Research On The Application Of Winter Thermal Comfort Model In Cold Region College Classrooms Based On Machine Learning

College education has become an indispensable part of education,and the classroom is the main place for college teaching activities.In colleges and universities in cold regions,a large part of classrooms use water as the heating medium for heating in winter.The advantage of this heating method is that the heat source is stable and can create a theoretical thermal comfort zone.However,according to field research,this heating method satisfies the existing thermal comfort evaluation index,but deviates from the evaluation of actual users.Therefore,how to construct an evaluation index model that fits the user,and then guide the subsequent heating adjustment has become the focus of this thesis.This paper adopts the methods of literature analysis,sampling analysis,mathematical statistics,field measurement,questionnaire survey and computer simulation.The related knowledge of thermal comfort evaluation index model is sorted out by literature analysis method:(1)The characteristics of each thermal sensation evaluation index model;(2)The characteristics and basic process of thermal engineering experiments;(3)The application of machine learning methods in the field of thermal comfort prediction How to use.Since thermal experiments need to meet the characteristics of small temperature fluctuations,small changes in personnel,and small changes in personnel activities,the overall temperature,personnel and personnel activities of the target building are measured through field pre-investigation and sampling analysis methods before the formal experiment begins.distribution.The standard deviation and variance of the temperature,personnel,and personnel activities of the sampled classrooms in the pre-investigation building were calculated by mathematical statistics,so as to screen out two classrooms suitable for the research experiment.The two classrooms were pre-investigated by field measurement method,and the number of valid survey samples was estimated by mathematical statistics.After calculation,it was found that when the number of samples exceeded 120,the results were theoretically valid.Then,through field measurement,the air temperature,air humidity,airflow velocity,and black ball temperature of the target classroom were measured;relevant human factors were collected through questionnaires: clothing types,activities,estimated clothing thermal resistance,and human metabolic rate.The data collected above are preprocessed:(1)use the phase line method to eliminate outliers;(2)calculate the variance of the data itself and calculate the Spearman coefficient between the features to eliminate redundant features.Use machine learning methods to model the collected data.The modeling methods used are: logistic regression model,decision tree model,neural network model,and ensemble learning model.The LightGBM method in ensemble learning performed well in the end by comparing the average accuracies to screen out the prediction models that performed well.Use LightGBM to model and save and design the deployment method.By comparing the LightGBM model with the PMV-PPD model on the collected data,it is found that the LightGBM model has relatively high scene adaptability.He also proposed his own conjectures on the possible scenarios: 1.Comparing the operating conditions of air conditioners guided by the two models;2.Proposing possible design process conjectures on how to guide the architectural design.The following conclusions are drawn from the above method:(1)The thermoneutral temperature calculated based on the actual vote value(TSV)collected in this article is 23.8°C,and the thermoneutral temperature range is[19.3°C,28.2°C];the thermoneutral value calculated based on the predicted value(PSV)The temperature is 24.2°C,and the thermoneutral temperature range is [20.6°C,27.9°C];the thermoneutral temperature calculated according to the PMV index is26.6°C,and the thermoneutral temperature range is [22.1°C,31.1°C];(2)While changing the temperature in the classroom,we should also consider optimizing the humidity,so as to improve the thermal comfort in the classroom;(3)The main way that the outdoor environment affects the thermal comfort of the human body is to affect the thermal sensation of the human body,which has little effect on the thermal response ability of the human body;(4)The integrated model LightGBM performs relatively well in wide applicability and thermal comfort prediction,so model fusion is an important idea for machine learning to improve the effect;(5)The PMV model cannot adjust the parameters according to the actual adaptability of the human body,which is the main reason that the actual prediction effect is lower than the original design expectation.