HIERARCHICAL FUZZY RULE-BASED CONTROL OF RENEWABLE ENERGY BUILDING SYSTEMS

Proper control of low energy buildings, which is more difficult than in conventional buildings due to their complexity and sensitivity to operating conditions, is essential for better performance. In this paper, a three-level hierarchical fuzzy rule-based supervisory control scheme is described that is capable of optimising the operation of a renewable energy building system. For the first level rules, a fuzzy decision tree is used to choose the appropriate set of rules according to weather and occupancy information; the second level fuzzy rules generate an optimal energy profile; and the third level fuzzy rules determine the mode of operation of the equipment and select the control variables so as to achieve the optimal energy profile in the most efficient way. The controller is developed using a computer simulation of a typical building using a renewable energy system. The optimization is partitioned into long-term energy profile optimization and hourly subsystem optimization. Fuzzy rules are learned from data generated by the above optimization. A hierarchical structure is used to reduce the number of rules, trim redundant information and reduce the computing time required for the optimization. The energy consumption and the thermal comfort when using the proposed control scheme is compared with those when using expert rule based control. INTRODUCTION The greatest energy consumption in buildings occurs during their operation rather than during their construction. Optimized control is vital for the performance of low energy buildings. There are some major differences between renewable energy systems and conventional systems: the existence of local energy generators; the intensive interaction with the natural environment; the different building insulation standards, the relationship between the renewable systems and the conventional back up systems. These factors introduce new problems that make the proper control both complicated and challenging. Some work has already been reported by several researchers. Predictive control is widely adopted to optimize building behaviour to save energy and improve comfort. The future load and environment are input into a building simulation model, an optimization algorithm is used to find optimum set points[1,2,3,4]. At the equipment level, the mode of operation is chosen so as to achieve short-term set points or sub goals[5,6]. Genetic algorithms, neural networks, dynamic programming and fuzzy logic are popular methods to solve the complex nonlinear problem of building energy system optimization [7]. Although some promising results have been presented, the high computational demands of on-line optimization limit the practical application of most of the proposed methods. In this paper, a method of partitioning the high dimensional problem into sub problems is introduced to reduce the optimization time. A novel hierarchical fuzzy supervisory controller