Power generation scheduling for hydropower plants using artificial neural network

Abstract

This thesis investigated the determination of electricity generation in hydropower plants using an artificial neural network in the central 1-network region of Electricite du Laos (EDL). The main goal was to design a neural network model in MATLAB to learn the optimized cost behavior concerning the total cost of each power plant, where electricity should be generated optimally for the supply and demand side. The neural network has trained with numerous problems, such as backpropagation algorithms to learn the optimal scheduling problems, optimization problems, and economic dispatch problems. One of the various methods used in computing to solve these problems is the Lambda method, which is based on quadratic function equations. In comparison, the results of the calculations have shown that the neural network method has better-operating costs than the Lambda method. It is indicated by an accuracy value represented with an R-value greater than 0.9, which is the input-output relationship. It represents the efficiency and accuracy of the neural network itself.

Comparison of the results between the neural network method and the Lambdas method in the electric generation of hydropower plants. It found that the neural network method can generate up to 7 megawatts at 0.0016% more power than the lambda method and saves $572.90 at 0.0019% per hour in production costs. Can saves times up to 420 seconds at 79.54%