Adaptive building day-ahead load prediction method based on transfer learning

Abstract

The invention discloses a self-adaptive building day-ahead load prediction method based on transfer learning, and relates to the technical field of building and environmental protection. The method comprises the following steps: S1, data acquisition and processing: dividing an original data set into a small data set of a target building and a big data set of a basic building group, and filling missing values of all the original data sets; s2, clustering energy consumption modes; s3, source domain data screening: screening a historical daily load curve of a load target building energy consumption mode, and respectively constructing a data migration training set and a model migration training set; s4, constructing a day-ahead load prediction model; and S5, adaptive model optimization: continuously adjusting model parameters by using Bayesian optimization to realize adaptive load prediction of the target building. According to the method, load prediction of the target building is realized through data migration and model migration methods of migration learning in combination with historical data of the building group with sufficient data.

Description

Technical field [0001] The present invention belongs to the field of building technology, for intelligent buildings, building energy management and other related systems, relates to adaptive building day load prediction method based on transfer learning. Background [0002] Building energy consumption in the total social energy consumption accounted for an increasing proportion, with the development of science and technology, buildings gradually from simple energy terminals to small, complex energy conversion equipment. Intelligent systems such as building energy management, intelligent buildings, and user behavior guidance effectively improve the efficiency of building energy consumption through rational allocation of energy consumption and energy storage equipment. The premise for the optimal management of building energy is accurate load forecasting. As a new model in recent years, deep learning algorithms have better nonlinear description capabilities and can effectively impr...

AI Technical Summary

Problems solved by technology

However, deep learning algorithms often require a large amount of historical data as support, otherwise it is very easy to fail to converge

However, in fact, due to the short running time and unsound data acquisition system of most buildings, they often cannot provide sufficient historical data to establish predictive models

Images