级联相关的神经网络模型在边际电价预测中的应用

在实际的电力市场运作中，电厂的报价反映了电厂的运行成本和市场供求，决定电厂机组能否上网发电和上网电量。而报价的一个重要指标是预测的系统边际电价。因此，电力市场中的边际电价预测在发电厂的市场化运营中处于重要的地位，特别是对电力供应商的决策有重要意义。文中应用神经网络理论中的级联相关模型对电力系统的边际电价进行预测，优点在于避免了对网络结构的估计，网络在训练的过程中能够自适应地增加隐含节点，同时提出了在训练集中增加特殊数据点以提高预测精度的方法。通过New EnglandISO数据算例预测第2天的24h边际电价说明了这种方法的可行性，并用3层BP神经网络做了对比研究。

APPLICATION OF THE CASCADED CORRELATION NETWORK TO FORECAST THE MARGINAL CLEARING PRICE IN POWER MARKET

In the power market, the bidding price of the supplier reflects the cost and the demand-supply of electricity and can decide the amount of the generation of the supplier. For making the bidding price, the forecasted marginal clearing price (MCP) is an important index and has a significant meaning to the supplier. In this paper the cascaded correlation network is applied to forecast the price and the results are compared with a three-layer BP network. The merit of using cascaded correlation network is to avoid the estimation of the structure of the network. Meanwhile a method of adding special data into the training set is used to improve the accuracy of forecasting. The example of forecasting MCP of the next 24 hours of New England ISO data shows this method promising.