基于动态模糊推理的舒适温度在线预测

On-line forecast for comfortable temperature based on dynamic fuzzy reasoning

摘要:

基于热感觉预测的室内热环境自动控制方法为解决基于传统温度设定值控制不满足用户舒适度问题提供新的途径。但建模过程中用户热感觉信息难以获取,因此开发了便捷的移动端智能交互系统以实时采集现场数据,建立用户学习样本；并针对热舒适的动态变化性特征,设计动态进化神经模糊推理系统(DENFIS)以建立用户热舒适在线预测模型。通过实时学习样本数据驱动,系统的模糊规则与模型输出函数系数可动态自校正,推理预测出用户偏好温度。实验结果表明所设计的DENFIS算法预测用户的舒适温度范围准确率高达90.5%,误差极小。证明了该算法所建立的在线预测模型用于智能空调温度控制,可解决现有的温度设定方式带来的温度设定值不合理的问题,在实际应用中具有可行性。

Abstract:

The automatic control method of indoor thermal environment based on thermal sensation prediction provides a new approach to solve the problem that the control based on traditional temperature set point does not meet the occupant comfort. However, it is a difficult task to acquire the thermal sensation of occupant during the modeling process. Therefore, a convenient intelligent interactive system of mobile terminal is developed to detect real-time field data to further establish learning samples. In addition, according to the dynamic variability feature of thermal comfort, the dynamic evolving neuro-fuzzy inference system (DENFIS) is designed so that the on-line forecasting model of occupant’s thermal comfort is established. Base on the data-driven of real-time learning samples, the system"s fuzzy rules and the coefficients of model output function can be dynamically self-corrected to predict the occupants’ preference temperature further. The result of experiment shows that the designed DENFIS algorithm predicts the occupant"s comfort temperature range as high as 90.5%, and the error was very small. It is proved that the on-line forecasting model established with the approach can be served as an intelligent air-conditioning temperature controller to solve the problem of unreasonable temperature set caused by the existing temperature set-point method, which is feasible in practical application.