Estimation of classrooms occupancy using a multi-layer perceptron

📝 Abstract

This paper presents a multi-layer perceptron model for the estimation of classrooms number of occupants from sensed indoor environmental data-relative humidity, air temperature, and carbon dioxide concentration. The modelling datasets were collected from two classrooms in the Secondary School of Pombal, Portugal. The number of occupants and occupation periods were obtained from class attendance reports. However, post-class occupancy was unknown and the developed model is used to reconstruct the classrooms occupancy by filling the unreported periods. Different model structure and environment variables combination were tested. The model with best accuracy had as input vector 10 variables of five averaged time intervals of relative humidity and carbon dioxide concentration. The model presented a mean square error of 1.99, coefficient of determination of 0.96 with a significance of p-value < 0.001, and a mean absolute error of 1 occupant. These results show promising estimation capabilities in uncertain indoor environment conditions.

💡 Analysis

This paper presents a multi-layer perceptron model for the estimation of classrooms number of occupants from sensed indoor environmental data-relative humidity, air temperature, and carbon dioxide concentration. The modelling datasets were collected from two classrooms in the Secondary School of Pombal, Portugal. The number of occupants and occupation periods were obtained from class attendance reports. However, post-class occupancy was unknown and the developed model is used to reconstruct the classrooms occupancy by filling the unreported periods. Different model structure and environment variables combination were tested. The model with best accuracy had as input vector 10 variables of five averaged time intervals of relative humidity and carbon dioxide concentration. The model presented a mean square error of 1.99, coefficient of determination of 0.96 with a significance of p-value < 0.001, and a mean absolute error of 1 occupant. These results show promising estimation capabilities in uncertain indoor environment conditions.

📄 Content

Energy for Sustainability International Conference 2017 Designing Cities & Communities for the Future Funchal, 8-10 February, 2017

ESTIMATION OF CLASSROOMS OCCUPANCY USING A MULTI-LAYER PERCEPTRON Eugénio Rodrigues1*, Luísa Dias Pereira1,
Adélio Rodrigues Gaspar1, Álvaro Gomes2, Manuel Carlos Gameiro da Silva1 1: ADAI, LAETA, Department of Mechanical Engineering Faculty of Sciences and Technology
University of Coimbra Rua Luís Reis Santos, Pólo II, 3030-788 Coimbra, Portugal

• e-mail: eugenio.rodrigues@gmail.com, web: http://www.adai.pt 2: INESC Coimbra, Department of Electrical and Computer Engineering Faculty of Sciences and Technology
  University of Coimbra Rua Luís Reis Santos, Pólo II, 3030-290 Coimbra, Portugal web: http://www.inescc.pt Keywords: occupancy, indoor environment, estimation, learning algorithm, neural network Abstract This paper presents a multi-layer perceptron model for the estimation of classrooms number of occupants from sensed indoor environmental data–relative humidity, air temperature, and carbon dioxide concentration. The modelling datasets were collected from two classrooms in the Secondary School of Pombal, Portugal. The number of occupants and occupation periods were obtained from class attendance reports. However, post-class occupancy was unknown and the developed model is used to reconstruct the classrooms occupancy by filling the unreported periods. Different model structure and environment variables combination were tested. The model with best accuracy had as input vector 10 variables of five averaged time intervals of relative humidity and carbon dioxide concentration. The model presented a mean square error of 1.99, coefficient of determination of 0.96 with a significance of p-value < 0.001, and a mean absolute error of 1 occupant. These results show promising estimation capabilities in uncertain indoor environment conditions. E. Rodrigues, L. Dias Pereira, A. R. Gaspar, Á. Gomes, M. C. Gameiro da Silva 2

1. INTRODUCTION Real-time occupancy data, by smart environmental control strategies, is definitely important to achieve energy-savings [1]. The importance of occupancy information for climate-control in buildings has been studied by different researchers [2]. Since measuring occupancy from sensors is not always feasible, as in commercial buildings, Liao and Barooah [3] developed an integrated approach (an agent-based model) “to simulate the behaviour of all the occupants of a building, and extract reduced-order graphical models from Monte-Carlo simulations of the agent-based model”. Different approaches, such as “an information technology enabled sustainability test-bed” [4], RFID based system [5], electricity consumption [6] and other different sensing strategies [7] [8] have been applied for the same purpose: space occupancy detection. Other authors tackled this issue using gas sensors/concentrations, mostly CO2, for occupancy estimations [9] [10]. Neural network models, have been less explored for this purpose [11] [12]. By enlarging the literature on space occupancy estimation, which as a fundamental role on indoor environmental control (IEC) decisions (either on energy simulation models or buildings in-use), and consequently on energy expenditures, this paper aims at contributing towards the challenge of modelling and estimation of occupancy, and therefore strengthening IEC decisions, ultimately improving indoor environmental quality conditions indoors and/or saving energy in buildings.
2. METHODOLOGY 2.1. Dataset The SD800 Datalogger by Extech recorded indoor environment data–relative humidity (RH), air temperature (Tin), and carbon dioxide (CO2) concentration—between 3 and 16 of April 2013 in two classrooms (classrooms 2.04 and 2.10) at the Secondary School in Pombal, Portugal. The measurements were taken in intervals of 60s. Due to normal class operation, the instrument was placed at a height of 2.70m above the floor in the middle of one of the rooms, thus not satisfying fully the ISO 7726 recommendations. The two classrooms have similar area and volume–around 50m2 and 141m3, respectively. The main difference was that one had its external wall facing northwest and the other southeast. Both classrooms were provided with centralized systems of thermal energy production and air renewal was ensured by air handling units with heating and cooling coil. The attendance reports of students classes were obtained. However, post-class occupancy was unknown. Therefore, only the periods of confirmed occupancy, weekends, and night time comprehended between midnight and 7:00 AM were considered in the modelling dataset. Also, when the data presented rapid variations readings of CO2 concentration that did not seem plausible, these were also disregarded in the modelling dataset. 2.2. Multi-layer perceptron model The model used to reconstruct the occupancy in the classrooms was a multi-layer perceptron

This content is AI-processed based on ArXiv data.