Continuous Emission Monitoring System Research Articles

Machine Learning for Advanced Emission Monitoring and Reduction Strategies in Fossil Fuel Power Plants

Fossil fuel power plants are a significant contributor to global carbon dioxide (CO2) and nitrogen oxide (NOx) emissions. Accurate monitoring and effective reduction of these emissions are crucial for mitigating climate change. This systematic review examines the current state of research on the application of machine learning techniques in evaluating the emissions from fossil fuel power plants. This review first briefly introduces the continuous emission monitoring (CEM) systems and predictive emission monitoring (PEM) systems that are commonly used in power plants and highlights that machine learning models can significantly improve PEM systems through their capability to process and interpret large datasets intelligently to transform traditional emission monitoring systems by enhancing their precision, effectiveness, and cost-efficiency. Compared to previously published review articles, the key contribution and innovation in this present review is the discussion of machine learning models in CO2/NOx emissions according to the different algorithms used, including their advantages and disadvantages in a systematic way, which aims to help future researchers to develop more effective machine learning models. The most popular machine learning model includes reinforcement learning, a forward neural network, a long short-term memory neural network, and support vector regression. While each model method has its own advantages and disadvantages, we noted that training data quality, as well as the proper selection of model parameters, plays an important role. The challenges and research gaps, such as model transferability, a deep understanding of the physics of CO2/NOx emissions, and the availability of high-quality data for training machine learning models, are identified, and recommendations as well as potential future research directions to address these challenges are proposed and discussed.

Leveraging long short‐term memory networks and transfer learning for the soft‐measurement of flue gas flowrate from coal‐fired boilers

AbstractThe dynamic operation and deep peak‐shaving of power‐generating units cause significant fluctuations in flue gas flowrate, thus affecting the accuracy of CO2 emissions measured by continuous emission monitoring systems (CEMS). This study established a long short‐term memory network with an attention mechanism (LSTM‐AM) for the soft measurement of the flue gas flowrate in real‐time. First, flue gas flowrate data and continuous operation parameters over 25 days were sampled from a typical 660 MW coal‐fired boiler in China. Then, a carbon balance model was established to verify the data reliability. The LSTM‐AM model was trained and testified at the 660 MW coal‐fired boiler. Results show that the LSTM‐AM model significantly surpassed the pristine LSTM model without attention, the convolutional neural network (CNN) with LSTM, and the static support vector regression (SVR) model in the real‐time prediction of flue gas flowrate. Finally, the LSTM‐AM model was generalized to a 630 MW coal‐fired power unit via transfer learning, which was further demonstrated to outperform the model re‐trained from scratch. This work manifests the feasibility of deep learning for the soft measurement of flue gas flowrate, which is promising to solve data‐lagging issues when measuring CO2 emissions from coal‐fired power plants.

Combined experimental and theoretical studies on adsorption characteristics of graphitic carbon nitride for flue gas molecules

Graphitic carbon nitride (g-C3N4) is a novel selective adsorbent for separation of Hg0 and Hg2+, which plays a pivotal role in mercury continuous emission monitoring systems (Hg-CEMS). However, some of the flue gas components may interfere with their binding characteristics with g-C3N4. This study employs the density functional theory (DFT) and experimental method to investigate the adsorption behavior of gas components on g-C3N4 and elucidates the intrinsic driving forces behind these interactions and assesses their impact on separation of Hg0 and HgCl2. The binding energies of H2O, SO2, CO2, NO, and O2 molecules on g-C3N4 are −0.575 eV, −0.530 eV, −0.309 eV, −0.246 eV, and −0.238 eV, respectively. The electronic structure and weak interaction analyses indicate that these interactions are primarily van der Waals forces caused by multi-atomic interactions. In contrast, the binding energy between HCl and g-C3N4 is −0.557 eV, primarily controlled by weak H-N covalent bonds. The co-adsorption results indicate that flue gas molecules exert negligible effects on the binding of HgCl2 to g-C3N4 because the changes in binding energy are within 0.15 eV. The experimental results verify that under practical operating conditions, the adsorption of H2O, HCl, SO2, CO2, NO, and O2 on g-C3N4 is below 0.1 %, demonstrating those components do not interfere with the separation of Hg0 and HgCl2 by g-C3N4, nor do they affect the regeneration performance of g-C3N4. This study provides new insights into the development and practical application of separation and conversion modules in Hg-CEMS technologies.

Soft sensing of NOx emission from waste incineration process based on data de-noising and bidirectional long short-term memory neural networks.

Continuous emission monitoring system is commonly employed to monitor NOx emissions in municipal solid waste incineration (MSWI) processes. However, it still encounters the challenges of regular maintenance and measurement lag. These issues significantly impact the accurate and stable control of NOx emissions. Therefore, developing a soft NOx emission sensor to complement hardware monitoring becomes imperative. Considering data noise, dynamic nonlinearity, time series characteristics and volatility in the MSWI process, this article introduces a soft sensor model for NOx emission prediction utilizing the complete ensemble empirical mode decomposition adaptive noise (CEEMDAN)-wavelet threshold (WT) method and bidirectional long short-term memory (Bi-LSTM). Firstly, the original data signal is decomposed into a group of intrinsic mode functions (IMFs) using the CEEMDAN. Subsequently, the WT processes the high-frequency IMFs that are noise-dominant. Then, all IMFs are reconstructed to obtain the denoized signal. Finally, the Bi-LSTM model is employed to predict NOx emissions. Compared to conventional modelling approaches, the model proposed in this article demonstrates the best predictive performance. The mean absolute percentage error, root-mean-squared error and average absolute error on the test set of the proposed model are 3.75%, 5.34 mg m-3 and 4.34 mg m-3, respectively. The proposed model provides a new method to soft sensing NOx emissions. It holds significant practical value for precise and stable monitoring of NOx emissions in MSWI processes and provides a reference for research on modelling key process parameters.

A comprehensive emission inventory of air pollutants from municipal solid waste incineration in China's megacity, Beijing based on the field measurements

The rising of municipal solid waste incineration (MSWI), constituting 5 % of NOx emissions in Beijing, poses a significant challenge to improving air quality. This study establishes a comprehensive historical inventory of air pollutants (APs) emitted from MSWI plants between 2004 and 2023. The inventory was developed using both the continuous emissions monitoring systems (CEMS)-based method and the EF (emission factors) -based method, incorporating detailed plant-level activity data and localized EF derived from field measurements. These include data from CEMS and manual monitoring. Analysis of CEMS data reveals high compliance rates with emission limits for MSW in Beijing, with 99.9 %, 99.5 %, 99.8 %, 98.7 %, and 99.5 % of units meeting standards for PM, SO2, NOx, CO and HCl, respectively. This suggests effective implementation of emission standards in Beijing, although further strengthening of policies, particularly for CO emissions, is warranted. Overall, total AP emissions have increased annually largely attributed to measures implemented for DeSOx, DeNOx, and DePM since 1998. Most MSWI facilities are located in suburban areas rather than urban cores. Emissions of SO2, HCl, CO, Hg, Cd + Ti, other metals, dioxins, VOCs, and NH3 exhibit a spatially homogeneous distribution at the district level, while PM and NOx emissions demonstrate heterogeneity. Scenario analysis underscores the importance of continuous improvement and upgrading of advanced air pollution control devices. This study contributes a methodological framework for estimating emissions, reducing uncertainties, and informing policy-making to mitigate APs emissions in megacities. It serves as a valuable reference for similar cities grappling with air quality challenges.

Data-Driven Modeling for the Prediction of Stack Gas Concentration in a Coal-Fired Power Plant in Türkiye

In this research, deep learning and machine learning methods were employed to forecast the levels of stack gas concentrations in a coal-fired power plant situated in Türkiye. Real-time data collected from continuous emission monitoring systems (CEMS) serves as the basis for the predictions. The dataset includes measurements of carbon monoxide (CO), sulfur dioxide (SO2), nitrogen oxides (NOx), oxygen (O2), and dust levels, along with temperatures recorded. For this analysis, deep learning methods such as multi-layer perceptron network (MLP) and long short-term memory (LSTM) models were used, while machine learning techniques included light gradient boosted machine (LightGBM) and stochastic gradient descent (SGD) models were applied. The accuracy of the models was determined by analysing their performance using mean absolute error (MAE), root means square error (RMSE), and R-squared values. Based on the results, LightGBM achieved the highest R-squared (0.85) for O2 predictions, highlighting its variance-capturing ability. LSTM excelled in NOx (R-squared 0.87) and SO2 (R-squared 0.85) prediction, while showing the top R-squared (0.67) for CO. Both LSTM and LGBM achieved R-squared values of 0.78 for dust levels, indicating strong variance explanation. Conclusively, our findings highlight LSTM as the most effective approach for stack gas concentration forecasting, closely followed by the good performance of LightGBM. The importance of these results lies in their potential to effectively manage emissions in coal-fired power plants, thereby improving both environmental and operational aspects.Graphical

Investigating the impact of control strategies on the sulfur dioxide emissions of South Korean industrial facilities using an aircraft mass balance approach

We present top-down SO2 emission estimates from four large point sources, including two power plants, a steel mill, and a petrochemical industrial facility in Taean, South Korea. The airborne observations were conducted over two years, with three intensive observations in Fall 2019, Fall 2020, and Spring 2021. During this period, an active policy implementation to reduce air pollutant emissions from large industrial point sources in the region had been exercised, in addition to activity disruptions from the global COVID-19 pandemic. We quantify the observed SO2 emission reduction over the time period resulting from the policy implementation. The Continuous Emission Monitoring System (CEMS) datasets from the coal power plants were 1) validated by comparing them with emissions calculated using a top-down airborne mass balance method and 2) compared with the annual bottom-up national SO2 emission inventory. The comparisons illustrate that the policy implementation resulted in a 35% reduction in SO2 emissions from increased implementation of SO2 reduction technologies such as scrubbers. Finally, the COVID-19 pandemic did not cause a meaningful disruption in SO2 emissions from industrial sectors, such as steel mills and petrochemical manufacturing facilities.

Full life cycle emission reduction potential of ultra-low emission transformation in China's cement industry

China accounts for more than 50% of the global total cement production, the emission of pollutants in the cement industry occupies a large proportion of the industries. Some provinces have first carried out in-depth treatment of the cement industry, and the national level will also introduce an implementation plan for ultra-low emission transformation of the cement industry to promote continuous improvement of China's air quality. In this work, a pollutant emission inventory covering the full life cycle of the cement industry, including organized emission, unorganized emission and transportation, was established based on pollutant discharge permit, Continuous Emission Monitoring System (CEMS) database and field investigation. And, the emission reduction potential of particulate matter (PM), sulfur dioxide (SO2) and nitrogen oxide (NOx) after the implementation of ultra-low emission transformation in cement industry was evaluated. The results show that the existing dedusting, desulphurization and denitration technologies can control the concentrations of PM, SO2 and NOx below 10 mg/m3, 35 mg/m3 and 50 mg/m3, respectively. The emission reduction potential of PM, SO2 and NOx in cement industry is 259.54 kt, 14.98 kt and 607.04 kt, corresponding emission reduction ratio is 37%, 24% and 64%, respectively. In addition, this work reveals the spatial characteristics of pollutant emission level and emission reduction potential of the cement industry, which is conducive to the rational layout of ultra-low emission transformation, and push forward the transformation task by area and stage.

Predicting air pollutant emissions of the foundry industry: Based on the electricity big data

Industrial enterprises are one of the largest sources of air pollution. However, the existing means of monitoring air pollutant emissions are narrow in coverage, high in cost, and low in accuracy. To bridge these gaps, this study explored a predicting model for air pollutant emissions from foundry industries based on high-accuracy electricity consumption data and continuous emission monitoring system (CEMS). The model has then been applied to the calculation of air pollutant emissions from foundries without CEMS and the optimization of air pollutant emission temporal allocation factors. The results reveal that electricity consumption and PM emissions during the 2022 Beijing Winter Olympics have the same ascending and descending relationship. Furthermore, a cubic polynomial model between electricity consumption and flue gas flow is established based on the whole year data of 2021 (R2 = 0.85). The relative errors between the PM emissions calculated by the model and the emission factor method are small (−17.09–24.12 %), and the results from the two methods revealed a strong correlation (r = 0.93, p < 0.01). In addition, the monthly PM emissions from foundries are mainly concentrated in spring and winter, and the daily emissions on weekends are significantly lower than those on workdays. These results can be useful for environmental regulation and optimization of air pollutant emission inventories of foundry industry.

Perhitungan Beban Emisi Particulate Matter berdasarkan Data CEMS dari PLTU Batu Bara Milik PT PLN (Persero)

In the Coal-Fired Power Plant (CFPP) operation, various pollutants will be generated, including Particulate Matter (PM), which can be monitored with the Continuous Emission Monitoring System (CEMS). The parameters monitored include PM concentration, volumetric flow rate, % O2, and the operation hours and/or CEMS monitoring hours. Emission load calculation by utilizing CEMS data is a Tier 3 calculation method that uses real operation data from a CFPP unit. Information related to PM emission load from CFPP can be considered for corporate decision making in developing emission control strategies. This research utilizes secondary data from January 2021 to July 2023 period, from 52 PT PLN (Persero)’s CFPP units with various boiler, installed capacity, and types of Air Pollution Control Devices specifications. The largest average/month emission load value is 139,632 kg PM/month for CFPP unit that uses PC boiler, have a capacity of &gt; 600 MW, and use low NOx burners and ESP as its APCD; while the smallest average/month emission load value is 525 kg PM/month for CFPP unit that uses PC boiler, have a capacity of 101-300 MW, and use ESP as its APCD. If the average emission load/month from January 2021 to July 2023 is compared with the maximum emission load/month with a PM quality standard of 100 mg/Nm3, all CFPP units do not exceed it and are generally ready to face the tightening of PM quality standard up to 75 mg/Nm3.

Penentuan Beban Emisi Karbon Dioksida PLTU Batubara Pulau Jawa dari Hasil Pengukuran CEMS

Climate change is a real phenomenon that occurs and is felt by all living things that live on earth. The increase in earth's surface temperature is one of the impacts of climate change that continues to occur and has the potential to threaten the sustainability of human life. Carbon dioxide is the main greenhouse gas that exacerbates this condition. One of the largest sources of carbon dioxide emissions comes from the energy sector, namely coal-fired power plants (CFPP). Java Island has CFPP’s with the largest total installed capacity in Indonesia, even the capacity will continue to be added by 8,5 GW or 39,4% until 2030. In its operation, the CFPPs have an air emission measuring device before being discharged into the atmosphere which works continuously called the Continuous Emission Monitoring System (CEMS). To be able to take appropriate climate change mitigation and adaptation steps, CO2 emission load data that has high accuracy and is analyzed directly is needed. The CO2 emission load generated from 20 CFPP units as the object of research is 88,56 million tons of CO2/year. The greater the generating capacity, the greater the CO2 emissions produced. The higher the quality of coal used, the lower the CO2 emissions tend to be. To support global efforts to combat climate change, mitigation actions are needed to reduce CO2 emissions into the atmosphere.

Development of Emission Factors from Indonesian Coal-Fired Power Plant Using Continuous Emission Monitoring Data

Continuous Emission Monitoring System (CEMS) is generally used for monitoring compliance with emission standards set by the government regulations and has not been optimally used for other additional purposes yet. If operated CEMS can produce reliable and accurate data, they can develop further specific data such as emission factors. These emission factors can be used for estimating pollutant emission loads from coal combustion activity in Coal-Fired Power Plants (CFPPs) without conducting direct source measurements. In this study, hourly 1 yr CEMS data from several units of CFPPs were processed to develop specific emission factors for principal air pollutants (SO2, NOx, particulates) and greenhouse gases (represented by CO2). Emission factors were determined by dividing the emission load of each pollutant by the amount of combusted coal during 1 yr. The results showed that emission factor ratings for this study could not be classified as A ratings due to the limited number of investigated CEMS facilities. According to the variability of the derived emission factor values, SO2 and CO2 emission factors can be rated as B or above average (with fewer variability values). In comparison, NOx and particulate emission factors can be placed as C or average (with more variability values).

Selective adsorption characteristics of g-C3N4 for Hg0 and HgCl2: A DFT study and experimental verification

The partitioning of Hg0 and HgCl2 with selective adsorbent is an essential technology for the mercury continuous emission monitoring system (Hg-CEMS), but the existing adsorbents fail to satisfy the prolonged operational demands of Hg-CEMS due to the irreversible deactivation. This work proposed a novel regenerable selective adsorbent material, g-C3N4, and delved into the selective adsorption behaviors of Hg0 and HgCl2 on the g-C3N4 using DFT calculations and experiments. The results show that g-C3N4 has a strong affinity for HgCl2, with the binding energy of −0.843 eV, whereas Hg0 is only weakly adsorbed. The electronic structure and weak interaction analysis reveal both Hg0 and HgCl2 are coupled to g-C3N4 via van der Waals. Further, COHP calculation demonstrates that the adsorption of Hg0 and HgCl2 on g-C3N4 is essentially the consequence of polyatomic co-binding and the parallel triatomic structure of HgCl2, which binds to more C/N atoms, thereby leading to the greater binding energy. The experiments confirmed the effective capability of g-C3N4 in separating Hg0 and HgCl2, along with its thermal regeneration potential. It maintains a breakthrough rate of over 98 % for Hg0 and less than 1.6 % for HgCl2. This work updates the conventional chemical adsorption-based methods for the separation of Hg0 and HgCl2, enhances the comprehension of separating both Hg0 and HgCl2 while concurrently broadening the understanding of the interaction between g-C3N4 and small molecules.

Particulate Matter Continuous Emission Monitoring System on Car Free Day Based on the Internet of Thing

Particulate Matter Continuous Emission Monitoring System on Car Free Day Based on the Internet of Thing

High temperature reduction of divalent mercury to elemental mercury for Hg-CEMS

The high temperature reduction (HTR) of flue gas divalent mercury to elemental mercury is the core technology for the mercury continuous emission monitoring system (Hg-CEMS). In this work, the high temperature reduction of HgCl2 to Hg0 was systematically investigated. The results show that HgCl2 can entirely decompose to Hg0 at 800 °C, but Cl, ClO, Cl2, and HCl can easily oxidize Hg0 at low temperature. A new efficient alkaline dechlorination agent for eliminating these Cl species was proposed firstly. The dechlorination agent can efficiently capture the HCl, Cl, and ClO but fail to completely adsorb Cl2. Consequently, a scheme was created that a pair of dechlorination agents were filled in inlet and outlet of the HTR furnace respectively. It features that the upper layer removes HCl and prevents the generation of Cl2 because HCl is the precursor of Cl2, and the lower layer adsorbs the Cl and ClO generated by the HTR process. The feasibility of the method was successfully verified in both a fixed-bed reactor and a Hg-CEMS pretreatment device. DFT calculation indicates that the adsorption of Cl, ClO, and HCl on CaO falls into the category of strong chemisorption while Cl2 is weakly chemisorbed.

Estimation of power plant SO2 emissions using the HYSPLIT dispersion model and airborne observations with plume rise ensemble runs

Abstract. The SO2 emission rates from three power plants in North Carolina are estimated using the HYSPLIT Lagrangian dispersion model and aircraft measurements made on 26 March 2019. To quantify the underlying modeling uncertainties in the plume rise calculation, dispersion simulations are carried out in an ensemble using a total of 15 heat release parameters. For each heat release, the SO2 emission rates are estimated using a transfer coefficient matrix (TCM) approach and compared with the Continuous Emissions Monitoring Systems (CEMS) data. An “optimal” member is first selected based on the correlation coefficient calculated for each of the six segments that delineate the plumes from the three power plants during the morning and afternoon flights. The segment influenced by the afternoon operations of Belews Creek power plant has negative correlation coefficients for all the plume rise options and is first excluded from the emission estimate here. Overestimations are found for all the segments before considering the background SO2 mixing ratios. Both constant background mixing ratios and several segment-specific background values are tested in the HYSPLIT inverse modeling. The estimation results by assuming the 25th percentile observed SO2 mixing ratios inside each of the five segments agree well with the CEMS data, with relative errors of 18 %, −12 %, 3 %, 93.5 %, and −4 %. After emission estimations are performed for all the plume rise runs, the lowest root mean square errors (RMSEs) between the predicted and observed mixing ratios are calculated to select a different set of optimal plume rise runs which have the lowest RMSEs. Identical plume rise runs are chosen as the optimal members for Roxboro and Belews Creek morning segments, but different members for the other segments yield smaller RMSEs than the previous correlation-based optimal members. It is also no longer necessary to exclude the Belews Creek afternoon segment that has a negative correlation between predictions and observations. The RMSE-based optimal runs result in much better agreement with the CEMS data for the previously severely overestimated segment and do not deteriorate much for the other segments, with relative errors of 18 %, −18 %, 3 %, −9 %, and 27 % for the five segments and 2 % for the Belews Creek afternoon segment. In addition, the RMSE-based optimal heat emissions appear to be more reasonable than the correlation-based values when they are significantly different for CPI Roxboro power plant.

Design of a continuous emission monitoring system (CEMS) for the NH3 analyzer at PT. Pusri Palembang

Design of a continuous emission monitoring system (CEMS) for the NH3 analyzer at PT. Pusri Palembang

Plants' emission behaviors under dual control of pollutant concentration and quantity

Quantity-based and concentration-based limits are two common environmental permitting approaches utilized by government worldwide in environmental management. While existing literature is still unambiguous about roles played by quantity-based versus concentration-based limits in environmental management, it becomes evident that relying exclusively on concentration-based or quantity-based limits to control plant emissions may not necessarily result in improved environmental quality. This paper leverages a unique opportunity arising from a recent reform in China's Pollutant Emission Permit System (PEPS) initiated in 2016 to analyze how the introduction of quantity-based limits in addition to concentration-based limits through the PEPS reform impact emissions at the plant level. Utilizing a unique plant-level continuous emission monitoring system data collected from Shaanxi Province (located in western China), the paper finds a significant reduction in air pollutant emissions as a result of the PEPS reform (nitrogen oxides (NOx) by 39%, sulfur dioxide (SO2) by 15% and particulate matter (PM) by 13%). The heterogeneity analyses show emission reductions in plants differ across those with varying quantity limits specified in their permits, distinct emission ratios and diverse ownership structures. Furthermore, plants that fall under the classified management system with more stringent regulations imposed, especially those operating in high-pollution sectors, situated within industrial parks, or classified as large-sized plants, attain higher pollutant quantity limits. Findings of the paper carry important implications for effective environmental management, particularly within developing countries, and shed some light on carbon emission reduction policies in China.

Machine learning for accurate methane concentration predictions: short-term training, long-term results

Abstract Although methane emissions from Alberta’s oil and gas sector have decreased in recent years, monitoring these emissions using Continuous Emission Monitoring Systems (CEMS) can be costly. Predictive Emissions Monitoring Systems (PEMS), powered by machine learning, offer an alternative to or can supplement CEMS. However, effective machine learning models for methane emissions prediction rely heavily on the amount of training data. To address this, we compare the prediction performance of different neural network models, including Long Short-Term Memory (LSTM), Stacked LSTM, Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM), using varying time intervals for training of methane concentration data from Alberta airshed stations. The results showed that the GRU model performed better with shorter datasets, whereas the LSTM and Stacked LSTM models outperformed the GRU and BiLSTM models when trained with more historical data. However, the study found that more training data did not necessarily result in significantly better prediction models.

Development of a sampling assembly for PM2.5 CEMS

For an extractive type particulate matter continuous emission monitoring system (PM CEMS), the sampling assembly acts as an important interface between the exhaust gas and sensing unit. It serves to transport the aerosol sample and pre-treat it to meet the working conditions required by the sensing unit. To obtain representative samples, the sampling assembly should be designed to minimize the loss of particles. In this study, a conditioner based on a dilution approach was proposed and experimentally validated for use with pipelines for PM CEMS targeting particles less than 2.5 μm in diameter (PM2.5). The results showed that for 3.625 μm particles, the transport efficiency through the conditioner was 98.30% at a dilution ratio of 10. Meanwhile, the conditioner could reduce the sample air temperature to below 50 °C without condensation at flue gas temperatures up to 150 °C and water contents below 10%. Finally, a sampling assembly consisting of the conditioner, a 90° bend sampling probe, and a straight tube was designed to obtain transport efficiency higher than 95%, a stated requirement in this study for particles with a diameter less than 3.625 μm. Overall, this study contributed to designing and verifying a sampling assembly for PM CEMS and provided new strategies for improving monitoring data quality using PM CEMS.