Dynamic Load Prediction Model Retraining for Building Energy Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Building energy systems face challenges in managing electricity consumption due to dynamic pricing schemes, as changes in occupancy, construction, or equipment usage can make it difficult to optimize energy usage from the electric grid.
Innovation Solution
A building energy system that includes an energy storage system (ESS) and a processing circuit to collect data, retrain load prediction models based on current conditions, and optimize energy storage and discharge schedules to minimize electricity costs, using models like generalized additive models or recurrent neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a static load prediction model is used, then the system structure is simple, but the system cannot adapt to changing building conditions such as occupancy changes, construction, or equipment upgrades
Solution Approach 1:
The patent implements a dynamic model retraining mechanism that automatically updates the load prediction model based on changing building conditions. The system monitors occupancy data, weather data, and load data to detect changes in building patterns, and triggers model retraining when changes are detected. This transforms a static model into a dynamic adaptive system without requiring complete model redesign, thus improving adaptability while controlling complexity.
Solution Approach 2:
The system incorporates feedback loops that continuously monitor building data (occupancy, weather, actual load) and use this information to determine when model retraining is needed. The feedback mechanism compares predicted load with actual load and detects pattern changes, triggering automated model updates. This feedback-driven approach enables the system to adapt to changing conditions automatically while maintaining manageable complexity through structured decision-making.
2Measurement precision
If the load prediction model is retrained frequently, then the prediction accuracy improves, but the computational resources and time required increase
Solution Approach 1:
The patent implements dynamic model retraining that adjusts the retraining frequency based on detected changes in building patterns. Rather than fixed periodic retraining or continuous retraining, the system monitors for pattern changes in occupancy, weather, and load data, and triggers retraining only when changes exceed predefined thresholds. This dynamic approach optimizes the balance between prediction accuracy and computational resource usage by retraining only when necessary.
Solution Approach 2:
The system changes the operational parameters of the prediction model based on detected building condition changes. When pattern changes are detected in occupancy, weather, or load data, the system adjusts the model retraining parameter from 'no retraining' to 'retrain model'. This parameter-based control mechanism enables flexible adaptation of model freshness to actual building conditions, improving accuracy when needed while minimizing unnecessary retraining computational overhead.
3Measurement precision
If the system collects and processes extensive building data for model retraining, then the prediction accuracy improves, but the data processing complexity and computational burden increase
Solution Approach 1:
The patent extracts and processes only the specific building data elements necessary for detecting pattern changes and retraining the model. The system focuses on collecting occupancy data, weather data, and load data - the key parameters needed for load prediction. By extracting only the essential data elements rather than processing all available building data, the system achieves accurate predictions while minimizing data processing complexity and computational burden.
Solution Approach 2:
The collected building data serves multiple functions within the system: it is used for real-time load prediction, for detecting pattern changes that trigger retraining, for monitoring building conditions, and for optimizing energy storage operations. This multi-functional use of the same data set eliminates the need for separate data collection and processing systems for each function, reducing overall data processing complexity while maintaining high prediction accuracy.
Data Source
AI summary
A building energy system for a building includes an energy storage system (ESS) configured to store energy received from an energy source and provide the stored energy to one or more pieces of building equipment to operate the one or more pieces of building equipment. The system includes a processing circuit configured to collect building data, determine whether to retrain a trained load prediction model based on at least some of the building data, retrain the trained load prediction model based on the building data in response to a determination to retrain the trained load prediction model, determine a load prediction for the building based on the retrained load prediction model, and operate the ESS to store the energy received from the energy source or provide the stored energy to the one or more pieces of building equipment to operate the one or more pieces of building equipment.