基于BP-神经网络的风速短期预报技术研究

Study on BP - neural - network - based Technology for Short - term Forecast of Wind Speed

针对风速预报中出现的资料获取困难、预报精度差等问题,文章提出采用临近历史数据的BP-神经网络风速短期预报模型,并重点对BP模型的输入层和隐含层参数进行估计。在一定范围内,枚举输入层和隐含层的参数,并采用大量数据进行模拟,同时采用SSE和MAE两种指标对模拟结果进行评价,得到了适合于风速预报的多个不同参数BP模型。同时将多个BP模型用于预报,发现预报结果精度都比较高,表明不同参数的BP模型均可用于预报且BP模型存在异参同效性。

Las the difficulty in data collection and poor precision in the wind speed forecast, the model based on BP-neural network for the short-term wind speed forecast by application of the recent historic data is proposed in the paper. Furthermore, estimates of parame- ters of the input layer and implication layer of the BP model are stressed. In a proper range, parameters of the input layer and implication layer are listed as well as data in a large quantity is applied for their simulation. Simultaneously, the simulation results are evaluated by application of SSE and MAE indicators. Therefore, the BP models with different parameters suitable for the wind speed forecast are gained. A couple of BP modes are applied for the wind speed forecast, showing that all the forecast preeisions are higher. This proves that the BP models with different parameters can be applied for the wind speed forecast. They cause the same results although the parameters are different.