Predictive Emissions Model for Sensor Validation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing continuous emissions monitoring systems (CEMS) face challenges in accurately replacing sensor readings that fall outside the operating envelope, as per the U.S. EPA's Performance Specification 16, which requires daily detection of defective sensors and operation within minimum and maximum values, with conventional methods not fully aligning with these standards.

Innovation Solution

A predictive emissions monitoring method that utilizes an emissions prediction model to determine emission levels based on input or modeled input values, evaluating the need for replacement and ensuring the effect of replacement is acceptable, using input prediction models to generate biased values that align with sensed inputs, and performing relative accuracy tests to verify compliance with regulatory standards.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If sensor readings outside the operating envelope are replaced with reconstructed values from training dataset, then the system maintains operation within the envelope, but this approach is inconsistent with PS-16 requirements and may compromise measurement accuracy

Engineering Contradiction:
Improvecompliance with PS-16 standardsVSAvoidsensor value accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent replaces the conventional mechanical approach of unconditional value substitution with a predictive emissions model that uses neural networks and multiple sensor inputs to calculate actual emissions, substituting the need for direct sensor measurements in problematic conditions

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

Solution Approach 2:

The system changes the parameter validation approach by evaluating whether sensor readings are 'grossly in error' based on multiple criteria including consistency with other sensors and predicted emissions, rather than simply checking against fixed training dataset ranges

Inventive Principle:
Principle #35Parameter changes

2Reliability

If strict operating envelope validation is enforced, then PS-16 compliance is achieved, but sensor drift detection and replacement capability is reduced

Engineering Contradiction:
ImprovePS-16 complianceVSAvoidsensor drift handling
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamic validation where the operating envelope is not fixed but adapts based on predicted emissions and multiple sensor readings, allowing the system to accommodate sensor drift while maintaining compliance through the predictive model's ability to identify and compensate for erroneous values

Inventive Principle:
Principle #15Dynamics

3Reliability

If multiple sensor inputs are used for emissions calculation, then measurement reliability improves, but system complexity increases

Engineering Contradiction:
Improveemissions measurement accuracyVSAvoidsensor validation system
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The predictive emissions model serves multiple functions simultaneously: it calculates emissions, validates sensor readings, detects sensor failures, and provides reconstructed values when needed, reducing overall system complexity despite using multiple sensors

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

Data Source

PatentEP2423768B1Sensor validation and value replacement for continuous emissions monitoring
Publication Date: 2020.02.12 ROCKWELL AUTOMATION TECH INC
  • EP2423768B1 patent drawingFigure 1~2
  • EP2423768B1 patent drawingFigure 3
  • EP2423768B1 patent drawingFigure 4

AI summary

A continuous emissions model system is described that employs an emissions model that can determine emissions values from a plant for use in place of sensed emission data in the event of failure or lack of communication from a sensor. The emissions model itself receives inputs that may be based upon sensed data. Each emissions model input may be substituted with a modeled input. The modeled inputs, as well as the emissions model outputs maybe biased to render them more accurate. If any one of the emissions model inputs fails (e.g., becomes unavailable or is clearly erroneous), the corresponding modeled input may be utilized so long as the modeled input passes an acceptability test.