CN-121996957-A - Medium-long-term wind-light resource prediction method and system considering climate remote correlation factors

CN121996957ACN 121996957 ACN121996957 ACN 121996957ACN-121996957-A

Abstract

The invention belongs to the technical field of electric power meteorology, and provides a medium-long term wind-solar resource prediction method and a system taking climate remote correlation factors into consideration, wherein the method comprises the steps of obtaining global climate index historical data, and obtaining standardized climate indexes after standardized treatment of the data; aiming at a target station historical resource sequence, calculating a mutual information value of a standardized climate index, taking a lag time corresponding to the maximum value of the mutual information as an optimal early warning window period to construct a prediction factor set, extracting wind and light resource actual measurement data in the same period of the historical year to construct a reference probability density function, taking the reference probability density function as prior distribution, and combining a preset power mode prediction result as a likelihood function to output a prediction result containing a deterministic value and a confidence interval. The invention breaks through the prediction limit of the atmospheric mode for 15 days by means of the characteristic of slow change of the ocean signals, successfully realizes the effective trend prediction from the quarter to the annual level, and meets the medium-long term resource evaluation requirement.

Inventors

SONG ZONGPENG
WANG ZHAO
ZHAO YANQING
JIANG WENLING
LIU XIAOLIN
WANG BO
WANG ZHENG
Jin shuanglong
WANG SHU
HUA SHENBING
MA ZHENQIANG
CHE JIANFENG

Assignees

中国电力科学研究院有限公司
国家电网有限公司
国网湖北省电力有限公司电力科学研究院

Dates

Publication Date: 20260508
Application Date: 20260115

Claims (14)

1. The medium-long term wind-light resource prediction method considering the climate remote correlation factor is characterized by comprising the following steps: acquiring global climate index historical data, and carrying out standardization treatment on the data to obtain a standardized climate index; calculating a mutual information value of the standardized climate indexes according to a historical resource sequence of the target station, taking the lag time corresponding to the maximum value of the mutual information as an optimal early warning window period, and constructing a forecasting factor set; Acquiring a current climate state vector based on the prediction factor set, searching a historical year closest to the Euclidean distance of the current state in a historical database, and extracting wind and light resource actual measurement data in the same period as the historical year to construct a reference probability density function; taking the reference probability density function as prior distribution, and combining a preset power mode prediction result as a likelihood function to output a prediction result containing a deterministic value and a confidence interval.
2. The method for predicting mid-long term wind and solar resources taking climate remote correlation factors into consideration according to claim 1, wherein the step of obtaining global climate index historical data and obtaining standardized climate indexes after standardized data processing comprises the following steps: acquisition of global climate index Historical data and construction of feature pool ; For characteristic pool Data were trended and normalized: Wherein, the For the normalized climate index, And Mean and standard deviation under sliding window.
3. The method for predicting mid-long term wind and solar resources taking climate remote correlation factors into consideration according to claim 2, wherein the feature pool Features include el nino-southern billows ENSO, north atlantic billows NAO, pacific annual oscillations PDO, indian dipoles IOD and north billows AO.
4. The method for predicting mid-long term wind-solar resources taking climate remote correlation factors into consideration according to claim 1, wherein for the target station history resource sequence, calculating a normalized mutual information value of the climate index, taking a lag time corresponding to a maximum value of the mutual information as an optimal early warning window period, and constructing a prediction factor set, wherein the method comprises the following steps: adopting a maximum correlation time lag search algorithm of mutual information, aiming at historical resource sequences of target stations Calculate it and each climate factor At different lag times , Mutual information value under month : Selecting Maximum value Constructing a set of predictive factors as the optimal early warning window period of the factors for the station, wherein Representing local resource values as joint probability distribution And the climate factor takes the value of The probability of the occurrence of the simultaneous occurrence, For the edge probability distribution of local resource data, i.e. taking into account the probability of occurrence of the resource data alone, The probability distribution of the edges of the climate factor data, i.e. the probability of occurrence of the climate factor values, is taken into account separately.
5. The method for predicting the mid-long term wind-solar resource taking climate remote correlation factors into consideration according to claim 1, wherein the steps of obtaining the current climate state vector based on the forecasting factor set, searching the historical database for the historical year closest to the current state euclidean distance, extracting wind-solar resource actual measurement data in the same period as the historical year, and constructing a reference probability density function comprise the following steps: Key factors screened based on predictor set Rate of change of Constructing a current climate state vector : Wherein the method comprises the steps of For the current moment of time, The total number of the key climate factors is selected; Searching a history database for the nearest Euclidean distance to the current state The historical years: extracting wind/light resource actual measurement data in the same period of the similar years, and constructing a predicted reference probability density function PDF, wherein: To the current year (forecast starting point) The state value of the individual climate factors, Is the first Chronology of historical year The state value of the individual climate factors, Climate state of the current year The euclidean distance between historical years, the smaller the distance, the more similar the climate modes for the two years, To assist in the total number of key climate factor features calculated, Index of climate factor features from 1 to , Is the first The weight coefficient of each climate factor reflects the importance degree of the factor on local resources, is used for weighting and calculating the similarity, Is the current year The state value of the individual climate factors, Is the first Chronology of historical year State values of individual climate factors.
6. The method for predicting the mid-long term wind-solar resource taking climate remote correlation factors into consideration according to claim 1, wherein the step of taking the reference probability density function as a priori distribution and combining a preset power mode prediction result as a likelihood function to output a prediction result containing a deterministic value and a confidence interval comprises the following steps: establishing a predictive equation in which nonlinear response terms of climate factors are introduced : Wherein the method comprises the steps of In order to predict the amount of resources, As the mean value of the climate state, For a nonlinear mapping function fitted by machine learning, , The ratio or contribution degree of the climate state mean value, the climate factor nonlinear term and the dynamic mode forecast term in the final forecast result are respectively, For a particular climate factor value (e.g., an ENSO index), here represents the key climate predictor of the input model, A seasonal power pattern forecast value; Is a residual error term or a random error term; the output results contain deterministic numbers and confidence intervals based on historical similarity annual distributions.
7. The medium-long term wind-light resource prediction system taking climate remote correlation factors into consideration is characterized by comprising: the data acquisition module is used for acquiring global climate index historical data, and standardized climate indexes are obtained after standardized processing is carried out on the data; The calculation module is used for calculating the mutual information value of the standardized climate indexes according to the historical resource sequence of the target station, taking the lag time corresponding to the maximum value of the mutual information as the optimal early warning window period, and constructing a forecasting factor set; The function construction module is used for acquiring a current climate state vector based on the forecasting factor set, searching a historical year closest to the Euclidean distance of the current state in the historical database, extracting wind and light resource actual measurement data in the same period as the historical year, and constructing a reference probability density function; And the prediction output module is used for taking the reference probability density function as prior distribution, combining a preset power mode prediction result as a likelihood function and outputting a prediction result containing a deterministic value and a confidence interval.
8. The mid-long term wind-solar resource prediction system considering climate remote correlation factors according to claim 7, wherein the obtaining global climate index history data, and the standardized climate index obtained by performing the standardized processing on the data, comprises: acquisition of global climate index Historical data and construction of feature pool ; For characteristic pool Data were trended and normalized: Wherein, the For the normalized climate index, And Mean and standard deviation under sliding window.
9. The mid-long term wind-solar resource prediction system considering climate remote correlation factors as claimed in claim 8, wherein the feature pool Features include el nino-southern billows ENSO, north atlantic billows NAO, pacific annual oscillations PDO, indian dipoles IOD and north billows AO.
10. The system for predicting mid-long term wind-solar resource taking climate remote correlation factors into consideration according to claim 7, wherein for the target station history resource sequence, calculating a normalized mutual information value of the climate index, taking a lag time corresponding to a maximum value of the mutual information as an optimal early warning window period, and constructing a prediction factor set, wherein the method comprises the following steps: adopting a maximum correlation time lag search algorithm of mutual information, aiming at historical resource sequences of target stations Calculate it and each climate factor At different lag times , Mutual information value under month : Selecting Maximum value Constructing a set of predictive factors as the optimal early warning window period of the factors for the station, wherein Representing local resource values as joint probability distribution And the climate factor takes the value of The probability of the occurrence of the simultaneous occurrence, For the edge probability distribution of local resource data, i.e. taking into account the probability of occurrence of the resource data alone, The probability distribution of the edges of the climate factor data, i.e. the probability of occurrence of the climate factor values, is taken into account separately.
11. The mid-long term wind-solar resource prediction system considering climate remote correlation factors according to claim 7, wherein the obtaining the current climate state vector based on the forecasting factor set searches the historical database for the historical year closest to the current state euclidean distance, and the extracting wind-solar resource actual measurement data in the same period as the historical year to construct the reference probability density function comprises: Key factors screened based on predictor set Rate of change of Constructing a current climate state vector : Wherein the method comprises the steps of For the current moment of time, The total number of the key climate factors is selected; Searching a history database for the nearest Euclidean distance to the current state The historical years: extracting wind/light resource actual measurement data in the same period of the similar years, and constructing a predicted reference probability density function PDF, wherein: To the current year (forecast starting point) The state value of the individual climate factors, Is the first Chronology of historical year The state value of the individual climate factors, Climate state of the current year The euclidean distance between historical years, the smaller the distance, the more similar the climate modes for the two years, To assist in the total number of key climate factor features calculated, Index of climate factor features from 1 to , Is the first The weight coefficient of each climate factor reflects the importance degree of the factor on local resources, is used for weighting and calculating the similarity, Is the current year The state value of the individual climate factors, Is the first Chronology of historical year State values of individual climate factors.
12. The system for predicting mid-long term wind and solar resources taking climate remote correlation factors into consideration according to claim 7, wherein said outputting a prediction result including a deterministic value and a confidence interval by taking a reference probability density function as a priori distribution and combining a preset power mode prediction result as a likelihood function comprises: establishing a predictive equation in which nonlinear response terms of climate factors are introduced : Wherein the method comprises the steps of In order to predict the amount of resources, As the mean value of the climate state, For a nonlinear mapping function fitted by machine learning, , The ratio or contribution degree of the climate state mean value, the climate factor nonlinear term and the dynamic mode forecast term in the final forecast result are respectively, For a particular climate factor value (e.g., an ENSO index), here represents the key climate predictor of the input model, As a predictive value for the seasonal power pattern, Is a residual error term or a random error term; the output results contain deterministic numbers and confidence intervals based on historical similarity annual distributions.
13. Computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, carries out the steps of the medium-long term wind and light resource prediction method taking into account climate remote correlation factors according to any of claims 1 to 6.
14. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the mid-long term wind and solar resource prediction method taking into account climate remote correlation factors according to any of claims 1 to 6.

Description

Medium-long-term wind-light resource prediction method and system considering climate remote correlation factors Technical Field The invention belongs to the technical field of electric power meteorology, and particularly relates to a medium-long term wind-light resource prediction method and system considering climate remote correlation factors. Background With the improvement of permeability of renewable energy sources in an electric power system, medium-long-term wind-solar resource assessment is important to electric power trade and maintenance planning and asset valuation. The existing medium-long term wind and light resource prediction technology has obvious limitation, so that the annual error of resource prediction in the region with obvious monsoon climate is extremely large. The current mainstream solutions include two types, namely numerical weather forecast (NWP) based on global climate pattern (GCM), and traditional statistical methods based on local historical data autocorrelation, such as ARIMA and markov chain, and in addition, qualitative application research of climate factors such as small amount of Elneno (ENSO), north atlantic wave motion (NAO) and the like exists in academia. The numerical weather forecast has the attenuation problem that the initial field error is amplified integrally along with time, the day-by-day deterministic forecast capability is obviously reduced after more than 15 days, the station power generation quantity forecast cannot be directly guided, the traditional statistical method has the defects of no memory, only relies on local data, cannot capture the wind speed and irradiance annual fluctuation and the abundant mutation which are controlled by the remote correlation of the sea interaction, the climate factors are rough to apply, quantitative engineering models aiming at specific stations and considering the lag time and the nonlinear response intensity are lacking, and the systematic method for effectively downscaling the large-scale climate long-period signal to the microscopic station resource fluctuation is integrally lacking, so that the annual error of resource forecast is extremely large in the region with obvious quaternary wind climate. Disclosure of Invention The invention aims to provide a medium-long-term wind-solar resource prediction method and a system taking climate remote correlation factors into consideration, so as to solve the problems of poor prediction capability and large prediction error in the prior art. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, the present invention provides a method for mid-long term wind and solar resource prediction taking into account climate remote correlation factors, comprising: acquiring global climate index historical data, and carrying out standardization treatment on the data to obtain a standardized climate index; calculating a mutual information value of the standardized climate indexes according to a historical resource sequence of the target station, taking the lag time corresponding to the maximum value of the mutual information as an optimal early warning window period, and constructing a forecasting factor set; Acquiring a current climate state vector based on the prediction factor set, searching a historical year closest to the Euclidean distance of the current state in a historical database, and extracting wind and light resource actual measurement data in the same period as the historical year to construct a reference probability density function; taking the reference probability density function as prior distribution, and combining a preset power mode prediction result as a likelihood function to output a prediction result containing a deterministic value and a confidence interval. Further, the obtaining global climate index historical data, and obtaining the standardized climate index after the standardized data is processed, includes: acquisition of global climate index Historical data and construction of feature pool; For characteristic poolData were trended and normalized: Wherein, the For the normalized climate index,AndMean and standard deviation under sliding window. Further, feature poolFeatures include el nino-southern billows ENSO, north atlantic billows NAO, pacific annual oscillations PDO, indian dipoles IOD and north billows AO. Further, for the target station history resource sequence, calculating a mutual information value of the normalized climate index, taking a lag time corresponding to the maximum value of the mutual information as an optimal early warning window period, and constructing a prediction factor set, including: adopting a maximum correlation time lag search algorithm of mutual information, aiming at historical resource sequences of target stations Calculate it and each climate factorAt different lag times,Mutual information value under month: SelectingMaximum valueConstructing a set of pre