Neural approximations for the optimal control of heating systems

This paper presents a case study of the HVAC system for a greenhouse. The optimal control of greenhouse climate entails operations on the control variables in order to maximize the economic profit of the grower and is heavily influenced by the accuracy of the greenhouse dynamics and the weather predictions, thus accurate dynamic model and good weather forecasts are required. We first address the problem of environmental temperature forecasting by using a classical linear model. We then suggest superior nonlinear prediction models based on neural networks. Finally, an optimal control is applied to the nonlinear greenhouse model and is approximately implemented by means of neural networks. >