Laboratory emulation and control of a sewer system with storageThe idea of using a Model Predictive Control for controlling wastewater flow was implemented and tested on a laboratory setup designed and built by Grundfos. The master thesis was done in collaboration with Grundfos A/S and Fredericia Spildevand og Energi A/S.Some of the key tasks of the project were:• Prediction of flow disturbances using Kalman Filter• Setup the laboratory equipment to emulate a sewer system• Setup the automation software to receive instrument readings and to send control signals• Implement the Model Predictive Control (MPC) algorithmResult: • Reduced the variance of wastewater flow by 80% • Built a real-time automation system that controlled and monitored the laboratory setup• Received the highest grade for the workLearning Outcomes: Data Analysis, Automation, Kalman Filter, Hands-on experience working with pumps, valves and sensors

Model Predictive Control of a sewer systemThe municipality of Fredericia wanted to investigate the possibilities of minimizing the discharge of untreated wastewater into nearby water bodies, while also minimizing the energy consumption at the wastewater treatment plant.To solve this issue, the intent of the project was to build special gates to redirect or hold back water where basins have the capacity to store wastewater giving the possibility to smooth out the flow of water and contaminants such that the wastewater plant is stressed as little as possible. The project was done in collaboration with Fredericia Spildevand og Energi A/S.Over the course of the project, the tasks we completed were:• Build a simplified model describing the dynamics of the main sewer system r• Develop a performance function and suitable constraints to form the basis for evaluation of controller performance• Use the model, performance function and constraints to develop a controller to operate the sewer systemLearning Outcomes: Modeling of Sewer Network, Storm Water Management (EPA SWMM), Model Predictive Control, Working of a Wastewater Treatment Plant

Data-driven approach to system modeling and performance optimizationIt is hard to develop a first principles based model that captures all the essential dynamics for a complex system. A new method was needed to obtain a sufficiently rich system model with the purpose of control optimization and diagnosis.This project addressed a data-driven approach to modeling and fault-tolerant control for a refrigeration system. A test setup situated in Aalborg University (AAU) and data obtained from supermarket test setup at Danfoss, Nordborg was used. System Identification or data-driven modeling with Subspace Identification and Auto-Regressive Moving Average with eXogenous input (ARMAX) was developed. The models were derived by persistently exciting the system with a PRBS input and recording the system response to the same. The obtained input-output data is then fit to a known model structure. Then, a fault-tolerant control scheme to detect and handle sensor faults was developed. This was done by residual generation and the use of CUSUM algorithm for detection, while a Kalman filter is used for estimation and fault tolerance. Learning Outcomes: Working of Refrigeration Systems, Modeling of Refrigeration System,System Identification, Fault Detection and DiagnosisThe project was done in collaboration with Danfoss A/S, Nordborg.

Company:An engineering, procurement an construction firm that provides project and technical services to energy, process and utility markets across the world.Areas of Responsibility:Instrument Database, Instrument Datasheets and Instrument Purchase Orders, Control Valve and Pressure Safety Valve Sizing, Vendor Document ReviewSkills Acquired: Interdisciplinary Coordination, Technical Writing, Industry Knowledge, Time Management, Accountability, HSE PracticesExperienced with:Multidisciplinary coordination with electrical, mechanical, chemical and civil engineering teams. Collaborative working with our international offices in Paris, Milan, Aberdeen and Alberta