Refrigerant amount determination device, method, and program
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining refrigerant amounts in air conditioning systems face challenges due to discrete parameter values affecting refrigerant amount index values, leading to poor prediction accuracy and increased data complexity, especially when dealing with multiple parameters.
Innovation Solution
A refrigerant amount determining device that includes an operation data acquiring unit, a calculating unit, an inferring unit, and a determining unit, which together calculate and correct refrigerant amount index values using continuous values, facilitating accurate determination even with multiple parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If discrete parameter values are used for prediction arguments in refrigerant amount determination, then the map data structure is simple, but the prediction accuracy deteriorates when step width is large
Solution Approach 1:
The patent changes the parameter type from discrete to continuous values. Specifically, it uses continuous values for refrigerant amount index values, outdoor temperatures, and compressor rotation speeds instead of discrete stepped values. This allows for more precise prediction without increasing map data complexity, as the neural network processes continuous parameters directly.
2Measurement precision
If the step width of argument parameters is reduced to increase prediction accuracy, then the prediction accuracy improves, but the amount of map data increases
Solution Approach 1:
The patent transitions from discrete parameters with small step widths to continuous parameters. This eliminates the need for fine-grained discrete steps while maintaining high prediction accuracy. The neural network model processes continuous values directly, avoiding the exponential growth of map data that would result from reducing discrete step widths.
3Reliability
If multiple parameter types are used for arguments, then the prediction comprehensiveness improves, but the map becomes multidimensional and data amount increases making implementation difficult
Solution Approach 1:
The patent introduces a neural network as an intermediary between the multiple input parameters and the prediction output. The neural network efficiently handles the multidimensional continuous parameter space, performing the complex correlations between refrigerant amount index values, outdoor temperatures, compressor rotation speeds, and other parameters without requiring explicit multidimensional map structures.
Solution Approach 2:
The patent uses continuous parameter representations that can naturally accommodate multiple variables. The neural network model processes multiple continuous inputs simultaneously, avoiding the combinatorial explosion that would occur with discrete multidimensional maps while maintaining comprehensive prediction capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A refrigerant amount determining device includes: an operation data acquiring unit configured to acquire operation data of an air conditioning system; a calculating unit configured to calculate a refrigerant amount index value from the operation data acquired; an inferring unit configured to infer information regarding correction of the refrigerant amount index value using a correction model and at least one of the acquired operation data or the calculated refrigerant amount index value; and a determining unit configured to determine a refrigerant amount of the air conditioning system based on the information regarding correction of the refrigerant amount index value.