Please use this identifier to cite or link to this item: http://hdl.handle.net/123456789/1835
Full metadata record
DC FieldValueLanguage
dc.contributor.authorNarayan Gaonkar, Dattatraya-
dc.date.accessioned2014-09-25T14:46:26Z-
dc.date.available2014-09-25T14:46:26Z-
dc.date.issued2007-
dc.identifierPh.Den_US
dc.identifier.urihttp://hdl.handle.net/123456789/1835-
dc.guidePatel, R. N.-
dc.guidePillai, G. N.-
dc.description.abstractThe research work aims at developing a complete data acquisition and control scheme for inferential control of distillation column through dynamic simulations and experimentation. To achieve this objective a laboratory set up of distillation column developed in the department is made use of. It has 9 trays (two bubble caps on each) with total height of the column being 1400 mm. Any one of the trays out of 4, 5 and 6 from bottom are available as the feed tray. The column has two feed tanks of 40 liters each. The column is fitted with a condenser for cooling overhead vapor and a reboiler to vaporise the binary mixture. The reboiler has 25 liters capacity with three electric heaters (one is of 4kW and two are of 2kW each) for adjusting the quiescent point of heat input and for applying heat input disturbance. The distillation column is interfaced with auxiliary components for instrumentation and control. Two pressure transmitters are used to sense pressure on top and first tray of the column. A total of 12 RTD temperature sensors are used to sense the temperature of liquid on each tray, condenser inlet and outlet and on reflux pipe. The rotameters are connected for the measurement of feed flow to the feed tray and inlet water flow to the condenser. A controller unit is interfaced to the reboiler heat input for controlling the temperature profile of distillation column. This controller unit can function in two modes namely (a) Manual and (b) PID controller mode. The manual control mode of the unit controls the heat input to the reboiler. In PID control mode, it is used for keeping the tray temperature constant. A reflux divider is also added to the existing distillation column set up to facilitate the control of reflux ratio. The upgraded laboratory setup of distillation column is interfaced with a PC through ethernet. Software based on Graphic User Interface (G.U.I) is developed for continuous data acquisition at specific time intervals using labview and National Instruments modules. Rigorous experimentation is performed on the developed setup for a binary mixture of methanol and water. For simulation studies a mathematical model of continuous distillation column is developed using the fundamental physical and chemical laws along with valid normal assumptions and is validated for the experimental setup. The simulation and experimental results are compared and observed for constant Murphree stage efficiency. Acorrelation for simulation and experimental results is developed and evaluated for different heat inputs assuming constant Murphree stage efficiency. Further the behavior of simulation results for different heat inputs at different Murphree stage efficiencies is observed. Murphree stage efficiency vs. heat input and pressure vs. heat input relationships are developed. The mathematical model of distillation column derived is validated using these relationships, thus resulting in a more realistic model for the experimental setup. The distillation control system should hold product compositions as near the set points as possible under all types of disturbances like disturbance in heat input, reflux ratio and feed flow. Distillation column is generally subjected to disturbances and the control of product quality is often achieved by maintaining a suitable tray temperature near its set point. Secondary measurements are used to adjust the values of the manipulated variables, as the controlled variables are not easily measured or not economically viable to measure. An Artificial Neural Network based estimator (ANNE) using Steepest Descent Back Propagation (SDBP) algorithm, is presented to estimate the composition of distillate using temperature profile as inferential variables. The presented ANNE is trained and tested for binary mixture of methanol and water. The results show that the estimation made by the estimator is in good agreement with that of simulation results and the actual composition observed from experimentation. The ANNE developed also gives the satisfactory results for multicomponent mixture. This has been verified by simulation results. The ANNE is trained rigorously for a large number of sets of temperature patterns observed on all trays by performing experimentation and is used for online estimation of composition of binary mixture. It consists of two groups. Group-1 is generated with varying the heat input and group-2 is generated with PID controller maintaining the bottom tray temperature constant at various levels and measuring the composition. The ANNE is first trained with group-1 patterns only thus conceiving estimator ANNE-1. A mixture of group-1 and group-2 patterns is then used to train the ANNE and thus conceiving estimator ANNE-2. The ANN structure with single and double hidden layers is also considered for these two estimators, thus providing four different estimators for online experimentation. These trained estimators are then tested for the set of test temperature patterns. It is observed that ANNE-1 with single hidden layer gives the best possible results. No significant improvement is observed from ANNE-2 modules. The experimentation is then performed for binary mixture using ANNE- 1 with single hidden layer module to estimate distillate composition online. The online estimated distillate composition is checked experimentally using refractometer. To maintain the product quality, a suitable tray temperature is maintained for the bottom tray with the help of PID controller near its set point depending upon the desired composition. An inferential control scheme is developed using ethernet based data acquisition system and the online estimator (ANNE-1). The distillation parameters are acquired by a PC interfaced to the network and the distillate composition is estimated online using ANNE-1. The response of inferential control scheme is also closely monitored for various disturbances. The effect of disturbances on distillate quality is observed experimentally. In case of any disturbance in the manipulated variables, (reflux ratio and heat input) the composition control action needs prior information to minimize the effect of disturbance on the distillate quality. Simulation study is performed for binary mixture of methanol and water to observe and analyze the effect of variation of reflux flow on the tray temperatures, composition of methanol in distillate and distillate flow rate. Two practical samples of mixtures namely (i) 10% methanol and 90%of water and (ii) 70% methanol and 30% of water, are used for these studies. The ANNE-1 developed does not take into account any disturbance. Therefore an enhanced model of ANN based estimator (EANNE) is developed. The EANNE is used for online predictive inferential control of distillation process; it detects the disturbances in manipulated variables and predicts the future distillate composition. An appropriate control action is then in initiated for online control of distillation process much earlier so that the effect of disturbance is minimal in distillate composition. This EANNE based predictive inferential control scheme gives improved performance as demonstrated through experimentation. The performance of EANNE is also studied for multi-component mixture through simulation and it is observed that Steepest Descent Back Propagation (SDBP) algorithm does not work properly for estimating the distillate composition and results in erroneous output because of outputs getting saturated. Hence a new approach known as Levenberg-Marquardt (LM) approach is used for training both ANNE and EANNE models resulting in new enhanced estimators NANNE and NEANNE respectively. NANNE and NEANNE are tested for binary and multicomponent mixtures. The estimator outputs and simulation results tally with each other within the tolerance band. The estimators are further fine tuned using pressure, reflux flow and heat input along with temperature profile of the column. The NEANNE developed senses the disturbance and hence predicts the product composition in advance. The NEANNE so developed and fine tuned gives the best possible estimate as demonstrated by comparison with simulation results. This NEANNE is proposed for predictive inferential control of distillation process. The sensitivity analysis is used for smooth and reliable operation so as to minimize the operation cost and maximize the profit. The sensitivity analysis is carried out on the basis of rigorous experimentation for binary mixture of methanol and water and supported by simulation study for the same mixture. The gain analysis is used to select the most sensitive tray to be controlled for single ended tray temperature control using disturbances in reflux, feed flow and heat input. The double-ended tray temperature control is also analyzed using two gains at a time with the help of Singular Value Decomposition (SVD) analysis, with a view to find out a best method for controlling distillate quality. Experimental results using single ended tray temperature control using the most sensitive tray is compared with the simulation results. The experiments conducted on the distillation column using the inferential control scheme controlling each tray temperature control separately one at a time also shows that the sensitive tray so IV selected using sensitivity analysis gives the best distillate quality. The suitable position of feed tray out of 4, 5 and 6 is selected by performing experiments and verified by simulated results. To summarize the research work, a complete data acquisition system is developed for distillation column. Inferential control scheme is implemented for the process using on-line artificial neural network estimator. A predictive inferential control scheme for distillation column is also implemented to improve the distillate quality. Experimental results are supported by Simulation studies. To estimate distillate composition an artificial neural network based estimator (ANNE) is developed using Steepest Descent Back Propagation (SDBP) algorithm for inferential control of distillation column. This estimator is further enhanced (EANNE) to detect the disturbances and used for predictive inferential control scheme for binary mixture of methanol and water. The performance of EANNE is not satisfactory for multicomponent mixture; hence a new approach known as Levenberg-Marquardt (LM) approach is used for training the ANN resulting NEANNE. The NEANNE is very well suited both for binary as well as multi component mixture. Sensitivity analysis is carried out experimentally for the single ended tray temperature control and results are supported by both simulation and experimental results. A suitable control scheme has been proposed for double ended tray temperature control to obtain better distillate quality.en_US
dc.language.isoenen_US
dc.subjectELECTRICAL ENGINEERINGen_US
dc.subjectDISTILLATION COLUMNen_US
dc.subjectDYNAMIC SIMULATIONSen_US
dc.subjectSTEEPEST DESCENT BACK PROPAGATION ALGORITHMen_US
dc.titleGRID INTERCONNECTION AND ISLANDING OPERATION OF DISTRIBUTED GENERATION SYSTEMSen_US
dc.typeDoctoral Thesisen_US
dc.accession.numberG14264en_US
Appears in Collections:DOCTORAL THESES (Electrical Engg)

Files in This Item:
File Description SizeFormat 
GRID INTERCONNECTION AND ISLANDING OPERATION OF DISTRIBUTED GENERATION SYSTEMS.pdf8 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.