Air-Conditioner Startup Control for Sparse-Data Time Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing air-conditioning control devices using machine learning models for estimating startup times face inaccuracies when insufficient training data is available, leading to inappropriate predictions.

Innovation Solution

An air-conditioning control device that employs a non-learning model for extrapolation in sparse data regions, using a training data group with both dense and sparse regions to improve estimation accuracy by determining whether data is in a sparse region and applying extrapolation techniques.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a machine learning model is used to estimate startup time, then estimation accuracy is improved when sufficient training data is available, but estimation accuracy deteriorates when training data is insufficient

Engineering Contradiction:
Improvestartup time estimation accuracyVSAvoidprediction reliability in sparse data regions
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the data space into dense regions and sparse regions based on training data distribution. The determination unit identifies whether input air-conditioning data falls into a sparse region, and the estimation unit selectively applies different estimation strategies (machine learning model for dense regions, extrapolation for sparse regions) accordingly. This segmentation resolves the contradiction by adapting the estimation method to the data availability in each region.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a dynamic estimation system that adapts its behavior based on the characteristics of the input data. The determination unit dynamically assesses whether the input data resides in a sparse region, and the estimation unit dynamically switches between the machine learning model and extrapolation methods. This dynamic adaptation ensures reliable estimation across varying data conditions, resolving the contradiction between ML accuracy and sparse data reliability.

Inventive Principle:
Principle #15Dynamics

2Reliability

If extrapolation using a non-learning model is applied in sparse regions, then estimation reliability is improved, but device complexity increases

Engineering Contradiction:
Improveestimation reliability in sparse regionsVSAvoidcontrol system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies different quality levels of estimation methods to different regions of the data space. In dense regions where sufficient training data exists, the simpler machine learning model is used. In sparse regions where reliability is critical, the more robust extrapolation method is applied. This local differentiation improves overall reliability without unnecessarily increasing complexity in all regions, resolving the contradiction between reliability and device complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The determination unit acts as an intermediary that bridges the machine learning model and the extrapolation method. It assesses the data region characteristics and directs the appropriate estimation method to the estimation unit. This intermediary structure enables the system to achieve high reliability in sparse regions through extrapolation while maintaining simplicity in dense regions, thus resolving the contradiction between reliability improvement and complexity increase.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4033169B1Air-conditioning control device, air-conditioning system, air-conditioning control method, and air-conditioning control program
Publication Date: 2023.07.19 MITSUBISHI ELECTRIC CORP
  • EP4033169B1 patent drawingFigure 1
  • EP4033169B1 patent drawingFigure 2
  • EP4033169B1 patent drawingFigure 3

AI summary

An air-conditioning control device with improved estimation accuracy regarding a startup time of an air conditioner in a case where a machine learning model learning of which is performed in a state in which sufficient data is not stored is used is acquired. An air-conditioning control device to estimate a startup time of an air conditioner on the basis of a machine learning model which has performed, by using a training data group including a dense region and a sparse region having less training data than the dense region, learning for estimating the startup time of the air conditioner from air-conditioning data which is information regarding control of the air conditioner includes: an air-conditioning data acquisition unit to acquire the air-conditioning data; a determination unit to determine whether or not the air-conditioning data is present in the sparse region; an estimation unit to, in a case where the determination unit determines that the air-conditioning data is present in the sparse region, apply extrapolation using a non-learning model for associating the air-conditioning data with the startup time to the machine learning model, and estimate the startup time of the air conditioner from the air-conditioning data by using the non-learning model; and a control unit to perform control to start up the air conditioner at the startup time estimated by the estimation unit.