Optimization of Pump Operations in a Complex Water Supply Network: New Genetic Algorithm Frameworks

In previous papers, a simple genetic algorithm (GA) was developed for the optimization of pump operations in water-distribution networks. Its application at the water supply network of Milano showed the possibility of a great improvement of its performance in terms of both energy and economic savings. In the present paper is now investigated the possibility of using different and improved GAs to obtain better results. Improvements concerned the description of the pump conditions with a real number (and therefore in continuous form) and the introduction of elitism and of a slightly modified form of mutation. Simulations were obviously performed with reference to the same model under the same assumptions of the previous papers. Results showed significant improvements in the passage from a discrete to a continuous description of the pumps functioning and a slight improvement using elitism and no differences using mutation. The latter result might need some more research: mutation is introduced to enlarge the space in which the ‘individuals’ perform their search, and there is the need to understand whether this little improvement is due to the poor performance of this mutation or instead, because the space of search is already well defined. The need for more in-depth investigations is also investigated in the present paper.