Energy Consumption Prediction Model And Energy Saving Strategy Research Of A University Library In Northwest China

As an important component of university buildings,library provides learning space for students.Because of a large flow of students and various forms of energy consumption(energy consumption from the lighting system,air conditioning system etc),it has great energy efficiency potential.The university library occupies only 5% of the building area but consumes 12% of the energy,which is one of the university buildings with the highest energy consumption per unit area.Therefore,accurate prediction of the energy consumption is critical to save energy under the premise of ensuring users' basic needs.The library building of a university in Northwest China is taken as the research object,and the findings are as follows:(1)The main factors that affect library building energy consumption are determined.Occupancy and energy consumption data was analyzed on a daily and monthly scale to determine the most significant factor,and to pave the way for later establishing prediction models and making energy conservation strategies.Occupancy and energy consumption data suggests that occupancy,when and what part of the library is open to students,and the operating status of air conditioning can all significantly affect energy consumption.(2)The electric energy prediction model of the library is built.Because the influence of different factors on the energy consumption of the library is often non-linear and coupled,and out of the requirements for prediction accuracy and speed,a LM neural network energy consumption projection model was established to project the energy consumption of the library building.The performance of the model was evaluated using MAPE,MSE and R-squared as perimeters,and traditional BP models as comparison.And LM has been shown to have some advantages over traditional BP models.(3)The LM prediction model of library energy consumption is optimized.Considering their differences in energy consumption profiles,separate prediction models were built for semesters and summer breaks.Building energy consumption is affected by an array of factors,forming distinct daily energy consumption profiles.Similar days were thus selected within a profile by soft clustering.The LM neural network optimized by the aforementioned modifications showed improved accuracy and had fewer outliers.(4)Based on the characteristics of the library and the data collected from it,two potential energy conservation strategies are proposed from a management standpoint and a technological standpoint,respectively.The management standpoint mainly concerns improvements on when and what part of the library should be open to students,and technological standpoint chiefly discusses improvements on illumination and air conditioning systems.The mean absolute percentage error,mean square error and fitting degree of the optimized prediction model are 10.81%,19.87 and 0.9525,respectively,indicating that the trained model was more accurate and had fewer outliers.At the same time,the energy saving strategy can provide basis and guidance for the optimal operation of library energy saving.