Neural Network Sensor Fusion for Missing HVAC Temperature Data Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Building heating and cooling systems face challenges in controlling temperature across large spaces due to sparse sensor data, leading to inefficient heating and cooling, as many areas lack simple ways to measure state values, resulting in temperature variations and difficulty in altering conditions.
Innovation Solution
A method using a neural network with test and target neurons to compute accuracy, where known sensor values are used to produce modeled values, compare them to actual values, calculate connection strengths, and determine the quality of test neuron values, enabling the determination of missing sensor values and improving data fusion and quality assessment in controlled systems like HVAC systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensors are installed throughout the building to measure temperature in all areas, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent introduces a neural network model as an intermediary that processes data from existing sensors to infer temperature values in areas without direct sensor measurement. The model acts as a mediator between sparse sensor data and comprehensive temperature mapping, enabling accurate temperature estimation without installing sensors everywhere.
Solution Approach 2:
The patent creates virtual copies of sensor measurements through the neural network model. Instead of physically installing sensors in every location, the system generates modeled temperature values that replicate what sensors would measure if they were present, based on patterns learned from existing sensor data.
2Measurement precision
If more sensors are deployed to cover large spaces, then measurement precision improves, but loss of energy increases due to more devices
Solution Approach 1:
The system replaces physical sensors with virtual sensor copies generated by the neural network model. These modeled measurements provide the necessary data coverage without the energy consumption of actual sensor hardware deployed throughout the building.
3Device complexity
If sensors are placed only near walls due to practical constraints, then device complexity is reduced, but measurement precision deteriorates for central areas
Solution Approach 1:
The neural network model serves as an intermediary that bridges the gap between wall-mounted sensors and central area temperature measurement. It processes the limited sensor data and infers temperature conditions in central areas where no physical sensors are installed.
Solution Approach 2:
The patent transitions from physical spatial distribution of sensors to a mathematical dimension through the neural network model. Instead of placing sensors physically throughout the space, the system uses computational modeling to extend measurement capability to all areas including centers, effectively moving the solution from physical to mathematical domain.
4Measurement precision
If a neural network model is used to infer missing sensor values, then measurement precision for uncovered areas improves, but device complexity increases due to modeling requirements
Solution Approach 1:
The neural network model is trained to be self-sufficient in generating accurate temperature predictions. Once trained on available sensor data, the model independently infers temperature values in uncovered areas without requiring additional sensors or complex external systems, serving its own purpose of comprehensive temperature mapping.
Data Source
AI summary
An unknown state value in a structure neuron value in a neural network, in one embodiment, is determined by using the difference between known values and output at an equivalent model location. The accuracy of model produced values with known values are determined compared to the known values. How much the known model produced locations were used to determine the unknown state value is determined. These amounts and accuracy of the model produced values are used to determine accuracy of the model produced value of the unknown state value.