Абстрактный

Neural Network Based Temperature Prediction

Ancy M Alias, Prof.Amey George, Anish Francis

Hydroelectric power contributes around 20% to the world electricity supply and is considered as the most important, clean, emission free and an economical renewable energy source. India is endowed with economically exploitable and viable hydro potential assessed to be about 84,000 MW at 60% load factor. Hydro electric power plants operating all over the world has been built in the 20th century and are running at a higher plant-factor. This is achieved by minimizing the failures in plant and operating the plant continuously for a longer period at a higher load. However, continuous operation of old plants have constrained with the failures due to bearing overheating. The aim of this research is to model and simulate the dynamic variation of temperatures of bearing of a hydro electric generating unit. Bearing heat exchanger system is a MIMO system with complex nonlinear characteristics, so it is difficult to model it using conventional modelling methods. Hence, in this research neural network (NN) technique has been used. Same approach is used to predict temperature rise in generator transformer in the plant. Temperature rise in transformers depend upon variety of parameters like ambient temperature, output current, type of core etc. Considering these parameters temperature rise estimation is a complicated procedure. This method avoids complication required for accurate estimation.