The Data-driven Energy Consumption Prediction Modeling Of The Hospital Building

As a special public building with complex structure and various functions,hospital building has high comprehensive energy consumption and seasonal differences.With the popularization of hospital energy monitoring systems,accurate prediction of energy consumption has gradually become one of the measures for refined management of hospital energy.Carrying out building energy management and energy consumption prediction has become the key research content of hospital building energy conservation.Therefore,taking a large-scale general hospital in Shanghai as the research object,the method of energy auditing is used to analyze the energy consumption of the hospital’s buildings,and based on this,the optimization suggestions for energy-saving renovation are put forward;the cleaning and preprocessing of hospital building data,the characteristics and influencing factors of building energy consumption,and the comparison and application of hospital building energy consumption prediction based on four machine learning algorithms are completed.This study adopts the method of energy auditing for analysis.The results show electricity is the main energy type.The power consumption of the hospital during the cooling period is high,accounting for about 45%of the total annual power consumption;the heating period is dominated by gas consumption,which is 2 to 4times the gas consumption during the non-heating period.The comprehensive energy consumption per unit area of the building is 25.64kgce/m~2,while the power consumption per unit area is 133.18k Wh/m~2,and the comprehensive energy consumption per patient is 1.13kgce/person and 5.86k Wh/person.Through the analysis of the energy consumption characteristics of the hospital,the energy saving optimization of the hospital can be carried out from three aspects:the energy-using equipment,the energy-using equipment control system and the hospital energy management system.Based on the Pandas library,boxplots,histograms and mean substitution methods,the data preprocessing is completed,and the energy consumption characteristics and operating laws of the hospital are mined.The research results show that the change trend of energy consumption between working hours and non-working hours of the hospital is inconsistent,and the hourly energy consumption operation state has seasonal differences;The daily total energy consumption is affected by the fluctuation of outdoor temperature,and the total energy consumption of hospital buildings in non-working days is slightly lower than that of working days;the factors affecting hospital energy consumption are determined as the energy consumption of sub-systems,outdoor temperature,whether it is working time,whether it is working day.Based on the determination of the influencing factors of hospital energy consumption,BP neural network,least squares support vector machine,random forest and multiple linear regression machine learning algorithm are used to complete the prediction model of hospital building.The outdoor temperature of the building,the total historical energy consumption in the first hour of the building,the number of 24hours,whether it is working time and whether it is a working day are used as input parameters,and the prediction model is optimized and selected by hyperparameters to determine the energy consumption prediction suitable for hospital buildings model.Using the above model to predict the hospital’s energy consumption and carry out the application research of the model.The results show that the random forest model is more suitable for the prediction of hospital building energy consumption in summer,and the long-term and short-term prediction errors are 3.93%and 2.99%respectively.The artificial neural network model is suitable for energy consumption prediction in winter and transition season.The long-term and short-term prediction errors in winter are 2.18%and 2.16%respectively,and the long-term and short-term prediction errors in transition season are 3.31%and 3.15%.The integrated design of hospital prediction models can more fully invoke the prediction capabilities of different machine learning models to improve performance.In addition,research on hospital building data analysis and predictive modeling can expand the application capabilities of the hospital energy efficiency management cloud platform,and provide hospitals with intelligent operation and maintenance management and operation optimization scheduling services.