Evaluation And Analysis Of Hospital Building Energy Based On Data Mining

An ocean of energy consumption data have been collected in energy audit of hospital buildings, which usually imply abundant information and knowledge. On the premise of energy statistics’ completion, this paper proposed the evaluation and analysis methods for hospital building energy based on data mining.Firstly, an evaluation method for end-user energy consumtion in hospital building was developed. The evaluation consists of overall evaluation and subentry evaluation, which are both based on k-means clustering algorithm. In overall evaluation, the energy consumtion amounts of end-users are evaluated, and the distinction of end-users behaviours are mined. In subentry evaluation, the subentry consumtions with energysaving potential are detected.Secondly, the analysis methods of hospital building energy were presented. In the analysis of energy consumption characteristics, 8 different parameters were summarized to indicate the general information of energy cosumption curves. In the analysis of energy-saving potential, two calculation methods were proposed: one is copmpared with historical data; the other is compared with the best end-user. In energy anomaly analysis, two different methods were put forward: one is used to find energy anomaly patterns through clustering algorithm; the other is able to make judgement on the reason of energy anomaly by artificial neural network(ANN).Finally, the developed methods were applied to case studies of a hospital in Harbin. The hospital departments’ energy consumption in February of 2014 was evaluated, 23 departments with high overall consumtion were mined, and the energy-saving directions for them were provided. Moreover, the electricity saving potential was calculated based on the evaluation results. Also, the characteristic of 3rd disinfection station’s steam consumption on November 27, 2013 was analysed, and the potential of load peak valley regulation was discovered. In addition to these, 6 typical energy anomaly patterns were mined with the anyasis method based on clustering; meanwhile, with the method based on ANN, the correct judgements were made on the reason of energy anomaly.