基于自组织映射神经网络的市场清算电价预测

市场清算电价预测是电力市场中交易决策的基础。人工神经网络是电价预测较为理想的方法,但依然存在一些问题,如样本训练有时需要很长时间,存在收敛问题,特别是当样本特征量不明显的时候,这种现象更为突出。针对这一问题,利用自组织映射的聚类特性将历史数据进行特征分类和筛选处理,处理后形成的新数据用于训练三层BP神经网络,仿真结果表明,经过这种数据处理后,网络的收敛速度得到了显著提高,且预测效果良好。

Forecasting market clearing price using self-organizing map neural network

Forecasting the market-clearing price (MCP) is the most essential task for any decision-making in electricity market. Artificial neural network (ANN) is a preferable forecasting method. However, there still exist some theoretic shortcomings in ANN method, such as the time-consuming sample training and convergence problem. Especially when the characteristic of sample is hard to capture, those phenomena will be more explicit. To solve the problem, based on the characteristic of self-organizing and clustering of self-organizing map (SOM), this paper proposes a method to deal with the sample dataset of the BP model, which can perform a data analysis, and then form a new training dataset. By using the BP network on the new dataset analyzed by SOM for the prediction, the efficiency is advanced remarkably and the prediction is satisfactory.