HYDRO GENERATION SCHEDULING USING ANN TECHNIQUE

Abstract

The continuous development in industrialization and the living standard of common people has led the electricity supply short of meeting the power demand. This increased demand of electric supply has greatly pressurized power utilities for generating more and more power to meet the demand. At the same time, the electric industry is also rapidly changing. The traditional monopolistic environment is giving way to increased competition among the power utilities. To extract more profit, power producers are trying to exploit the full advantages of the available resources. The Country like India, where the hydro energy shares a significant portion of total power production, optimized hydro generation scheduling, can provide an effective solution for meeting power demand, without losing other advantages. Hydro Generation Scheduling, problem is a complex constrained optimization problem. Its aim, basically, is to find out the optimum periodic water release from reservoirs through turbines so as to maximize the total benefits incurred from hydro energy, while meeting the physical and operational constraints of the system. Mostly all major water resource systems are multi-purpose projects and have the capacity of providing a number of water related benefits such as irrigation, domestic & industrial use, recreation etc. When these benefits are considered in optimization problem, the problem becomes Multi-Objective Optimization Problem. The availability of limited water in system reservoir makes the problem more complex, because the discharge at a time affects the future discharge, and hence, the performance of hydro electric system. In this dissertation, Bhakra-pong-Dehar hydro power system is considered for hydro generation scheduling. The problem is a multi-objective, constrained optimization problem in nature whose aim is to optimize the total monthly hydro power generation while considering the irrigation demand. The priority is given to hydro power generation over the irrigation demand by setting 0.9 weightage to power generation function and 0.1 weightage to the square function of shortage of irrigation demand. To find the solution of the problem is a tough task. Various computational procedures such as maximum principle, variational calculus, dynamic programming, non-linear programming, decomposition method etc. have been reported for finding the optimal hydro schedule in the literature, but these methods suffer from various computational difficulties such as dimensionality difficulty, large memory requirement, non optimal results and large computation time. In this work, Hybrid Neural Network is proposed for this particular Bhakra-Pong-Dehar system. This neural network takes advantage of two types of neural networks, Primary Neural Network (PNN) ad Secondary Neural Network (SNN). Both PNN and SNN are Multilayer Feed-Forward~Nleural Networks An algorithm is also suggested for finding the configuration of Hybrid Neural Network, using PNNs and SNNs. The difference between PNN and SNN is with the way targets are assigned. The Primary neural network is trained against the direct normalized output patterns, while secondary neural network is trained against the error generated in the earlier stage of training, it may either be the training error of PNN or SNN. This method is then tested and results are compared with the optimal solution obtained form iterative Decomposition method. The results, obtained from the proposed hybrid neural network, are found quite promising for the Bhakra-Pong-Dehar hydro power system. ii