考虑多重周期性的短期电价预测

考虑到电价各时段变化以及周末与工作日变化的差异,提出了区分周末的分时段短期电价预测模型。该模型首先将各日中同一时段的电价形成该时段的电价序列,再将各时段电价序列分为工作日电价序列和周末电价序列。这样形成了多个消除了日周期性和星期周期性的子电价序列,分别对各子电价序列进行预测以得到预测日电价。采用基于小波分析的广义回归神经网络对这些子电价序列分别进行提前一天的预测,各子电价序列的预测电价就形成了下一天的预测电价。采用该方法对西班牙电力市场电价进行了长时间的连续预测,并与已有的预测方法进行了详细的比较分析,研究表明该方法能够提供更准确的预测电价。

Short-term Electricity Price Forecasting Based on Price Subsequences

The features of electricity prices in one period are different from that in another period. Furthermore, the price of weekday is generally higher than that of weekend. A period-decoupled short-term electricity price forecasting method based on separating weekend is presented. The prices of the same time period in each day form the period-decoupled price sequence, then each period-decoupled price sequence is separated into one weekday price sequence and one weekend price sequence. Thus some electricity price subsequences are obtained. The generalized regression neural network (GRNN) based on wavelet analysis is applied to forecast the day-ahead price for each electricity price subsequence respectively, and the hourly price of the forecasted day is obtained by seriating the forecasted prices of each subsequence. The proposed method is applied to forecast electricity price for Spanish electricity market in numerical examples, and it is compared with four existing methods. The comparisons of forecasting results indicate that the proposed method can provide more accurate predictions.