Building Heat Generator Thermal Power Prediction With Machine Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing thermal power prediction systems for heat generators in buildings do not account for the actual operating conditions of the generation plant, leading to inefficiencies and energy waste due to mismatched climate correlations and lack of closed-loop regulation.

Innovation Solution

A computer-implemented method using machine learning to predict thermal power by detecting environmental and generator parameters, applying linear and non-linear regression models to calculate coefficients, and optimizing generator settings for energy efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If standard climate correlation curves are used for thermal power prediction, then the control system is simple to implement, but the prediction accuracy does not match the actual generator performance requirements

Engineering Contradiction:
Improvecontrol system implementationVSAvoidthermal power prediction accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The system transforms the prediction approach by changing from using fixed standard climate correlation curves to dynamically adjusting prediction parameters based on actual generator operating conditions. The machine learning model learns optimal parameter transformations from historical data, adapting the correlation between external temperature and required thermal power to match specific generator characteristics and performance curves.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system implements closed-loop feedback by continuously comparing predicted thermal power with actual generator output and using this information to refine future predictions. The machine learning model is trained on historical operational data that includes actual generator responses, creating a feedback mechanism that improves prediction accuracy over time while maintaining system simplicity.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If closed-loop regulation based on real generator performance is implemented, then the thermal power prediction accuracy is improved, but the system complexity increases

Engineering Contradiction:
Improvethermal power prediction accuracyVSAvoidcontrol system structure
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system employs self-service by automatically training the machine learning model using historical operational data from the generator itself. The model learns generator-specific characteristics and performance curves directly from the plant's own historical data, eliminating the need for complex manual calibration or external expertise while achieving high prediction accuracy tailored to the specific generator.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The machine learning model serves multiple functions simultaneously: it predicts thermal power requirements, characterizes generator performance, and adapts to different operating conditions. This multi-functionality reduces overall system complexity by consolidating what would otherwise require separate systems into a single integrated solution.

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

3Adaptability or versatility

If machine learning models are trained on historical operational data, then the prediction adapts to actual generator performance, but the data processing and model training complexity increases

Engineering Contradiction:
Improveprediction adaptation to generator performanceVSAvoiddata processing system
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system replaces complex mechanical calibration and adjustment procedures with automated machine learning algorithms. Instead of manually tuning system parameters to match generator performance, the ML model automatically learns the relationships from historical data, substituting automated computational processes for manual mechanical adjustment and reducing operational complexity.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4266134B1A computer-implemented method for training a machine learning system to predict the thermal power supplied by a heat generator located in a predefined environment
Publication Date: 2025.07.09 IMMERGAS
  • EP4266134B1 patent drawingFigure 1A
  • EP4266134B1 patent drawingFigure 1B
  • EP4266134B1 patent drawing

AI summary

Computer-implemented method (100) for training a machine learning system to predict the thermal power supplied by a heat generator located in a predefined environment, particularly in a building, the method (100) comprising the following steps: detecting a set of temperature values of the air of the predefined environment (101); detecting a set of delivery and return temperature values of a heat transfer fluid (102, 103); calculating hourly average values on the sets of values detected (110); calculating values of hourly average power supplied by the heat generator (111); applying a linear regression model (112) to the calculated average power values; applying a non-linear regression model (113) to the average hourly values of the temperature of the air of the predefined environment, to the average hourly values of the delivery temperature of the heat transfer fluid, and to the calculated average power values.