Indoor occupancy estimation from carbon dioxide concentration (1607.05962v1)

Abstract: This paper presents an indoor occupancy estimator with which we can estimate the number of real-time indoor occupants based on the carbon dioxide (CO2) measurement. The estimator is actually a dynamic model of the occupancy level. To identify the dynamic model, we propose the Feature Scaled Extreme Learning Machine (FS-ELM) algorithm, which is a variation of the standard Extreme Learning Machine (ELM) but is shown to perform better for the occupancy estimation problem. The measured CO2 concentration suffers from serious spikes. We find that pre-smoothing the CO2 data can greatly improve the estimation accuracy. In real applications, however, we cannot obtain the real-time globally smoothed CO2 data. We provide a way to use the locally smoothed CO2 data instead, which is real-time available. We introduce a new criterion, i.e. $x$-tolerance accuracy, to assess the occupancy estimator. The proposed occupancy estimator was tested in an office room with 24 cubicles and 11 open seats. The accuracy is up to 94 percent with a tolerance of four occupants.

• The paper proposes a non-intrusive method to estimate indoor occupancy using existing CO2 sensors and the Feature Scaled Extreme Learning Machine (FS-ELM) technique.
• The proposed FS-ELM model achieved high accuracy, up to 94% within four occupants, significantly outperforming standard ELM in trials conducted in a real office environment.
• This CO2-based estimation method has practical implications for optimizing HVAC and lighting systems, contributing to building energy efficiency and paving the way for smarter building management.

Evaluating Indoor Occupancy Estimation Using Carbon Dioxide Concentration and Machine Learning

The paper "Indoor occupancy estimation from carbon dioxide concentration" by Chaoyang Jiang et al. presents a novel approach for estimating indoor occupancy by analyzing carbon dioxide (CO2) concentration levels. The research introduces the Feature Scaled Extreme Learning Machine (FS-ELM) technique as an improvement over traditional Extreme Learning Machine (ELM) methods for this purpose. This manuscript offers significant contributions to the field of occupancy estimation, primarily through non-intrusive means using existing environmental sensors, which could be integrated into standard HVAC systems.

Summary of Methodology

The work is built upon the need for a robust, privacy-preserving, and cost-effective solution to track indoor occupancy—an essential determinant for optimizing HVAC systems and thereby improving energy efficiency in buildings. Unlike intrusive methods such as video monitoring or terminal-based solutions that infringe on privacy, the proposed method employs the already available CO2 sensors in HVAC systems, which do not require additional hardware investments.

The paper describes the model that estimates real-time indoor occupancy directly as a dynamic model function of CO2 concentrations. FS-ELM enhances the robustness of the estimation process by introducing a feature scaling mechanism that prepares the input data to improve the learning process and accuracy of the model. The feature layer integrates various CO2 data aspects such as past concentrations, integrated differentials, and differences over specific time intervals, together with past occupancy and ventilation levels.

To combat the noise and spikes inherently present in the CO2 concentration measurements, the paper advocates for smoothing techniques, with a particular focus on global smoothing during the training phase and local smoothing during real-time application.

Key Results

The model was validated in a 9.3m x 20m office space containing 24 cubicles and 11 open seats, crucially demonstrating the practical applicability of the approach in realistic environments with more than 20 occupants. The paper reports a high accuracy level of up to 94% for estimates within a tolerance of four occupants. The FS-ELM significantly outperformed the standard ELM, with lower Root Mean Squared Error (RMSE) values and improved false detection rates in terms of estimating accurate room occupancy states (FPR and FNR).

Implications and Future Work

By successfully implementing a machine-learning-based estimation system driven by environmental CO2 data, this research represents an essential step towards more sustainable and efficient use of energy in buildings. The methodology allows for HVAC and lighting systems to function more reactively and intelligently, offering immediate practical benefits in terms of reduced carbon footprints and optimized energy consumption.

Future work can extend this approach by exploring deeper integrations into building management systems where dynamic changes and predictions might be automated further. Moreover, challenges such as optimal sensor placement and enhancing the model's responsiveness to rapidly fluctuating occupant numbers remain an area for future investigation. Furthermore, while FS-ELM has shown remarkable potential, examining other advanced machine learning paradigms, possibly incorporating deep learning techniques, could yield even more refined and robust models for real-time occupancy estimation.

In conclusion, Jiang et al.'s work presents a compelling case for leveraging CO2 sensors already present in commercial buildings to provide high-accuracy occupant estimations. This development holds promise not only for energy management but also for broader applications in intelligent building systems.