Development And Research On Statistical Methods For Short-term Wind Speed Forecasting

Due to conventional resource depletion and global environmental pollution issues, renewable energy (such as wind, solar, and hydropower) has received increasing attention in the world. Wind power is one of the cleanest renewable energy sources, and has the rich deposit, renewable, widely distributed, and pollution-free characteristics. Recently, the amount of energy generated by wind power has rapidly increased. With the rapid development of wind power, it occupies an important position in the electric power development. However, due to the intermittent and stochastic nature of wind power, and thus it is still a less reliable source and difficult to be integrated into power grid systems. Accurate forecasting of wind power is a critical issue for adjusting the scheduling plan, ensuring the quality of electric energy,reducing system reserve capacity, and reducing the system cost. However, the wind speed prediction is the basis of the wind power prediction. As a result, accurate forecasting of wind speed plays a crucial role for the development of wind power.According to the difference of forecast periods, wind speed prediction can be classified into three categories: long-term prediction, medium-term prediction, and short-term prediction. For long-term prediction, prediction accuracy is important for wind farm planning and design. For medium-term prediction, prediction accuracy is important for power balance and scheduling, trading, and transient stability assessment in power system. For short-term prediction, accurate prediction are important in power system control, and it is important for correcting the deviation in power grid and electric plan, improving the power grid and electric plan, wind energy utilization, reducing the deviation in medium-term and long-term predictions, and ensuring the safety of power grid. However, accurate forecasting of the short-term wind speed is still insufficient, and it is a critical issue for improving the short-term wind speed prediction. The focus of this study is the development and research statistical methods for short-term wind speed prediction.Based on the wind speed of hexi area as the research object, the statistical law of wind speed and direction has been systematically studied, and the wind change general rule has been discussed. According to its characteristics, three types of novel approaches are developed for short-term wind speed forecasting. The first type is a novel prediction approach for wind speed based on the seasonal adjustment method(SAM), the second type is a novel prediction approach for wind speed based on the Empirical Mode Decomposition (EMD), and the third model is a novel prediction approach for wind speed based on the ideas of combination. The research achievements were expected to offer guidance for developing the short-term wind power prediction system.The main achievements are as follows:1) Considering the complex cycles of wind speed, the SAM was used to pre-process raw wind speed data in this study. The first type method is developed for short-term wind speed prediction based on SAM. Two novel prediction models for wind speed were proposed in this method. The first model (known as SAM-ESM) is a novel prediction approach for wind speed based on the SAM and exponential smoothing method (ESM),the second model (known as SAM-GA-WNN) is a novel prediction approach for wind speed based on the SAM, genetic algorithm (GA), and wavelet neural networks (WNN). The mean hourly wind speed data in the Hexi Corridor of China were used as examples to evaluate the performance of the proposed approach. Numerical results show that the proposed approach can outperform the other conventional methods. It is concluded that the proposed approach is an effective way to improve the short-term prediction accuracy;2) Considering the intermittent and stochastic nature of wind speed, the EMD was used to pre-process raw wind speed data in this study. The second type method is developed for short-term wind speed prediction based on EMD. Two novel prediction models for wind speed were proposed in this method. The first model (known as EMD-ARMA) is a novel prediction approach for wind speed based on the SAM and auto-regressive and moving average (ARMA), the second model (known as EMD-PSO-BPNN) is a novel prediction approach for wind speed based on the EMD,Particle Swarm Optimization (PSO), and back-propagation neural networks (BPNN).The mean hourly wind speed data in the Hexi Corridor of China were used as examples to evaluate the performance of the proposed approach. Numerical results show that the proposed approach can outperform the other conventional methods. It is concluded that the proposed approach is an effective way to improve the short-term prediction accuracy;3) Considering the different variation characteristics of wind speed, the ideas of combination was used to predict wind speed. The third type method is developed for short-term wind speed prediction based on the ideas of combination. Two kinds of combined prediction models for wind speed were proposed in this method. The first model (known as ESM-GA-WNN) is a novel prediction approach for wind speed based on the ESM, GA, and WNN, the second model (known as ARMA-PSO-BPNN)is a novel prediction approach for wind speed based on the ARMA, PSO, and BPNN.The mean hourly wind speed data in the Hexi Corridor of China were used as examples to evaluate the performance of the proposed approach. Numerical results show that the proposed approach can outperform the other conventional methods. It is concluded that the proposed approach is an effective way to improve the short-term prediction accuracy;4) The prediction results based on the above three types of short-term wind speed statistical prediction methods were compared and analyzed, and their applicability was studied in this section. On the whole, both SAM-GA-WNN and EMD-PSO-BPNN models are the better models.