Continuous learning compressor input power predictor
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing HVAC&R systems lack the ability to efficiently and reliably detect potential problems and performance degradations early, requiring extensive data acquisition and processing over long periods, making early detection impractical.
Innovation Solution
A monitoring system employing continuous machine learning and a temperature map to establish a relation between compressor input power parameters and condenser and evaporator intake fluid temperatures, allowing for immediate prediction of power parameter values and early detection of performance degradation, while minimizing false positives and negatives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine learning systems are used to detect performance degradations, then reliability of detection is improved, but the time required for data acquisition and processing increases significantly
Solution Approach 1:
The system performs preliminary actions by continuously learning and updating the compressor input power predictor model during normal operation. The model is trained in advance on newly maintained system characteristics, so when performance degradation occurs, the system can immediately compare actual values against the pre-established model without requiring lengthy data collection periods. This preliminary modeling enables rapid detection while maintaining reliability.
2Measurement precision
If extensive data acquisition over extended periods is performed, then detection accuracy is improved, but system complexity and implementation difficulty increase
Solution Approach 1:
The invention extracts only the essential features needed for detection by focusing specifically on the relationship between compressor input power and system performance. Rather than analyzing extensive multi-dimensional data sets, the system isolates the critical predictor variables and maintains a simplified model that captures the essential degradation patterns. This extraction of key features maintains detection accuracy while dramatically reducing system complexity.
3Productivity
If the system begins making predictions immediately with minimal commissioning, then productivity and early detection capability are improved, but measurement precision may be compromised
Solution Approach 1:
The system employs dynamic adaptation by continuously updating the compressor input power predictor model as more data becomes available. The model transitions from an initial state with minimal commissioning to an increasingly refined state over time. This dynamic learning process allows the system to begin predictions immediately with acceptable accuracy, then progressively improve measurement precision as the model adapts to the specific system characteristics through continuous operation.
Data Source
AI summary
System and method for monitoring and detecting potential problems early in a VCC based HVAC&R system employs a monitoring application or agent that uses continuous machine learning and a temperature map to derive or “learn” a relation between a measured input power parameter of one or more system compressors, and condenser and evaporator intake fluid temperatures, based on observations of the temperatures and the input power parameter when the HVAC&R system is new or in a “newly maintained” condition. The monitoring agent can then use the learned relation to determine, based on subsequent observations of the condenser and evaporator intake fluid temperatures, the input power parameter values that should be expected if the HVAC&R system were operating in the “newly maintained” condition. The agent can thereafter compare the expected compressor input power parameter values with observed input power parameter values to determine early whether the system is experiencing performance degradation.