HVAC Predictive Control Using Semantic State Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current HVAC management techniques face challenges in optimizing energy consumption while maintaining user comfort, especially under changing weather and occupancy conditions, as they are limited by the use of raw data and fail to distinguish between operation modes and occupancy conditions effectively.
Innovation Solution
A smart HVAC management technique that segments historic data into semantic states independent of time-of-day, using self-trained modules to compute optimal on and off periods and temperature set points, considering weather forecasts and occupancy schedules, to optimize energy consumption and adapt to varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If classical logic based on rules and thresholds is used for climatization control, then the system is simple to implement, but it cannot effectively adapt to changing weather conditions and user requirements
Solution Approach 1:
The patent implements dynamic adaptability by using trainable models (neural networks, genetic algorithms) that continuously learn from historical data and adjust control strategies in real-time based on changing weather conditions and user requirements, transforming the static rule-based system into a dynamic adaptive system
Solution Approach 2:
The system employs self-learning capabilities where the trainable models automatically improve their performance by learning from historical climatization data without requiring manual reprogramming, enabling the system to adapt to new patterns and conditions autonomously
2Measurement precision
If trainable models based on neural networks are used to learn from historic data, then the system can optimize climatization conditions, but it is limited by the use of raw data and unable to distinguish between operation modes and occupancy conditions
Solution Approach 1:
The patent segments historical data into distinct operational categories (operation modes and occupancy conditions) before feeding it to the trainable models. This segmentation allows the system to process different types of data separately, preserving the ability to distinguish between various operating contexts while still leveraging machine learning for optimization
Solution Approach 2:
The system performs preliminary processing and categorization of historical data before training the models. By pre-segmenting data into meaningful categories (operation modes, occupancy conditions), the system prepares the data in advance to ensure the models can effectively learn from structured, context-rich information rather than raw unprocessed data
3Use of energy by moving object
If exhaustive optimization techniques are used for HVAC systems, then energy consumption can be minimized, but the system requires a more exhaustive set of measurements and does not address secondary air loop problems
Solution Approach 1:
The patent develops a universal optimization framework that can handle multiple HVAC system configurations and components (including both primary and secondary air loops) through a single trainable model architecture. This multi-functional approach allows the system to optimize energy consumption across different system types without requiring separate exhaustive optimization procedures for each configuration
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A climatization management technique is disclosed, which combines automated data retrieval with smart data segmentation and energy consumption optimization. The set of Regression models (Reg1, Reg2) are trained with historic data segmented by semantic states instead of by time-of-day, enabling greater adaptability and optimization. The set of regression models (Reg1, Reg2) are then applied to current boundary conditions in order to sequentially compute the optimal ON/OFF periods of of a HVAC system -and optionally on/off mechanical ventilation on/off periods (ON_OFF_MV)- and temperature set points (Tsp(t)).