Developing An Artificial Algorithm To Predict Energy Consumption In A Traditional House Adapting To External Environmental Conditions And Occupation

Abstract

The aim of the research was to development a prediction model that could control electricity consumption in a traditional house, adapting to external environmental conditions and occupation. As identified high-end users of electricity, a backup geyser element in a solar geyser solution, including a swimming pool pump, was identified to be controlled more efficient than a time-based controller, as an end result with the same or higher customer satisfaction index to the occupant. A system was developed, recording multiple remote sensor readings from Internet of Things (IoT) devices to a central SQL database which included the hot water usage and heating patterns. Official weather predictions replaced physical sensors, recording environmental conditions. Fuzzifying the sensor recordings into four linguistic terms simplified hot water usage. Partitioning clustering was used to determine relationship patterns between weather predictions and solar heating efficiency. A Hidden Markov Model was used to profile the system, with the Viterbi algorithm calculating the geyser and pool solar heating predictions, while the Baum-Welch algorithm trained the system. The solar geyser’s standby element was controlled by a time-based controller heating water to pre-set value if required. During this investigation the time-based controller was simulated with a controller profiling weather conditions and hot water usage. Adjusting pre-set temperatures accordingly without human intervention. The same algorithm was used to simulate a timer controlling a swimming pool pump, but excluding the hot water usage. Simulations found that profile base prediction controller can increase user satisfactions, including a maximum of twenty percent electrical power savings against time-based controllers.