Building control system with peer analysis based on weighted outlier detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building control systems face challenges in accurately assessing the performance of HVAC assets due to insufficient or uncertain data, leading to skewed outlier detection results, which can result in inaccurate identification of faulty or underperforming units.
Innovation Solution
A system and method for performing peer analysis using weighted outlier detection, where peer metrics are generated and weighted based on the amount of operation data, allowing for accurate identification of outliers among HVAC assets by comparing their performance within peer groups, employing techniques like the Generalized Extreme Studentized Deviate test and modified Sequential Application of Wilk's Multivariate Outlier Test.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If peer analysis is performed using available operation data, then performance evaluation of HVAC assets can be conducted, but insufficient or uncertain data leads to skewed analysis results and inaccurate outlier identification
Solution Approach 1:
The patent applies parameter changes by introducing a weighting mechanism that transforms the quality assessment of data into a quantitative parameter. Each peer metric value is assigned a weight based on data sufficiency and certainty, allowing the system to adjust the influence of each data point in the outlier detection process. This resolves the contradiction by enabling accurate outlier identification even when data quantity varies across assets.
Solution Approach 2:
The system performs preliminary evaluation of data quality and sufficiency before conducting peer analysis. By pre-assessing whether sufficient data is available for each HVAC asset and calculating appropriate weights in advance, the system prevents skewed analysis results from insufficient data. This preliminary action ensures that only reliable data contributes meaningfully to the outlier detection process.
2Reliability
If traditional outlier detection methods are used without data weighting, then the analysis process is simpler, but the results are skewed when data availability varies across assets
Solution Approach 1:
The patent modifies the traditional outlier detection process by introducing a weighting parameter that reflects data quality and sufficiency. This parameter change transforms the simple unweighted comparison into a more sophisticated weighted analysis, improving reliability while adding manageable complexity through systematic weight calculation and application in the detection algorithm.
Solution Approach 2:
The system incorporates feedback mechanisms where the results of data quality assessment feed into the weighting process, which in turn influences the outlier detection outcomes. This feedback loop ensures that the analysis continuously adapts to the actual data quality, improving reliability by automatically adjusting the influence of each asset's data based on its sufficiency and certainty levels.
3Ease of operation
If peer metrics are calculated for all HVAC assets uniformly, then the peer analysis can be standardized, but assets with insufficient data produce inaccurate metric values that skew the overall analysis
Solution Approach 1:
The patent applies local quality by treating each HVAC asset's data differently based on its specific data quality characteristics. Instead of uniform treatment, the system calculates individual weights for each asset's peer metric values based on their data sufficiency and certainty. This allows standardization to be maintained in the overall process while accommodating local variations in data quality to ensure accurate metric values.
Solution Approach 2:
The system dynamically changes the weighting parameter for each peer metric based on the specific data quality of each HVAC asset. This parameter adjustment ensures that assets with sufficient and certain data have appropriate influence, while assets with insufficient or uncertain data contribute less to the overall analysis, maintaining both standardization and accuracy.
Data Source
AI summary
A system for assessing relative performance among a plurality of heating, ventilation, or air condition (HVAC) assets, the system comprising a processing circuit configured to identify a peer group comprising two or more of the plurality of HVAC assets having a common characteristic, generate values of a peer metric for the HVAC assets in the peer group based on corresponding operation data associated with the HVAC assets, perform a weighted outlier detection process using the values of the peer metric, and initiate an automated action in response to detecting an outlier HVAC asset.