Energy Consumption Prediction And Model Optimization Of Residential Buildings Based On Ensemble Learning

Building energy conservation is an indispensable step in the process of energy saving and emission reduction.The accurate prediction of building energy consumption plays a good guiding role for building energy optimization and building energy management.The data-driven energy consumption prediction model has higher precision and shorter time.Such advantages have been widely promoted in the energy field.In this thesis,a residential building with heating station for winter heating is taken as the research object.Each single machine learning algorithm and integrated learning algorithm are used,according to the data collected by the data acquisition system,the data collected in the weather station,and the total energy consumption of the community.An energy consumption prediction model has been established,combining the measures of outlier processing,feature selection and grid search to improve the accuracy of the energy consumption prediction model.The final building energy consumption prediction model needs fewer feature variables while having higher accuracy,and improving the generalization ability of the model.The original data set contains unit operation data,weather station acquisition data,time data and energy consumption data.The scatter plot and box plot are used to detect outliers.The median fill method is used to replace the outliers.The maximum information coefficient between the variables is used to judge the correlation of each variable,and then the Bortua algorithm is used to judge the importance of each variable and select the best input characteristic variable set.The feature extracted data is divided into training set and test set.The energy consumption prediction model is established by using multiple linear regression algorithm,extreme learning machine algorithm,limit gradient descent algorithm and support vector regression.Cross-validation and network search methods for parameter optimization.Finally,the integrated learning algorithm is used to combine the four single machine learning models to improve the prediction accuracy.In the result,the average absolute error of the test set decreases by 4.36% ~ 71.70%,and the root mean square error decreases by 3.80% ~ 49.73%.In view of the residential building characteristics,the new historical energy consumption characteristic variable EWMA is added to the input variable set of the integrated learning model.The results show that the new characteristic variable set can effectively improve the accuracy of the prediction model,and the average absolute error of the test set is compared with the original characteristic variable set.The reduction was 10.36% and the root mean square error was reduced by 19.89%.