函数联接神经网络在电力系统短期在线负荷预报中的应用

Application of Functional Link Network for Short-term Online Load Forecasting

提出了 2个新的用于实时在线短期负荷预报的函数联接神经网络 (FLN)模型 .2个模型都把负荷与气象参数结合起来构成非线性ARMA过程 ,并应用FLN的函数逼近能力获得了 2个模型的参数 .测试与在线操作表明2个模型的预测效果是令人满意的 ,2 4h向前负荷预报的平均绝对百分误差 (MAPE)对HFLN来说几乎都在 3%以下 ,而对DFLN来说几乎都在 5%以下 .

Two new functional link network(FLN) based short term load forecasting models for real time implementation are presented.The load and weather parameters are modeled as a nonlinear ARMA process and parameters of these models are obtained using the functional approximation capabilities of FLN.Testing and online operation has shown satisfactory performance with mean absolute percentage error(MAPE) mostly less than 3% by HFLN and less than 5% by DFLN for a 24 hour ahead forecast.