Prediction Model Adaptation Thanks to Control Chart Monitoring - Application to Pollutants Prediction

Abstract

Indoor air quality is a major determinant of personal exposure to pollutants in today's world since people spend much of their time in numerous different indoor environments. The Anaximen company develops a smart and connected object named Alima, which can measure every minute several physical parameters: temperature, humidity, concentrations of COV, CO2, formaldehyde and particulate matter (pm). Beyond the measurement aspect, Alima presents some data analysis feature named 'predictive analytics', whose primary aim is to predict the evolution of indoor pollutants in time. In this article, the neural network (NN) model, embedded in this object and designed for pollutant prediction, is presented. In addition with this NN model, this article also details an approach where batch learning is performed periodically when a too important drift between the model and the system is detected. This approach is based on control charts.

Relearning procedure to adapt pollutant prediction neural model: Choice of relearning algorithm

“…With the same philosophy, forecasting model has been proposed to predict the IAQ. Yu and Lin [26] use an ARIMA model and exploit temperature, humidity and CO2 concentration data when Thomas et al [6] use an ANN model and exploit temperature, humidity and CO2 concentration as well, but also VOC and pm.…”

“…The Levenberg-Marquardt approximation of the Hessian is given by (6) In comparison with the two other algorithms, this algorithm is not iterated, and don't need the tuning of a parameter.…”

“…These models use also as inputs the past values of the five collected physical quantities: temperature (T°), humidity (Hu), concentration of VOC (VOC), concentration of CO2, formaldehyde and fine particles (pm). A first work [6] have allowed to design these prediction neural network models by using Foobot® 1 (Kitchen) dataset. The optimal structure of the models is obtained by using pruning step [33].…”

“…Neural models have been proposed in the past to detect the current situation (cooking, sleeping as example) [5]. These current situations are used as inputs in another neural model which can predict the evolution of pollutants levels (30 minutes forecast) [6].…”