Model predictive control of district heating system

Abstract

District heating system (DHS) is a widely used and increasingly popular energy source in cities. The uncertainty in the heat load (HL) due to customer demand fluctuations makes unit commitment (UC) and heat production unit (HPU) control a complex task. This case study of the DHS at Fortum Oslo Varme AS (FOV) aims to find a strategy to optimize and fully automate UC and HPU. Our results suggests this can be accomplished by using model predictive control (MPC) to control HPU power and flow rate, mixed integer linear programming (MILP) optimization to solve UC problem, and multiple linear regression (MLR) model to predict the HL. We also show that the fuel cost can be reduced significantly.