Predictive Control Model Construction via Data Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing predictive control technologies face challenges in accurately modeling complex real-world systems, particularly in environments where physical laws are not applicable, and struggle to adapt to changes over time, leading to decreased tracking capability and difficulty in selecting proper control strategies.

Innovation Solution

An information processing device that accumulates control-target information, learns and generates a prediction-equation set, and constructs a predictive-control model to determine operation quantities, enabling automatic model construction and adaptation to environmental changes without requiring a pre-defined predictive-control model.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If a predictive-control model is constructed using physical laws and equations, then the model can be derived systematically, but it cannot thoroughly reflect complicated correlations in the real world, leading to deviation between the model and actual behavior

Engineering Contradiction:

Improveease of model constructionVSAvoidmodel accuracy

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent replaces the traditional mechanical approach of constructing predictive-control models based on physical laws with a machine learning-based approach. The learning device automatically learns the predictive-control model from input-output data without requiring explicit physical equations, thereby substituting the equation-based methodology with data-driven learning to achieve better model accuracy while maintaining ease of construction

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The learning device performs self-learning to automatically construct the predictive-control model without human intervention in model formulation. The system serves itself by autonomously identifying patterns and relationships from data, eliminating the need for manual model construction based on physical laws and enabling automatic adaptation to complex real-world correlations

Inventive Principle:

Principle #25Self-service

2Reliability

If a predictive-control model is designed to reflect all physical laws, then the model should be complete, but it is difficult to derive a model that thoroughly reflects complicated correlations in the real world

Engineering Contradiction:

Improvemodel completenessVSAvoidmodel construction complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent substitutes the complex manual process of deriving models from physical laws with an automated machine learning system. The learning device processes input-output data to automatically generate the predictive-control model, replacing the intricate analytical work required to account for all physical laws and complicated correlations with an automated learning process

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the fundamental parameters of model construction from equation-based parameters (physical laws, mathematical relationships) to data-based parameters (input-output patterns, learned correlations). This parameter transformation allows the system to capture complicated real-world relationships without requiring explicit formulation of all underlying physical laws

Inventive Principle:

Principle #35Parameter changes

3Reliability

If a predictive-control model is constructed manually by a designer, then the model can be created with domain knowledge, but the process is time-consuming and the model becomes incomplete

Engineering Contradiction:

Improvemodel qualityVSAvoidmodel development time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the manual model construction process with an automated learning device that generates predictive-control models from data. This substitution eliminates the time-consuming manual derivation process while maintaining or improving model quality through automated pattern recognition and relationship identification

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The learning device performs preliminary learning from historical data to pre-establish the predictive-control model before actual control operations begin. This preliminary action allows the system to be ready for deployment without requiring time-consuming manual model construction at the point of use

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10048658B2Information processing device, predictive control method, and recording medium

Publication Date: 2018.08.14 NEC CORP

AI summary

An information processing device according to the present invention, includes: an information accumulation unit which receives and accumulates control-target information that includes information related to a control target and an environment including the control target; a prediction-equation-set learning and generation unit which learns and generates a prediction-equation set to be used for determination of an operation quantity of the control target based on the control-target information accumulated in the information accumulation unit; and an operation-quantity determination unit which receives input information needed for determination of an operation quantity of the control target, constructs a predictive-control model of the control target based on the prediction-equation set, the control-target information accumulated in the information accumulation unit, the control-target information received, and the input information, and determines an operation quantity used for control of the control target.