Research On Short-term Wind Power Prediction Based On Cloud Computing And Intelligent Algorithm

With the full tension of fossil energy and the increasing environmental pollution,the development and utilization of renewable energy has become the world’s important energy development strategy.Wind power is the most rapid development of renewable energy clean energy,wind power is also the most large-scale development and commercial development prospects of power generation.Due to the randomness,volatility and intermittent of wind power itself,wind power output is instability.At present,the main problem is how to improve the accuracy of wind power prediction,especially in the next 24 hours forecast.Based on the above background,this article mainly focuses on the following aspects.(1)The real data of wind power generation system is the basis of wind power forecasting,and in the wind power generation system or data acquisition,measurement,transmission,conversion and other sectors,especially the power of abandoned wind artificially,abnormal data is inevitable in historical data.Based on the analysis of the characteristics of wind field anomaly data,the four-point method is used to pre-process the wind power system and improve the accuracy of historical data.(2)Compared with other intelligent prediction algorithms,the performance of the artificial neural network in self-learning habits,adaptability,robustness,fault tolerance and generalization ability is outstanding.The artificial neural network which is the most used in wind power prediction is a static neural network,and the static neural network is used to predict the time-varying characteristics of the wind power sequence,so the prediction accuracy is not high.Therefore,this article chooses the Elman neural network which can better reflect the dynamic characteristics of the wind power,and proposes the Elman neural network wind power prediction algorithm.(3)The network parameters of Elman neural network used will affect the network performance,and in the current study stage of the neural network,gradient direction change fixed gradient descent method is widely used.This method has defects of slow convergence and easy to fall into local optimal solution,these defects can restrict the network optimization ability.Therefore,this article has the global optimization performance improved cuckoo search algorithm to optimize Elman neural network weights and thresholds,aim to improve the stability and generalization ability of Elman neural network.(4)In this article,the improved cuckoo search algorithm and Elman neural network are designed in parallel,and the performance of the algorithm is tested on the Spark cloud platform,which can not meet the actual demand of the short-term wind power forecasting.The experimental results show that the prediction accuracy and the real-time performance are superior to the traditional single-power prediction algorithm.