Совершенствование алгоритмической базы эволюционного синтеза нейросетевых моделей потребления электрической энергии в коммунально-бытовом секторе

Abstract

A novel diploid genetic algorithm (DGA) scheme is proposed. An original modification of the crossover operator and an additional genetic operator are introduced. For Alekseevka residential district, the DGA parameters chosen for solving a real-world problem prove to provide fast and stable synthesis of a neural-network-based model of mutual electricity and heat consumption in the district. The accuracy of the obtained model is enough to apply it to controlling quality of heat supply for domestic consumers.