The statistical hybrid approach we use is a unique empirical predictive system that requires only a fixed sample of paired process and emissions data. A statistical hybrid PEMS provides the following features:
- A robust model that is highly accurate across the full load range of the unit
- Valid for normal operations and during transitional states such as startup and shutdown
- Equivalent accuracy to that of a CEMS with superior reliability – tied to the plant control system
- Flexibility of implementation to use existing process instrumentation and data interfaces
- Certified as an alternative system under S. EPA for continuous compliance monitoring
- Can be assessed using quality control procedures required under programs of the S. EPA
- Can be developed and maintained by non-technical on-site staff or consultants
- Can be tested against EPA reference methods
- Has been demonstrated under 40 CFR Part 60, Performance Specification 16
- Has been demonstrated under 40 CFR Part 75, Subpart
The statistical hybrid model exploits the existing statistical relationships of the historical training data set depending on the input parameters available and how they are represented in the empirical data. The historical data set is fixed prior to certification when used in compliance monitoring. This allows the PEMS to calculate a model envelope that defines the operating conditions represented in the historical training data set.
Alarms can be configured to detect when the process is operating outside the model envelope. All normal operating conditions including startups, shutdowns, and transitional states can be tested such that envelope excursions are minimized. This type of historical training data set (containing all normal operating conditions) is deemed to be ‘robust’. Robust statistical hybrid models produce minimal monitor downtime over extended periods.
SmartCEMS® can be installed to replace an older CEMS or in lieu of CEMS on a new installation. SmartCEMS® was designed to work as a stand-alone solution to provide data acquisition, display of real-time and averaged emission data, and report generation to meet regulatory compliance requirements. In one mode of operation, the CEMS provides a critical role in the optimization of the process, but not a mandatory role in compliance monitoring of the exhaust emissions.
In most instances, however, SmartCEMS® is deployed using historical CEMS or temporary mobile CEMS data. SmartCEMS® provides real- time analysis of process efficiency and predictive capability for any failed input in addition to emissions monitoring for compliance purposes. SmartCEMS® can operate without a CEMS when certified as a primary continuous monitoring system for the source through a petition for approval of an alternate monitoring system (40 CFR Part 75, Subpart E) or utilizing a performance testing specification promulgated by U.S. EPA (40 CFR Part 60, Performance Specification 16).
If you have an existing CEMS, we would recommend that the CEMS not be removed, but that it is turned off during most of the year to save in operational, maintenance, and support costs. The CEMS would be used to retrain the model if new operating or ambient conditions are encountered following a major maintenance event and for tuning or optimization of the combustion controls.
The plant can certify the CEMS along with SmartCEMS®-75 as a backup system if a testing team is mobilized for a relative accuracy test audit. If the CEMS is operated according to 40 CFR Part 60 Reference Methods, the existing CEMS can be used by the source owner to recertify the SmartCEMS®-60 system each year in lieu of an annual certification test (typically performed by an independent third-party testing team) or for performing periodic audits. CMC Support, L.L.C. is available to provide support to both an existing CEMS and the SmartCEMS®-75 or SmartCEMS®-60 PEMS.
The accuracy of the system is entirely dependent on the quality of the training data set including the emission testing data and the site configuration of the SmartCEMS®. The accuracies achieved under the demonstration on gas turbines and boilers to date have been comparable to a CEMS, achieving better than 10% relative accuracy even at very low levels of NOx (less than 10 ppmv).
SmartCEMS® has consistently met the EPA requirement of less than 7.5% relative accuracy for NOx compliance under 40 CFR Part 75. In many cases, SmartCEM® can achieve accuracy levels less than 5% and maintain this level of accuracy for many years running the same model. SmartCEMS® reliability approaches 100% in our installations, typically better than the majority of CEMS installed to date.
The minimum data required to validate a SmartCEMS® PEMS model for compliance with 40 CFR Part 75 is 720 operating hours. There is an additional requirement to provide at least 24 hours under each type of fuel combusted as part of the 720 hour demonstration required of Part 75, Subpart E. The statistical analysis includes the F-test, t-test, correlation, and variance analysis along with a bias test of the 720 hour Subpart E demonstration dataset. CMC provides a Subpart E package that guarantees Part 75 Subpart E certification results that are successful and includes all statistical analyses, data quality assurance, and onsite support along with the required submittals and reports. The final decision for approval of all Part 75, Subpart E petitions resides with the Administrator of U.S. EPA.
The minimum data required for a SmartCEMS® PEMS model is from 24 hours to 72 hours of data to build a successful model. The entire range of operations at various loads and including a startup and/or shutdown is typically required. The initial certification under PS-16 requires a 27-run relative accuracy test audit and statistical analysis of the data. CMC provides a certification guarantee with each SmartCEMS® product such that the initial testing will meet the requirements of PS-16 and CMC remains involved with the required submittals and reports until notification is received by the appropriate regulatory agencies.
Quality assurance requirements under PS-16 include the initial 27 run RATA test and three quarterly audits during the first year of operation of the PEMS. During the second and subsequent years, the annual RATA test is conducted at a single load (normal) and only one audit is required during the year. The relative accuracy audit (RAA) is typically conducted in the opposing quarter of the year as the RATA.
The validity of the SmartCEMS® PEMS model is entirely dependent on the training data set and the quality assurance program in place at the site. If a comprehensive model is developed initially, there would be no reason to limit the range of time that it would be valid.
An initial model that includes all normal operations, startups, shutdowns, and transitional operating states is deemed ‘robust’ and will provide valid predictions for compliance purposes for many years. The model can easily be retrained, if required, for any extreme, ambient, or other operating conditions not encountered during the initial training data collection.
The system keeps track of when the model is valid for a given set of process data in real-time and provides alarms for tracking excursions from the established model envelope and records the status of the model prediction automatically. Requirements for a periodic or annual recertification test would be dependent on the local regulatory requirements and the facility quality assurance program.
We would recommend retraining the PEMS model (and recertification if required) following any major process change, such as the addition of pollution controls or for changes in fuel and fuel quality. Validation of the PEMS model can be conducted following tuning or annual maintenance activities (such as an extensive overhaul) but this would not be mandatory. The PEMS can be retuned at any time (periodically or continuously) using existing CEMS equipment or by mobilizing temporary or mobile emission monitoring equipment and collecting the process data concurrently with the target pollutant emission rates.
The accuracy of the prediction for NOx mass emission rates from gas turbines and gas-fired boilers have been the most extensively studied and demonstrated to date. Other pollutant emission rates that have been established to be accurately modeled by SmartCEMS®-60 including CO, SO2, and hydrocarbon emission rates, as well as, exhaust gas diluent O2 and CO2.
Other turbine or boiler parameters such as exhaust temperature and flow rate have also been evaluated and have shown good accuracy. Hydrocarbon, CO, SO2, NOx, and O2 have also been demonstrated to be accurately modeled from sewage sludge incinerators and industrial boilers.
The prediction is valid for any emission or process data that can be continuously measured and included in the initial training data set that can be correlated with the available process data. The model is initially developed with NOx, CO, CO2, O2, SO2, and total hydrocarbon data. Particulate, H2S, NH3, PM 10, PM 2.5, opacity, and other emission parameters can also be predicted with the statistical hybrid engine.
SmartCEMS® utilizes any quality assured process input parameter that correlates with emission data for model generation. Typically, around a dozen parameters are utilized in a simple cycle gas turbine or gas-fired boiler PEMS model with more parameters required associated with any add-on pollution control technologies.
Ambient data is utilized when it is available. There are no mandatory inputs required for a given model (with the exception of the unit load in megawatts and the fuel flow rate). The system analyzes the training data set to determine which of the available inputs or control system data is relevant to the model. Instrumentation and field devices that are determined to be used in the emission model may be subjected to additional quality controls per the local regulatory requirements and the site quality assurance program.
Interfacing with existing control and emission monitoring systems can be accomplished using serial communications, Ethernet connectivity, or through hard wiring to remote I/O devices. CMC can provide the hardware and equipment required or the system can be configured to utilize any standard ‘Windows’ based technology such as OPC, ODBC, OLE, DDE, Modbus, etc. to extract the data from the turbine control system.
This interface between existing control and monitoring systems with the PEMS server deployed on site can eliminate the need for hard wiring I/O to the SmartCEMS® system. Hard wire interfaces range from economical Ethernet-ready devices to a custom programmable controller with data-loggers and data buffering with uninterruptible power supply guaranteeing near 100% data availability.
The model is valid during normal operation of the turbine or boiler, during start-ups and during shutdowns, as well as, trips, interruptions, etc. Each of the deployed SmartCEMS® PEMS to date have been certified to continuously monitor emissions during all normal operations including startups and shutdowns.
The accuracy of the model during startups and shutdowns is demonstrated, however, the accuracy is improved as more transition data is included in the historical training data set. It is important to include startup and shutdown data along with other data from transient operations in the initial historical training data set to ensure accuracy of predictions for these operating conditions.
The model was developed for the purpose of providing operators with real- time feedback as to actual pollutant emission rates, unit efficiency, combustion efficiency, and compliance status. The rate in change of unit efficiency normalized for ambient and other operating conditions can be used to determine when maintenance should be scheduled, and following completion of maintenance activities, if the unit was restored to optimal combustion efficiency.
Yes, SmartCEMS® PEMS provides more than just efficient and cost-effective compliance emissions monitoring. The statistical hybrid engine can be used for plant and combustion process optimization and for tracking noncompliance emissions for waste minimization purposes or for process monitoring in applications where continuous monitoring with gas analyzers is extremely difficult or expensive to maintain.
CMC has certified over one hundred sources at various sites in the U.S. and abroad including power generation, automotive, university, municipal, governmental, and research facilities. SmartCEMS® PEMS has been applied to gas turbines (both simple cycle and combined cycle) of various sizes from microturbine, small turbines, and mid-size turbines all the way up to the largest turbines manufactured.
These turbines are of varying configuration with all types of NOx controls including dry low NOx, steam and water injection, solid catalytical reduction, and custom firing controls. Boiler applications include small to large size units of various configuration and those equipped with post combustion controls.
In addition, CMC has provided PEMS at a number of ethanol facilities and automotive or university steam plants. Please contact CMC for a detailed list of references and contacts from the regulatory agencies who are familiar with the CMC product line and the performance of the statistical hybrid PEMS.
Contact Mr. Brian Swanson or Brad Wessel at CMC Solutions (U.S. 248-960-1632 or 808-679-8767) or via email at Sales@CMCSolutions.org or you can fill out a request online. The first step in the process will be to develop a site-specific datasheet for each of the sources to be monitored to identify the specific needs of the customer and the regulatory program.
CMC will provide a proposal that may include computer hardware, temporary CEMS, mobile emission testing services, training, onsite support, certification testing, commissioning, warrantee, and quality assurance programming onsite. Together with our business alliance partners, CMC can provide a turnkey quotation for products and services as part of a facility wide emissions tracking program that can meet the requirements of your air permit and applicable federal regulations.
The software will be delivered (typically within 30 to 60 days of receipt of purchase order) and certification can be scheduled anytime following startup and collection of the training data. Training data is typically collected for a period of three days and up to 30 days during normal operation to develop the model. Sites with existing CEMS may have available data to build a model without additional testing or training. Certification can typically be achieved within 60 to 90 days of receipt of your purchase order for a SmartCEMS® PEMS.
The engine is a unique statistical hybrid model. It is neither a neural network nor a first principle type, but it is an empirical method. All CMC PEMS run the same core module – the statistical hybrid engine. The model does not use a theoretical methodology such as a first principle formula nor does it require an iterative model development and testing regimen with experts onsite. There is no specialized staff required to build or maintain the emission model. Continuous Emissions Monitoring (CEM) or Reference Method (RM) test data is used to build the initial model. The empirical SmartCEMS®model utilizes historical data, paired emissions data and process data, to generate predictions in real-time. The predictions are derived directly from the historical training dataset using input parameters from the process that are available from the existing control system and are configured in the PEMS model.
Unlike more complicated empirical systems such as neural network and first principle formulations, the statistical hybrid model can be developed for any given process without any knowledge of the underlying combustion chemistry or of the pollutional controls. The system uses the historical data collected during normal operations and during startups, shutdowns, or transitional states to accurately predict emissions for compliance purposes over the full load range of the unit. This feature allows the system to predict the value of any failed input (or pollutant emission data in lieu of CEMS) with the accuracy of the system entirely dependent on the range and quality of the original data in the training dataset. The system is deterministic in that a given set of inputs (process variables or lack of them) along with a given fixed training dataset will yield a single result for each value to be predicted. The model is unique in its ability to be developed by non-specialized staff that have no familiarity with the process, pollution control devices, or the methodology used by the model. Customers and third party consultants can update the model without support of the manufacturer’s engineering staff.
Once every day, at a minimum though this could be done more often, a test vector is presented to the PEMS. This represents the process during some normal load condition that matches data we have taken from RATA testing. We know what the emissions should be, so we can evaluate the resulting prediction against the target (RM) value. They should be very close and not change day to day.
This tests the entire prediction cycle and the database such that any change in either would result in a change in the prediction for the test vector (analogous to a calibration drift). In actuality, we do not make such changes between RATAs. Thus, the response from the PEMS to the test vector should not deviate at all. Only when we make adjustments to the model just prior to RATA, will the value change.
This is noted in the QA manual and after the RATA the daily validation is compared to the new target. This allows us to treat this daily validation just like a CEMS calibration. We also present a zero vector (offline) to get zero pollutants and ambient O2 values also. This completes a zero/span calibration record and allows the data acquisition system to process and handle the calibration much like a CEMS daily calibration.
Each of the sensors are validated for each cycle of the prediction (every 20 seconds). Sensors that are deemed to have failed (the readings are outside the envelope or not normal for a given load condition) are flagged and not used in the prediction. In some cases, we can generate a prediction for the failed sensor and then predict the emissions.
This is not typically done by CMC; however, as we usually have redundant inputs for any critical parameters, this second level of predictions are not needed. The real-time display of the PEMS (inputs page) will show colors indicating the status of any failed sensors.
Non-critical parameters are, by definition, not important in maintaining accuracy at the required level (usually this is within 10%). We determine the critical parameters using an input failure test and analysis. This is done each time the model is adjusted. The input failure test is conducted similar to the daily calibration, only we fail various combinations of inputs and look at the deviation in the prediction (usually on a mass emission basis) taking into account the pollutant and diluent.
If a single input failure produces greater than 10% deviation in the predicted mass emission rate, it is deemed critical. Failures in the critical inputs, will generate an alarm and invalid flag that can be used to screen the data and can account for monitor downtime. If a combination of inputs yields a deviation greater than 10%, these inputs are also deemed critical.
It is important to note that we have much flexibility in the use of inputs in our model, so by properly designing the model and using reliable inputs, we can minimize the critical inputs and reduce monitor downtime. When the process goes outside of the envelope (or finds a gap in the data within the envelope), the PEMS status is flagged invalid even if the prediction is accurate.
Correction to this problem is either to operate within the envelope or add data to the model to extend the envelope (or fill the gap) such that the PEMS status is valid. We always try to build models up front that are robust – that includes all of the normal operating conditions such that monitor downtime is less than 1% per quarter.
No, but it is important to have startup and shutdown data in the model in order for it to be robust. This allows the PEMS to provide accurate predictions for these normal operations. The PEMS will process the data like any other load condition and generate the appropriate prediction.
There are some methods we use in building the model that account for process hysteresis and other time related emission profiling that is typically associated with startup and shutdown. As an example, let us assume a gas turbine typically operates at 40 to 100 MW load and the megawatt output is used as a critical input in the model.
During startup, the unit is ramped up at 10 MW, then after 15 minutes the load is applied and brought up rapidly to 40 MW. If the database only contains data from 40 MW to 100 MW and the envelop reflects this, we would be unable to generate valid predictions when the load was 0 MW. If we include startup data (with 0 MW) in the database, the envelope would be extended such that 0 to 100 MW is covered.
If this turbine runs at 20 MW, the model will not show an envelope excursion, but will also not generate valid data at that point (the PEMS status will be invalid). Again, our solution is to add the 20 MW data to the database. It is always better to add data in correcting a model deficiency, as this allows the model to get progressively better over time with a wider envelope and more robust database underlying the predictions.
The same is true for dual fuels or multiple fuels. Each different fuel combination uses the corresponding data in the database. We do not need to treat the dual fuel condition differently than the single fuel or startup data. We just need to ensure we have the data in the database to handle all the different normal fuel firing conditions.