Method and system for predicting wind speed

Abstract

The invention discloses a method and system for predicting wind speed. The method includes: obtaining the original sequence of wind speed data; using a particle swarm algorithm to determine the optimal preset scale parameter and optimal bandwidth parameter of the variational mode decomposition method, and converting the original sequence Decompose into several modal function subsequences; use the differential evolution algorithm to determine the kernel parameters of the least squares support vector machine model of each modal function subsequence, the variation factor of the mutation operation decreases with the increase of the evolutionary algebra, and the generated mutant individuals It is related to the optimal individual of the previous generation, and the crossover probability factor of the crossover operation increases with the increase of the evolutionary algebra; according to the autocorrelation of each modal function subsequence and each kernel parameter, determine the least squares of each modal function subsequence Support vector machine wind speed prediction sub-model, and predict the decomposition wind speed of each sub-sequence through each wind speed prediction sub-model; determine the final wind speed prediction value according to each decomposition wind speed. The method and system provided by the invention can accurately predict wind speed.

Description

technical field [0001] The invention relates to the field of time series prediction, in particular to a method and system for predicting wind speed. Background technique [0002] As a renewable and clean energy, wind power has been developed on a large scale in my country in recent years. At the same time, the randomness, intermittence and volatility of wind power have brought security risks to the stability and economic operation of the power grid. Accurate wind power forecasting can provide an important basis for power dispatching and effectively reduce the impact of wind power on the grid. Since wind power has a direct relationship with wind speed, wind power can be predicted through wind speed prediction, so it is very necessary to accurately predict the wind speed of wind farms. Since the wind speed time series is preprocessed by signal analysis method and then predicted separately and then integrated, it can reduce the influence of wind speed non-stationarity on the p...

AI Technical Summary

Problems solved by technology

[0003] There are three main methods of decomposing wind speed series in the prior art: (1) wavelet decomposition, this decomposition method has good time-frequency localization characteristics, but the decomposition effect depends on the selection of basis functions, and its adaptability is poor

(2) Empirical mode decomposition, this decomposition method has strong adaptability, but there are problems such as endpoint effect and over-envelope

(3) Local mean decomposition. This decomposition method has fewer iterations and lighter endpoint effects, but the conditions for judging pure FM signals need to be tried and tested. If the sliding span is not selected properly, the function will not converge, resulting in over-smoothing and affecting the algorithm. Accuracy

However, the performance of the least squares support vector machine depends on the selection of kernel parameters. At present, the selection methods of kernel parameters include grid search algorithm and intelligent population optimization algorithm. The grid search algorithm can only achieve rough optimization, and often cannot find Optimal value; intelligent population optimization is easy to fall into local optimum due to the design flaws of the algorithm itself

It can be seen that due to improper selection of kernel parameters in the prior art, the prediction accuracy of the prediction model based on the least squares support vector machine is low

Images