Dynamic heat consumption benchmarking and carbon reduction potential mining of heating stations based on improved decision tree and association rule mining

The dynamic heat consumption benchmark is an important basis for the management and regulation of heating stations in. Existing benchmarks are generally based on uniform static values for climate zones. However, due to variations in building insulation, heating demand, and system aging across different heating stations, energy consumption benchmarks may be different. Establishing dynamic benchmarks for each heating station is costly and challenging. This study introduces a data-driven approach to calculate the dynamic energy consumption benchmarks for heating stations and evaluate their carbon reduction potential. The study proposes a three-step feature selection method, including filtering, embedding, and wrapping techniques, to identify the key features that influence energy consumption. An improved decision tree classification model for heating stations is developed and optimized through Bayesian and cost-complexity post-pruning. The dynamic energy consumption benchmarks for each type of heating station are established using Apriori association rule mining. The carbon reduction potential during the operation of heating stations is evaluated based on the carbon emission factor method. Through case analysis, the studied heating stations are categorized into 13 types. The annual energy consumption benchmark range for each level of heating stations is 0.16–0.43 GJ/(m2·a), with a maximum carbon reduction potential reaching 1.53 kgCO2/(m2·a).