Multi-time scale wind power prediction method and system

Abstract

The invention discloses a multi-time-scale wind power prediction method and system, and the method comprises the steps: collecting meteorological data, carrying out the seasonal decomposition of the meteorological data, carrying out the seasonal adjustment after the decomposition, constructing a hybrid model through employing a statistical model, a physical model and a machine learning model, carrying out the seasonal time series prediction through employing an SARIMA model, and carrying out the prediction of a multi-time-scale wind power. An SARIMA model and a hybrid model are combined; a prediction model and algorithm parameters can be further optimized; compared with the prior art, the influence of seasonality on prediction can be effectively reduced, and the prediction precision is improved; a hybrid model is adopted, a physical model prediction model, a statistical model prediction model and a machine learning model prediction model; the hybrid model combines the advantages of a physical model, a statistical model and a machine learning model, and the overall prediction precision of the multi-time-scale wind power prediction method and system is improved by integrating prediction results of different models.

Description

Technical Field

[0001] The present invention relates to a multi-time scale wind power prediction method and a system thereof, and in particular to a multi-time scale wind power prediction method and a system thereof. Background Art

[0002] With the increasing attention paid to renewable energy around the world, the development and utilization of wind energy as a clean and renewable energy form has become an important research direction in the energy field. Wind power prediction, as an important basis for wind farm operation and maintenance and grid dispatching, is of great significance for improving wind power utilization and ensuring stable grid operation. However, wind power is affected by many factors, such as meteorological conditions such as wind speed, wind direction, temperature, humidity, and air pressure, as well as factors such as terrain, wind farm layout, and wind turbine performance. These factors show different change patterns and influence degrees on different time scales, making wind power prediction face huge challenges.

[0003] In the field of wind power prediction, existing technologies mainly include prediction methods based on physical models, statistical models and machine learning models.

[0004] Prediction method based on physical model: This method uses physical principles such as computational fluid dynamics (CFD) and combines the meteorological conditions of the wind farm and the performance parameters of the wind turbine to predict wind power; Prediction method based on statistical model: Such as time series analysis methods such as ARIMA and SARIMA, through statistical analysis of historical wind power data, to establish a prediction model; Prediction method based on machine learning model: Such as machine learning algorithms such as LSTM and SVM, through training a large amount of historical data, automatically learn the characteristics and laws in the data to predict wind power.

[0005] The traditional multi-time scale wind power forecasting method uses a relatively single model and has some limitations. At the same time, seasonal meteorological factors will affect the forecast accuracy. Summary of the invention

[0006] The present invention overcomes the deficiencies of the prior art and provides a multi-time scale wind power prediction method and system.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a multi-time scale wind power prediction method and system thereof, comprising: S1, collecting meteorological data, pre-processing the data as data input;

[0008] S2, apply the STL decomposition method to seasonal decomposition of meteorological data;

[0009] S3. According to the results of seasonal decomposition, seasonal adjustment is performed;

[0010] S4, construct hybrid models using statistical models, physical models and machine learning models;

[0011] S5 uses the SARIMA model for seasonal time series forecasting and combines the SARIMA model with a hybrid model;

[0012] S6, using the meteorological data pre-processed in S1 as data input, and selecting different prediction models according to the prediction purpose;

[0013] In a preferred embodiment of the present invention,

[0014] The meteorological data include wind speed data, wind direction data, temperature data, humidity data and air pressure data.

[0015] In a preferred embodiment of the present invention,

[0016] Meteorological data have long time spans.

[0017] In a preferred embodiment of the present invention,

[0018] In step S2, the STL decomposition method is used to perform seasonal decomposition on meteorological data. The specific steps are as follows:

[0019] The seasonal decomposition method STL is used to decompose the wind power time series into trend components, seasonal components and random components; the decomposition process can be expressed as:

[0020] Y t =T t +S t +R t

[0021] Among them, Y t is the wind power observation value at time t, T t is the trend component at time t, S t is the seasonal component at time t, R t is the residual at time t.

[0022] In a preferred embodiment of the present invention,

[0023] In step S3, seasonal adjustment is specifically to smooth or remove seasonal components through seasonal adjustment factors according to the decomposition results to reduce the impact of seasonality on the forecast results; the specific steps for calculating the seasonal adjustment factor are:

[0024] First, define the historical wind power average value F[t] for each season, where t represents the season. These values ​​can be obtained by averaging the wind power data of the same season in the past few years.

[0025] Calculate the seasonal adjustment factor F seasonal[t], which reflects the deviation of the forecast value for the current season from the historical average;

[0026]

[0027] Among them, F[m] is the average of the average wind power in all seasons, that is, the annual average power, which is used to normalize the seasonal average power; α is an adjustment coefficient, which is used to adjust the influence of seasonal factors according to actual conditions;

[0028] Use the calculated seasonal adjustment factor F seasonal [t] to adjust the original seasonal component S t :

[0029] Adjusted_S t =S t ×F seasonal [t],

[0030] Adjusted_S t is the adjusted seasonal component.

[0031] In a preferred embodiment of the present invention,

[0032] In step S4, constructing the hybrid model includes:

[0033] The hybrid model combines the advantages of physical models, statistical models and machine learning models, and improves the multi-time scale wind power prediction method and the overall prediction accuracy of the system by integrating the prediction results of different models. The hybrid model includes; physical model prediction model, statistical model prediction model and machine learning model prediction model; more specifically;

[0034] Physical model prediction model:

[0035] Using the CFD simulation method physical model, the basic wind power forecast value p is calculated based on meteorological data and wind turbine characteristics. phys [n];

[0036] Statistical Model Prediction Model:

[0037] The ARIMA time series analysis method is applied to model the wind power time series and consider seasonal factors to obtain the statistical model prediction value p stat [n];

[0038] Machine Learning Model Prediction Model:

[0039] Using the LSTM machine learning algorithm, combined with a large amount of historical data and meteorological data, the model is trained and the predicted value p is obtained. ml [n].

[0040] Hybrid prediction model:

[0041] The prediction results of the physical model, statistical model and machine learning model are weighted and integrated, while considering the seasonal adjustment factor:

[0042] p hybrld [n] = w1·p phys [n]+w2·(p stat [n]·F seasonal [t])+w3·p ml [n]

[0043] Among them, w1, w2, w3 are the weights of the predicted values ​​of different models.

[0044] In a preferred embodiment of the present invention,

[0045] In step S5, the SARIMA model is used to perform seasonal time series forecasting, including;

[0046] The SARIMA model is used for seasonal time series forecasting. The SARIMA model is a seasonal autoregressive integrated moving average model, which is particularly suitable for time series data with seasonal changes. The specific algorithm is:

[0047] Φ(B S )Φ(B)Y[n]=Θ(B S )Θ(B)Z[n],

[0048] Φ(B) and Θ(B) are the non-seasonal autoregressive and moving average operators, respectively;

[0049] Φ(B S ) and Θ(B S ) are seasonal autoregressive and seasonal moving average operators respectively, S is the seasonal cycle, B is the backshift operator; Z[n] is a white noise sequence.

[0050] Parameter estimates:

[0051] ACF and PACF plots were used for preliminary identification of model parameters.

[0052] Use AIC or BIC criteria to optimize model parameters.

[0053] Prediction formula:

[0054] The model parameters are determined, and the future wind power is predicted by the recursive formula:

[0055]

[0056] in, is the predicted wind power value, φ i ,θ j, Φ I , Θ J are the parameters of the SARIMA model, p, q, P, Q, d, D, S are model parameters;

[0057] More specifically, the model parameters include:

[0058] Non-seasonal part parameters: p, d, q.

[0059] Seasonal part parameters: P, D, Q, S.

[0060] In a preferred embodiment of the present invention,

[0061] In step S5, the SARIMA model and the hybrid model are combined to include:

[0062] The prediction value of the hybrid model and the prediction value of the SARIMA model are weighted and fused to obtain the final comprehensive prediction value:

[0063]

[0064] Among them, β is the weight of the predicted value of the hybrid model, and 1_β is the weight of the predicted value of the SARIMA model.

[0065] In a preferred embodiment of the present invention,

[0066] The forecasting models include: short-term forecasting model, medium-term forecasting model and long-term forecasting model.

[0067] In a preferred embodiment of the present invention,

[0068] Data acquisition module:

[0069] This module is responsible for collecting various data required for wind power forecasting, including but not limited to historical wind speed data, wind direction data, temperature data, air pressure data, terrain data, wind farm operation data, and numerical weather forecast data;

[0070] Data sources include local databases of wind farms, meteorological observation stations, and NWP model outputs;

[0071] Data processing module:

[0072] Preprocess the acquired data, including data cleaning, data conversion and data fusion;

[0073] Prediction Model Module:

[0074] This is the core module of the system, which contains multiple prediction models for wind power prediction at different time scales;

[0075] Prediction models include models based on statistical methods, physical models, machine learning models, and hybrid models;

[0076] Each model will make predictions based on the input data and its own algorithm logic, and output the prediction results;

[0077] Prediction result fusion module:

[0078] For systems that include multiple prediction models, this module is responsible for fusing the prediction results of different models to improve the accuracy of multi-time scale wind power prediction methods and overall system prediction;

[0079] Fusion methods include weighted averaging, Bayesian fusion, machine learning fusion, etc.

[0080] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0081] (1) The present invention collects meteorological data, decomposes the meteorological data seasonally, and performs seasonal adjustments after the decomposition. Then, a hybrid model is constructed using a statistical model, a physical model, and a machine learning model. The SARIMA model is used to predict seasonal time series, and the SARIMA model and the hybrid model are combined. The prediction model and algorithm parameters can be further optimized. Compared with the prior art, the influence of seasonality on prediction can be effectively reduced, and the prediction accuracy can be improved.

[0082] (2) The present invention adopts a hybrid model, which includes: a physical model prediction model, a statistical model prediction model and a machine learning model prediction model; the hybrid model combines the advantages of the physical model, the statistical model and the machine learning model, and improves the multi-time scale wind power prediction method and the overall prediction accuracy of the system by integrating the prediction results of different models.

[0083] (3) The present invention adopts the SARIMA model for seasonal time series prediction and combines the SARIMA model with the hybrid model; compared with the prior art, the accuracy of wind power prediction can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0085] Figure 1 is a flow chart of a multi-time scale wind power prediction method according to a preferred embodiment of the present invention; DETAILED DESCRIPTION

[0086] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0087] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0088] like Figure 1 As shown, a multi-time scale wind power prediction method and system thereof include:

[0089] S1. Collect meteorological data and pre-process the data as data input;

[0090] S2, apply the STL decomposition method to seasonal decomposition of meteorological data;

[0091] S3. Make seasonal adjustments based on the results of seasonal decomposition;

[0092] S4, construct hybrid models using statistical models, physical models and machine learning models;

[0093] S5. Use SARIMA model for seasonal time series forecasting and combine SARIMA model with hybrid model;

[0094] S6, using the meteorological data pre-processed in S1 as data input, and selecting different prediction models according to the prediction purpose;

[0095] Specifically, in step S1, the meteorological data includes wind speed data, wind direction data, temperature data, humidity data and air pressure data;

[0096] Meteorological data have long time spans (such as several years) to capture seasonal changes.

[0097] Specifically, in step S1, data preprocessing includes: data cleaning: removing outliers and missing values, and smoothing the data; and standardizing / normalizing the data: that is, converting data of different dimensions into a unified range to facilitate model processing.

[0098] Specifically, in step S2, the STL decomposition method is used to perform seasonal decomposition on meteorological data. The specific steps are as follows:

[0099] The seasonal decomposition method STL is used to decompose the wind power time series into trend components, seasonal components and random components; the decomposition process can be expressed as:

[0100] Y t =T t +S t +R t

[0101] Among them, Y t is the wind power observation value at time t, T t is the trend component at time t, S t is the seasonal component at time t, R t is the residual (random component) at time t.

[0102] Specifically, in step S3, seasonal adjustment is to smooth or remove seasonal components through seasonal adjustment factors according to the decomposition results to reduce the impact of seasonality on the forecast results; the specific steps for calculating the seasonal adjustment factor are:

[0103] First, define the historical wind power average value F[t] for each season, where t represents the season. These values ​​can be obtained by averaging the wind power data of the same season in the past few years.

[0104] Calculate the seasonal adjustment factor F seasonal [t], which reflects the deviation of the forecast value for the current season from the historical average;

[0105]

[0106] Among them, F[m] is the average of the average wind power in all seasons, that is, the annual average power, which is used to normalize the seasonal average power; α is an adjustment coefficient, which is used to adjust the influence of seasonal factors according to actual conditions;

[0107] Use the calculated seasonal adjustment factor F seasonal [t] to adjust the original seasonal component S t :

[0108] Adjusted_S t =S t ×F seasonal [t],

[0109] Adjusted_S t is the adjusted seasonal component.

[0110] Specifically, in step S4, constructing the hybrid model includes:

[0111] The hybrid model combines the advantages of physical models, statistical models and machine learning models, and improves the multi-time scale wind power prediction method and the overall prediction accuracy of the system by integrating the prediction results of different models. The hybrid model includes; physical model prediction model, statistical model prediction model and machine learning model prediction model; more specifically;

[0112] Physical model prediction model:

[0113] Using the CFD simulation method physical model, the basic wind power forecast value p is calculated based on meteorological data and wind turbine characteristics. phys [n];

[0114] Statistical Model Prediction Model:

[0115] The ARIMA time series analysis method is applied to model the wind power time series and consider seasonal factors to obtain the statistical model prediction value p stat [n];

[0116] Machine Learning Model Prediction Model:

[0117] Using the LSTM machine learning algorithm, combined with a large amount of historical data and meteorological data, the model is trained and the predicted value p is obtained. ml [n].

[0118] Hybrid prediction model:

[0119] The prediction results of the physical model, statistical model and machine learning model are weighted and integrated, while considering the seasonal adjustment factor:

[0120] p hybrld [n] = w1·p phys [n]+w2·(p stat [n]·F_seasonal[t])+w3·p ml [n]

[0121] Among them, w1, w2, w3 are the weights of the predicted values ​​of different models.

[0122] Specifically, in step S5, the SARIMA model is used to perform seasonal time series forecasting, including:

[0123] The SARIMA model is used for seasonal time series forecasting. The SARIMA model is a seasonal autoregressive integrated moving average model, which is particularly suitable for time series data with seasonal changes. The specific algorithm is:

[0124] Φ(B S )Φ(B)Y[n]=Θ(B S )Θ(B)Z[n],

[0125] Φ(B) and Θ(B) are the non-seasonal autoregressive and moving average operators, respectively;

[0126] Φ(B S ) and Θ(B S ) are seasonal autoregressive and seasonal moving average operators respectively, S is the seasonal cycle, B is the backshift operator; Z[n] is a white noise sequence.

[0127] Parameter estimates:

[0128] ACF and PACF plots were used for preliminary identification of model parameters.

[0129] Use AIC or BIC criteria to optimize model parameters.

[0130] Prediction formula:

[0131] The model parameters are determined, and the future wind power is predicted by the recursive formula:

[0132]

[0133] in, is the predicted wind power value, φ i ,θ j , Φ I , Θ J are the parameters of the SARIMA model, p, q, P, Q, d, D, S are model parameters;

[0134] More specifically, the model parameters include:

[0135] Non-seasonal parameters: p (number of non-seasonal autoregressive terms), d (number of non-seasonal difference orders), and q (number of non-seasonal moving average terms).

[0136] Seasonal parameters: P (number of seasonal autoregressive terms), D (order of seasonal differences), Q (number of seasonal moving average terms), and S (seasonal cycle).

[0137] Specifically, in step S5, the SARIMA model and the hybrid model are combined to include:

[0138] The prediction value of the hybrid model and the prediction value of the SARIMA model are weighted and fused to obtain the final comprehensive prediction value:

[0139]

[0140] Among them, β is the weight of the predicted value of the hybrid model, and 1_β is the weight of the predicted value of the SARIMA model.

[0141] Specifically, in step S6, different prediction models are selected according to the prediction purpose;

[0142] The prediction models are divided into short-term prediction models, medium-term prediction models and long-term prediction models; specifically;

[0143] Short-term prediction models mainly use LSTM algorithms in statistical models and machine learning models. These models can capture the rapid changes of meteorological factors such as wind speed and wind direction in a short period of time, as well as instant feedback on the operating status of wind farms;

[0144] For the medium-term forecast model, a hybrid model and a SARIMA model are used;

[0145] The long-term forecast model uses physical models and analysis of macroeconomic and policy factors, while referring to the changing trends of historical wind farm operation data and meteorological data.

[0146] A multi-time scale wind power prediction system, comprising:

[0147] Data acquisition module:

[0148] This module is responsible for collecting various data required for wind power forecasting, including but not limited to historical wind speed data, wind direction data, temperature data, air pressure data, terrain data, wind farm operation data (such as wind turbine status, power generation, etc.) and numerical weather forecast (NWP) data.

[0149] Data sources include wind farm local database, meteorological observation station, NWP model output, etc.

[0150] Data processing module:

[0151] The acquired data is preprocessed, including data cleaning (removing outliers, filling missing values, etc.), data conversion (such as normalization, standardization, etc.) and data fusion (integrating data from different sources together).

[0152] Prediction model module: contains multiple prediction models, which are used for wind power prediction at different time scales.

[0153] Prediction models include models based on statistical methods (such as time series analysis and regression analysis), physical models (such as CFD simulation and power curve method), machine learning models (such as artificial neural networks, support vector machines, random forests), and hybrid models (combining multiple methods).

[0154] Each model will make predictions based on the input data and its own algorithm logic, and output the prediction results.

[0155] Prediction result fusion module:

[0156] For systems that contain multiple prediction models, this module is responsible for fusing the prediction results of different models to improve the accuracy of multi-time scale wind power prediction methods and overall system prediction.

[0157] Fusion methods include weighted averaging, Bayesian fusion, machine learning fusion, etc.

[0158] Prediction output module: Use the constructed model to predict wind power at multiple time scales and output the prediction results.

[0159] The above is based on the ideal embodiment of the present invention. Through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. A multi-time scale wind power prediction method, characterized in that: include: S1. Collect meteorological data and pre-process the data as data input; S2, apply the STL decomposition method to seasonal decomposition of meteorological data; S3. According to the results of seasonal decomposition, seasonal adjustment is performed; S4. Use statistical models, physical models and machine learning models to build hybrid models, and combine the hybrid models with seasonal adjustments; S5. Use SARIMA model for seasonal time series forecasting and combine SARIMA model with hybrid model; S6. Use the meteorological data pre-processed in S1 as data input, and select different prediction models for prediction and output according to the prediction purpose.

2. A multi-time scale wind power prediction method according to claim 1, characterized in that: The meteorological data include wind speed data, wind direction data, temperature data, humidity data and air pressure data.

3. The multi-time scale wind power prediction method according to claim 1, characterized in that: Meteorological data have long time spans.

4. The multi-time scale wind power prediction method according to claim 1, characterized in that: In step S2, the STL decomposition method is used to perform seasonal decomposition on meteorological data. The specific steps are as follows: The seasonal decomposition method STL is used to decompose the wind power time series into trend components, seasonal components and random components; the decomposition process can be expressed as: Y t =T t +S t +R t , Among them, Y t is the wind power observation value at time t, T t is the trend component at time t, S t is the seasonal component at time t, R t is the residual at time t.

5. The multi-time scale wind power prediction method according to claim 1, characterized in that: In step S3, seasonal adjustment is specifically to smooth or remove seasonal components through seasonal adjustment factors according to the decomposition results to reduce the impact of seasonality on the forecast results; the specific steps for calculating the seasonal adjustment factor are: Define the historical average wind power F[t] for each season, where t represents the season. These values ​​can be obtained by averaging the wind power data of the same season in the past few years. Calculate the seasonal adjustment factor F seasonal [t], which reflects the deviation of the forecast value for the current season from the historical average; Among them, F[m] is the average of the average wind power in all seasons, that is, the annual average power, which is used to normalize the seasonal average power; α is an adjustment coefficient, which is used to adjust the influence of seasonal factors according to actual conditions; Use the calculated seasonal adjustment factor F seasonal [t] to adjust the original seasonal component S t : Adjusted_S t =S t ×F seasonal [t], Adjusted_S t is the adjusted seasonal component.

6. A multi-time scale wind power prediction method according to claim 1, characterized in that: In step S4, the method for constructing a hybrid model includes: Determine the physical model of the CFD simulation method and calculate the basic wind power prediction value p according to meteorological data and wind turbine characteristics phys [n]; Determine the statistical model prediction model, apply the ARIMA time series analysis method, consider the seasonal factor, model the wind power time series, and obtain the statistical model prediction value p stat [n]; Determine the machine learning model prediction model, use the LSTM machine learning algorithm, combine a large amount of historical data and meteorological data, train the model and obtain the predicted value p ml [n]; The hybrid forecasting model combines the forecast results of the physical model, statistical model and machine learning model in a weighted manner, while taking into account the seasonal adjustment factor: p hybrld [n]=w1·p phys [n]+w2·(p stat [n]·F seasonal [t])+w3·p ml [n], Among them, w1, w2, w3 are the weights of the predicted values ​​of different models.

7. The multi-time scale wind power prediction method according to claim 1, characterized in that: In step S5, the SARIMA model is used to perform seasonal time series forecasting, including; The SARIMA model is used for seasonal time series forecasting. The SARIMA model is a seasonal autoregressive integrated moving average model, which is particularly suitable for time series data with seasonal changes. The specific algorithm is: Φ(B S )Φ(B)Y[n]=Θ(B S )Θ(B)Z[n], Φ(B) and Θ(B) are the non-seasonal autoregressive and moving average operators, respectively; Φ(B S ) and Θ(B S ) are seasonal autoregressive and seasonal moving average operators, S is the seasonal cycle, B is the backshift operator; Z[n] is a white noise sequence; Parameter estimates: Use ACF and PACF plots for preliminary identification of model parameters; Apply AIC or BIC criteria to optimize model parameter selection; Prediction formula: The model parameters are determined, and the future wind power is predicted by the recursive formula: in, is the predicted wind power value, φ i ,θ j , Φ I , Θ J are the parameters of the SARIMA model, p, q, P, Q, d, D, S are model parameters; More specifically, the model parameters include: Non-seasonal parameters: p, d, q; Seasonal part parameters: P, D, Q, S.

8. The multi-time scale wind power prediction method according to claim 1, characterized in that: In step S5, the SARIMA model and the hybrid model are combined to include: The prediction value of the hybrid model and the prediction value of the SARIMA model are weighted and fused to obtain the final comprehensive prediction value: Among them, β is the weight of the hybrid model prediction value, and 1-β is the weight of the SARIMA model prediction value.

9. The multi-time scale wind power prediction method according to claim 1, characterized in that: The forecasting models include: short-term forecasting model, medium-term forecasting model and long-term forecasting model.

10. A multi-time scale wind power prediction system, based on a multi-time scale wind power prediction system according to any one of claims 1 to 9, characterized in that: include: Data acquisition module: This module is responsible for collecting various data required for wind power forecasting, including but not limited to historical wind speed data, wind direction data, temperature data, air pressure data, terrain data, wind farm operation data, and numerical weather forecast data; Data sources include local databases of wind farms, meteorological observation stations, and NWP model outputs; Data processing module: Preprocess the acquired data, including data cleaning, data conversion and data fusion; Prediction Model Module: This is the core module of the system, which contains multiple prediction models for wind power prediction at different time scales; Prediction models include models based on statistical methods, physical models, machine learning models, and hybrid models; Each model will make predictions based on the input data and its own algorithm logic, and output the prediction results; Prediction result fusion module: For systems that contain multiple prediction models, this module is responsible for fusing the prediction results of different models to improve the accuracy of multi-time scale wind power prediction methods and overall system predictions; fusion methods include weighted average, Bayesian fusion, machine learning fusion, etc.

Images