Modelling and optimization of residential heating system using random neural networks

In this paper, a novel random neural network (RNN) model based optimization process for radiator-based heating system is proposed to maintain a comfortable indoor environment in a living room of a single storey residential building. The predictive model of the living room is developed by training a feed forward RNN and then optimisation algorithms are used to calculate the optimal flowrate for the radiators. Three optimisation algorithms: Genetic Algorithm (GA), Particle swarm optimization (PSO) algorithm, and Sequential quadratic programming (SQP) optimization algorithm are investigated to calculate the optimal control input. The accuracy of the control scheme is verified by simulations using International Building Physics Toolbox (IBPT). It was found that mean squared error (MSE) for PSO is 38.87% less than GA and the MSE for PSO is 21.19% less than SQP. The RNN model based optimization technique is further compared with model predictive controller (MPC) designed for the radiator based heating system. The comparison results showed that the proposed RNN technique minimize the energy consumption and maintains accurate room thermal comfort according to the predicted mean vote (PMV) based setpoints.