Building Energy Prediction Model Updates for Adaptive Equipment Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Operating building equipment based on standard control schemes can lead to unnecessary energy consumption and increased costs due to changes in energy prices, as these systems do not adapt to changing energy requirements effectively.
Innovation Solution
A building management system that utilizes an energy prediction model (EPM) to predict energy needs over time, retrain based on new data, and adjust operations by identifying and deploying updated hyper-parameters to optimize energy usage, switching to older models if they provide more accurate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If standard control schemes are used to operate building equipment, then the system is simple and easy to operate, but energy consumption increases and the system cannot adapt to changing energy requirements
Solution Approach 1:
The patent implements dynamic adaptability by continuously monitoring triggering events and dynamically updating hyper-parameters of the energy prediction model. The system transitions from static standard control schemes to dynamic adaptive control that automatically adjusts to changing energy requirements, building characteristics, and energy prices through automated model retraining and hyper-parameter optimization.
Solution Approach 2:
The patent changes key parameters of the energy prediction model by identifying and deploying updated hyper-parameters when triggering events occur. This involves modifying model parameters such as training data windows, regularization coefficients, and other hyper-parameters to optimize energy predictions and reduce consumption while adapting to new conditions.
2Measurement precision
If the energy prediction model is continuously retrained with updated hyper-parameters, then prediction accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent implements periodic model retraining triggered by specific triggering events rather than continuous retraining. The system monitors for triggering events and retrains the energy prediction model only when these events occur, balancing prediction accuracy with computational efficiency by avoiding unnecessary retraining operations.
Solution Approach 2:
The patent uses feedback mechanisms to monitor model performance and determine when retraining is necessary. By implementing feedback loops that track prediction accuracy and system performance, the system can intelligently decide when to retrain the model and update hyper-parameters, optimizing the balance between accuracy and computational resource usage.
3Reliability
If older energy prediction models are used, then the system is more stable, but the predictions become less accurate for current building conditions
Solution Approach 1:
The patent prepares for model obsolescence by implementing a version management system that retains older models and allows switching between versions. This preliminary preparation enables the system to revert to stable older models when needed while having the capability to deploy updated models when they demonstrate superior accuracy, thus maintaining both stability and adaptability.
Solution Approach 2:
The patent manages the transition between older and newer models by carefully adjusting hyper-parameters and training configurations. When updating from older to newer models, the system modifies parameters such as training data selection, hyper-parameter values, and model architecture to ensure smooth transitions that maintain stability while improving prediction accuracy for current building conditions.
Data Source
AI summary
A building management system including building equipment operable to affect a variable state or condition of a building. The building management system includes a controller including a processing circuit. The processing circuit is configured to obtain an energy prediction model (EPM) for predicting energy requirements over time. The processing circuit is configured to monitor one or more triggering events to determine if the EPM should be retrained. The processing circuit is configured to, in response to detecting that a triggering event has occurred, identify updated values of one or more hyper-parameters of the EPM. The processing circuit is configured to operate the building equipment based on the EPM.