WO2020056812A1 - 用于评价室内环境质量的环境参数权重确定方法及*** - Google Patents

用于评价室内环境质量的环境参数权重确定方法及*** Download PDF

Info

Publication number
WO2020056812A1
WO2020056812A1 PCT/CN2018/109858 CN2018109858W WO2020056812A1 WO 2020056812 A1 WO2020056812 A1 WO 2020056812A1 CN 2018109858 W CN2018109858 W CN 2018109858W WO 2020056812 A1 WO2020056812 A1 WO 2020056812A1
Authority
WO
WIPO (PCT)
Prior art keywords
weight
data
environmental
environmental parameter
model
Prior art date
Application number
PCT/CN2018/109858
Other languages
English (en)
French (fr)
Inventor
卞春
孙宝石
曹石
Original Assignee
苏州数言信息技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 苏州数言信息技术有限公司 filed Critical 苏州数言信息技术有限公司
Publication of WO2020056812A1 publication Critical patent/WO2020056812A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01DMEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00Measuring or testing not otherwise provided for
    • G01D21/02Measuring two or more variables by means not covered by a single other subclass

Definitions

  • the invention belongs to the technical field of indoor environmental quality monitoring, and particularly relates to a method for determining environmental parameter weights for evaluating indoor environmental quality, and a weight determination system.
  • the invention provides a method for determining environmental parameter weights for evaluating indoor environmental quality, which fills a gap in the current industry, automatically generates environmental parameter weights for evaluating indoor environmental quality, and meets requirements for operation stability and accuracy.
  • the present invention provides an environmental parameter weight determination method for evaluating indoor environmental quality, including:
  • the environmental parameters include, but are not limited to, temperature, humidity, illuminance, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise;
  • the method further includes obtaining the weight model through model training.
  • the training process is:
  • the weighted model is obtained by training the characterized data samples based on a related algorithm.
  • the training process that further includes the weight model further includes:
  • the optimal weight model is used to perform weight analysis on the characteristics of the environmental parameters to obtain the weight of the environmental parameters.
  • the matching degree of the optimal weight model is not less than 95%.
  • further related algorithms for training to obtain weight models include, but are not limited to, machine learning algorithms, convolutional neural network algorithms, recurrent neural network algorithms, decision trees, and Bayesian decision theory-based Classification algorithms and deep learning algorithms.
  • the data sample in the current test data group is merged into the training data group, and the new data sample enters the test data group.
  • the present invention also provides an environmental parameter weight determination system for evaluating indoor environmental quality, including:
  • a data acquisition module for collecting indoor environmental parameter data includes, but are not limited to, temperature, humidity, illumination, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise;
  • Feature extraction module which is used to extract the environmental parameter data in combination with the use scenario of the indoor environment to obtain the environmental parameter characteristics
  • the weight analysis module uses a weight model to perform weight analysis on the characteristics of the environmental parameters to obtain the weight of the environmental parameters.
  • a weight model training module that trains environmental parameter data samples to obtain the weight model, including:
  • a data sample collection unit configured to collect data samples of indoor environmental parameters, including, but not limited to, temperature, humidity, illumination, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise;
  • a feature extraction unit configured to perform feature extraction on a sample of environmental parameter data to obtain a characterized data sample
  • a model training unit is used to train a characterized data sample using a related algorithm to obtain a weight model.
  • the weight model training module further includes a data grouping unit, which randomly groups the characteristic data samples into a training data group and a test data group; the model training unit The data samples in the training data set are trained to obtain a weight model; the data samples in the test data are used to test and verify the matching degree of the weight model.
  • the model training unit further uses several related algorithms to separately train the data samples in the training test group to obtain several weight models;
  • the model training unit uses the data samples in the test data set to test and verify the performance of the several weight models, and selects the one with the highest matching degree among the several weight models as the optimal weight model;
  • the weight analysis module uses the optimal weight model to perform weight analysis on the characteristics of the environmental parameters to obtain the weight of the environmental parameters.
  • the method for determining the weight of environmental parameters for evaluating the quality of an indoor environment of the present invention obtains indoor environment parameter data and performs feature extraction, and characterizes the extracted environmental parameters, and uses a weight model to perform weight analysis on the characteristic environmental parameter data to obtain an environment.
  • the weight of the parameter Both operational stability and accuracy meet requirements, filling gaps in the current industry.
  • FIG. 1 is a structural block diagram of a weight determination system in a preferred embodiment of the present invention.
  • FIG. 3 is a flowchart of training a weight model.
  • this embodiment discloses an environment parameter weight determination system for evaluating indoor environmental quality, particularly an environment parameter weight determination system for evaluating indoor environmental quality, a data acquisition module, and feature extraction.
  • Module weight analysis module and weight training module.
  • the weight training module trains the environmental parameter data samples used as samples to obtain the weight module.
  • the weight model obtained through training is used to perform weight analysis on the actual environmental parameter data, and directly output the weight corresponding to the environmental parameter.
  • the weight model training module includes a data sample collection unit, a feature extraction unit, a data grouping unit, and a model training unit.
  • a data sample acquisition unit is used to collect indoor environmental data samples.
  • Environmental parameters include, but are not limited to, temperature, humidity, illuminance, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise.
  • Collect various environmental parameter data in the current state through various sensors distributed in the environment, for example, the temperature sensor collects the current temperature of 28 ° C, the humidity sensor collects 47% of humidity, etc. Data samples of various environmental parameters in the current state.
  • a feature extraction unit is used for feature extraction of environmental parameter data samples in combination with an indoor environment usage scenario to obtain a characterized data sample.
  • abnormal data samples and non-working time data samples are excluded.
  • the collected data is abnormal (far greater than or far less than the normal value), or "-" is output when the power is off, and this abnormal data sample is eliminated by the feature extraction unit.
  • the time of an indoor environment is normally from 9:00 to 17:00, and data samples other than this working time are eliminated by a feature extraction unit.
  • a data grouping unit that randomly divides the characterized data samples into a training data group and a test data group. Specifically, 80% of all the data samples after the characterization are classified as a training data group, and 20% are classified as a test data group.
  • a model training unit is configured to use a related algorithm to train data samples in the training array to obtain a weight model.
  • related algorithms include, but are not limited to, machine learning algorithms, convolutional neural network algorithms, recurrent neural network algorithms, decision trees, classification algorithms based on Bayesian decision theory, and deep learning algorithms.
  • the model training unit uses the above several related algorithms to separately train data samples to obtain several weight models.
  • the data samples in the test data set are used to test and verify the performance of several weight models.
  • the one with the highest matching degree is selected as the optimal weight model.
  • the data samples in the test data set are imported into each weight model to see whether the weights output by each weight model match the actual situation, and the weight model with the highest matching degree is selected as the optimal weight model.
  • the matching degree of the optimal weight model is not less than 95%. After testing and verification, if the matching degree of all weight models is less than 95%, the weighting model with the highest matching degree is iteratively optimized until the matching degree is not less than 95%.
  • the optimal weight model After obtaining the optimal weight model through training and screening, use the optimal weight model to perform weight analysis on the environmental parameters in the current state, and obtain the weight of each environmental parameter in the current state, which includes the data acquisition module, feature extraction module, and weight analysis. Module.
  • a data acquisition module is used to collect indoor environmental data; environmental parameters include, but are not limited to, temperature, humidity, illumination, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise. Collect various environmental parameter data in the current state through various sensors distributed in the environment, for example, the temperature sensor collects the current temperature of 28 ° C, the humidity sensor collects 47% of humidity, etc. Data of various environmental parameters in the current state.
  • a feature extraction module is used to extract environmental parameter data in combination with an indoor environment usage scenario to obtain environmental parameter characteristics. Specifically, in combination with an indoor environment usage scenario, the normal environmental parameter data and the non-working time environmental parameter data are excluded. For example, when the sensor is in a fault or power failure state, the collected data is abnormal (much larger than or far less than the normal value), or "-" is output when the sensor is powered off, and this abnormal environmental parameter data is eliminated by the feature extraction module. In another case, for example, the indoor environment time is 9: 00-17: 00 under normal circumstances, and the environmental parameter data other than this working time is eliminated by the feature extraction module.
  • a weight analysis module which uses the optimal weight model obtained through training and screening to perform weight analysis on the characteristics of the environmental parameters to obtain the weight of the environmental parameters. Specifically, the characteristic environmental parameter data is imported into the optimal weight model, and the optimal weight model performs weight analysis on the environmental parameter data to directly output the weight of each environmental parameter.
  • this embodiment discloses a method for determining an environmental parameter weight for evaluating indoor environmental quality, including:
  • environmental parameter data include, but are not limited to, temperature, humidity, illuminance, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise.
  • environmental parameter data include, but are not limited to, temperature, humidity, illuminance, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide, and noise.
  • Collect various environmental parameter data in the current state through various sensors distributed in the environment, for example, the temperature sensor collects the current temperature of 28 ° C, the humidity sensor collects 47% of humidity, etc. Data of various environmental parameters in the current state.
  • the normal environmental parameter data and the non-working time environmental parameter data are excluded.
  • the collected data is abnormal (much larger than or far less than the normal value), or "-" is output when the sensor is powered off, and this abnormal environmental parameter data is eliminated by the feature extraction module.
  • the indoor environment time is 9: 00-17: 00 under normal circumstances, and the environmental parameter data other than this working time is eliminated by the feature extraction module.
  • the foregoing weight model is obtained through training.
  • the training process is as follows:
  • Environmental parameters include, but are not limited to, temperature, humidity, illumination, color temperature, PM2.5, PM10, formaldehyde, TVOC, carbon dioxide and noise.
  • Collect various environmental parameter data in the current state through various sensors distributed in the environment, for example, the temperature sensor collects the current temperature of 28 ° C, the humidity sensor collects 47% of humidity, etc. Data samples of various environmental parameters in the current state.
  • the feature parameters of the environmental parameter data samples are extracted in combination with the use scenarios of the indoor environment to obtain the characterized data samples. Specifically, in combination with the use scenario of the indoor environment, except for normal data samples and non-working time data samples. For example, when the sensor is in a fault or power failure state, the collected data is abnormal (far greater than or far less than the normal value), or "-" is output when the power is off, and this abnormal data sample is eliminated by the feature extraction unit. In another case, for example, the time of an indoor environment is normally from 9:00 to 17:00, and data samples other than this working time are eliminated by a feature extraction unit.
  • the characteristic data samples are randomly divided into training data groups and test data groups. Specifically, 80% of all characteristic data samples are classified as training data groups and 20% are classified as test data groups.
  • the characteristic data samples are trained based on the correlation algorithm to obtain a weight model.
  • related algorithms include, but are not limited to, machine learning algorithms, convolutional neural network algorithms, recurrent neural network algorithms, decision trees, classification algorithms based on Bayesian decision theory, and deep learning algorithms.
  • the model training unit uses the above several related algorithms to separately train data samples to obtain several weight models.
  • the data samples in the test data set are used to test and verify the performance of several weight models. Among the several weight models, the one with the highest matching degree is selected as the optimal weight model. For example, the data samples in the test data set are imported into each weight model to see whether the weights output by each weight model match the actual situation, and the weight model with the highest matching degree is selected as the optimal weight model.
  • the matching degree of the optimal weight model is not less than 95%. After testing and verification, if the matching degree of all weight models is less than 95%, the weighting model with the highest matching degree is iteratively optimized until the matching degree is not less than 95%.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Testing Resistance To Weather, Investigating Materials By Mechanical Methods (AREA)

Abstract

一种用于评价室内环境质量的环境参数权重确定方法,包括获取环境参数数据;所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;结合室内环境的使用场景对所述环境参数数据进行特征提取,获得环境参数特征;使用经训练获得的权重模型对所述环境参数特征做权重分析,获得环境参数的权重。所述方法运行稳定性和准确性均达到较高的要求。

Description

用于评价室内环境质量的环境参数权重确定方法及*** 技术领域
本发明属于室内环境质量监测技术领域,具体涉及一种用于评价室内环境质量的环境参数权重确定方法,及权重确定***。
背景技术
测量室内环境质量的客观物理指标有很多,包括照度、色温、温度、湿度、PM2.5浓度、二氧化碳浓度等等。目前可以通过各种传感器测量这些物理指标,然后将各项指标的数值报告给相关信息使用者。这样做的问题在于相关信息使用者通常无法直观地理解这些物理指标与身心健康的关系。目前的室内环境质量指标缺少对环境质量的综合量化评价,缺少易于理解的指标。为了解决技术问题,业内正在尝试开发既能定性又能定量的评估方案来评估室内环境质量,将包括照度、色温、温度、湿度、PM2.5浓度、二氧化碳浓度等等的环境参数均囊括进来,每个环境参数对环境质量的影响程度不一样,为此需要对每个环境参数引入权重,如何确定环境参数的权重成为目前需要进一步突破的技术难题。
发明内容
本发明提供一种用于评价室内环境质量的环境参数权重确定方法,填补目前行业内的空白,自动生成用于评价室内环境质量的环境参数权重,运行稳定性和准确性均达到要求。
为了解决上述技术问题,本发明提供了一种用于评价室内环境质量的环境参数权重确定方法,包括,
获取环境参数数据;所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
结合室内环境的使用场景对所述环境参数数据进行特征提取,获得环境参数特征;
使用经训练获得的权重模型对所述环境参数特征做权重分析,获得环境参数的权重。
本发明一个较佳实施例中,进一步包括通过模型训练获得所述权重模型,其训练过程为,
获取用作样本的环境参数数据样本;
结合室内环境的使用场景对环境参数数据样本进行特征提取,获得特征化后的数据样本;
基于相关算法训练特征化后的数据样本,获得所述权重模型。
本发明一个较佳实施例中,进一步包括所述权重模型的训练过程还包括,
对特征后的数据样本进行随机分组,分为训练数据组和测试数据组,使用训练数据组中的数据样本进行模型训练获得所述权重模型;
使用若干个相关算法分别训练所述训练数据组中的数据样本,获得若干个权重模型;
使用所述测试数据组中的数据样本分别测试验证所述若干个权重模型的性能,在所述若干个权重模型中选取匹配度最高的一个为最优权重模型;
使用最优权重模型对环境参数特征做权重分析,获得环境参数的权重。
本发明一个较佳实施例中,进一步包括所述最优权重模型的匹配度不低于95%。
本发明一个较佳实施例中,进一步包括用于训练获得权重模型的相关算法包括但不局限于机器学习算法、卷积神经网络算法、循环神经网络算法、决策树、基于贝叶斯决策理论的分类算法和深度学习算法。
本发明一个较佳实施例中,进一步包括在有新的数据样本导入时,当前测试数据组内的数据样本合并至训练数据组中,新的数据样本进入测试数据组。
为了解决上述技术问题,本发明还提供了一种用于评价室内环境质量的环境参数权重确定***,包括,
数据采集模块,用于采集室内的环境参数数据;所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
特征提取模块,用于结合室内环境的使用场景对环境参数数据进行提取,获得环境参数特征;
权重分析模块,其使用权重模型对环境参数特征做权重分析,获得环境参数的权重。
本发明一个较佳实施例中,进一步包括其还包括权重模型训练模块,所述权重模型训练模块对环境参数数据样本进行训练,获得所述权重模型,其包括,
数据样本采集单元,用于采集室内的环境参数数据样本,所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
特征提取单元,用于对环境参数数据样本进行特征提取,获得特征化后的数据样本;
模型训练单元,用于使用相关算法训练特征化后的数据样本,获得权重模型。
本发明一个较佳实施例中,进一步包括所述权重模型训练模块还包括数据分组单元,所述数据分组单元将特征后的数据样本随机分组为训练数据组和测试数据组;所述模型训练单元训练训练数据组中的数据样本获得权重模型;所述测试数据中的数据样本用于测试验证所述权重模型的匹配度。
本发明一个较佳实施例中,进一步包括所述模型训练单元使用若干个相关算法分别训练所述训练测试组中的数据样本,获得若干个权重模型;
所述模型训练单元使用所述测试数据组中的数据样本分别测试验证所述若干个权重模型的性能,在所述若干个权重模型中选取匹配度最高的一个为最优权重模型;
权重分析模块使用最优权重模型对环境参数特征做权重分析,获得环境参数的权重。
本发明用于评价室内环境质量的环境参数权重确定方法,获取室内环境参数数据后进行特征提取,并对提取的环境参数特征化,使用权重模型对特征化后的环境参数数据做权重分析获得环境参数的权重。运行稳定性和准确性均达到要求,填补目前行业内的空白。
附图说明
图1是本发明优选实施例中权重确定***的结构框图;
图2是本发明优选实施例中权重确定方法的流程图;
图3是训练获得权重模型的流程图。
具体实施方式
下面结合附图和具体实施例对本发明作进一步说明,以使本领域的技术人员可以更好地理解本发明并能予以实施,但所举实施例不作为对本发明的限定。
实施例一
如图1所示,本实施例公开了一种用于评价室内环境质量的环境参数权重确定***,尤其是一种用于评价教室室内环境质量的环境参数权重确定***,数据采集模块、特征提取模块、权重分析模块和权重训练模块。权重训练模块对用作样本的环境参数数据样本进行训练,获得权重模块。使用经训练获得的权重模型对实际的环境参数数据做权重分析,直接输出环境参数对应的权重。
具体的,通过模型训练获得权重模型的过程如下:
如图3所示,权重模型训练模块包括数据样本采集单元、特征提取单元、数据分组单元和模型训练单元。
数据样本采集单元,用于采集室内的环境参数数据样本,环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音。通过分布在环境中的各种传感器采集当前状态下的各个环境参数数据,比如,温度传感器采集当前状态下的温度为28℃、湿度传感器采集的湿度为47%等等,以此来获取环境中当前状态下各个环境参数的数据样本。
特征提取单元,用于结合室内环境的使用场景对环境参数数据样本进行特征提取,获得特征化后的数据样本。具体的,结合室内环境的使用场景,剔除 非正常的数据样本和非工作时间的数据样本。比如,传感器处于故障或者断电状态时,采集的数据异常(远大于或者远小于正常值)、或者断电状态时输出“--”,通过特征提取单元剔除这种非正常的数据样本。另一种情况下,比如,正常情况下室内环境的时间是9:00-17:00,通过特征提取单元剔除这一工作时间以外的数据样本。
数据分组单元,其将特征化后的数据样本随机分为训练数据组和测试数据组。具体的,将特征化后的所有数据样本中的80%归为训练数据组,20%归为测试数据组。
模型训练单元,用于使用相关算法训练上述训练数组中的数据样本,获得权重模型。具体的,相关算法包括但不局限于机器学习算法、卷积神经网络算法、循环神经网络算法、决策树、基于贝叶斯决策理论的分类算法和深度学习算法。模型训练单元使用以上的若干个相关算法分别训练数据样本,获得若干个权重模型。并使用测试数据组中的数据样本分别测试验证若干个权重模型的性能,在若干个权重模型中选取匹配度最高的一个为最优权重模型。比如,将测试数据组中的数据样本导入各个权重模型中,看各个权重模型输出的权重是否与实际情况相匹配,选取匹配度最高的的权重模型为最优权重模型,本实施例技术方案中,限定最优权重模型的匹配度不低于95%。测试验证后,如果所有的权重模型的匹配度都低于95%,则对匹配度最高的权重模型进行迭代优化,直至其匹配度不低于95%。
需要注意的是,在有新的数据样本导入时,当前测试数据组内的数据样本合并至训练数据组中,新的数据样本进入测试数据组,以此不断优化权重模型。
通过训练、筛选获得最优权重模型后,使用最优权重模型来对当前状态下 的环境参数做权重分析,获得当前状态下各个环境参数的权重,其包括数据采集模块、特征提取模块、权重分析模块。
数据采集模块,用于采集室内的环境参数数据;环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音。通过分布在环境中的各种传感器采集当前状态下的各个环境参数数据,比如,温度传感器采集当前状态下的温度为28℃、湿度传感器采集的湿度为47%等等,以此来获取环境中当前状态下各个环境参数的数据。
特征提取模块,用于结合室内环境的使用场景对环境参数数据进行提取,获得环境参数特征。具体的,结合室内环境的使用场景,剔除非正常的环境参数数据和非工作时间的环境参数数据。比如,传感器处于故障或者断电状态时,采集的数据异常(远大于或者远小于正常值)、或者断电状态时输出“--”,通过特征提取模块剔除这种非正常的环境参数数据。另一种情况下,比如,正常情况下室内环境的时间是9:00-17:00,通过特征提取模块剔除这一工作时间以外的环境参数数据。
权重分析模块,其使用经训练和筛选获得的最优权重模型对环境参数特征做权重分析,获得环境参数的权重。具体的,将特征化后的环境参数数据导入最优权重模型中,最优权重模型对环境参数数据做权重分析直接输出各个环境参数的权重。
实施例二
如图2所示,本实施例公开了一种用于评价室内环境质量的环境参数权重确定方法,包括,
(1)获取环境参数数据;环境参数包括但不限于温度、湿度、照度、色温、 PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音。通过分布在环境中的各种传感器采集当前状态下的各个环境参数数据,比如,温度传感器采集当前状态下的温度为28℃、湿度传感器采集的湿度为47%等等,以此来获取环境中当前状态下各个环境参数的数据。
(2)结合室内环境的使用场景对环境参数数据进行特征提取,获得环境参数特征。具体的,结合室内环境的使用场景,剔除非正常的环境参数数据和非工作时间的环境参数数据。比如,传感器处于故障或者断电状态时,采集的数据异常(远大于或者远小于正常值)、或者断电状态时输出“--”,通过特征提取模块剔除这种非正常的环境参数数据。另一种情况下,比如,正常情况下室内环境的时间是9:00-17:00,通过特征提取模块剔除这一工作时间以外的环境参数数据。
(3)使用经训练获得的权重模型对环境参数特征做权重分析,获得环境参数的权重。
本实施例技术方案中,上述权重模型通过训练获得,如图3所示,其训练过程为:
(3.1)获取用作样本的环境参数数据样本。环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音。通过分布在环境中的各种传感器采集当前状态下的各个环境参数数据,比如,温度传感器采集当前状态下的温度为28℃、湿度传感器采集的湿度为47%等等,以此来获取环境中当前状态下各个环境参数的数据样本。
(3.2)结合室内环境的使用场景对环境参数数据样本进行特征提取,获得特征化后的数据样本。具体的,结合室内环境的使用场景,剔除非正常的数据 样本和非工作时间的数据样本。比如,传感器处于故障或者断电状态时,采集的数据异常(远大于或者远小于正常值)、或者断电状态时输出“--”,通过特征提取单元剔除这种非正常的数据样本。另一种情况下,比如,正常情况下室内环境的时间是9:00-17:00,通过特征提取单元剔除这一工作时间以外的数据样本。
(3.3将特征化后的数据样本随机分为训练数据组和测试数据组。具体的,将特征化后的所有数据样本中的80%归为训练数据组,20%归为测试数据组。
(3.4)基于相关算法训练特征化后的数据样本,获得权重模型。具体的,相关算法包括但不局限于机器学习算法、卷积神经网络算法、循环神经网络算法、决策树、基于贝叶斯决策理论的分类算法和深度学习算法。模型训练单元使用以上的若干个相关算法分别训练数据样本,获得若干个权重模型。并使用测试数据组中的数据样本分别测试验证若干个权重模型的性能,在若干个权重模型中选取匹配度最高的一个为最优权重模型。比如,将测试数据组中的数据样本导入各个权重模型中,看各个权重模型输出的权重是否与实际情况相匹配,选取匹配度最高的的权重模型为最优权重模型,本实施例技术方案中,限定最优权重模型的匹配度不低于95%。测试验证后,如果所有的权重模型的匹配度都低于95%,则对匹配度最高的权重模型进行迭代优化,直至其匹配度不低于95%。
需要注意的是,在有新的数据样本导入时,当前测试数据组内的数据样本合并至训练数据组中,新的数据样本进入测试数据组,以此不断优化权重模型。
(4)使用经训练、筛选获得的最优权重模型对环境参数特征做权重分析,获得环境参数的权重。具体的,将特征化后的环境参数数据导入最优权重模型 中,最优权重模型对环境参数数据做权重分析直接输出各个环境参数的权重。
以上所述实施例仅是为充分说明本发明而所举的较佳的实施例,本发明的保护范围不限于此。本技术领域的技术人员在本发明基础上所作的等同替代或变换,均在本发明的保护范围之内。本发明的保护范围以权利要求书为准。

Claims (10)

  1. 一种用于评价室内环境质量的环境参数权重确定方法,其特征在于:包括,
    获取环境参数数据;所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
    结合室内环境的使用场景对所述环境参数数据进行特征提取,获得环境参数特征;
    使用经训练获得的权重模型对所述环境参数特征做权重分析,获得环境参数的权重。
  2. 如权利要求1所述的用于评价室内环境质量的环境参数权重确定方法,其特征在于:通过模型训练获得所述权重模型,其训练过程为,
    获取用作样本的环境参数数据样本;
    结合室内环境的使用场景对环境参数数据样本进行特征提取,获得特征化后的数据样本;
    基于相关算法训练特征化后的数据样本,获得所述权重模型。
  3. 如权利要求2所述的用于评价室内环境质量的环境参数权重确定方法,其特征在于:所述权重模型的训练过程还包括,
    对特征后的数据样本进行随机分组,分为训练数据组和测试数据组,使用训练数据组中的数据样本进行模型训练获得所述权重模型;
    使用若干个相关算法分别训练所述训练数据组中的数据样本,获得若干个权重模型;
    使用所述测试数据组中的数据样本分别测试验证所述若干个权重模型的性能,在所述若干个权重模型中选取匹配度最高的一个为最优权重模型;
    使用最优权重模型对环境参数特征做权重分析,获得环境参数的权重。
  4. 如权利要求3所述的用于评价室内环境质量的环境参数权重确定方法,其特征在于:所述最优权重模型的匹配度不低于95%。
  5. 如权利要求3所述的用于评价室内环境质量的环境参数权重确定方法,其特征在于:用于训练获得权重模型的相关算法包括但不局限于机器学习算法、卷积神经网络算法、循环神经网络算法、决策树、基于贝叶斯决策理论的分类算法和深度学习算法。
  6. 如权利要求3所述的用于评价室内环境质量的环境参数权重确定方法,其特征在于:在有新的数据样本导入时,当前测试数据组内的数据样本合并至训练数据组中,新的数据样本进入测试数据组。
  7. 一种用于评价室内环境质量的环境参数权重确定***,其特征在于:包括,
    数据采集模块,用于采集室内的环境参数数据;所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
    特征提取模块,用于结合室内环境的使用场景对环境参数数据进行提取,获得环境参数特征;
    权重分析模块,其使用权重模型对环境参数特征做权重分析,获得环境参数的权重。
  8. 如权利要求7所述的用于评价室内环境质量的环境参数权重确定***, 其特征在于:其还包括权重模型训练模块,所述权重模型训练模块对环境参数数据样本进行训练,获得所述权重模型,其包括,
    数据样本采集单元,用于采集室内的环境参数数据样本,所述环境参数包括但不限于温度、湿度、照度、色温、PM2.5、PM10、甲醛、TVOC、二氧化碳和噪音;
    特征提取单元,用于对环境参数数据样本进行特征提取,获得特征化后的数据样本;
    模型训练单元,用于使用相关算法训练特征化后的数据样本,获得权重模型。
  9. 如权利要求8所述的用于评价室内环境质量的环境参数权重确定***,其特征在于:所述权重模型训练模块还包括数据分组单元,所述数据分组单元将特征后的数据样本随机分组为训练数据组和测试数据组;所述模型训练单元训练训练数据组中的数据样本获得权重模型;所述测试数据中的数据样本用于测试验证所述权重模型的匹配度。
  10. 如权利要求9所述的用于评价室内环境质量的环境参数权重确定***,其特征在于:所述模型训练单元使用若干个相关算法分别训练所述训练测试组中的数据样本,获得若干个权重模型;
    所述模型训练单元使用所述测试数据组中的数据样本分别测试验证所述若干个权重模型的性能,在所述若干个权重模型中选取匹配度最高的一个为最优权重模型;
    权重分析模块使用最优权重模型对环境参数特征做权重分析,获得环境参数的权重。
PCT/CN2018/109858 2018-09-21 2018-10-11 用于评价室内环境质量的环境参数权重确定方法及*** WO2020056812A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811107677.XA CN109297534B (zh) 2018-09-21 2018-09-21 用于评价室内环境质量的环境参数权重确定方法及***
CN201811107677.X 2018-09-21

Publications (1)

Publication Number Publication Date
WO2020056812A1 true WO2020056812A1 (zh) 2020-03-26

Family

ID=65163710

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/109858 WO2020056812A1 (zh) 2018-09-21 2018-10-11 用于评价室内环境质量的环境参数权重确定方法及***

Country Status (2)

Country Link
CN (1) CN109297534B (zh)
WO (1) WO2020056812A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110009245A (zh) * 2019-04-12 2019-07-12 阳江职业技术学院 基于神经网络的室内空气质量预测方法、装置及电子设备
CN110675006A (zh) * 2019-11-07 2020-01-10 桂林理工大学 一种室内空气质量实时监测与甲醛衰减预测***
CN112561244B (zh) * 2020-11-26 2021-08-03 清华大学 一种结合室内人员信息的建筑环境评价方法及***
CN112903028A (zh) * 2021-03-03 2021-06-04 中国建筑股份有限公司 一种室内环境人员满意度评价方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107477772A (zh) * 2017-07-21 2017-12-15 天津大学 基于室内监测数据的住宅新风***控制方法
CN108053054A (zh) * 2017-11-22 2018-05-18 中国农业大学 一种河流水质预测方法
CN108319134A (zh) * 2018-01-15 2018-07-24 沈阳大学 一种基于极限学习机网络的温室环境智能控制方法
CN108363294A (zh) * 2017-01-26 2018-08-03 安徽东方果园生物科技有限公司 一种农产品运输封闭和半封闭厢体的环境监测预警方法和环境监测预警***
CN108492044A (zh) * 2018-04-01 2018-09-04 安徽大学江淮学院 基于人工神经网络模型的室内舒适度综合评价***及方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105698861A (zh) * 2016-03-04 2016-06-22 广东轻工职业技术学院 一种室内环境舒适度评价方法
CN105929804A (zh) * 2016-06-07 2016-09-07 江苏长宁电力科技有限公司 建筑环境自适应节能管理***
CN105974869B (zh) * 2016-06-28 2019-05-21 江苏长宁电力科技有限公司 一种应用于建筑环境自适应节能管理***的节能监控中心
CN109060023B (zh) * 2018-08-08 2021-02-05 广东盈峰科技有限公司 一种微型环境监测的数据质控方法及***

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108363294A (zh) * 2017-01-26 2018-08-03 安徽东方果园生物科技有限公司 一种农产品运输封闭和半封闭厢体的环境监测预警方法和环境监测预警***
CN107477772A (zh) * 2017-07-21 2017-12-15 天津大学 基于室内监测数据的住宅新风***控制方法
CN108053054A (zh) * 2017-11-22 2018-05-18 中国农业大学 一种河流水质预测方法
CN108319134A (zh) * 2018-01-15 2018-07-24 沈阳大学 一种基于极限学习机网络的温室环境智能控制方法
CN108492044A (zh) * 2018-04-01 2018-09-04 安徽大学江淮学院 基于人工神经网络模型的室内舒适度综合评价***及方法

Also Published As

Publication number Publication date
CN109297534A (zh) 2019-02-01
CN109297534B (zh) 2020-12-29

Similar Documents

Publication Publication Date Title
WO2020056812A1 (zh) 用于评价室内环境质量的环境参数权重确定方法及***
KR101771053B1 (ko) 실시간 통합 실내 공기질 알림지수 결정 방법 및 장치
CN106557546B (zh) 一种对特高压在线监测数据进行挖掘并评价的方法及***
KR20170122043A (ko) 실시간 실내 공기질 이상치 제거 방법 및 장치
CN108090515B (zh) 一种基于数据融合的环境等级评估方法
CN106500754A (zh) 传感器的检测方法和传感器的检测装置
WO2018184304A1 (zh) 一种网元健康状态的检测方法及设备
CN109543743B (zh) 一种基于重建预测残差的制冷机组多传感器故障诊断方法
CN109061387A (zh) 一种基于圆环定理的电网异常状态判别方法
KR101953558B1 (ko) 스마트 기기 결함 관리 장치 및 방법
CN111964909A (zh) 滚动轴承运行状态检测方法、故障诊断方法及***
CN111130890A (zh) 一种网络流量动态预测***
WO2020056811A1 (zh) 表征室内环境综合质量的综合指标计算方法
CN115204227A (zh) 基于深度学习的设备故障诊断中不确定性量化校准方法
CN108470570B (zh) 电机异音检测方法
CN101237357B (zh) 工业无线传感器网络故障在线检测方法
CN110668280A (zh) 一种电梯称重传感器的检测方法及相关设备
KR101828456B1 (ko) 부분 고장 데이터를 이용한 소프트웨어 신뢰도 평가모델 선택 시스템 및 방법
CN114236272B (zh) 一种电子产品的智能检测***
CN112699609B (zh) 一种基于振动数据的柴油机可靠性模型构建方法
CN104794148B (zh) 煤种化验数据合规性自动监控***及方法
CN114186586A (zh) 基于二维卷积神经网络的损伤识别方法及设备
CN111456915A (zh) 风机机舱内部部件的故障诊断装置及方法
CN113947504B (zh) 一种基于随机森林法的窃电分析方法及其***
CN116522248B (zh) 基于机器学习的核酸异常数据智能研判***

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18934156

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18934156

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 18/11/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18934156

Country of ref document: EP

Kind code of ref document: A1