A novel hybrid model based on particle swarm optimisation and extreme learning machine for short-term temperature prediction using ambient sensors. (August 2019)

• Record Type: Journal Article
• Authors: Kumar, Sachin; Pal, Saibal K; Singh, Rampal
• Is Part Of: Sustainable cities and society. Volume 49(2019)
• Publication Date: 2019-08
• Subjects: Indoor temperature prediction -- Energy consumption -- Extreme learning machine -- Hybrid model -- Online sequential extreme learning machine -- Particle swarm optimization -- Sensors; Sustainable urban development -- Periodicals; Sustainable buildings -- Periodicals; Urban ecology (Sociology) -- Periodicals; 307.76
• DOI: 10.1016/j.scs.2019.101601
• Languages: English
• ISSNs: 2210-6707

Abstract

Highlights: This paper proposed the dynamic methods for optimised feature selection for energy and temperature prediction using Particle Swarm Optimization(PSO). Proposes the hybrid model based on PSO and ELM that minimizes the forecasting error of indoor temperature and energy consumption for 30 min ahead and 4 h ahead. Proposes the hybrid model based on PSO and OSELM for adaptive control of forecasting temperature by reading data in online fashion for 30 min ahead and 4 h ahead prediction. Comparative study of proposed methods on the same data set. Abstract: In modern buildings which are becoming smart day by day, indoor temperature can be forecasted with the data obtained from outfitted sensors. Predictive models based on data can accurately forecast temperature which further saves energy by optimisation of resources and technologies such as heating, ventilation and air conditioners for making the atmosphere conducive and comfortable in the ambient environment. This paper discusses an experiment from such house fitted with sensors for different parameter affecting indoor temperature. We propose a hybrid model which is based on particle swarm optimisation and emerging extreme learning machine to forecast the temperature to optimise the use of energy further. The proposed hybrid model also includes the variant online sequential extreme learning machine(OSELM) which accepts data online and is adaptive to the changing environmental conditions. We perform experiment based on many sensors combinations affecting temperature using particle swarm optimization and statistical tools to determine their relevance and correlation with temperature and compare results with conventional methods. Proposed methods improved the accuracy of the forecasting and also generalisation performance over other methods on the same dataset.