Model Learning Apparatus for Steady-State Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model learning techniques fail to accurately predict steady-state values due to reliance on transient state data, resulting in discrepancies between estimated and actual steady-state outputs.
Innovation Solution
A model learning apparatus that learns a nonlinear equation of state using both steady-state and transient state data, incorporating a bijective mapping to ensure unique determination of steady-state values, with the equation of state defined by expressions that include time derivatives and multilayer neural networks for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a model is learned based on time series data in the transient state, then the model can capture dynamic changes, but the prediction accuracy of steady-state values deteriorates
Solution Approach 1:
The patent segments the learning data into two distinct components: transient state data for capturing dynamic behavior and steady-state data for ensuring accurate steady-state predictions. The learning portion processes these segmented data separately and combines them to create a comprehensive model that maintains both adaptability and precision.
Solution Approach 2:
The patent changes the parameters used in model learning by introducing steady-state data as an additional parameter set. By modifying the learning input to include both transient and steady-state parameters, the model achieves improved steady-state prediction accuracy while maintaining its ability to capture dynamic changes.
2Measurement precision
If the model uses a bijective mapping with output variable as input, then the steady-state value can be determined uniquely, but the model complexity increases
Solution Approach 1:
The patent applies inversion by using the output variable y as an input to the bijective mapping φ, rather than the conventional approach where inputs map directly to outputs. This inverted structure with φ(y, u) enables unique determination of steady-state values by creating a reversible mapping relationship.
Solution Approach 2:
The bijective mapping φ serves as an intermediary function that mediates between the input variable u and output variable y. This intermediary structure with the property of bijectiveness ensures unique steady-state determination while managing the complexity through a well-defined mathematical function.
Data Source
AI summary
A model learning apparatus is configured to learn a model that shows a relationship between an input variable u input into a system and an output variable y output from the system. The model learning apparatus includes a storage that stores store a model used to learn a nonlinear equation of state for predicting the output variable y by using the input variable u; and a processor programmed to learn the equation of state by using the model and an input-output data set including a set of data of a steady-state value of the output variable y and data of the input variable u corresponding to the data of the steady-state value. The model is an equation of state including a bijective mapping ϕ that uses the output variable y as an input thereof.