31 / 2025-03-26 20:32:02

A modified virtual sensor modeling method and its application in heating, ventilation, and air-conditioning system

Virtual sensor modeling,irregular interval sampling,long short-term memory,maximal information coefficient,condition monitoring

Realizing sensor redundancy is of great significance to ensure the normal operation of the heating, ventilation, and air-condition (HVAC) system. Developing a virtual sensor is an effective method to achieving sensor redundancy. Traditional virtual sensor models typically focus on feature extraction from uniformly sampled data. However, in actual HVAC system, the collected data may have irregular time intervals, posing challenges for conventional methods. To address this issue, this paper proposes a virtual sensor modeling method based on maximal information coefficient (MIC) and sample interval attention-based long short-term memory (SIA-LSTM). First of all, MIC is employed to identify process variables that exhibit a strong correlation with the target variable. Subsequently, SIA-LSTM introduced an attention mechanism to adaptively estimate varying sampling intervals is utilized to model the virtual sensors for supply and return water temperatures in the chiller unit of the HVAC system. Finally, the effectiveness and superiority of the proposed method are validated and verified on an actual HVAC system.