Energy System Parameter Validation Using Confidence Bounds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for validating system parameters in energy systems, especially in model-predictive control, are inadequate due to reliance on root-mean-squared error (RMSE) and cross-validation, which fail to provide robust validation, especially when measured values are limited or have multicollinearity, leading to increased uncertainty outside the recorded value range.
Innovation Solution
A method that calculates a confidence bound based on the standard deviation of system parameters, setting them as valid if the quotient of the confidence bound and model function is less than or equal to a specified threshold within a defined value range, allowing for robust validation and extrapolation beyond the recorded value range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual parameter identification is used, then effort and costs are reduced, but the error rate increases and precision decreases
Solution Approach 1:
The system performs automated parameter identification using measured values and a model function, allowing the system to identify its own parameters without external manual intervention. The computing device automatically calculates system parameters from recorded measured values, eliminating the need for manual identification while maintaining high accuracy through mathematical modeling.
2Productivity
If RMSE-based validation is used, then validation is simple and quick, but robustness decreases and reliability is compromised
Solution Approach 1:
The system implements a feedback-based validation approach where the identified system parameters are used to generate predicted measured values, which are then compared with actual measured values. This closed-loop feedback mechanism allows the system to validate parameters across the entire working range, not just the recorded value range, significantly improving robustness and reliability while maintaining computational efficiency.
3Ease of manufacture
If measured values are limited to a specific value range, then data collection is easier, but validation reliability decreases outside this range
Solution Approach 1:
The validation method using confidence intervals serves multiple functions: it validates parameters within the recorded value range and simultaneously assesses their reliability across the entire working range of the component. This universal validation approach allows the same measured data to provide confidence in both local (recorded range) and global (entire working range) parameter accuracy, eliminating the need for separate validation procedures.
4Reliability
If cross-validation is used, then validation is more comprehensive, but it cannot detect errors when measured data have multicollinearity
Solution Approach 1:
The system transforms the validation problem from directly analyzing correlated measured values to analyzing the distribution and confidence intervals of the identified system parameters themselves. By changing the parameter space from input variables to model parameters, the method becomes insensitive to multicollinearity in the original measured data, while still providing comprehensive validation across all operating conditions.
Data Source
Figure 1~2
AI summary
The invention relates to a method for validating system parameters (41) determined by means of measurement data for a model function ŋ (10) of at least one component of an energy system, the model function ŋ (10) characterising at least one dependency of at least one output variable of the component of at least one input variable of the component, taking into account the system parameters (41). The claimed method is charaterised by at least the following steps: - calculating a standard deviation {σ η of the system parameters (41) determined from the measurement data; - calculating a confidence limitation Ψ (42) according to the calculated standard deviation σ η ; and determining the system parameters (41) as valid if the quotient of the confidence limitation Ψ (42) and the model function ŋ (10) within a value range (22) defined for the input variable is less than or equal to a defined threshold value δ. The invention also relates to a method for operating an energy system and to an energy management system for an energy system.