based on different correction strategies in a preferred embodiment of the present application.

[0025] Figure 8 is a structure diagram of a new energy power prediction error decoupling analysis system in a preferred embodiment of the present application.

[0026] Figure 9 is a structure diagram of an electronic device in a preferred embodiment of the present application.

[0027] Among them, the above-mentioned drawings include the following reference signs: 11, first confirmation module; 12, second confirmation module; 13, third confirmation module; 14, fourth confirmation module; 15, fifth confirmation module; 201, calculation unit; 202, ROM; 203, RAM; 204, bus; 205, I / O interface; 206, input unit; 207, output unit; 208, storage unit; 209, communication unit. DETAILED DESCRIPTION

[0028] Embodiments of the present application will be described in more detail by referring to the accompanying drawings. Although certain embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms, and should not be interpreted as being limited to the embodiments set forth herein, but rather these embodiments are provided to more thoroughly and completely understand the present application. It should be understood that the drawings and embodiments of the present application are for exemplary purposes only, and are not intended to limit the scope of protection of the present application.

[0029] Referring to Figure 1 , the embodiment of the present application provides a new energy power prediction error decoupling analysis method. The new energy power prediction error decoupling analysis method comprises the steps of:

[0030] According to the target meteorological forecast data and the target historical time power generation data, a first true prediction error is determined.

[0031] According to the target meteorological measured data, the target historical time period power generation data, the target irradiance data, the target meteorological forecast data, the target historical moment power generation data and the first real prediction error, the error of a power prediction data input link is calculated to obtain a first prediction error.

[0032] According to the target meteorological forecast data, the target historical moment power generation data and the first real prediction error, the error of a power prediction model construction link is calculated to obtain a second prediction error.

[0033] According to the target meteorological forecast data, the target historical moment power generation data and the first real prediction error, the error of a power prediction correction strategy link is calculated to obtain a third prediction error.

[0034] According to the first prediction error, the second prediction error and the third prediction error, the error decoupling analysis result of the power prediction is determined.

[0035] In the embodiment of the present application, the photovoltaic power generation ultra-short-term power step-by-step prediction is mainly aimed at. For example, meteorological forecast data is first input into an irradiance prediction model to obtain irradiance prediction data. Then, combined with the irradiance prediction data, the historical power generation data, the meteorological measured data, the measured power data and other meteorological forecast data except for irradiance, a preliminary prediction result is obtained by using a meteorological power conversion model, and the result is corrected to obtain the final prediction power. The meteorological power conversion model belongs to the prior art, and its specific working principle will not be described here.

[0036] On this basis, the embodiment of the present application mainly analyzes the error sources of the photovoltaic power generation ultra-short-term power step-by-step prediction from three links of data input, model construction and correction strategy.

[0037] Specifically, it includes: in the data input link, the power prediction error caused by the accuracy of the meteorological forecast data, the time sequence length of the input data and the characteristic dimension of the input variable; in the model construction link, the power prediction error caused by the hyperparameter setting of the model and the modeling method of the model; in the correction strategy link, the power prediction error caused by the selection of the prediction result error correction strategy.

[0038] Of course, in addition to the errors of the above links, there are also power prediction errors caused by other factors. For example, factors such as equipment failure and grid dispatching instructions, which are not in the prediction link, will cause power prediction errors. However, these factors usually belong to uncontrollable external interference, or can be monitored and processed independently, so they are not included in the decoupling framework. Specifically, the relationship between the error of each link and the total error satisfies the following formula:

[0039] .

[0040] In the formula, is the total error, is the error of the data input link, is the error of the model construction link, is the error of the correction strategy link, is other error, is the predicted power, is the corresponding measured power.

[0041] Among them, the person skilled in the art can select appropriate target meteorological forecast data, target historical time instant power generation data, target meteorological measured data, target historical time period power generation data, and target irradiance data according to actual needs to carry out error decoupling analysis.

[0042] The target meteorological forecast data mainly includes wind speed, wind direction, temperature, humidity, etc., which are prediction values; while the target meteorological measured data is the measured value corresponding to the target meteorological forecast data. The target historical time instant power generation data is the power data at a certain time (i.e. time point), while the target historical time period power generation data is the average value of power in a certain time period. The above data acquisition methods all belong to the prior art and will not be described in detail.

[0043] Among them, the power prediction of new energy is highly dependent on meteorological conditions, and systematic deviation or random error of meteorological forecast data will be directly transmitted to the power prediction result. Decoupling the error of meteorological forecast data can evaluate the reliability of external data sources and provide a basis for subsequent data assimilation or correction.

[0044] The time sequence length of input data will affect the ability of the meteorological power conversion model to capture historical regularities. Too short time sequence may lead to underfitting, and too long time sequence may introduce noise or redundancy.

[0045] The feature dimension of input variables will also affect the prediction accuracy. Insufficient feature dimension may ignore key meteorological elements, leading to information loss, while too high dimension may cause "dimension disaster", increasing the complexity of the model. Through error contribution analysis, dimension reduction or feature enhancement strategies in feature engineering can be guided.

[0046] The hyperparameters of the model directly affect the fitting ability of the model. Improper setting may lead to overfitting or underfitting, significantly affecting the prediction accuracy. Decoupling the error of hyperparameters can provide a target function for automatic parameter tuning, reducing the cost of manual parameter tuning.

[0047] Different modeling methods have different data processing methods. By comparing the error contributions of different modeling methods, model selection can be supported to further improve the prediction accuracy.

[0048] The applicability of the correction strategy is affected by the error distribution characteristics. Quantifying the error reduction effect of the correction strategy can customize the optimal correction scheme for different scenarios.

[0049] After obtaining the first prediction error, the second prediction error and the third prediction error, the error decoupling analysis result can be obtained by using the errors. For example, the error decoupling analysis result can be a percentage value of one of the errors in the sum of the three errors, such as a percentage value of the first prediction error in the sum of the three errors, a percentage value of the second prediction error in the sum of the three errors, or a percentage value of the third prediction error in the sum of the three errors; or the largest error, for example, if the first prediction error is the largest, the error decoupling analysis result is the first prediction error.

[0050] In the prior art, the analysis and introduction of the prediction method and the prediction index are mainly focused on, and the error sources of each link of the new energy power prediction are not analyzed in detail. The new energy power prediction error decoupling analysis method disclosed in the present application discusses the influence of different influencing factors under different links on the total error, so as to guide the new energy station to trace the deviation of the power prediction result from the practical level, and to formulate the corresponding prediction accuracy improvement strategy.

[0051] In summary, the new energy power prediction error decoupling analysis method disclosed in the present application compares the first prediction error of the power prediction data input link, the second prediction error of the model construction link and the third prediction error of the correction strategy link, so as to calculate the error decoupling analysis result according to the three errors. In this way, the key influencing factors of the link power prediction error can be determined, so as to guide the improvement of the subsequent power prediction accuracy. On this basis, accurate power prediction can facilitate the optimization of the power system, and ensure the safe and stable operation of the power system. At the same time, the consumption capacity of new energy is significantly improved, the balance pressure of the power grid is reduced, and the adaptability and operation resilience of the power system to high proportion of new energy access are enhanced.

[0052] In some embodiments, the new energy power prediction error decoupling analysis method disclosed in the present application comprises the following steps:

[0053] Inputting the target meteorological forecast data and the target historical time power generation data into a meteorological power conversion model to determine a first original prediction power .

[0054] Subtracting the first original prediction power from the measured power corresponding to the first original prediction power and taking an absolute value to obtain a first true prediction error . It is expressed as:

[0055]

[0056] Preferably, in the embodiments of the present application, the first original predicted power is calculated by the weather power conversion model according to the future four-hour weather forecast data and the historical 96-time power measured data.

[0057] The new energy power prediction error decoupling analysis method provided by the present application, in some embodiments, calculates the error of the power prediction data input link according to the target weather measured data, the target historical time period power data, the target irradiance data, the target weather forecast data, the target historical time power data and the first true prediction error, to obtain the first prediction error, including the steps of:

[0058] In the first aspect, the first prediction sub-error is determined according to the target weather measured data, the target historical time power data and the first true prediction error . The first prediction sub-error is the power prediction error caused by the weather forecast error in the power prediction data input link.

[0059] In the second aspect, the second prediction sub-error is determined according to the target weather forecast data, the target historical time period power data and the first true prediction error . The second prediction sub-error is the power prediction error caused by the data time sequence length error in the power prediction data input link.

[0060] In the third aspect, the third prediction sub-error is determined according to the target irradiance data, the target historical time power data and the target weather forecast data . The third prediction sub-error is the power prediction error caused by the feature vector dimension error in the power prediction data input link.

[0061] Finally, the first prediction sub-error , the second prediction sub-error and the third prediction sub-error are summed to obtain the first prediction error . It is expressed as:

[0062] .

[0063] This way can accurately identify the contribution degree of factors such as weather forecast accuracy, historical data coverage length and feature construction dimension in the data input process to the prediction error, thereby providing clear guidance for data quality control and feature engineering optimization, effectively improving the input data reliability of the power prediction model, and laying a solid foundation for subsequent prediction accuracy improvement.

[0064] The new energy power prediction error decoupling analysis method, in some embodiments, according to the target meteorological measured data, the target historical time generating power data and the first real prediction error, determines the first prediction sub-error, including steps of:

[0065] Firstly, the target meteorological measured data and the target historical time generating power data are input into a meteorological power conversion model to determine the first theoretical prediction power .

[0066] Secondly, the first theoretical prediction power is subtracted from the measured power corresponding to the first theoretical prediction power and the absolute value is taken to obtain the first theoretical prediction error . It is expressed as:

[0067] .

[0068] Finally, the first real prediction error is subtracted from the first theoretical prediction error and the absolute value is taken to obtain the first prediction sub-error . It is expressed as:

[0069] .

[0070] Preferably, based on the idea of control variable, in the embodiments of the present application, the first theoretical prediction power is calculated through the same meteorological power conversion model as the first original prediction power . Specifically, the first original prediction power is input into the corresponding meteorological power conversion model together with the meteorological measured data at the corresponding time and the same historical generating power measured data at multiple times to obtain the first theoretical prediction power .

[0071] By using the same meteorological power conversion model to calculate the first theoretical prediction power and the first original prediction power, the influence of meteorological prediction data error on power prediction is effectively isolated, thereby ensuring the first prediction sub-error Purely reflect the error contribution of meteorological forecast link. This mode scientifically excludes the interference introduced by model inconsistency, so that the error decomposition result is more reliable. It can also quantitatively determine the specific proportion of meteorological data inaccuracy in the total prediction error, providing a direct basis for the selection and correction of meteorological data sources, helping the operation unit to optimize the data input quality, improve the overall accuracy and stability of new energy power prediction from the source, and enhance the response capability of the power system to renewable energy fluctuations. The new energy power prediction error decoupling analysis method described in the application, in some embodiments, determines a second prediction sub-error according to target meteorological forecast data, target historical time period power generation data, and a first real prediction error, including the steps of:

[0072] First, input the target meteorological forecast data and the target historical time period power generation data into the meteorological power conversion model to determine the second theoretical prediction power . Among them, the second theoretical prediction power is the theoretical prediction power under the optimal time scale.

[0073] Second, the second theoretical prediction power is subtracted from the measured power corresponding to the second theoretical prediction power , and the absolute value is taken to obtain the second theoretical prediction error .

[0074] .

[0075] Finally, the first real prediction error and the second theoretical prediction error are subtracted and the absolute value is taken to obtain the second prediction sub-error . It is expressed as:

[0076] .

[0077] Preferably, based on the idea of control variable, in the embodiments of the application, the second theoretical prediction power and the first original prediction power are calculated by the same meteorological power conversion model. Specifically, the meteorological forecast data used to calculate the first original prediction power , and the measured data of power generation in different historical time periods are input into the corresponding meteorological power conversion model, so that a plurality of theoretical prediction powers can be calculated. Among the plurality of theoretical prediction powers, the theoretical prediction power under the optimal time scale is obtained as the second theoretical prediction power by traversing.

[0078] Based on the idea of control variable, the second theoretical prediction power and the first original prediction power are derived from the same meteorological power conversion model, and the core role is to ensure the comparability of the difference between the power prediction results and the accuracy of the attribution. By fixing the model itself, the interference introduced by the different model structures or parameters is effectively excluded, so that the difference between the two powers is purely attributed to the difference in input data, that is, the time scale effect of different historical measured data and meteorological forecast data. Further, the optimal time scale under the theoretical prediction power is screened in a traversal manner, and the effect is to accurately identify and lock the most effective historical data time window for the current prediction. This process not only improves the prediction accuracy, but also deepens the understanding of the correlation between data timeliness and prediction performance, and provides a clear and reliable basis for model optimization. The new energy power prediction error decoupling analysis method described in the application, in some embodiments, determines a third prediction sub-error according to target irradiance data, target historical time power generation data, and target meteorological forecast data, including the steps of:

[0079] In one aspect, first, the target irradiance data and the target historical time power generation data are input into the meteorological power conversion model to determine the first original prediction power .

[0080] Then, the first original prediction power is subtracted from the measured power corresponding to the first original prediction power and the absolute value is taken to obtain the first true prediction error . It is expressed as:

[0081] .

[0082] In another aspect, first, the optimal feature vector combination in the feature combination data is determined according to the feature combination data and the target historical time power generation data.

[0083] Second, the optimal feature vector combination and the target historical time power generation data are input into the meteorological power conversion model to determine the third theoretical prediction power .

[0084] Subsequently, the third theoretical prediction power is subtracted from the measured power corresponding to the third theoretical prediction power and the absolute value is taken to obtain the third theoretical prediction error . It is expressed as:

[0085] .

[0086] Finally, the first true prediction error and the third theoretical prediction error Difference and absolute value, get the third predictor error . Represented as:

[0087] .

[0088] The feature combination data is the target weather forecast data of different feature vector combinations. For example, the feature combination data is the combination of irradiance and wind speed, the combination of irradiance, wind speed and temperature, etc. Among them, the optimal feature vector combination is determined by traversal.

[0089] Preferably, based on the idea of control variable, in the embodiment of the present application, the second original predicted power , the third theoretical predicted power are calculated by the same meteorological power conversion model. Specifically, the irradiance data at the corresponding moment and the same, historical multiple moment power measured data are input into the corresponding meteorological power conversion model, so that the second original predicted power . Similarly, when calculating the third theoretical predicted power , the feature matrix to be input can be first arranged, and historical power, irradiance, wind speed, wind direction, temperature, humidity, and pressure are selected as candidate features.

[0090] Secondly, the feature matrix and the predicted power are input into the trained random forest regression model for fitting. Among them, the random forest regression model can be trained using default parameters or optimized parameters. For example, the optimized parameters are: tree number 200, maximum depth 10.

[0091] Subsequently, the contribution of each feature to the prediction error is calculated by arranging importance, and the features are sorted from high to low according to the importance value to obtain a feature priority list. The optimal subset is obtained by using the forward feature selection method, including: initializing parameters, setting an empty feature subset; add the first ranked feature to the feature subset, train the LSTM model, input the feature as 1, and record the model performance; add the second ranked feature to the feature subset, train the LSTM model, at this time the input feature becomes 2, compare the performance difference with the last round; constantly repeat the above process to obtain the optimal feature vector combination with the optimal performance.

[0092]

[0093] ​​Therefore, by combining permutation importance evaluation and forward feature selection, the most effective feature combination for improving the prediction accuracy of the LSTM model can be systematically screened, and the balance between dimension reduction and performance optimization can be achieved. Through permutation importance ranking, the key factors that have the greatest impact on errors can be identified from a large number of features, establishing a scientific priority for subsequent screening. On this basis, the forward selection strategy incrementally iterates to gradually include features into the subset and evaluate the performance change of the model. This process can effectively capture the synergistic effect between features, avoid missing important interaction information, and strictly follow the 'performance-driven' principle. The optimal feature subset obtained finally not only significantly reduces the complexity and overfitting risk of the model, but also enables the LSTM model to focus on the most information-rich inputs by eliminating redundant noise features, thereby improving the robustness and computational efficiency of the prediction.

[0094] The new energy power prediction error decoupling analysis method provided by the application, in some embodiments, calculates the error of the power prediction model construction link according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, obtains a second prediction error, including the steps of:

[0095] First, according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error , a fourth prediction sub-error is determined. The fourth prediction sub-error is the power prediction error caused by the optimal hyperparameter error in the power prediction model construction link.

[0096] Second, according to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error , a fifth prediction sub-error is determined. The fifth prediction sub-error is the power prediction error caused by the optimal modeling method in the power prediction data input link.

[0097] Finally, the fourth prediction sub-error and the fifth prediction sub-error are summed to obtain the second prediction error . It is expressed as:

[0098] .

[0099] By quantifying the errors caused by the "optimal hyperparameters" and "optimal modeling methods" respectively, the main link of error generation and its specific contribution can be accurately identified. This not only reveals the sources of uncertainty at different levels of the model, and turns the overall error which is originally vague into a clear structured analysis. Finally, the sum of the two sub-errors is the second prediction error, which means that a more complete and more explanatory error evaluation framework is constructed, providing a clear direction and quantitative basis for subsequent targeted optimization of model structure and adjustment of input data strategy, thereby improving the prediction accuracy.

[0100] The new energy power prediction error decoupling analysis method provided by the application, in some embodiments, determines a fourth prediction sub-error according to target meteorological forecast data, target historical time power generation data, and a first real prediction error, and includes the following steps:

[0101] First, according to the target meteorological forecast data and the target historical time power generation data, the optimal parameters of the meteorological power conversion model are determined, and the fourth theoretical prediction power corresponding to the optimal parameters is determined. .

[0102] Secondly, the fourth theoretical prediction power is subtracted from the measured power corresponding to the fourth theoretical prediction power , and the absolute value is taken to obtain the fourth theoretical prediction error . It is expressed as:

[0103] .

[0104] Finally, the first real prediction error is subtracted from the fourth theoretical prediction error , and the absolute value is taken to obtain the fourth prediction sub-error . It is expressed as:

[0105] .

[0106] Preferably, based on the idea of control variables, in the embodiments of the application, the fourth theoretical prediction power uses the same input data as the first original prediction power , but the hyperparameters of the meteorological power conversion model are different. The hyperparameters of the meteorological power conversion model used by the fourth theoretical prediction power are the optimal parameters.

[0107] Specifically, before inputting the data, the number of hidden layer units, the number of LSTM layers, the learning rate, the batch size, and the number of training epochs are selected as key hyperparameters, and their ranges are defined. Next, the absolute difference between the predicted power and the actual power is used as the objective function, and a Gaussian process is used to model the relationship between the hyperparameters and the objective function. Then, the acquisition function is set and the sampling points are initialized, and iteration begins. In each iteration, the next set of hyperparameters is selected based on the surrogate model and the acquisition function, the model is trained, the error is calculated, and the surrogate model is updated. Finally, the combination of hyperparameters corresponding to the minimum error is extracted from the optimizer as the optimal parameters.

[0108] In this way, the optimal hyperparameter combination of the LSTM model can be found efficiently and automatically, significantly improving the accuracy of power prediction. By modeling the hyperparameter tuning process as a black-box optimization problem and using a Gaussian process as a surrogate model, the effect is to intelligently simulate the complex nonlinear relationship between hyperparameters and prediction error. Compared with traditional grid search or random search, Bayesian optimization does not blindly try, but actively recommends the "most promising" new parameter combination based on feedback information from existing sampling points through a sampling function. This "belief-based" iterative strategy can quickly approach the global optimum with as few trials as possible, greatly saving computational resources and time costs. The final hyperparameter combination ensures that the structure and training process of the LSTM model are adjusted to the optimal state, thereby fully realizing its prediction potential and improving the model's accuracy and stability.

[0109] The new energy power prediction error decoupling analysis method of the present invention, in some embodiments, determines a fifth prediction sub-error based on target weather forecast data, target historical power generation data, and a first true prediction error, including the following steps:

[0110] First, based on the target weather forecast data and the target historical power generation data, determine the fifth theoretical predicted power corresponding to the meteorological power conversion model with different modeling methods. For example, meteorological power conversion models can be modeled in the following ways: single-input single-output, multiple-input multiple-output, and multiple-input single-output.

[0111] Secondly, the predicted power of each fifth theory will be... Compare with the corresponding measured power By subtracting the values ​​and taking the absolute value, we obtain the corresponding fifth theoretical prediction error. 。 is represented as:

[0112] .

[0113] Among them, the sixth theoretical prediction error The fifth theoretical prediction error The minimum value in.

[0114] Finally, the first true prediction error And the sixth theory prediction error The difference is taken and the absolute value is used to obtain the error of the fifth predictor. 。 is represented as:

[0115] .

[0116] Preferably, based on the idea of ​​controlling variables, in this embodiment of the invention, the fifth theoretical prediction power... With the first original predicted power The same input data was used.

[0117] This invention effectively solves the problem of accuracy loss in power prediction caused by improper model structure selection through the error decoupling analysis method in the model construction stage. By systematically comparing the prediction performance of meteorological power conversion models under different modeling methods such as single-input single-output, multi-input multi-output, and multi-input single-output, and selecting the minimum theoretical prediction error corresponding to the optimal modeling method as the benchmark, the specific impact of model structure factors on the total prediction error can be accurately isolated. This process makes the fifth prediction sub-error purely reflect the error component caused by improper model construction, thus providing a clear quantitative basis for selecting the optimal prediction model most suitable for specific station characteristics and data conditions. In this way, the scientificity and pertinence of power prediction model construction are significantly enhanced, which helps to fundamentally improve prediction accuracy from the model level and provides key technical support for accurate power prediction of new energy power plants and reliable dispatch of power systems. In some embodiments, the new energy power prediction error decoupling analysis method of this invention calculates the error of the power prediction correction strategy stage based on the target meteorological forecast data, the target historical power generation data, and the first true prediction error, to obtain the third prediction error, including the following steps:

[0118] First, the first original predicted power is adjusted according to different correction strategies. After making corrections, the corresponding sixth theoretical predicted power is obtained. The first raw predicted power is obtained by inputting the target weather forecast data and the target historical power generation data into the weather power conversion model.

[0119] Secondly, the sixth theory predicts the power. , and the corresponding sixth theoretical predicted power Measured power The difference is taken and the absolute value is used to obtain the seventh theoretical prediction error. 。 is represented as:

[0120] .

[0121] Among them, the eighth theoretical prediction error The seventh theory prediction error The minimum value in.

[0122] Finally, the first true prediction error And the prediction error of the eighth theory The third prediction error is obtained by subtracting the two values ​​and taking their absolute values. 。 is represented as:

[0123] .

[0124] For example, an evaluation matrix for correction strategies is established, which includes three correction strategies: Kalman filtering, moving average, and machine learning correction.

[0125] The first original predicted power was corrected using a Kalman filter strategy. During the correction process, firstly, a state-space model is established, with power as the state variable and the observation equation representing LSTM predictions. Secondly, the covariance matrix and noise parameters are initialized. Finally, the Kalman gain is recursively updated, the predictions are corrected, and the corrected error is calculated.

[0126] The first original predicted power was adjusted using a moving average correction strategy. The correction process involves several steps. First, determining the sliding window size. Second, using the original values ​​as a starting point, a moving average is applied to the original predicted sequence. Finally, the smoothed sequence is obtained, and the corrected error is calculated.

[0127] Using machine learning correction strategies to improve the first original predicted power When making corrections, firstly, a secondary model (such as XGBoost or Random Forest) is trained, with historical errors, meteorological data, and temporal features as inputs. Secondly, the prediction error correction amount is obtained. Finally, the original prediction power is summed with the prediction error correction amount to obtain the corrected prediction value, and the corrected error is calculated.

[0128] Thus, the strategy of using machine learning models for error correction establishes a mapping relationship between errors and various influencing factors, enabling precise compensation for systematic biases in the original forecast, thereby improving the final forecast accuracy. This allows the correction target to focus directly on the "error" of the original forecast itself, rather than attempting to reconstruct the entire power sequence. Training a secondary machine learning model to predict errors is equivalent to having the model specifically learn and capture the inherent bias patterns and regularities of the original forecast model. By incorporating multi-dimensional information such as historical errors, meteorological data, and temporal characteristics, the correction model can gain insight into the complex conditions that cause biases, such as systematic overestimation or underestimation under specific weather patterns, diurnal variations, or seasonal changes.

[0129] On the other hand, embodiments of the present invention also provide a new energy power prediction error decoupling analysis system, applied to the new energy power prediction error decoupling analysis method described in any of the above embodiments. The new energy power prediction error decoupling analysis system includes a first confirmation module 11, a second confirmation module 12, a third confirmation module 13, a fourth confirmation module 14, and a fifth confirmation module 15.

[0130] The first confirmation module 11 is used to determine the first true prediction error based on the target meteorological forecast data and the target historical power generation data.

[0131] The second determining module 12 is used to calculate the error of the power prediction data input link based on the target meteorological measured data, the target historical power generation data for a certain period of time, the target irradiance data, the target meteorological forecast data, the target historical power generation data at a certain time, and the first true prediction error, so as to obtain the first prediction error.

[0132] The third confirmation module 13 is used to calculate the error in the power prediction model construction process based on the target meteorological forecast data, the target historical power generation data, and the first true prediction error, to obtain the second prediction error.

[0133] The fourth confirmation module 14 is used to calculate the error of the power prediction correction strategy based on the target meteorological forecast data, the target historical power generation data and the first true prediction error, and obtain the third prediction error.

[0134] The fifth confirmation module 15 is used to determine the error decoupling analysis results of power prediction based on the first prediction error, the second prediction error and the third prediction error.

[0135] This invention also provides a non-transitory machine-readable medium storing a computer program. When executed by a computer's processor, the computer program causes the computer to perform the new energy power prediction error decoupling analysis method described in any of the above embodiments.

[0136] This invention also provides a computer program product, including a computer program. When executed by a computer's processor, the computer program causes the computer to perform the new energy power prediction error decoupling analysis method described in any of the above embodiments.

[0137] This invention also provides an electronic device, including at least one processor and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the new energy power prediction error decoupling analysis method described in any of the above embodiments.

[0138] Reference Figure 9 The diagram illustrates a structural block diagram of an electronic device that can serve as an embodiment of the present invention, representing an example of a hardware device applicable to various aspects of the invention. The term "electronic device" is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0139] Reference Figure 9 As shown, the electronic device includes a computing unit 201, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 202 or a computer program loaded from a storage unit 208 into a random access memory (RAM) 203. The RAM 203 may also store various programs and data required for the operation of the electronic device. The computing unit 201, ROM 202, and RAM 203 are interconnected via a bus 204. An input / output (I / O) interface 205 is also connected to the bus 204.

[0140] Multiple components in the electronic device are connected to I / O interface 205, including: input unit 206, output unit 207, storage unit 208, and communication unit 209. Input unit 206 can be any type of device capable of inputting information into the electronic device. Input unit 206 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 207 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 208 may include, but is not limited to, disks and optical discs. Communication unit 209 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, and / or wireless communication transceivers, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0141] The computing unit 201 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 201 include, but are not limited to, CPUs, graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 201 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention can be implemented as a computer program tangibly contained in a machine-readable medium, such as storage unit 208. In some embodiments, part or all of the computer program can be loaded and / or installed on an electronic device via ROM 202 and / or communication unit 209. In some embodiments, the computing unit 201 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0142] Computer programs for implementing the methods of embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0143] In the context of embodiments of the present invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, or infrared systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0144] It should be noted that the term "comprising" and its variations used in the embodiments of the present invention are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "a plurality" mentioned in the embodiments of the present invention are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0145] The steps described in the method embodiments provided by the present invention can be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0146] The term "embodiment" in this specification refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply independence or alternativeity from other embodiments. The various embodiments in this specification are described in a related manner, with reference to each other for similar or identical parts. In particular, for apparatus, device, and system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and relevant details are referred to in the description of the method embodiments.

[0147] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A new energy power prediction error decoupling analysis method, characterized in that, The method comprises the steps of: determining a first real prediction error according to target meteorological forecast data and target historical time period power generation data; calculating the error of the power prediction data input link according to the target meteorological measured data, the target historical time period power generation data, the target irradiance data, the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error, to obtain a first prediction error; calculating the error of the power prediction model construction link according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error, to obtain a second prediction error; calculating the error of the power prediction correction strategy link according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error, to obtain a third prediction error; determining the error decoupling analysis result of the power prediction according to the first prediction error, the second prediction error, and the third prediction error.

2. The new energy power prediction error decoupling analysis method according to claim 1, characterized in that, The method comprises the steps of: inputting the target meteorological forecast data and the target historical time period power generation data into a meteorological power conversion model to determine a first original predicted power; subtracting the first original predicted power from the measured power corresponding to the first original predicted power and taking the absolute value to obtain the first real prediction error.

3. The new energy power prediction error decoupling analysis method according to claim 1, characterized in that, The method comprises the steps of: determining a first prediction sub-error according to the target meteorological measured data, the target historical time period power generation data, and the first real prediction error; the first prediction sub-error is the power prediction error caused by the meteorological prediction error in the power prediction data input link; determining a second prediction sub-error according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error; the second prediction sub-error is the power prediction error caused by the data time sequence length error in the power prediction data input link; determining a third prediction sub-error according to the target irradiance data, the target historical time period power generation data, and the target meteorological forecast data; the third prediction sub-error is the power prediction error caused by the feature vector dimension error in the power prediction data input link; summing the first prediction sub-error, the second prediction sub-error, and the third prediction sub-error to obtain the first prediction error.

4. The new energy power prediction error decoupling analysis method according to claim 3, characterized in that, The method comprises the steps of: inputting the target meteorological measured data and the target historical time period power generation data into a meteorological power conversion model to determine a first theoretical predicted power; determining a first theoretical prediction error by subtracting the measured power corresponding to the first theoretical prediction power from the first theoretical prediction power and taking an absolute value; determining the first prediction sub-error by subtracting the first theoretical prediction error from the first real prediction error and taking an absolute value.

5. The new energy power prediction error decoupling analysis method according to claim 3, characterized in that, determining a second prediction sub-error according to the target meteorological forecast data, the target historical time period power generation data, and the first real prediction error, including the steps of: inputting the target meteorological forecast data and the target historical time period power generation data into a meteorological power conversion model to determine a second theoretical prediction power; the second theoretical prediction power is a theoretical prediction power under an optimal time scale; determining a second theoretical prediction error by subtracting the measured power corresponding to the second theoretical prediction power from the second theoretical prediction power and taking an absolute value; determining the second prediction sub-error by subtracting the first real prediction error from the second theoretical prediction error and taking an absolute value.

6. The new energy power prediction error decoupling analysis method according to claim 3, characterized in that, determining a third prediction sub-error according to the target irradiance data, the target historical time point power generation data, and the target meteorological forecast data, including the steps of: inputting the target irradiance data and the target historical time point power generation data into a meteorological power conversion model to determine a first original prediction power; determining a first real prediction error by subtracting the measured power corresponding to the first original prediction power from the first original prediction power and taking an absolute value; determining an optimal feature vector combination in the feature combination data according to the feature combination data and the target historical time point power generation data; the feature combination data is the target meteorological forecast data under different feature vector combinations; inputting the optimal feature vector combination and the target historical time point power generation data into a meteorological power conversion model to determine a third theoretical prediction power; determining a third theoretical prediction error by subtracting the measured power corresponding to the third theoretical prediction power from the third theoretical prediction power and taking an absolute value; determining the third prediction sub-error by subtracting the first real prediction error from the third theoretical prediction error and taking an absolute value.

7. The new energy power prediction error decoupling analysis method according to claim 1, characterized in that, calculating the error of a power prediction model construction link according to the target meteorological forecast data, the target historical time point power generation data, and the first real prediction error to obtain a second prediction error, including the steps of: determining a fourth prediction sub-error according to the target meteorological forecast data, the target historical time point power generation data, and the first real prediction error; the fourth prediction sub-error is a power prediction error caused by an optimal hyperparameter error in the power prediction model construction link; determining a fifth prediction sub-error according to the target meteorological forecast data, the target historical time point power generation data, and the first real prediction error; the fifth prediction sub-error is a power prediction error caused by an optimal modeling method in the power prediction data input link; summing the fourth prediction sub-error and the fifth prediction sub-error to obtain the second prediction error.

8. The new energy power prediction error decoupling analysis method according to claim 7, characterized in that, According to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, a fourth prediction sub-error is determined, including the steps of: According to the target meteorological forecast data and the target historical time instant power generation data, an optimal parameter of a meteorological power conversion model is determined, and a fourth theoretical prediction power corresponding to the optimal parameter is determined; The fourth theoretical prediction power is subtracted from a measured power corresponding to the fourth theoretical prediction power, and an absolute value is taken to obtain a fourth theoretical prediction error; The first real prediction error and the fourth theoretical prediction error are subtracted, and an absolute value is taken to obtain the fourth prediction sub-error. 9.The new energy power prediction error decoupling analysis method according to claim 7, characterized in that, According to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, a fifth prediction sub-error is determined, including the steps of: According to the target meteorological forecast data and the target historical time instant power generation data, fifth theoretical prediction powers corresponding to meteorological power conversion models of different modeling modes are determined; Each of the fifth theoretical prediction powers is subtracted from a corresponding measured power, and an absolute value is taken to obtain a corresponding fifth theoretical prediction error; A sixth theoretical prediction error is the minimum value in the fifth theoretical prediction errors; The first real prediction error and the sixth theoretical prediction error are subtracted, and an absolute value is taken to obtain the fifth prediction sub-error.

10. The new energy power prediction error decoupling analysis method according to claim 1, characterized in that, According to the target meteorological forecast data, the target historical time instant power generation data and the first real prediction error, an error of a power prediction correction strategy link is calculated to obtain a third prediction error, including the steps of: According to different correction strategies, a first original prediction power is corrected to obtain a corresponding sixth theoretical prediction power; the first original prediction power is obtained by inputting the target meteorological forecast data and the target historical time instant power generation data into a meteorological power conversion model; The sixth theoretical prediction power is subtracted from a measured power corresponding to the sixth theoretical prediction power, and an absolute value is taken to obtain a seventh theoretical prediction error; An eighth theoretical prediction error is the minimum value in the seventh theoretical prediction errors; The first real prediction error and the eighth theoretical prediction error are subtracted, and an absolute value is taken to obtain the third prediction error.

Images