Improved evolutionary algorithms for economic load dispatch optimization problems

We describe and evaluate a new evolutionary algorithm (EA) approach for solving the optimal power flow problem for economic load dispatch in the electricity generation industry. The method combines a standard evolutionary algorithm (EA) with `smart mutation´ and hill-climbing. The key aspect of the method that improves performance is the smart mutation operator, which targets mutation of genes according to their respective contributions to the cost function. We consider benchmark instances of the economic load dispatch problem, which involve minimum/maximum generation limits, power balance, ramp rates and prohibited operating zones. Violation of either of these constraints introduces the concept of penalties, and these in turn provide the basis for the smart mutation operator. Our `smart´ EA (SEA) is compared with a basic EA (BEA), and with reported results for other recent algorithms, on three benchmark cases involving 6, 15 and 20 generating units. On the larger two of these problems we find better solutions than have so far been reported in the literature, and (where sufficient information is available) in some cases statistical tests confirm the superiority of SEA on these problems to other recent algorithms.