modellering og regulering af et stoker system

Modeling and Control of a Stoker Firing System

Hansen, Heine ; Lantow, Louis Schultz

10. term

Elektronik og IT, Kandidatuddannelsen (Spec. Intelligent Autonomous Systems)

2008

150

Rapporten undersøger, hvordan man kan styre et Benekov/Liagro stokeranlæg, stillet til rådighed af Techno-Matic A/S, som normalt udvikler regulatorer til Benekov/Liagro. For at beskrive anlægget er der udviklet to ikke-lineære modeller: en for træpiller og en for hvede. Modellerne bygger på seks differentialligninger af første orden udledt fra termodynamik og fysiske love og er verificeret i Simulink med måledata fra anlægget. For at designe en regulator er modellerne lineariseret ved hjælp af en førsteordens Taylor-approksimation. Regulatoren er konstrueret ud fra en state-space model (en matrixbaseret fremstilling af den lineære model). Den indeholder to optimale regulatorer baseret på lineær kvadratisk regulator (LQR) teori, en til hvert brændsel, for at balancere ydeevne og styreindsats. Derudover indeholder den et brændselsestimat, som skal afgøre hvilket brændsel der afbrændes, designet ud fra de to modeller, Kalman-filtre (til at estimere tilstande under støj) og Bayes' sandsynlighedsregel (til at sammenligne sandsynligheder). De tre dele er først verificeret hver for sig i simuleringer med tilfredsstillende resultater og derefter testet på anlægget. Simulationen af de to optimale regulatorer bruger den ikke-lineære model som anlæg. Resultaterne opnået i simuleringerne kunne ikke reproduceres på det virkelige anlæg, og regulatorerne opfylder derfor ikke de ønskede krav. De tre dele er samlet til en adaptiv regulator, som også er testet på anlægget. På grund af tidsmangel blev der ikke opnået tilfredsstillende resultater for den adaptive regulator.

This report investigates how to control a Benekov/Liagro stoker system provided by Techno-Matic A/S, which normally develops controllers for Benekov/Liagro. To describe the plant, two nonlinear models were developed: one for wood pellets and one for wheat. The models are based on six first-order differential equations derived from thermodynamics and physical laws and were validated in Simulink using measurement data from the installation. To enable controller design, the models were linearized using a first-order Taylor approximation. The controller was built from a state-space model (a matrix-based representation of the linear model). It includes two optimal controllers based on linear quadratic regulator (LQR) theory, one for each fuel, to balance performance and control effort. It also contains a fuel estimator to determine which fuel is being burned, designed using the two models, Kalman filters (to estimate states under noise), and Bayes' rule (to compare probabilities). The three parts were first verified individually in simulations with satisfactory results and then tested on the plant. The simulation of the two optimal controllers used the nonlinear model as the plant. However, the results achieved in simulation were not obtained on the real installation, so the controllers do not meet the desired requirements. The three parts were combined into an adaptive controller, which was also tested on the plant. Due to time constraints, satisfactory results were not achieved for the adaptive controller.

[This abstract was generated with the help of AI]

Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
Download
View record in AAU Student Projects