Indoor environment model creation device
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing air-conditioning system prediction methods require numerous parameters, including the number of people and building specifics, making accurate load prediction challenging, and previous methods like CO2 concentration-based estimation suffer from low accuracy.
Innovation Solution
An air-conditioning control device that integrates CO2 and humidity sensors with a data storage unit and model parameter learning unit to create an indoor environment model using physics models for heat, moisture, and CO2 concentration, learning common and unique parameters simultaneously to predict indoor conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based models with multiple parameters (heat load model, ventilation model) are used to predict indoor environment, then prediction accuracy is improved, but the number of required setting parameters increases significantly
Solution Approach 1:
The patent combines heat load prediction model and ventilation prediction model into a unified indoor environment prediction system. By merging these models and using integrated learning, the system reduces the number of separate parameter settings required while maintaining comprehensive prediction capability across multiple environmental factors.
Solution Approach 2:
The system performs self-learning by automatically acquiring setting parameters through machine learning from operational data and measurement data. This eliminates the need for manual parameter configuration and reduces complexity while improving prediction accuracy through data-driven parameter optimization.
2Ease of operation
If CO2 concentration measurement alone is used to estimate number of people and predict heat load, then parameter setting is simplified, but prediction accuracy deteriorates
Solution Approach 1:
The patent creates a multi-functional prediction system that simultaneously predicts indoor temperature, humidity, and CO2 concentration using a unified model. This integrated approach allows the system to use multiple data sources (operation data and measurement data) to improve heat load prediction accuracy while maintaining ease of operation through automated learning.
3Ease of manufacture
If multiple physics models are learned separately, then each model can be optimized independently, but overall prediction accuracy and consistency deteriorate
Solution Approach 1:
The patent implements integrated learning that simultaneously optimizes parameters across heat load prediction model, ventilation prediction model, and indoor environment prediction model. This unified optimization approach ensures parameter consistency and improves overall prediction accuracy while maintaining the flexibility to develop each model component separately.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An indoor environment model creation device is configured to create an indoor environment model of an indoor space in which air-conditioning equipment configured to condition air, an indoor humidity sensor configured to measure an indoor humidity of the indoor space, and a CO2 sensor configured to measure a CO2 concentration in the indoor space are installed. The indoor environment model includes a plurality of physics models in which heat, moisture, and CO2 concentration parameters are included. The indoor environment model creation device includes: a data storage unit configured to store operation data of the air-conditioning equipment in a learning target period as learning-use input data, and store measurement data measured by the CO2 sensor and the humidity sensor; and a model parameter learning unit configured to comprehensively learn the plurality of physics models with use of the learning-use input data and the measurement data, which are stored in the data storage unit.