Marian Popescu

A comparative study of computational intelligence techniques applied to PM2.5 air pollution forecasting

The paper presents the results of a comparative study performed between two computational intelligence techniques, artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFIS) applied to particulate matter (fraction PM2.5) air pollution forecasting. The experiments were realized on datasets from the Airbase databases with PM2.5 hourly measurements. The main statistical parameters that were computed are root mean square error (RMSE) and mean absolute error (MAE).

PID Controller optimal tuning

A controller is the most important part of an automatic control system. Its main purpose is to receive the signals from a transmitter and to compare it with a reference value. Depending on the error the controller sends a control signal to an actuator, so that the controlled variable to be equal with the reference value. The most used algorithm by a controller is the PID algorithm. In this paper, two methods of determining the optimum values of the tuning parameters for the PID controller will be presented, using case studies. The optimization criteria are the integral criteria, the square of the error between reference and controlled value being taken into consideration. The results consisted in the optimal values of the PID tuning parameters (gain and reset) for the analyzed cases associated to a temperature control system.

Particulate matter prediction using ANFIS modelling techniques

Recent studies on air pollution emphasized particulate matter impact on human health and climate changes. This impact generated a trend for developing research projects which deal with monitoring and forecasting air quality. This paper fits into this trend and presents an ANFIS (adaptive neuro-fuzzy inference system) modelling approach to predict particulate matter concentration for short terms. The ANFIS technique was tested for three data sets covering all seasons specific to urban areas from Romania. The data sets type imposed the fuzzy inference system generating method and the optimization method. The resulted prediction model can be used to warn the population when the PM concentration exceeds standard limits, and also to extract useful data for knowledge-based modelling.

Applying Artificial Neural Networks to Short-Term PM2.5 Forecasting Modeling

Air pollution with suspended particles from PM2.5 fraction represents an important factor to increasing atmospheric pollution degree in urban areas, with a significant potential effect on the health of vulnerable people such as children and elderly. PM2.5 air pollutant concentration continuous monitoring represents an efficient solution for the environment management if it is implemented as a real time forecasting system which can detect the PM2.5 air pollution trends and provide early warning or alerting to persons whose health might be affected by PM2.5 air pollution episodes. The forecasting methods for PM concentration use mainly statistical and artificial intelligence-based models. This paper presents a model based protocol, MBP – PM 2.5 forecasting protocol, for the selection of the best ANN model and a case study with two artificial neural network (ANN) models for real time short-term PM2.5 forecasting.

Data Mining and ANFIS Application to Particulate Matter Air Pollutant Prediction. A Comparative Study.

The paper analyzes two artificial intelligence methods for particulate matter air pollutant prediction, namely data mining and adaptive neuro-fuzzy inference system (ANFIS). Both methods provide predictive knowledge under the form of rule base, the first method, data mining, as an explicit rule base, and ANFIS as an internal fuzzy rule base used to perform predictions. In order to determine the optimal number of prediction model inputs, we have perform a correlation analysis between particulate matter and other air pollutants. This operation imposed NO2 and CO concentrations as inputs of the prediction model, together with four values of PM10 concentration (from current hour to three hours ago), the output of the model being the prediction of the next hour PM10 concentration. The two prediction models are investigated through simulation in different structures and configurations using SAS® and MATLAB® respectively. The results are compared in terms of statistical parameters (RMSE, MAPE) and simulation time.

Models of particulate matter concentration forecasting based on artificial neural networks

The paper presents some models based on artificial neural networks for particulate matter concentration forecasting. A methodology framework is proposed for selecting the best forecasting model from a set of neural networks models. First, two artificial neural network types (feed forward and radial basis) are analyzed for concentration forecasting of the particulate matter with diameter less than 10 μm, based on the proposed methodology. Also, a forecasting model for particulate matter with diameter less than 2.5 μm is developed, and then tested on real time data provided by two air quality monitoring microstations built within the ROKIDAIR project. In both cases, statistical indicators are calculated in order to assess the performance of the forecasting models.

Modelling Missing Data for PM2.5 Time Series Forecasting with Computational Intelligence

The paper presents two missing data filling methods which can be applied to time series forecasting. The basic idea of the proposed methods is that usually, the forecasted parameter (in this case PM2.5 air pollutant concentration) is dependent on some related parameters that influence its value. When the parameter time series have missing data due to various reasons (e.g. faulty measurement instruments), the time series of other parameters (if available) can be used to fill in the missing values. One method is based on an artificial neural network that has as input the values of the other related parameters measured at time t and as output the value of the missing value of the forecasted parameter at time t. The other method is Holt-Winters which uses as inputs previous values of the forecasted parameter. These methods are proper for cases with larger gaps in the time series (more than several days). These filling methods are compared in terms of statistical indicators (e.g. RMSE). Also, a comparative study was performed for PM2.5 forecasting accuracy analysis with two forecasting methods: a feed forward artificial neural network and Holt-Winters.

Development of ANFIS models for PM short-term prediction. case study

The growing rate of urban and industrial development leads to high levels of air pollution in most countries around the world. Because air pollution has a major impact on human health, monitoring and forecasting of the most important pollutants concentrations are very important. The modelling of the non-linear and complex phenomena associated to air pollution is successfully performed using artificial intelligence-based methods. This paper aims to develop a model based on adaptive neuro-fuzzy inference system (ANFIS) technique for short-term prediction of particulate matter (PM) concentration. There are proposed three models, one that uses only PM concentrations as inputs, and the other two that have as additional inputs meteorological parameters. All models have as output the prediction of the next hour PM concentration. The simulation results for the three proposed models are compared using statistical indices, the best model being the one that takes into account the current hour temperature as additional input besides the PM concentrations.

Control of propylene - propane distillation process using Unisim ® design

The paper addresses the control problem of a binary distillation processes. There are analyzed the quality specifications of separated products and also the distributed and multivariable character of the process. The robustness of the control system lies in its simplicity versus the investment. The paper presents the studies made for development and numerical testing of three types of control structures for propylene-propane separation process: the feedback control structure, the feedforward control structure and the optimal control structure.

Robust Control of Industrial Propylene-Propane Fractionation Process

Abstract The paper presents some results in the field of the industrial propylene-propane separation process control. The process inputs and outputs, the quality specifications for the separated products and the possible pairs of control agents, based on the relative gain array, are analyzed. The analysis emphasized the fact that the control structure based on L and B control agents is the most suitable. Starting from this result two robust control structures for the quality of the separated products are developed. The first structure refers to the feedback control of the distillate quality, and the second structure is associated with feedforward control of top and bottom products quality. Each structure was tested using dynamic simulation for the same step input of the feed flowrate disturbance. Test results confirmed the superiority of feedforward control.