A hidden markov model and fuzzy logic forecasting approach for solar geyserwater heating
| Autor: | William Hurst, Daniel N. de Bruyn, Ben Kotze |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Technology
Profile usage Computer science 020209 energy Real-time computing Profile weather conditions 02 engineering and technology Viterbi algorithm Fuzzy logic symbols.namesake Esp8266 Backup Control theory 0202 electrical engineering electronic engineering information engineering Partial clustering General Materials Science Solar geyser Hidden Markov model Cluster analysis Civil and Structural Engineering 020208 electrical & electronic engineering Toegepaste Informatiekunde Building and Construction Energy consumption Geotechnical Engineering and Engineering Geology Computer Science Applications Algorithmic efficiency Hidden markov model symbols Information Technology |
| Zdroj: | Infrastructures Volume 6 Issue 5 Infrastructures, 6(5) Infrastructures, Vol 6, Iss 67, p 67 (2021) Infrastructures 6 (2021) 5 |
| ISSN: | 2412-3811 |
| Popis: | Time-based smart home controllers govern their environment with a predefined routine, without knowing if this is the most efficient way. Finding a suitable model to predict energy consumption could prove to be an optimal method to manage the electricity usage. The work presented in this paper outlines the development of a prediction model that controls electricity consumption in a home, adapting to external environmental conditions and occupation. A backup geyser element in a solar geyser solution is identified as a metric for more efficient control than a time-based controller. The system is able to record multiple remote sensor readings from Internet of Things devices, built and based on an ESP8266 microcontroller, to a central SQL database that includes the hot water usage and heating patterns. Official weather predictions replace physical sensors, to provide the data for the environmental conditions. Fuzzification categorises the warm water usage from the multiple sensor recordings into four linguistic terms (None, Low, Medium and High). Partitioning clustering determines the relationship patterns between weather predictions and solar heating efficiency. Next, a hidden Markov model predicts solar heating efficiency, with the Viterbi algorithm calculating the geyser heating predictions, and the Baum–Welch algorithm for training the system. Warm water usage and solar heating efficiency predictions are used to calculate the optimal time periods to heat the water through electrical energy. Simulations with historical data are used for the evaluation and validation of the approach, by comparing the algorithm efficiency against time-based heating. In a simulation, the intelligent controller is 19.9% more efficient than a time-based controller, with higher warm water temperatures during the day. Furthermore, it is demonstrated that a controller, with knowledge of external conditions, can be switched on 728 times less than a time-based controller. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|