Prediction of Performance of Coal-Based KWU Designed Thermal Power Plants using an Artificial Neural Network

摘要

Thermal power plants of different capacities have been installed in India. However, the coal-based power plant having 210 MW capacity turbo-generators with KWU design became popular and the plants were installed at various places across India. As they are operated continuously and maintained as and when needed, it is essential to evaluate and analyze thermal performance of each of the components of the plant. Energy and exergy analysis are the most frequently used techniques to evaluate the thermal performance of a power plant. As power plants are operated round the clock, it is essential to know the energetic and exergetic performance of the plants at various loads and understand the effect of critical parameters on them. Instant estimation of energetic and exergetic performance of the plant can be predicted by combining theoretical evaluation with an artificial neural network (ANN). Sixty five ANN models have been prepared and trained using forty emperical data sets of the thermal plants located in the western region of India. Seven parameters have been considered as input neurons and forty four parameters have been considered as output neurons. Trained ANN models have been employed to predict the performance of another power plant having same capacity and design using separate data sets. Errors in prediction have been evaluated in terms of MSE, NMSE, MAE, MARD and MRE. On comparing the error in prediction, it has been found that ANN model E1 (generalized regression network having spread constant of one) yields minimum value of error. Thus, it has been proposed as a suitable model to predict the thermal performance of the 210 MW KWU designed power plant while changing the operating load and deciding the parameters at the required instant.