Building dynamic cooling/heating load prediction method based on hyperball CMAC neural network

It is difficult to timely predict dynamic loads of green buildings in order to optimize operation of its energy supply systems. In this paper, HCMAC (Hyperball CMAC) neural networks are used to build load prediction models of buildings. The model inputs are outdoor meteorological parameters and the personnel distribution, and outputs cold / heat load and electricity load. A Novel fuzzy C-means clustering algorithm is proposed to overcome the drawback that the node number of HCMAC neural network increases exponentially with the increasing of input dimensions, effectively reducing the number of the network nodes, and decreasing the computational burden of neural network parameter learning. Load characteristics of a building are analyzed applying software TRNSYS, and the simulating operation data used for building load models are obtained. Simulation results demonstrated that the presented building load prediction method is an effective data-driven method to be universally applied to modeling of buildings.