Predictive 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.

CO and NOx emissions prediction in gas turbine using a novel modeling pipeline based on the combination of deep forest regressor and feature engineering

The main objective of this study is to estimate carbon oxides (CO) and nitrogen oxides (NOx) emissions from a gas turbine using the predictive emission monitoring systems dataset. First, four methods were developed for feature generation: Principal Component Analysis, t-Distributed Stochastic Neighbor Embedding, Uniform Manifold Approximation and Projection, and Potential of Heat-diffusion for Affinity-based Trajectory Embedding. Then Feature Relevance-based Unsupervised Feature Selection was evaluated for ranking features. With all features generated, the regression models Ridge Regression, Least Absolute Shrinkage and Selection Operator, k-Nearest Neighbor, Cubist Regression, Random Forest, Light Gradient Boosting Machine, Categorical Boosting, and Deep Forest Regression (DFR) were evaluated. The hyperparameters were tuned with Randomized Search Cross Validation using a 5-fold cross-validation (CV) procedure. In this paper, an innovative modeling pipeline based on the mentioned regressors, feature engineering, and effective hyperparameters tuning was proposed. The developed regression models based on feature engineering and hyperparameters tuning were compared using performance criteria, including mean 5-fold CV values of the coefficient of determination (R2) and root mean square error (RMSE). The obtained best results based on mean CV values using the DFR presented R2 values equal to 0.9647 and 0.5355 and RMSE values equal to 0.4245 and 1.4508 for CO emissions, for the training and validation datasets, respectively. On the other hand, DFR presented R2 equal to 0.9866 and 0.4737 and RMSE equal to 1.3876 and 7.6528 for NOx emissions, for the training and validation datasets, respectively.The results show that the proposed DFR model has higher potential for CO and NOx emissions prediction and that feature engineering and hyperparameter tuning has demonstrated impacts on the predictive capacity of the evaluated regression models.

Multiple Regression and Big Data Analysis for Predictive Emission Monitoring Systems

Predictive Emission Monitoring Systems (PEMS) offer a cost-effective and environmentally friendly alternative to Continuous Emission Monitoring Systems (CEMS) for monitoring pollution from industrial sources. Multiple regression is one of the fundamental statistical techniques to describe the relationship between dependent and independent variables. This model can be effectively used to develop a PEMS, to estimate the amount of pollution emitted by industrial sources, where the fuel composition and other process-related parameters are available. It often makes them sufficient to predict the emission discharge with acceptable accuracy. In cases where PEMS are accepted as an alternative method to CEMS, which use gas analyzers, they can provide cost savings and substantial benefits for ongoing system support and maintenance. The described mathematical concept is based on the matrix algebra representation in multiple regression involving multiple precision arithmetic techniques. Challenging numerical examples for statistical big data analysis, are investigated. Numerical examples illustrate computational accuracy and efficiency of statistical analysis due to increasing the precision level. The programming language C++ is used for mathematical model implementation. The data for research and development, including the dependent fuel and independent NOx emissions data, were obtained from CEMS software installed on a petrochemical plant.

Data Analysis and Symbolic Regression Models for Predicting CO and NOx Emissions from Gas Turbines

Predictive emission monitoring systems (PEMS) are software solutions for the validation and supplementation of costly continuous emission monitoring systems for natural gas electrical generation turbines. The basis of PEMS is that of predictive models trained on past data to estimate emission components. The gas turbine process dataset from the University of California at Irvine open data repository has initiated a challenge of sorts to investigate the quality of models of various machine learning methods to build a model for predicting CO and NOx emissions depending on ambient variables and the parameters of the technological process. The novelty and features of this paper are: (i) a contribution to the study of the features of the open dataset on CO and NOx emissions for gas turbines, which will enable one to more objectively compare different machine learning methods for further research; (ii) for the first time for the CO and NOx emissions, a model based on symbolic regression and a genetic algorithm is presented—the advantage of this being the transparency of the influence of factors and the interpretability of the model; (iii) a new classification model based on the symbolic regression model and fuzzy inference system is proposed. The coefficients of determination of the developed models are: R2=0.83 for NOx emissions, R2=0.89 for CO emissions.

Development of Рredictive Еmission Мonitoring Systems

An overview of world experience in the development and implementation of emission monitoring systems for industrial enterprises, based on mathematical models is presented. The main problems of such systems have been analyzed, their advantages and disadvantages have been revealed. The authors have demonstrated that at Russian enterprises, the introduction of predictive emission monitoring systems was possible at the initial stages of transition to BAT as part of the digital technologies’ integration in production processes.

Predicting CO and NOxemissions from gas turbines: novel data and abenchmark PEMS

Predictive emission monitoring systems (PEMS) are important tools for validation and backing up of costly continuous emission monitoring systems used in gas-turbine-based power plants. Their implementation relies on the availability of appropriate and ecologically valid data. In this paper, we introduce a novel PEMS dataset collected over five years from a gas turbine for the predictive modeling of the CO and NOx emissions. We analyze the data using a recent machine learning paradigm, and present useful insights about emission predictions. Furthermore, we present a benchmark experimental procedure for comparability of future works on the data

ODOUR POLLUTION ASSESSMENT THROUGH INDIRECT METHODS BASED ON THE MONITORING OF TECHNOLOGICAL PARAMETERS - CASE STUDY

The odours in the ambient air, through the discomfort that they induce, represent an actual problem for the communities located nearby companies with technological processes that emit in the air different strong and unpleasant odorous substances. The standard method for odour assessment involves measuring the concentration through dynamic olfactometry according with SR EN 13725:2003, a very complex method that requires specialized working staff and expensive equipment. The paper presents an indirect method for odour assessment in the ambient air, based on monitoring the process and meteorological data - Predictive Emission Monitoring Systems (PEMS) and it is applied to a livestock farm. Using the multiple regression analysis of the monitoring data for the most important specific technological and meteorological parameters it can be developed a mathematical model that could be used for the calculation of odour concentration in air, without the necessity of direct measurement, after the initial step. For the case study presented in the paper, the distance between the slurry lagoon was identified as a significant statistical parameter that can determine in a proportion of 72% the concentration of odour in the ambient air nearby the farm; the margin of error for odour concentration assessment, according to the model validation tests, is ± 8%, acceptable value for an estimation method by mathematical modelling.

Development of Predictive Emissions Monitoring System Using Open Source Machine Learning Library – Keras: A Case Study on a Cogeneration Unit

The study provides an overview of Predictive Emissions Monitoring System's (PEMS) research, application, installation, and regulatory framework as well as develops predictive models for NOx emissions from a natural gas fired cogeneration unit using an open source machine learning library, Keras, and open source programming languages, Python and R. Nine neural network based predictive models were trained with 12 086 examples and tested with 3020 examples. The neural network-based models use eight process parameters as inputs to predict NOx emissions. All models meet the regulatory requirements for precision. The best model (32-64-64-64) has four hidden layers and uses the Nadam method for optimization. The best model has a mean absolute error of 0.5982, r-value of 0.9451, and a difference of 0.14% between the measured and predicted emission values using the test dataset. The study demonstrated the feasibility of using open source machine learning library in PEMS development. It also provides guidance to facility operators to develop their own PEMS models for monitoring emissions.

Enhanced PEMS Performance and Regulatory Compliance through Machine Learning

<p><em>Modeling technologies</em><em> can pro</em><em>vide strong support to existing emission management systems, by means of what is known as a Predictive Emission Monitoring System (PEMS). These systems do not measure emissions through any hardware device, but use computer models to predict emission concentrations on the ground of process data (e.g., fuel flow, load) and ambient parameters (e.g., air temperature, relative humidity). They actually represent a relevant application arena for the so-called Inferential Sensor technology which has quickly proved to be invaluable in modern process automation and optimization strategies (Qin et al., 1997; Kadlec et al., 2009). While lots of applications demonstrate that software systems provide accuracy comparable to that of hardware-based Continuous Emission Monitoring Systems (CEMS), virtual analyzers are able to offer additional features and capabilities which are often not properly considered by end-users. Depending on local regulations and constraints, PEMS can be exploited either as primary source of emission monitoring or as a back-up of hardware-based CEMS able to validate analyzers’ readings and extend their service factor. PEMS consistency (and therefore its acceptance from environmental authorities) is directly linked to the accuracy and reliability of each parameter used as input of the models. While environmental authorities are steadily opening to PEMS, it is easy to foresee that major recognition and acceptance will be driven by extending PEMS robustness in front of possible sensor failures. Providing reliable instrument fail-over procedures is the main objective of Sensor Validation (SV) strategies. In this work, the capabilities of a class of machine learning algorithms will be presented, showing the results based on tests performed actual field data gathered at a fluid catalytic cracking unit.</em></p>

Assessment and testing of sensor Validation algorithms for environmental monitoring applications

The increasing attention devoted to air quality by legislative, scientific, industrial and public sectors has led to the development of different control strategies for the emission level monitoring. In this scenario, Predictive Emission Monitoring System (PEMS) is able to predict emission concentrations thanks to empirical or first principles models fed by real-time process data provided by measurement sensors. It follows that PEMS consistency (and, crucially, its acceptance from regulations-enforcing agencies) strictly depends on input accuracy and that reliable Sensor Validation (SV) strategies are fundamental. In this work, the capability of two different SV techniques, Feed Forward Neural Networks and Locally Weighted Regression, is tested exploiting a commercial software package (ABB’s IMP) on actual field data from a fluid catalytic cracking unit. The results showed that both techniques are suitable as complement to PEMS applications, but Locally Weighted Regression results are preferable for performance, economic and operating reasons.

B. Indonesia

In January, the minister of environment and forestry and the Danish ambassador for Indonesia conducted a meeting to discuss future outcomes for co-operation between Indonesia and Denmark involving public and private entities. Indonesia and Denmark, through the Danish International Development Agency, have pursued co-operation that focuses on the Environmental Support Program since 2015. Denmark has supported Indonesia in the areas of strategic environmental assessment, environmental impact assessment, and environmental taxes and fines. The program has allowed Indonesia to develop instruments on strategic environmental assessment and an environmental tax and has included Act no. 32/2009 on Environmental Protection and Management. Indonesia and Denmark’s future co-operation will focus on safeguard instruments that include planning and evaluating effective programs and developing strategic environmental assessment, environmental impact assessment, and economic instruments. These instruments are to be implemented in 2018 (<http://www.menlh.go.id/pertemuan-menteri-lhk-ri-dengan-dubes-denmark/>). In March, Indonesia’s Ministry of Environment and Forestry, the Embassy of Denmark, and the Hyprowira Company conducted a focus group discussion on the use of a predictive emission monitoring system in the ministry’s office. The participants of the focus group included government officers, industrial workers, and academic communities. The focus group aimed at increasing compliance with the emission monitoring system and at informing the public on the results of a pilot project on the implementation of emission monitoring in Indonesia, which is supported by Denmark. It also informed the participants on new rules and technology related to the monitoring of emissions (<http://www.menlh.go.id/pemantauan-emisi-industri-melalui-teknologi-predictive-emission-monitoring-system-pems/>).

Modeling NO x emissions from coal-fired utility boilers using support vector regression with ant colony optimization

Modeling NO x emissions from coal fired utility boiler is critical to develop a predictive emissions monitoring system (PEMS) and to implement combustion optimization software package for low NO x combustion. This paper presents an efficient NO x emissions model based on support vector regression (SVR), and compares its performance with traditional modeling techniques, i.e., back propagation (BPNN) and generalized regression (GRNN) neural networks. A large number of NO x emissions data from an actual power plant, was employed to train and validate the SVR model as well as two neural networks models. Moreover, an ant colony optimization (ACO) based technique was proposed to select the generalization parameter C and Gaussian kernel parameter γ. The focus is on the predictive accuracy and time response characteristics of the SVR model. Results show that ACO optimization algorithm can automatically obtain the optimal parameters, C and γ, of the SVR model with very high predictive accuracy. The predicted NO x emissions from the SVR model, by comparing with the BPNN model, were in good agreement with those measured, and were comparable to those estimated from the GRNN model. Time response of establishing the optimum SVR model was in scale of minutes, which is suitable for on-line and real-time modeling NO x emissions from coal-fired utility boilers.

A Feasibility Study of a Predictive Emissions Monitoring System Applied to Taipower’s Nanpu and Hsinta Power Plants

The Hsinta and Nanpu Power Stations are located in southern Taiwan. The Hsinta Power Station consists of five combined-cycle gas turbines (CCGT), whereas the Nanpu Power Station consists of four. A project was undertaken to develop and deploy a predictive emissions monitoring system (PEMS) on CCGT unit 3 of Hsinta Power Station (HT-3) and CCGT unit 1 of Nanpu Power Station (NP-1) with the long-term goal of developing a universal model for this kind of power plant. After the first-year PEMS project at the Hsinta power plant, one goal of the second-year PEMS project was to set up a second PEMS at the Nanpu power plant and compare the PEM models applied the to two gas-fired combined cycle power generation units. Consequently, the second and third PEMS of Taiwan at CCGT HT-3 and NP-1 were finished. After comparing the differences among HT-1, HT-3, and NP-1 PEMS models, the pattern of model functionality indicated that this model could be applied to the other units of the same type and size. However, the PEMS function constant or parameter coefficients must be modified on a case-by-case basis. With regard to the PEMS model developed for HT-3, the relative accuracy (RA) of the 15-variable model with start-up mode is only 7.43% and met the criteria of draft PS-16. With regard to the PEMS model developed for NP-1, the RA of the 10-variable model with start-up mode was only 7.76% and also met the criteria of draft PS-16.

A performance study of PEMS applied to the Hsinta power station of Taipower

Abstract Predictive emission monitoring systems (PEMS) have been shown to be a promising technology for calculating the emissions from gas-fired stationary sources. One of the first demonstration PEMS in Taiwan was installed and operated successfully at the Hsinta power plant of the Taiwan Power Company (Taipower) in 2000. To improve the accuracy of this PEMS model, further studies were conducted at the same combined cycle gas turbine (CCGT) unit #1 of the Hsinta power plant. To optimize the NOX PEMS model, several functions were progressively developed including a log model, a one-level linear model, a three-level linear model, a 15-variable linear model, and a 15-variable with start-up mode model. The latter gave the relative accuracies (RA) and correlation coefficients between PEMS and a continuous emission monitoring system (CEMS) in the range of 3.38–14.77% and 0.955–0.989, respectively, which fulfill the criteria of the USEPA draft on PEMS performance specifications. Similarly, this model demonstrated RAs for low, medium, high level, and a three-level average less than 20% which meet CEMS requirements set out by Taiwan EPA.

Applicability of the PEMS technique for simplified NO monitoring on board ships

Abstract The performance of a predictive emission monitoring system (PEMS) as a technique for NOx monitoring on medium speed marine diesel engines has been evaluated for 16 similar engines on four different ships. The PEMS function tested measured O2 concentration in the exhaust gas, engine load, combustion air temperature and humidity, and barometric pressure to calculate the NOx concentration. Emission measurements were carried out by means of a conventional continuous emission monitoring system (CEMS) and the measured NOx concentrations were compared with those calculated by the PEMS function. For 11 of the 16 engines, the average error between measured and calculated NOx concentration was Bearing in mind the performance criteria for using PEMS at land-based installations, the results from this study are promising. Viewed as a single data set of 56 h with 16 separate engine comparisons between CEMS and PEMS, the data set shows a relative accuracy of 14.5% i.e. within the 20% requirement of the US Environmental Protection Agency. In light of the increased interest and international guidelines for continuous NOx monitoring on board ships, the PEMS technique can offer a simple but cost-effective option.

A Feasibility Study on the Predictive Emission Monitoring System Applied to the Hsinta Power Plant of Taiwan Power Company

The continuous emission monitoring system (CEMS) can monitor flue gas emissions continuously and instantaneously. However, it has the disadvantages of enormous cost, easily producing errors in sampling periods of bad weather, lagging response in variable ambient environments, and missing data in daily zero and span tests and maintenance. The concept of a predictive emission monitoring system (PEMS) is to use the operating parameters of combustion equipment through thermodynamic or statistical methods to construct a mathematic model that can predict emissions by a computer program. The goal of this study is to set up a PEMS in a gas-fired combined cycle power generation unit at the Hsinta station of Taiwan Power Co. The emissions to be monitored include nitrogen oxides (NOx) and oxygen (O2) in flue gas. The major variables of the predictive model were determined based on the combustion theory. The data of these variables then were analyzed to establish a regression model. From the regression results, the influences of these variables are discussed and the predicted values are compared with the CEMS data for accuracy. In addition, according to the cost information, the capital and operation and maintenance costs for a PEMS can be much lower than those for a CEMS.

Reduced Kinetics Schemes for Oxides of Nitrogen Emissions from a Slow-Speed Marine Diesel Engine

A number of reduced chemical kinetics schemes are compared for prediction of NOx emissions from a slow-speed marine diesel engine, using a zero-dimensional model. The kinetic evolution of NO is tracked in 10 representative parcels of burnt gas, formed sequentially during combustion. Dilution of the burnt gas by unburnt air is accounted for. The model is developed for use in a machinery space simulator for training marine engineers and in a predictive emissions monitoring system. It runs in real time on a standard PC and requires experimental data for calibration. Kinetics schemes modeled include the extended Zeldovich mechanism and five schemes involving nitrous oxide. The addition of nitrous oxide reactions to the extended Zeldovich mechanism increases predicted NOx by up to 15%. The N2O reactions which give the most significant contribution to NOx in the context of a large marine diesel engine have been identified. NO from fuel bound nitrogen is likely to be significant for engines operating on residual fuel oil.

Predictive NO x emission monitoring on board a passenger ferry

NO x emissions from a medium speed diesel engine on board a servicing passenger ferry have been indirectly measured using a predictive emission monitoring system (PEMS) over a 1-yr period. Conventional NO x measurements were carried out with a continuous emission monitoring system (CEMS) at the start of the study to provide historical data for the empirical PEMS function. On three other occasions during the year the CEMS was also used to verify the PEMS and follow any changes in emission signature of the engine. The PEMS consisted of monitoring exhaust O 2 concentrations (in situ electrochemical probe), engine load, combustion air temperature and humidity, and barometric pressure. Practical experiences with the PEMS equipment were positive and measurement data were transferred to a land-based office by using a modem data communication system. The initial PEMS function (PEMS1) gave systematic differences of 1.1–6.9% of the calibration domain (0–1725 ppm) and a relative accuracy of 6.7% when compared with CEMS for whole journeys and varying load situations. Further improvements on the performance could be obtained by updating this function. The calculated yearly emission for a total engine running time of 4618 h was 316 t NO x ±38 t and the average NO x emission corrected for ambient conditions 14.3 g kWh corr −1. The exhaust profile of the engine in terms of NO x , CO and CO 2 emissions as determined by CEMS was similar for most of the year. Towards the end of the study period, a significantly lower NO x emission was detected which was probably caused by replacement of fuel injector nozzles. The study suggests that PEMS can be a viable option for continuous, long-term NO x measurements on board ships.

An accurate Predictive Emissions Monitoring System (PEMS) for an ethylene furnace

AbstractThe Predictive Emissions Monitoring System (PEMS) has been recently proposed as an alternative to hardware continuous emissions monitors (CEM). PEMS has two distinct advantages over hardware CEM: much lower costs to install and maintain, and ability to provide information on emissions under various conditions. Using a state‐of‐the‐art computer program, an accurate PEMS was developed for an ethylene furnace at Exxon Chemical Company's Baytown Olefins Plant, Baytown, Texas. First, a test plan was carefully designed to collect furnace process and stack emissions data over the total operating range of the furnace. The emissions data (NOx, O2 and CO) were collected and measured by an independent company for 5 days. The related process data were recorded in the DCS system during the same 5‐day period. A dataset with some 7,000 patterns of related process and emissions data was generated. The PEMS system contains 2 parts: emissions prediction and sensor validation models. The main purpose of the sensor validation model is to validate the input sensor's raw data and reconstruct new sensor values, if necessary, before they are used in the prediction model. Its usage provides a high degree of confidence for the PEMS system. Two separate emissions prediction models were built, one for NOx and O2, and the other for CO. Both final models have the same 9 input sensors. A PEMS relative accuracy test audit (RATA) was performed by an independent company to test the PEMS system under the three firing rate conditions (low, medium, and high). Test results indicate that the PEMS system satisfied both the relative accuracy (RA) and the stringent statistical test requirements including the F‐test, t‐test and r‐correlation.