Home HVAC Control Using Occupancy Habit Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for controlling heating/cooling systems in domestic environments, such as autonomous and centralized systems, face challenges in efficiently predicting and adapting to user occupancy habits, leading to energy waste and comfort issues due to the need for manual adjustments and the limitations of GPS tracking and sparse data analysis.
Innovation Solution
A method utilizing a Bayesian topic modeling algorithm, like Latent Dirichlet Allocation (LDA), to forecast home occupancy routines by analyzing occupancy-related events across a population, allowing for precise setting of heating/cooling systems based on shared habits, even with sparse data, and enhancing prediction accuracy through combined user data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If manual thermostat settings are used based on user occupancy habits, then energy waste is reduced, but the system requires continuous user intervention and timely modification of settings
Solution Approach 1:
The system automatically learns and adapts to user occupancy patterns without requiring manual thermostat adjustments. The learning algorithm continuously monitors occupancy data and autonomously optimizes temperature settings, allowing the system to serve itself rather than requiring continuous user intervention while maintaining energy efficiency
Solution Approach 2:
The system implements a feedback loop where occupancy information is continuously collected, analyzed, and used to adjust temperature settings. This closed-loop control enables the system to respond dynamically to actual occupancy patterns, reducing energy waste while eliminating the need for manual user adjustments
2Productivity
If GPS tracking is used to determine user location and predict occupancy, then just-in-time heating and cooling is enabled, but the system becomes invasive and requires constant user presence tracking
Solution Approach 1:
The system extracts occupancy information from passive sensor data already present in the environment (motion sensors, door sensors, appliance usage patterns) rather than relying on active GPS tracking. This removes the invasive tracking component while maintaining the ability to predict occupancy and enable just-in-time climate control
Solution Approach 2:
The system uses intermediate sensors and inference algorithms that indirectly determine occupancy status without directly tracking user location. These intermediaries process environmental data to infer presence patterns, achieving the same goal as GPS tracking but with reduced complexity and user intrusion
3Measurement precision
If individual user occupancy data is analyzed separately, then personalized predictions are achieved, but the system produces unreliable predictions due to sparse data
Solution Approach 1:
The system merges occupancy data from multiple users into a collective dataset that the learning algorithm processes together. By combining sparse individual data streams into a richer aggregated dataset, the system achieves reliable prediction patterns that would be impossible to derive from any single user's limited data alone, while still providing personalized predictions
4Ease of operation
If centralized heating systems are used with individual room control, then user comfort is improved, but energy waste occurs when flats are heated or cooled when not occupied
Solution Approach 1:
The system dynamically adjusts temperature settings based on real-time and predicted occupancy status. Rather than static pre-set temperatures, the system continuously adapts heating/cooling levels to match actual occupancy patterns, ensuring comfort when occupied while eliminating energy waste when unoccupied
Data Source
Figure 1
Figure 2
Figure 3
AI summary
It is disclosed a method for controlling a heating/cooling system located at home of a user. For a population of users, occurrences of occupancy-related events occurring in the users' homes (e.g. user enters or exits home) are detected. Then, the occurrences of such occupancy-related events in different timeslots are counted separately for each user of the population. The counts of the various users are then merged and a topic model is applied thereto, so as to provide probabilities that latent habits give raise to certain sequences of occupancy-related events and proportions of each habit in the home occupancy routine of each user of the population. This way, a reliable forecast of the home occupancy routine of each user of the population may be provided. This allows setting the heating/cooling system of each user of the population to fit such reliable forecast, thereby minimizing energy waste and/or maximizing user's comfort.