Prediction And Analysis Of Energy Consuming Equipment In Buildings Based On Measured Data

The problem about lack of energy has been bothering people all the time. Recently, in China, the energy consumption in public buildings is increasing sharply especially electricity consumption which accounts for 22% of the total. Among this, heating(or cooling) equipment consumption occupies 50%-70% which generates during the equipment operates. Generally, a heating(or cooling) equipment's operation is consisted of cycles. Each cycle is divided into runtime and idle states. The runtime process is further decomposed into three phases: startup, stable operation and shutdown. Aiming at analyzing the performance of heating equipment's operations, this research studies the cyclic characteristics of typical gas burning furnaces and the idling time is chosen to be the output of the predicting model. As a result, the study can make a contribution to equipment diagnostics and reducing the power consumption in buildings. The main achievements and methods are summarized as follows.(1) By analyzing measured usage data and meteorological data of surrounding environment comprehensively, a program was designed to realize the coupling of the two automatically. By which, the power consumption data in every minute and related weather data during a special time can be collected. And on this basis, the equipment operating cycles is able to be gleaned. All these lay a solid foundation for the later prediction.(2) The idling time was chosen to be the predicted target on the base of analyzing the measured data. What's more, four Data Mining Algorithms(DMAs) including k-Nearest Neighbors(KNN), Naive Bayes(NB), Support Vector Machines(SVM) and Artificial Neural Networks(ANN) are used to analyze the relationship between idling time and weather conditions under No.3 condition. The obtained results show that SVM and ANN provide more accurate predictions of idling time so SVM algorithm and ANN algorithm was chosen to forecast the equipment performance. In the process, the cycle data was divided into training set and prediction set which were treated by using the method of nonlinear normalization before data mining. In addition, the prediction model parameters were decided through experiment and cross validation. At last, the optimal prediction model under different conditions was able to be decided on the base of error analysis.(3) Select a certain density of the three-dimensional model for grid subdivision and collect the temperature of wall by using Infrared camera and position it to the building model. And then, obtain the temperature of blocks at center based on inverse distance weighted interpolation algorithm by using MATLAB. Eventually, an indoor 3D temperature field model could be successfully built. Furthermore, according to the model and thermal comfort theory, the performance of the equipment was evaluated at the macro level.