Tensor Table Linearization for Nonlinear Model Simulation Speed

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing models for integrated energy systems, which incorporate nonlinear physical processes, face challenges in real-time simulation due to the low running speed of nonlinear models.

Innovation Solution

A linearization processing method and device are proposed to process nonlinear models by determining input parameter value ranges, dividing them into subintervals, and generating a tensor table for interpolation, enabling faster simulation and real-time performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If nonlinear models are directly used to run the simulation of an integrated energy system, then the accuracy is relatively high, but the running speed is relatively low which affects the real-time performance

Engineering Contradiction:
Improvemodel accuracyVSAvoidrunning speed
Core Design Contradiction:
Measurement precisionVSSpeed

Solution Approach 1:

The patent divides the value range of each input parameter into multiple subintervals based on interpolation points, creating a segmented tensor table structure. This segmentation allows the system to precompute and store results for different parameter ranges, enabling fast lookup and interpolation during simulation while maintaining the accuracy of the underlying nonlinear model.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary computation by pre-calculating the tensor table offline using the nonlinear model across all parameter ranges. This preliminary action stores the computed results in a structured format, so that during real-time simulation, the system only needs to perform fast interpolation queries rather than running the full nonlinear model computation.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If nonlinear models with many nonlinear physical processes are established for equipment, then the accuracy is relatively high, but the modeling process becomes complicated

Engineering Contradiction:
Improvemodel accuracyVSAvoidmodeling process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent creates a simplified copy of the nonlinear model in the form of a tensor table. Instead of directly implementing complex nonlinear physical processes during simulation, the system uses a pre-computed tabular representation that captures the essential behavior of the nonlinear model, significantly reducing computational complexity while preserving accuracy.

Inventive Principle:
Principle #26Copying

3Adaptability or versatility

If a universal integrated energy system is established to meet requirements of random changes in terminal loads, then the adaptability is improved, but the system complexity increases due to multiple energy sources and components

Engineering Contradiction:
Improveadaptability to load changesVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent develops a universal linearized model framework that can handle multiple energy sources, storage facilities, and load types through a single tensor table structure. This universal approach allows the same modeling methodology to be applied across different equipment and energy carriers (electricity, heating, cooling, gas), reducing overall system complexity while maintaining adaptability to random load changes.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4024259B1Linearization processing method and device for nonlinear model, and storage medium
Publication Date: 2025.06.04 SIEMENS AG
  • EP4024259B1 patent drawingFigure 1
  • EP4024259B1 patent drawingFigure 2
  • EP4024259B1 patent drawingFigure 3~4

AI summary

A linearization processing method and device for a nonlinear model, and a storage medium. Said method comprises: for a nonlinear model of each device, determining a value range of each input parameter of the model; dividing, on the basis of multiple interpolation points, the value range of each input parameter into multiple sub-intervals; evenly determining multiple input sample values in each sub-interval; traversing input sample value combinations of various input parameters of the model, and obtaining an output sample value combination corresponding to each input sample value combination by using the nonlinear model; and generating a tensor table by using all input sample value combinations and the corresponding output sample value combinations thereof. Said method can realize linearization processing of a nonlinear model for some models comprising nonlinear physical processes.