Hybrid State Prediction for Short- and Long-Term System Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately predict both short-term and long-term behavior of technical systems without compromising precision, often leading to inaccuracies or loss of information.
Innovation Solution
A hybrid modeling approach combining a learning-based model for short-term predictions and a physical model for long-term predictions, using complementary filters like high-pass and low-pass filters to integrate both models effectively, ensuring reliable behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning-based model is used for predicting short-term behavior, then short-term prediction precision is improved, but long-term prediction accuracy deteriorates
Solution Approach 1:
The prediction task is segmented into two distinct components: short-term behavior prediction handled by a learning-based model and long-term behavior prediction handled by a physical model. Each model is specialized for its respective time horizon, with the learning-based model capturing short-term dynamics and the physical model ensuring long-term plausibility. The final prediction is formed by combining these segmented predictions through a fusion mechanism.
2Reliability
If a physical model is used for predicting long-term behavior, then long-term prediction accuracy is improved, but short-term prediction precision deteriorates
Solution Approach 1:
The prediction task is segmented into two distinct components: short-term behavior prediction handled by a learning-based model and long-term behavior prediction handled by a physical model. Each model is specialized for its respective time horizon, with the learning-based model capturing short-term dynamics and the physical model ensuring long-term plausibility. The final prediction is formed by combining these segmented predictions through a fusion mechanism.
3Reliability
If hybrid modeling is used to combine both models, then overall prediction reliability is improved, but device complexity increases
Solution Approach 1:
A fusion mechanism acts as an intermediary between the learning-based model and the physical model. This fusion mechanism integrates the predictions from both models in a computationally efficient manner, combining the short-term accuracy of the learning-based model with the long-term reliability of the physical model. The fusion mechanism handles the complexity of integration while maintaining the benefits of both approaches.
Data Source
AI summary
A device and computer-implemented method for predicting a state of a technical system. A state of the technical system is detected and a time series is provided which comprises values which characterize a course of the detected state of the technical system. Using a learning-based model for predicting the short-term behavior of the technical system, a first value for the prediction is determined as a function of the values of the time series, and, using a physical model for predicting the long-term behavior of the technical system, a second value for the prediction is determined as a function of the values of the time series, and wherein a value of the prediction is determined as a function of the first value for the prediction and the second value for the prediction.