Study On Power Prediction Of Small Hydropower Generation In Multi Small Hydropower Area

As an important part of renewable energy,small hydropower also has the advantages of low investment cost,short construction period and stable electricity price.Many provinces in China are rich in small hydropower resources,which can not only supply power to local residents,but also send part of the power to the large power grid.However,most of the small hydropower stations are run of river hydropower stations with small reservoir capacity,and their power generation is easily affected by the weather,which makes it difficult to predict the output of small hydropower stations and increases the difficulty of local power grid management of small hydropower stations.Therefore,it is of far-reaching significance to study the prediction model of small hydropower generation in multi small hydropower areas.This paper first introduces the influencing factors of small hydropower generation power in multi small hydropower areas,and analyzes them in detail through internal and external laws,and then preprocesses the power data of small hydropower generation.The accuracy of small hydropower output prediction largely depends on whether there is a typical training set.Therefore,the fuzzy c-means clustering method is used to cluster the small hydropower generation power data set,and the training samples with large correlation with the prediction day are selected.Generalized regression neural network has strong advantages in regression prediction and learning speed,so this paper takes generalized regression neural network as a prediction method,but the value of its smoothing factor determines its prediction performance.An improved gray wolf algorithm is proposed to optimize the value of the smoothing factor,so as to improve the prediction performance of generalized regression neural network.The example proves the feasibility.Due to the strong non-stationary of small hydropower generation power,it presents the characteristics of large randomness.In order to extract the local characteristics of small hydropower generation power,this paper uses the variable mode decomposition method to decompose the original sequence into a limited bandwidth subsequence,and uses the zero crossing rate method to reconstruct the subsequence to obtain the trend curve and fluctuation curve of small hydropower,and uses different prediction methods to deal with the different characteristics of the two curves.The trend curve of small hydropower is predicted by the neural network model of gating cycle unit with simple structure,and the fluctuation curve of small hydropower is predicted by the generalized regression neural network model optimized by the Improved Grey Wolf algorithm.Finally,the final power prediction curve of small hydropower is obtained by superposition.Finally,this paper uses the power data of a small hydropower station in Hubei Province for example analysis.The simulation results show that the two models proposed in this paper achieve good prediction results,and verify the effectiveness and feasibility of the proposed model.