Chiller Plant Machine Learning Model for Energy Efficiency Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chiller plants face inefficiencies over time, making them costly to operate, and existing tools fail to accurately assess the energy efficiency of individual chillers within the plant, hindering optimal management and maintenance.
Innovation Solution
A system utilizing machine learning models trained on data from sensors to analyze chiller performance, determining the best model based on power consumption, enabling efficient chiller management and maintenance scheduling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If multiple chillers work together to cool a building, then cooling capacity is improved, but energy efficiency becomes difficult to assess and manage
Solution Approach 1:
The system segments the chiller plant into individual chiller units, creating separate machine learning models for each chiller. This segmentation allows the system to assess energy efficiency at the individual component level while maintaining the collective cooling capacity of the entire plant, resolving the information loss problem in multi-chiller systems.
Solution Approach 2:
The patent introduces machine learning models as intermediary tools between the physical chillers and the management system. These models act as mediators that capture and process energy efficiency data from multiple chillers, providing actionable insights without requiring direct complex measurements from each chiller unit.
2Loss of energy
If chiller efficiency is monitored individually, then energy loss identification is improved, but system complexity increases
Solution Approach 1:
The patent implements a universal machine learning framework that can be applied to any chiller unit regardless of its specific characteristics. This universal approach allows individual chiller monitoring through a standardized system, reducing overall complexity while maintaining the ability to identify energy losses in each unit.
Solution Approach 2:
Each chiller's machine learning model operates autonomously to monitor and assess its own energy efficiency. The models self-train on historical data and automatically identify energy losses without requiring complex external intervention, simplifying the monitoring system while improving energy loss detection capabilities.
3Measurement precision
If machine learning models are trained on historical data, then prediction accuracy is improved, but data requirements and processing time increase
Solution Approach 1:
The system performs preliminary actions by continuously collecting and pre-processing chiller data during normal operation. This preliminary data preparation reduces the time required for model training when predictions are needed, as the foundational data infrastructure is already in place and organized for analysis.
Data Source
AI summary
A method for evaluation of chiller plant operation economy is disclosed. The method comprises receiving a first set of data associated with each chiller over a training interval; generating a ML model for each chiller; deploying the generated ML model to a model of each chiller over a ranking interval; receiving a second set of data associated with each chiller, from the sensors; averaging second set of data for each type of sensor across chillers; determining a power consumption of the ML model of each chiller and comparing it over the ranking interval to determine a best ML model; deploying the ML models to an non-degraded chiller plant model and the best ML model to an ideal chiller plant model; comparing the sum of measured actual power consumptions of chillers with calculated consumptions of a non-degraded chiller plant model and ideal chiller plant model to determine estimated potential of savings.