Search

A new method is proposed for systematic generation of conceptual design of reactor networks. Here, non-isothermal conditions are considered. Optimal distribution of heat transfer area, in addition to optimal mixing and distribution of feeds are targeted. Given temperature and composition of feed streams, the objective is to find the sequence and size of ideal reactors, the distribution of extra feed streams, the heat transfer distribution along the reactor path and the total reaction time that maximizes the space–time yield of the key product component. The method can be applied to solve problems of any number of components and reactions. The method is applied to industrially important processes such as the methanol synthesis and the steam methane reforming process. Interesting results are obtained with respect to the required heat exchange area.

Membraner aksepteres i økende grad som metode for naturgassrensing ettersom etterspørselen etter naturgass øker og strengere klima- og utslippsregler innføres. I denne rapporten, har et hybrid membransystem for fjerning av vann og CO2 i naturgass blitt studert. En spiralmodul med en MTR Pebax® co-polymer-basert sammensatt membran ble simulert for dehydrering, og to fibermoduler med polyimid-derivert karbonmembran ble simulert i serie for fjerning av CO2. Målet var å evaluere kost-nytte-effekten som et forstudie for systemet. Videre ble en membranmodell utviklet, sammenlignet med eksperimentelle data og analysert for bruk i simuleringsverktøy for å forutsi prestasjonen til membraner for gasseparasjon. Et utvalg av ulike designkonfigurasjoner ble vurdert med hensyn til antallet og utformingen av kompressorer. Hver designkonfigurasjon ble optimalisert ved bruk av Nelder-Mead-metoden med en penalty-funksjon. Optimaliseringsprogrammet ble utført eksternt ved å opprette en COM automatasjonsserver (ActiveX) som koblet det eksterne programmet til brukergrensesnittet for simuleringsprogrammet. Likningssettet for membranmodellen ble diskretisert ved å bruke en ortogonal kollokasjonsmetode som transformerte likningene til ikke-lineære algebraiske likninger som videre ble løst ved bruk av Newton-Rhapson-metoden. Membranmodellen ble deretter integrert i simuleringsprogrammet med den optimale designkonfigurasjon som en Cape-Open enhetsoperasjon basert på COM-teknologi. Det optimale designet besto av to kompressorer i serie mellom enheten for dehydrering og enheten for CO2-fjerning, med kompresjon av den resirkulerte retentatstrømmen fra CO2-enheten, og en trykkventil for å redusere trykket i den resirkulerte permeatstrømmen fra CO2-enheten. Den spesifikke kostnaden for renset naturgass var 2.192E−3 $/m^3 og kravene til gassens sammensetning og utslipp ble innfridd. I det optimale designet med ChemBrane-modellen var duggpunktet til vann -52.5 grader celsius, den rensede naturgassen besto av 97.5% metan og det var 95% CO2 i strømmen for videre bruk i EOR. Gassens duggpunkt var -11.5 grader celsius. Det optimale systemdesignet med membranmodellen utviklet i dette arbeidet ble ikke optimalisert på grunn av tilkoblingsproblemer med COM-serveren og simuleringsprogrammet. Membranmodellen ble validert av ett av to eksperimenter med et relativt avvik ≤ 3.65%, og kunne derfor brukes til simulering av gasseparasjon i fibermoduler. Sammenlignet med data fra det andre eksperimentet, ble membranmodellen funnet å være ugyldig på grunn av store relative avvik i permeatstrømmene og renheten. Membranmodellen viste derimot tilnærmet like resultater som ChemBrane-modellen, men ble vurdert å vise mer realistisk gasseparasjon i fibermoduler. Studiet viser at membraner for naturgassresning er både teknisk og økonomisk mulig, men dehydreringsenheten må forbedres med hensyn til design, modulkonfigurasjon eller membranytelse. Testing av membranene i anleggsskala er nødvendig for å bestemme levetiden og for videre validering av membranmodellen. COM-teknologien og Cape-Open enhetsoperasjon i simuleringsprogrammet må utforskes for ekstern optimalisering av bruker-definerte modeller. Til slutt, en detaljert evaluering av kostnaden, inkludert en økonomisk analyse og en sensitivitetsanalyse, er nødvendig for å forbedre kostnadsberegningene. Membranes for natural gas treatment are gaining increasing acceptance with the increasing natural gas demand and stricter environmental regulations. In this work a hybrid membrane system for natural gas dehydration and CO2 removal was studied. A Spiral-wound MTR Pebax® co-polymer based composite membrane for dehydration was simulated, and a hollow fiber polyimide derived carbon membrane was simulated in series for removal of CO2. The aim was to evaluate the technical end economical feasibility with respect to pipeline and sales specifications and specific cost of natural gas processing. Furthermore a membrane model was developed, compared to experimental data and analyzed for use in the process simulation to predict the membrane performance. Various design configurations were proposed with respect to the number and arrangement of compressors. Each design configuration was optimized using the Nelder-Mead simplex algorithm with a penalty function. The optimization routine was carried out externally by creating a COM automation server (ActiveX) to connect with the simulation interface. The set of membrane model equations were dicretized using the orthogonal collocation method to form a set of non-linear algebraic equations which was solved applying the Newton-Rhapson method. The membrane model was then integrated in the simulation environment with the optimal design configuration as a Cape-Open unit operation based on the COM technology. The optimal design configuration contained two compressors in series between the dehydration unit and the sweetening unit, with compression of the recycled retentate from the sweetening unit, and a pressure relief valve instead of a turbine for expanding the sweetening unit permeate before entering the dehydration unit as sweep gas. The specific natural gas processing cost for the optimal design was 2.192E−3 $/m^3, and the pipeline and sales specifications were met. For the optimal design with the ChemBrane model the water dew point was -52.5 centigrades, the methane purity was 97.5%, the CO2 purity in the gas for enhanced oil or gas recovery was 95%, and the gas dew point was -11.5 centigrades. The design with the model developed in this work was not optimized due to connection issues between the COM automation server and the simulation case. The model was found suitable for gas transport through hollow fiber membranes by one of two experiments with a relative deviation ≤ 3.65%. The other experiment deemed the model invalid due to large deviations in the permeate flow rate and permeate purity. Compared to the ChemBrane model, which was initially applied in the process design, the model showed similar but more realistic performance. In conclusion, the study shows that it is technically and economically feasible to use membranes for gas treatment, but the dehydration unit must be improved in terms of design, module configuration or membrane performance. Testing of the membranes in plant scale is necessary to determine the membrane lifetime and for further model validation. Investigation of the COM-technology and Cape-Open unit operation in the simulation environment is necessary for external optimization of user-defined models. Lastly, a detailed economical evaluation including an economic analysis and a sensitivity analysis is required to improve the cost estimations.

In this project the Power-to-Methane process will be designed and evaluated that consists of an alkaline electrolyser, multi-tubular methanation reactor and polyimide membrane. The feed gas CO2 is modelled in accordance with specifications given from the amine absorber located at the Biokraft plant in Skogn and the product is required to be of liquefaction quality. By applying the Power-to-Methane process the productivity of the plant can be increased while decreasing the CO2 emissions of the plant. The process has been designed and evaluated in HYSYS V10. An advanced MATLAB (R2019a) code for the multi-tubular methanation reactor was connected by using a CAPE-OPEN unit operation. Such an advanced way of modelling the methanation reactor was required because of a high runaway temperature from the highly exothermic methanation reaction. As a consequence of this high runaway temperature, the catalyst can be deactivated which is strongly unwanted. The design work and results are in accordance with similar work in literature. The Power-to-Methane process is rather novel and has become more interesting in industry over the last years. The process becomes more novel in this project since the thought is to include the liquefaction step to produce bio-LNG from the high quality biomethane that is produced. The results of the model suggest that the Power-to-Methane process is favoured at high operating pressure. However, this results in a higher reaction rate which in turn increases the temperature runaway. To solve this challenge, it is possible to reduce the dilution factor to oppose the reaction rate. This suggests that there must be an optimum point at which the pressure can be maximised while maintaining a stable process. The alkaline electrolyser for the production of hydrogen has a high investment cost and electrical energy demand. Therefore, it is the main contributor to the cost for the Power-to-Methane process. In fact, the electrical energy demand is found to be so high, that the cost for producing methane are almost as high as the production cost of hydrogen making the process not profitable for now. However, the electricity cost and investment cost electrolyser are expected to decrease in the next years before the plant is expected to be build.

Målet med dette prosjektet er å studere implementering av ulineær modellprediktiv regulering (NMPC) for en reaktor for emulsjonpolymerisasjon. En modell utledes ved bruk av fysiske prinsipper, og brukes til å erstatte målte data fra prosessanlegget. Den brukes også som modell for NMPC'en. Størsteparten av denne modellen ble utviklet i et tidligere prosjekt, men noen tilpasninger utføres for å gjøre den anvendelig for NMPC. Ytelsen til kontrollsystemet testes og demonstreres ved simuleringer. De teoretiske aspektene ved polymerer og emulsjonspolymerisasjon er spesifikt utdypet i denne oppgaven. Modellikningene som beskriver denne prosessen er implementert i C-kode, i en av Cyberneticas maler for polymerprosesser. Prosessen som studeres er emulsjonspolymerisasjonen av vinyl klorid (VCM), til det ønskede produktet poly-vinyl klorid (PVC). Modellen valideres og modifiseres mot prosessdata fra Inovyn Norge. Modellen ga akseptable resultater sammenlignet med prosessdata. Det understrekes at modellen kan gjøres anvendelig for andre systemer enn PVC, med litt anstrengelse og forståelse av prosessen. Konseptene bak optimalisering, MPC (Modellprediktiv regulering) og tilstands- og parameterestimering er gitt. I tillegg er noen spesifikke aspekter ved regulering og NMPC implementering av en semi-batch reaktor for emulsjonspolymerisasjon presentert. Noen avvik mellom modellene som ble brukt til å erstatte målte data for prosessanlegget, og i NMPC'en, ble bevisst implementert. Dette skal etterligne det faktum at en modell aldri vil være en eksakt representasjon av virkeligheten. Et KF (Kalman Filter) ble implementert som tilstand- og parameterestimator. Applikasjonen testes så både med og uten estimatoren. Implementeringen av NMPC, i kombinasjon med estimering av tilstand og parametere, ansees som vellykket i dette prosjektet. Resultatene indikerer at kontrollsystemet er i stand til å utnytte kjølekapasiteten ved å føde reaktanter og kontrollere temperaturen på en effektiv måte. Sikker drift og kvalitetskrav er da opprettholdt. The objective of this thesis is to study the implementation of NMPC (Nonlinear Model Predictive Control) on a reactor for an emulsion polymerization process. A model is derived using first principles and serves as a plant replacement model, in addition to an integrated part of the controller. The majority of this model was previously developed, and some customization is done to make it applicable for NMPC. The control system is tuned, and the performance is demonstrated by simulations. The theoretical aspects of polymers and emulsion polymerization are explicitly provided in this thesis. Necessary model equations describing this process is implemented in C code, in one of Cyberneticas templates for polymer processes. The process studied is the emulsion polymerization of vinyl chloride (VCM) to the desired product poly-vinyl chloride (PVC). The model is validated and modified against process data provided by Inovyn Norge. It is concluded that the model gave satisfactory results compared to process data. It is emphasized that the model can be done applicable for other systems than PVC, with some effort and an understanding of the process. The concepts behind optimization, MPC (Model Predictive Control) and state- and parameter estimation is provided. In addition, is some specific aspects of control and NMPC implementation on this process, presented. Some deviations between the model used as plant replacement, and in the controller, is deliberately implemented. This is done to imitate the fact that a model never will be an exact representation of reality. A KF (Kalman Filter) was implemented as the state- and parameter estimator, and was tuned and tested. The implementation of NMPC, in combination with state- and parameter estimation, on the reactor, is considered successful in this project. The results indicate that the control system can exploit the cooling capacity by feeding reactants and controlling the temperature in an efficient matter. This is done by still ensuring safe operation and meeting quality requirements.