Designing an Expert Climate Improvement System with Pattern Finding of Cold Climate Residential Buildings based on Learning Models

Purpose: With the increasing growth of energy consumption in the modern world and the position of the construction industry as one of the major factors contributing to the creation and exacerbation of this crisis, a complex issue known as “energy supply and consumption management” has emerged. This issue, in addition to affecting the sustainable development of societies, is of particular importance due to its significant impact on the healthy lives of human beings. The present study aims to design a decision‑making system for the maximum utilization of solar energy (passive) in supplying the energy required by residents of a residential building (information entity), while considering the minimum amount of energy loss. This objective is pursued based on the appropriate design of the building’s outer shell without the need to install additional electrical equipment.
Method: The research method used in the present study is a combination of descriptive, experimental, analytical, and comparative approaches. It also possesses characteristics such as being indigenous, relying on Iranian data and regulations, and having the capability to be developed based on climatic data. The integrated research approach is influenced by the concepts of knowledge management, decision‑making systems, expert system rule construction, and the application of learning models in the specialized field of residential building architecture, with the aim of addressing a real‑world problem.
In order to construct the required knowledge set, characteristics related to the site of the proposed project—particularly geographical coordinates—were examined. In addition, characteristics related to the building shell, including building dimensions and openings, as well as characteristics associated with energy loss, were investigated as criteria and sub‑criteria for information production. In line with
the aforementioned objective and considering the environmental comfort of residents in a climate with high heating demands (Zanjan city, located in a cold climate), a technical investigation of the issue was conducted by modeling the condition of residential buildings in ten districts of Zanjan city. This analysis included 100 buildings selected through a simple random sampling method. Subsequently, the potential for improving the climatic performance of the building envelope was assessed from the perspective of three passive solar strategies: direct gain, Trombe wall, and solar space (solar greenhouse). For the desired modeling process, a dataset containing more than 3,500 information fields was used as a statistical sample representing the population. To collect the required data, a combination of field data collection methods, review of descriptive information layers in the detailed urban plan, and information obtained from aerial maps of selected residential plots was employed.
The validity criterion for the output status of the models was calculated based on the climate energy label criterion, with the aim of evaluating the ability of the residential building form to receive solar energy. The complexity of the problem was managed through a multi‑stage process that included analyzing climatic conditions in terms of temperature variations and solar radiation changes, examining fixed urban development characteristics centered on the site plan, and utilizing the relevant decision matrix.
Achieving the lowest level of energy loss (through thermal transfer and heat dissipation) together with the highest solar absorption—based on the implementation status of passive solar systems—was defined as the main operational criterion through the introduction of a new concept called the climate energy label (for generating training and testing data). Subsequently, three learning models—decision tree, association rules, and Bayesian theory—were applied and analyzed in order to produce the rules required for designing the knowledge base of the proposed decision‑making expert system.
Findings: An analysis of the feasibility of implementing passive solar energy strategies indicated that among the ten districts studied in the city of Zanjan (a cold climate region), the implementation of the solar greenhouse method had the highest priority. This was followed by the direct gain method and the Trombe wall method. In addition to prioritizing features within the structured resolution of the problem space and exerting a positive influence on the forward inference chains used in the inference engine of the proposed expert system, the research findings demonstrated appropriate performance, with a confidence factor exceeding 85 percent. Considering the high complexity of the problem, this result indicates the acceptable accuracy of the prediction models and also implicitly guarantees the reliability of the rules derived from them in forming the inference database.
Conclusion: The results of this research can serve as an auxiliary knowledge system for improving existing residential buildings and can support decision‑making processes for planning and policymaking organizations. Furthermore, the findings can contribute to generating design recommendations for addressing the complex problem of existing energy imbalances in residential buildings.