Ultra-Short-Term Wind Power Prediction By SSA-ELM And Its Grid Connected Scheduling

Wind power generation has the characteristics of green,clean,environmental protection,and the wind energy reserve in nature is huge,if the full use of wind resource will have an important impact on the economic development of society and people's living.In addition,China has a vast territory and abundant wind energy resources.The windy regions account for about two thirds of the country's total area.The regions with abundant wind energy resources are mostly concentrated in north China,northwest China,northeast China and southeast China,covering more than 20 provinces,cities and autonomous regions.However,the large randomness,volatility and intermittently of wind speed have a huge impact on the safety and stable operation of the power grid.When the wind speed is too large,and the wind power is large,it will seriously affect the safety of the power system.In the actual wind power grid connection and power system dispatch,reliable wind power prediction can not only reduce the influence of wind power volatility on economy,but also ensure the safety and stability of power system operation after wind power is connected.Based on the above background,this paper through the research and analysis from the two aspects of improving the ultra-short-term wind power prediction accuracy and considering the economic optimal dispatch of the power system with wind power prediction error,including the following contents:(1)This paper divides the prediction of wind power from time scale and method,and decides to use the heuristic algorithm to optimize the extreme learning machine to predict the ultrashort-term wind power.According to the formula,physical content and other aspects of the wind power output factors are analyzed in detail to determine the main influencing characteristics of wind speed,wind direction,temperature and density as the prediction of ultrashort-term wind power.(2)Simulate the salp swarm foraging behavior,optimize the input weight matrix and the hidden layer deviation value of the extreme learning machine by using the salp swarm algorithm to optimize the parameters in the iterative process,so as to improve the adaptability and accuracy of the prediction model.Salp swarm algorithms optimization extreme learning machine(SSA-ELM)and Particle swarm optimization extreme learning machine(PSO-ELM)were used to predict the output power of the wind farm.Through simulation analysis,the number of nodes in the hidden layer of SSA-ELM is determined to be 30 and the number of iterations to be 300.The simulation training of PSO-ELM was carried out to lay a foundation for example analysis.(3)According to the historical data of a wind farm in Henan province,data were preprocessing.The sample set was selected by samples of different sample sizes and similar days of FCM clustering,and the 5 historical days with high predicted daily similarity were determined as the training set.Using ELM,SSA-ELM,PSO-ELM and BP neural network to establish ultra-short-term wind power prediction model,four days in the four seasons were selected as the prediction days respectively,combined with the error evaluation indexes of mean absolute percentage error(MAPE),root mean square error(RMSE),mean absolute error(MAE)and decision coefficient(R2),it was found that the prediction effect of SSA-ELM model was better.And according to the relevant documents specified in the wind farm report every 15 minutes as required rolling over the next 15 minutes to four hours of wind power prediction data,puts forward to four hours for time scale of the short-term wind power prediction error analysis of the rolling,the mean values of MAPE,RMSE and MAE of SSA-ELM were 0.24,2.8 and 0.18 respectively,which again proved the high accuracy of the prediction model.(4)According to the influence of wind power grid connection on the economic dispatching of power system,combined with the management methods and requirements of grid-connected wind power dispatching,on the premise of improving the prediction accuracy of wind power,a dispatching model incorporating wind power generation into the objective function is established,and PSO algorithm is adopted to solve the optimal dispatching model of power system considering the prediction error of wind power.Through data analysis,the cost of power generation with wind power grid connection is $4,017.67 lower than that of traditional power generation.The results show that the ultra-short-term wind power prediction with high accuracy not only improves the utilization rate of the output power of the wind farm,but also effectively reduces the generation cost of the power system,which is of great significance for the research of wind power grid connection.