Environmental Economic Load Dispatch Using Hopfield Neural Network

Abstract

The economic load dispatch (ELD) is one of the most important optimization problems from the view point of power system to derive optimal economy. Classically, it is to identify the optimal combination of generation level of all generating units which minimizes the total fuel cost while satisfying the load. This classical ELD formulation have been solved by various methods like Lagrange method, Newton’s method etc. As the time progresses, the environmental constraints are becoming foremost important in deciding the operation of thermal units. Therefore, conventional load dispatch problem has to be solved to find the generation level that minimise the cost or minimise the emission level or their combination subjected to load balance. Even the environmental emission can be taken as the constraints in cost optimisation problem. Therefore, an efficient and diversified model is needed to handle the above variations in the problem. The solutions to the above problems is attempted using Modified Hopfield Neural Network (HNN), which works on the principal of minimizing the energy function as conventional HNN and therefore sure to converge but differs from conventional HNN. In the conventional HNN, equality constraint on load is combined into objective cost after assigning suitable weightage factors. The computational procedures include selection of weighting factors and thus the convergence depends on the weight selection. In the modified HNN, there is flexibility of taking objective function and the constraints separately. The internal parameters of neural networks are computed using valid subspace approach, which guarantee the convergence of solution at equilibrium points. In this thesis, the environmental economic dispatch is considered and the following has been attempted 1. Cost optimisation with load balance constraint 2. NOx emission optimisation with load balance constraint 3. SOx emission optimisation with load balance constraint 4. Cost optimisation with SOx and NOx emissions and load balance as constraint The results are presented for the above using Modified HNN. The results are also compared with classical λ and classical HNN methods for cost optimisation.