Fuzzified neural network approach for load forecasting

In load forecasting, the operator or the concerned person uses his or her e xperience and intuitions to obtain a good guess of the load demand. This gu ess is normally supported by sophisticated mathematical prediction techniqu es. The short term load not only varies from hour to hour, but is also infl uenced by the nature of events, load demand, the type of the load considere d, seasonal variations, weekend day or holidays, and also by sudden demand and loss of load. Accordingly, it is quite clear that the electrical load-f orecasting problem is quite difficult to model with mathematical difference or differential equations. In this paper the short term load forecasting p roblem has been formulated using artificial neural networks model. But the existing neural networks have various drawbacks like large training time, h uge data requirement to train for a non linear complex load forecasting pro blem, the relatively larger number of hidden nodes required etc. Hence, an attempt has been made to develop a non linear load forecasting model using fuzzified neuron models to overcome the above mentioned problems. These mod els would have the capability of representing operators' experience and com plex mathematical formulation needed for short term load forecasting.