Fusion of Neural Computing and PLS Techniques for Load Estimation

M. Lu, H. Xue, X. Cheng, and W. Zhang (PRC)


Computational intelligence, short term load forecasting, neural computing, partial least squares.


Power system usually uses statistical modeling methods and Computational Intelligence (CI) techniques such as neural computing for Short Term Load Forecasting (STLF). Statistical modeling methods often need dubious distributional assumption to be claimed while neural computing is weak in the determination of topology. To overcome the drawback of conventional techniques, we developed a CI hybrid model based on neural computation and PLS techniques, which is appropriate for nonlinear modeling and latent structure extracting. The CI hybrid model can automatically determine the optimal topology to maximize the generalization. Due to the adoption of load conversion method and novel transfer functions, we can make the best of historical information and obtain a quicker convergence. The effectiveness of the CI hybrid model is demonstrated through an application to the data set of the Puget Sound Power and Light Company. Compared with the abductive networks model, the CI hybrid model reduces the forecast error by 32.37% on workday, and 27.18% on weekend.

Important Links:

Go Back