Integrating evolving fuzzy neural networks and tabu search for short term load forecasting

An integrated evolving fuzzy neural network and tabu search (IEFNN-TS) for short term load forecasting method is presented in this paper. In this paper, short-term load forecasting is presented first using fuzzy hyperrectangular composite neural networks (FHRCNNs). Then, we use evolutionary programming (EP) and tabu search (TS) to find the optimal solution of the parameters of the FHRCNNs (that parameters include such as synaptic weights (w/sub jk/), biases (/spl theta//sub jk/), membership function (m/sub j/(_/sub t/)), sensitivity factor in membership function (s/sub j/) and adjustable synaptic weight (M/sub ij/ and m/sub ij/). We know that the EP has a good capability at search globe optimal value, but has poor capability at search local optimal. But the TS has good capability at local optimal search. So, here, we combine this two methods advantages to improve the shortcoming of the tradition ANN training that the weights and biases always trapped into a local optimal. Finally, we use this (IEFNN-TS) to improve the solution quality. Actually, we can reduce the error of load forecasting. The proposed IEFNN-TS load forecasting scheme was test using data obtained from a sample study include one year, one month and 24 hours. The result demonstrated the accuracy of the proposed load forecasting scheme.