Electrical Distribution Modeling for Nonlinear Asset Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for asset and energy management in electrical distribution systems rely on linear data modeling and machine learning, which fail to accurately predict outcomes due to the complex non-linear relationships between electrical and mechanical variables, leading to inaccurate or error-prone predictions.
Innovation Solution
A method and system that utilize sparse regression to generate coefficient matrices from measurement data, determining system representations that describe the relationships between input and output variables, and use these representations to train machine learning models for improved prediction accuracy in asset and energy management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If linear data modeling and machine learning are used for prediction, then the system is simple to implement, but the prediction accuracy is poor due to inability to capture non-linear relationships
Solution Approach 1:
The patent transforms the modeling approach by changing from linear parameter relationships to non-linear parameter relationships through sparse regression. This allows the system to capture complex non-linear correlations between electrical and mechanical variables while maintaining computational efficiency through sparsity constraints, thereby improving prediction accuracy without proportionally increasing complexity
Solution Approach 2:
The patent replaces conventional linear machine learning models with a sparse regression-based non-linear modeling system. This substitution enables the system to inherently capture non-linear relationships in the data without requiring complex neural networks or ensemble methods, achieving better prediction accuracy with controlled complexity
2Reliability
If conventional linear modeling is used, then computational resources are minimized, but the system fails to account for complex non-linear correlations between variables
Solution Approach 1:
The patent changes the fundamental parameter relationships from linear to non-linear through sparse regression modeling. This transformation allows the system to reliably capture complex correlations between multiple variables while the sparsity constraint prevents overfitting, maintaining model interpretability and computational tractability
Solution Approach 2:
The patent introduces dynamic non-linear relationships between variables through the sparse regression framework. The model can adaptively capture changing relationships between electrical and mechanical variables over time, improving reliability while the sparsity constraint keeps the computational complexity manageable
Data Source
AI summary
A method for performing asset and energy management for an electrical distribution system includes receiving measurement data of plurality of input variables of electrical distribution system; generating coefficient matrix for each output variable used for asset and energy management, based on effects of plurality of input variables on corresponding output variable, using sparse regression. The coefficient matrix comprises one or more input variables from plurality of input variables. Furthermore, the method comprises determining plurality of system representations for each output variable, based on corresponding coefficient matrix. Each of plurality of system representations indicates relationship between one or more input variables and the corresponding output variable. Thereafter, the method comprises identifying system representation from plurality of system representations to train machine learning model for predicting value of output variable, for performing asset and energy management for electrical distribution system.