Research On The Optimization Design Of Office Building Surfaces In Cold Regions Based On Machine Learnin

With the continuous development of China’s economy and the improvement of people’s living standards,energy supply and demand are becoming increasingly tense.In order to effectively reduce energy consumption,the implementation of energy-saving and environment-friendly energy solutions has become an important issue to be considered in the14 th Five-Year Plan period.The proportion of energy consumption in the construction industry is increasing in all industries,so it is important to carry out energy-efficient building design.The study explores office buildings in cold regions as an object to optimize the building skin for sustainable development,and the study covers the following aspects:First,the article conducted a detailed research and analysis on the types of office buildings and skin forms,determined the types of buildings and skin forms to be studied,and also conducted a systematic research on the range of architectural parameters and skin form parameters involved in the research object,and obtained the parameter optimization intervals for buildings and skins,so that the whole optimization process meets the actual situation and can be carried out for practical projects.The whole optimization process meets the actual situation and can be applied in practical projects.Secondly,the optimization of the building and the skin is carried out,focusing on the building skin optimization,using a number of techniques such as parametric modeling,performance simulation,machine learning,multi-objective optimization and objective decision making.The parametric modeling part relies on the building morphology model,and obtains equidistant points by dividing the four sides equally,moving the points into lines,extruding the lines into surfaces,and pushing out the surfaces into bodies to form a rectangular mosaic building skin,and modeling the four southeast and northwest sides separately to realize independent control of each orientation of the skin model.The performance simulation part simulates each orientation separately,and in order to carry out effective operation,the experiment divides the simulation object into four light and heat partitions,and the four regions form a trapezoid with the traffic nucleus as the upper edge,and carry out simulation of lighting and indoor thermal comfort respectively.The machine learning part and the multi-objective optimization part compare the most commonly used supervised machine learning prediction models SVM support vector machine,ANN neural network and BP neural network to find a prediction model with high accuracy,high speed and high adaptability,and use multi-objective optimization to optimize the trained model instead of direct optimization in the conventional case.The objective decision part uses a decision tool to filter out the optimization solutions and finally gives the optimal solution.Finally,for the completed optimization experiments,corresponding strategies are proposed for prediction model selection,skin parameter design and performance target optimization,and the completed method and process are practiced based on the actual project to verify the feasibility and effectiveness of the process and method,and compare with the original scheme to demonstrate the technical advantages of the method,and apply the skin to obtain the desired performance index.The research adapts to the current development trend of green building,combines the outstanding problems in cold regions,and obtains the performance-driven building skin design method,using computer simulation optimization tools,machine learning methods and decision making methods,and proves its correctness and feasibility through practice,especially for the design of office building skins in cold regions,and proposes design strategies to maximize energy saving,improve the light and heat environment,and these strategies can provide guidance for the skin design of this type of building and contribute to the realization of sustainable development.At the same time,an integrated platform operation process is built to integrate modeling-simulation-prediction-optimization in Grasshopper platform,which solves the drawbacks of inefficient and problematic multi-platform collaboration,and integrates machine learning prediction models into Grasshopper platform to maximize the elimination of platform barriers and improve optimization efficiency.