6 Conclusion
At present, the forecasting problems of short-term wind speed are mainly solved by machine learning methods such as BP neural network, SVM and LSSVM. However, the SVM largely depends on the sample data. Especially the efficiency of solving the problem becomes lower with large number of samples. The application scope of BP neural network is limited as slow to learn and easy to fall into the local minimum problem. LSSVM method reduces the unknown parameters compared with SVM method, and reduces the complexity of the solution. On the other hand, the choice of parameters is extremely crucial when LSSVM is used to forecast the shortterm wind speed. If the parameters are chosen improperly, it will often cause the model to owe learning or over learning, which directly affects the forecasting effect on wind speed. To solve this problem, a short-term wind speed forecast model with optimized parameter LSSVM model based on improved ant colony algorithm is presented in this paper. The simulation results show that the proposed algorithm is more accurate than the non-optimized LSSVM model and the BP neural network forecast model. Consequently, the proposed algorithm is more effective in short-term wind speed forecast. In the future, the parameters optimization method of LSSVM based on improved ant colony algorithm will continue to be used in other areas on predictive control.
