Adaptive AI Building Control With MPC-PID Confidence Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional control systems for building management, such as PID control, are not predictive and require time to respond to changing conditions, leading to inefficiencies and potential instability, while adaptive methods like PRAC need frequent re-tuning and can be sluggish in moderate conditions.
Innovation Solution
Implementing an adaptive AI device model that integrates model predictive control (MPC) with PID control, using neural networks and autoregressive models to generate predictions and adjust control strategies based on confidence thresholds, while detecting equipment tampering and degradation through adaptive AI models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If PID control is used in building management systems, then the system is simple to implement and operate, but the response time is slow and the system cannot predict changing conditions
Solution Approach 1:
The AI model performs preliminary actions by predicting future system states and optimal control inputs before actual changes occur. The model is trained offline on historical data to learn system dynamics, enabling it to anticipate disturbances and generate proactive control commands that improve response time without requiring complex real-time computations during operation.
Solution Approach 2:
The patent introduces an AI model as an intermediary between the simple PID controller and the building systems. This AI intermediary processes sensor data, predicts future states, and generates control commands that are then executed by actuators. The AI model acts as a mediator that adds predictive capabilities while maintaining compatibility with existing simple control infrastructure.
2Adaptability or versatility
If adaptive control methods like PRAC are used, then the system can adapt to changing conditions, but the system becomes sluggish in moderate conditions and requires frequent re-tuning
Solution Approach 1:
The control system dynamically adapts its behavior based on the AI model's confidence in its predictions. When the AI model is highly confident (above threshold), the system operates in predictive mode with full adaptability. When confidence drops below the threshold, the system automatically transitions to a more conservative mode, blending AI predictions with traditional control to maintain stability and responsiveness.
Solution Approach 2:
The system changes the confidence threshold parameter to control the level of adaptability. By adjusting this threshold, the system can optimize performance for different operating conditions - higher thresholds provide more stability and responsiveness but reduce adaptability, while lower thresholds increase adaptability but may reduce responsiveness. This parameter adjustment allows the system to balance adaptability and productivity based on current conditions.
3Productivity
If model predictive control is used with high confidence AI predictions, then the system responds quickly and proactively to disturbances, but the system may become unstable when the AI model confidence is low
Solution Approach 1:
The system implements feedback by continuously monitoring the AI model's confidence level and using this information to adjust control strategy. When confidence is high, the system executes AI-generated control commands for efficient predictive control. When confidence drops below the threshold, the system provides feedback to switch to a more conservative control mode, preventing instability while maintaining overall system reliability.
Solution Approach 2:
The system prepares for potential AI model failures by having a fallback control strategy ready in advance. The confidence threshold acts as a pre-established safety mechanism that cushions against the risks of using AI predictions when the model is uncertain. This beforehand cushioning ensures that even if the AI model produces inaccurate predictions, the system remains stable by switching to proven traditional control methods.
Data Source
AI summary
A controller for equipment that operates to affect a variable state or condition of a building including one or more processors and non-transitory computer-readable media storing instructions that, when executed by the processors, cause the processors to perform operations. The operations include performing model predictive control and proportional, integral, derivative control using adaptive artificial intelligence, performing tampering prediction using adaptive artificial intelligence, or detecting degradation using adaptive artificial intelligence.