Gradient-Based Parameter Estimation for Technical Installations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining parameters of technical models, such as those describing battery behavior, require extensive computational resources and hardware, especially when measurements are sequential, as they often rely on recursive implementations of algorithms like the least-square method.
Innovation Solution
A gradient-based method is used to efficiently estimate model parameters by updating them with each measurement, approximating the least-square algorithm with a lower computational burden, allowing for real-time adaptation and prediction of system responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If recursive implementations of least-square algorithm are used to adapt parameters with each measurement, then the ability to further adapt parameters with each sequential measurement is improved, but the computational outlay and hardware requirements increase
Solution Approach 1:
The patent transforms the recursive least-square algorithm into a gradient-based approach by changing the mathematical parameters and computational methodology. Instead of using the computationally intensive recursive least-square formulas, the patent employs gradient descent optimization with simplified update rules that require fewer arithmetic operations per measurement while maintaining the ability to adapt parameters sequentially
Solution Approach 2:
The patent replaces the mechanical computation structure of recursive least-square algorithms with a gradient-based optimization mechanism. This substitution uses a different computational paradigm (gradient descent) that achieves the same adaptive parameter estimation goal with reduced computational complexity and simpler hardware requirements
2Measurement precision
If all measurements are collected before adapting parameters, then measurement completeness is improved, but the time delay and loss of real-time capability increase
Solution Approach 1:
The patent implements preliminary gradient computation and parameter update mechanisms that can operate with partial measurement data. The gradient-based approach allows the system to perform preliminary parameter adaptations as measurements arrive, rather than waiting for complete data sets, thus reducing time delay while maintaining estimation accuracy through iterative refinement
Solution Approach 2:
The patent introduces dynamic parameter adaptation where the model parameters are continuously updated as new measurements become available. The gradient-based algorithm dynamically adjusts parameters in real-time based on incoming data, transforming the static batch-processing approach into a dynamic sequential adaptation system that responds to measurements as they arrive
Data Source
AI summary
Methods and devices for determining at least one parameter of a model of a technical installation are provided. In this case, the parameters are updated on the basis of measurements as a function of an observation matrix. The observation matrix is prescribed as a function of the model and being able, if appropriate, to depend on a time variable which can be measured.