Building system with automatic chiller anti-surge control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
HVAC chiller systems often enter a surge state, leading to compressor damage and reduced lifespan due to inability to overcome pressure differences, causing refrigerant to flow backwards and result in vibrations and noise, and existing control systems struggle to predict and prevent these events accurately.
Innovation Solution
A method using machine learning models, trained with historical chiller operating data, to predict future surge events and adjust controllable variables such as variable speed drive frequency, pre-rotational vane position, and variable geometry diffuser position to prevent surges by setting thresholds based on probability predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the chiller operates close to the surge state to maximize efficiency, then cooling capacity and energy efficiency are improved, but the risk of entering surge state increases causing compressor damage and reduced reliability
Solution Approach 1:
The machine learning model predicts future surge events before they occur by analyzing historical and real-time operating data. The system takes preliminary action by adjusting controllable variables (VSD frequency, PRV/VGD position) proactively to prevent surge events, rather than reacting after surge has already damaged the compressor. This allows the chiller to operate closer to the surge state safely.
Solution Approach 2:
The system continuously monitors chiller operating data and uses machine learning models to predict surge probability. Based on this feedback, the control system dynamically adjusts controllable variables to maintain operation near the optimal point while preventing surge entry. The feedback loop enables real-time optimization of the reliability-productivity tradeoff.
2Reliability
If traditional control systems are used to prevent surge events, then compressor protection is provided, but the system cannot accurately predict future surge events and operates conservatively away from optimal efficiency
Solution Approach 1:
The patent replaces traditional mechanical/electrical control systems with machine learning-based predictive control. Instead of using conventional sensors and controllers that react to current surge conditions, the system uses ML models trained on historical data to predict future surge events and preemptively adjust operating parameters, enabling both high reliability and high efficiency.
Solution Approach 2:
The system changes the control parameters from reactive threshold-based control to predictive probability-based control. By using machine learning to estimate surge probability and optimizing controllable variables based on predicted outcomes, the system can maintain operation near the surge state for maximum efficiency while ensuring reliability through probabilistic safety margins.
3Productivity
If the chiller operates at high capacity near surge state, then cooling output is maximized, but vibrations and noise increase and compressor damage occurs
Solution Approach 1:
The machine learning model predicts surge events before they occur and the control system takes preliminary anti-action by adjusting controllable variables to counteract the conditions that lead to surge. This prevents the harmful vibrations and noise associated with surge operation while maintaining high cooling output through optimized operation near the surge boundary.
Data Source
AI summary
A method of operating a chiller includes applying chiller operating data associated with the chiller as an input to one or more machine learning models, generating a boundary for a controllable chiller variable based on an output of the one or more machine learning models, and affecting operation of the chiller based on the boundary.