The Research Of Indoor Temperature Simulation Based On Multi-factor Grey Neural Network

The core principle of green architectures is to reduce building energy consumption without affecting the comfort of building occupants.And design and optimal control of indoor thermal environment is one of the main directions in the field of green architectures' research.As an important part of the design and evaluation of indoor thermal environment,the indoor temperature of the building is not only highly correlated with the evaluation level of building occupant's thermal comfort,but also an important factor to be considered in the control of building energy consumption.The simulation of building room temperature can provide an important reference for building thermal environment evaluation,regulation and energy saving.The combination of various mechanisms causes a complex nonlinear relationship between the indoor temperature and the surrounding environmental factors,and the changing of indoor temperature is characterized by slowness,hysteresis,and vulnerability to external influences,make it very difficult to completely factors analysis simulation to such question.At present,the analysis of building room temperature changes mainly includes two major directions: mechanism model analysis and data-driven model analysis.In practical applications,the mechanism model has key factors and data missing problems in the application of complex mechanisms,while the data-driven model has poor adaptability or insufficient simulation accuracy.Based on the idea of data-driven,this paper constructs three types of building room temperature simulation models.They are respectively the building room temperature simulation model based on the improved gray prediction OGM(1,N),the building room temperature simulation model based on Elman neural network,and the building room temperature simulation model constructed by combining gray prediction and neural network.The mathematical structure of the building room temperature simulation model based on the combination of grey prediction and neural network has both the superior processing ability of the gray system for poor information and the adaptability of the neural network system.The theoretical model of building room temperature simulation constructed in this paper has theoretical innovations in dealing with the poor information problem caused by insufficient input conditions in traditional mathematical models,and has the reality significance for exploring the relationship between limited engineering environmental factors and indoor temperature.The case study,taking an office building in Hangzhou as an example,measured the indoor temperature and outdoor climate parameters.Based on the measured data,the three room temperature simulation models constructed in this paper were used to predict the room temperature.The model was validated from the actual results to simulate the indoor temperature under small data conditions,and it is feasible to explore the potential relationship between indoor temperature and various environmental factors.The results of the measured data in the case study show that the gray prediction-neural network combination models have the highest fitness and accuracy in these three building room simulation models constructed in this paper.They have better adaptability and robustness to data changes.It can give a more accurate prediction of the short-term changes of the indoor temperature in the actual building under small data conditions,and provide reference for the design of building thermal environment and optimization of energy consumption control strategy.