LITERATURE SURVEY
2.1 INTRODUCTION
The proliferation of urbanization and industrialization causes large fluctuations in air quality due to pollution. The air pollution can be minimized effectively by defensive measures through efficient air pollution/quality prediction techniques. The air quality prediction models are tedious and time-consuming. The soft Computing paradigm has emerged out to be more flexible, less assumption dependent and adaptive methodology. Hence this chapter proposes a hybrid model through soft computing techniques for air quality prediction. The rapid increase of urbanization and industrialization causes large fluctuations in air quality. These unmanaged rapid growths of urbanization, manufacturing, and anthropogenic activities have led to the change of the chemical composition of the atmosphere. All the biotic and abiotic components of the environment are directly or indirectly correlated to the Quality of air. Since there is a varied fluctuation in the composition of constituents of the environment as a result of which air quality is degrading at a rapid rate. Air Pollution has many significant consequences, such as the detrimental effect on human health and other animals, considerable harm to the vegetation, and also to the monuments and an immense impact on climate changes. Thus, it is a matter of great concern since air pollution is leading to such harmful effects. Setting up monitoring stations, application of air pollution control technology, implementation of various environmental management plans, etc. are some necessary steps taken by governments to minimize the effect of air pollution. A lot of research has carried out to solve this problem, but these steps are not adequate for today’s extreme climatic situation. The fusion of soft computing techniques generates the Hybrid Soft Computing model, and these types of models share robust in the modelling of real-world systems.
Air quality monitoring often used to determine air pollution levels in urban or rural environments. A monitoring network produces concentration measurements that can then compared with the national and international guideline values. The prevalence of risk factors can use to estimate population health risk from exposure data. One of the essential functions of the monitoring is to provide necessary information for evaluating:
- The level and distribution of exposure in the population;
- The population groups with high exposure; and
- The risks of potential health effects.
On-line air quality monitoring can used in warning and alert systems during episodic pollution events. These systems use mass media to inform people of the current air quality and, if necessary, to give instructions aimed at reducing excess pollution and minimizing exposure. Monitoring ambient air quality also provides important input data for epidemiological studies, which are crucial in establishing associations between health outcomes and concentrations of ambient air pollution. In general, exposure assessment requires both monitoring and modelling to identify target sources for reduced emissions and to implement an effective programme of air quality management for protecting human health. Air quality monitoring is the main source of information in assessing the exposure of the population to ambient air pollution. Exposure determined by the concentrations people experience in their living environments. Thus, the monitoring should measure the concentrations in the places the population is, taking into account both the areas with maximum concentrations and the areas with a high population density
2.2 Purpose of the air quality monitoring system
The activities of human society – the economy, production of goods, transport and consumption – all affect the environment. All stages of these activities contribute directly or indirectly to creating air pollution. Air quality management includes all activities aimed at managing air quality in the environment. Air quality management aims to keep the ambient air clean enough so that it is safe for public health and the environment. This process is leading from the functions of the economy and society in general to the health effects of air pollution described in the driving force–pressure–state–exposure–effect–action chain. The chain emphasizes the action society can take at each link of the chain to minimize the adverse health effects. The role of air quality monitoring is to provide information on the concentrations of pollution in the environment. These are then used to assess the population exposure and adverse health effects caused by pollution. If the health risks considered being too high, the action is needed to control the emissions and to improve the environment.
In this chapter, we are proposing a hybrid model, which is a fusion of Artificial Neural Networks. The reason for selecting these two techniques is their robustness in complex data systems modelling. The prediction of air quality, effectively addressed by the prediction of various air pollutants like Sulphur, carbon monoxide, nitrogen, ozone, suspended particulate matter (SPM) by divided the data set into training, validation and verification further simulation using ANN. Artificial Neural Networks (ANN) has an inbuilt advantage in modelling complex systems without prior knowledge. This is a wonderful tool in the modelling of decision-making systems where the multi-class data exist. The model consists of 4 layers; the first layer represents the input variables; the middle layer consists of the ANN model comprising of the hidden layers and the output layer. The weights and bias updated in the hidden layer; the output of the ANN layer will given as input to the next layer; this layer generates the quality of air. The range of output parameters will generate either Low or High as the final output. The difficulty in choosing a suitable architecture and the tendency to overfit the training data, leading to poor generalization, particularly in situations where limited labelled data are available. High-quality air pollutant concentration prediction is an important basis of air pollution early warning and effective emergency management by which the air pollution based datasets stored effectively along with the use of a microprocessor. The IoT enabled data storage done in the HTTP server to activate the different sensor modules and the effective fog computing approach.
2.3 Literature survey on Air Pollution Monitoring and prediction System
(Al-Janabi, Mohammad, & Al-Sultan, 2020) author proposed the intelligent predictor for the concentrations of air pollutants over the next two days based on deep learning techniques using a recurrent neural network (RNN). The operation determined using a particle swarm optimization (PSO) algorithm. The new predictor based on intelligent computation relying on unsupervised learning, i.e., long short-term memory (LSTM) and optimization (i.e., PSO), is called the smart air quality prediction model (SAQPM). The main goal is to predict six concentrations of six types of air pollution contaminants.
(Cabaneros, Calautit, & Hughes, 2019) author proposed the dynamics between air pollution concentration levels and other explanatory contaminants. Lastly, powerful and less-complicated computing tools that can develop and implement air pollution control. The Artificial Neural Networks (ANNs) predicting and forecasting ambient air pollution. As poor air quality in urban areas has attributed to chronic diseases and premature mortalities of vulnerable members of the public. The involved factors, including the scale and quality of the parameters involved, computationally expensive, and dependent on large databases of air quality.
(Kala, Joshi, Agrawal, Yadav, & Joshi, 2020) author reported the Population growth, urbanization, and air pollutants (AP) in the environment are increasing significantly, posing severe health hazards. AP data provide information about the quality of air and health risk in the surrounding, which is important for environmental management. The entire world is having a severe problem of an energy shortage, and the challenge of air pollution is increasing. Thus, the growth of renewable energy is the highest priority of all other preventive resources.
(Yadav & Nath, 2020) author stated the air pollution produces major influences on human health and the environment. Therefore, it is necessary to forecast air pollutant data to provide prior information about its concentration in the environment for health monitoring. Most urban areas mainly experienced the severe effect of air pollution rapidly along with the problem of energy shortage.
(Maleki et al., 2019) author reviewed the ability of an artificial neural network (ANN) algorithm to predict hourly criteria air pollutant concentrations and two air quality indices, air quality index (AQI) and air quality health index (AQHI). Air pollution also has detrimental effects on the environment, socioeconomics, agriculture, and politics. Meteorology and emissions sources of pollutants are two basic factors influencing the air above quality indices and can be used in computational approaches to predict spatiotemporal pollutant profiles and air quality index.
(Kelly et al., 2019) author stated the national ambient air quality standards (NAAQS) are useful for characterizing exposure in regulatory assessments. Projecting air quality just to meet NAAQS for many scenarios is challenging due to the computational expense of the photochemical grid models (PGMs) commonly used for air quality projection. The emission decreases only, whereas simulations for air quality monitoring prediction for both emission decreases and increases would facilitate broader application. The method also neglects nonlinear interactions in chemistry when applied for combinations of emission reduction cases, which stated the health risk assessment of air pollution via joint toxicity prediction illustrated for prediction of pollution system.
(Maciąg, Kasabov, Kryszkiewicz, & Bembenik, 2019) proposed the several thousand deaths caused by excessive air pollution. The investigation of air pollution and its impact on the health of the population of people make a significant correlation between pollution concentration and increased mortality rate. To predict air pollution data with high effectiveness, one might consider using ensemble models, which regarded more predicting/classifying with much higher quality than other existing prediction models.
(Wen et al., 2019) stated the prediction of mass concentration has a pivotal role in making atmospheric management decisions. These cause negative health effects, such as excessive morbidity and mortality from cardiovascular and respiratory diseases. Therefore, predicting air pollutant concentration in advance is fundamental to strengthen air pollution prevention and achieve comprehensive environmental management, which is of great significance to people’s daily health and government decision-making of controlling the air pollution problem.
(Ghasemi & Amanollahi, 2019) stated the growth and seed production reduced under O3 effect through the air pollutant. Numerous studies have highlighted the potential concerns of the trans-boundary air pollution in different parts of the world. The air pollution monitoring model results for the air quality forecasting are not yet fully convincing. Most recently, attention has focused on the comparison results of different models in forecasting air quality.
(Honarvar & Sami, 2019) proposed the air pollution influences on urban sustainability, therefore measuring air pollution and exploiting the resulting information to predict and discover the relationships between different urban problems is very important. The real-time air pollution data (including ozone, particulate matter, carbon monoxide, sulphur dioxide, and nitrous oxide) are paramount to control air pollution and protect humans against their damages.
(Li, Dong, Zhu, Li, & Yang, 2019) stated the air pollutant concentration exhibits the sensitivity analysis shows that the accomplished optimization process commendably. The dilemma of different air quality standards internationally and avoids the breakpoint effect of AQI, which can provide the guidance of human health against exposure to air pollution.
(Liu et al., 2019) stated the people are suffering from severe air pollution. However, in terms of the current technology level, there are no efficient ways to solve heavy smog fundamentally. Therefore, the best way to prevent people from being harmed by heavy smog is to predict the concentration of PM2.5 and remind people of taking actions in time. Air quality data are sequential, so models that are good at processing sequential data such as a recurrent neural network (RNN).
(Lei, Monjardino, Mendes, Gonçalves, & Ferreira, 2019) proposed the exposure to particulate air pollution and NO2 annually. Background levels of air pollution also have implications for morbidity. Air pollution exposure has been associated with several physiological diseases. In contrast, exposure to even low concentrations of indoor pollutants, such as carbon monoxide (CO), has been linked with neurological symptoms. Air pollution exposure, with building characteristics such as geometry and design, permeability, and ventilation components impacting on the infiltration of outdoor pollution indoors and the removal of internally generated pollution.
(Xu, Shan, Li, & Zhang, 2020) stated the popular air quality evaluation indexes include criteria air pollutants and comprehensive indexes (Zhu et al., 2018b). Criteria air pollutants including PM2.5, PM10, SO2, NO2, CO and O3 are expressed in micrograms per cubic meters or part per million and estimated concerning the new ambient air quality standards and higher the number, the greater the health risks and the need for preventive measures.
(Wu & Lin, 2019) stated the air pollution might be incomplete and it would be convenient to include other relevant aspects of this impact, such as the issue of waste generated by the ships’ visit to the ports, secondly the effect on the waters of the port, the pollution observed in this environment. The particulate matter, the term for a mixture of solid particles and liquid droplets found in the air.
(Ruiz-Guerra, Molina-Moreno, Cortés-García, & Núñez-Cacho, 2019) performed several numerical experiments using the monitoring of various air pollution predictions through soft computing techniques and the best possible configuration for real-time operation. soft computing techniques such as Artificial Neural Network (ANN), Fuzzy Logic (FL), Genetic Algorithm (GA) and Genetic Programming (GP) are popular and been widely used to address highly complex, dynamic, and nonlinear systems.
(Sahoo & Bhaskaran, 2019) proposed the air quality data is temporal data as the concentration of air pollutants patterns at various locations from a large quantity of air quality and metrological datasets. The soft computing techniques and hybrid approaches provide the different monitoring systems along with the consideration as input parameters and output parameters. As a result, the neural network model provides a better estimation of parameters in comparison with the linear and non-linear regression model.
(Bhowmik & Ray, 2019) stated its simplicity and reliability in the analysis of problems at a near-perfect performance rating projected soft computing methods as a handy tool for solving air pollution system. Compared to other statistical methods such as the neural network, it has proved to show better performance accuracies in terms of prediction and forecasting of air pollution through the AQI of various areas datasets.
(J. Li et al., 2019) stated the air quality early-warning plays a vital role in improving air quality and human health, especially multi-step ahead air quality early-warning, which is significant for both citizens and environmental protection departments through the soft computing approach. This developed to extract the characteristics of air quality data effectively and to further the forecasting performance.
(Jain, Saini, & Mittal, 2019) stated that for linear and nonlinear regression problems, the performance plot of various soft computing–based models for both training and testing stages of air pollution prediction. To improve air quality in a specific region is the dependency upon the contribution of educated individuals who know the regional and national level air pollution-related problems. It has scientifically proven that certain air pollutants deteriorate air quality.
(Sihag, Kumar, Afghan, Pandhiani, & Keshavarzi, 2019) proposed the ANN used soft computing technique to unravel complex nonlinear problems. This technique offers the flexibility of learning the mapping between the input factors and the process responses to sort out complicated air quality pollution problems. The neural network consists of immensely interconnected neural soft computing elements. The neural elements have a competency to learn and extract air pollution information.
(Mookherji & Sankaranarayanan, 2019 proposed the soft computing models such as artificial intelligence (AI) provide an excellent and reliable technique for modelling surface and underground air quality of the system. That is highly dimensionally spaced but does not involve nonlinear transformations, thereby making data to be indispensable and linearly separable since there is no room for assumption during the functional transformation according to the quality of air pollution and air pollution monitoring systems.
(Guo, Ren, Jin, & Ding, 2019) stated the fruit fly optimization algorithm (FFOA)-based water quality modelling method and studies a eutrophication risk assessment method of lakes and reservoirs based on the mechanism model, which is based on fruit fly optimization algorithm (FFOA) to model the water quality mechanism of lakes and reservoirs while a behaviour to seek a globally optimal solution.
(Ding, Dong, & Zou, 2019) stated the FOA is a novel global optimization technique on the foundation of foraging behaviours. For fruit flies, the divine nature in smell and vision is conducive to smell collection in the air pollution and correct flight to the food or gathering. To build a power consumption forecasting model using various machine learning algorithms. They propose two electric load forecasting models using artificial neural network and support vector regression analysis.
(Ding et al., 2019) proposed the Modern Machine Learning (ML) techniques such as the Artificial Neural Network, the Genetic Programming, by which clean air is essential for human health and well-being. However, many people around the world live in places where the constant air pollution expose them to a higher risk of pulmonary disease or lung cancer, which is the different pollution particles present in the air along with the levels of small pollutants particles,
(Cassano, Casale, Regina, Spadafina, & Sekulic, 2019) stated the dynamic artificial neural network (ANN). Previous forecasting applications of the NARX model include electricity prices, oil prices, and air pollution peaks, but it has not previously applied to CO2 emissions. In addition to capturing nonlinear and dynamic aspects of CO2 emissions, Nonlinear Auto Regressive model with exogenous inputs (NARX) and other ANN modes have self-learning capabilities, using both past input and output values to predict future CO2 output levels on reducing CO2 emissions.
(Leong, Kelani, & Ahmad, 2019) reviewed the API is a number tha