ABSTRACT. The industry is currently seeing a rapid development of cyber-physical system products containing integrated software, hardware, and communication components. The increasing system complexity in the automotive and aerospace industries are some examples. The systems that are developed have increasing demands of dependability and usability. Moreover, lead time and cost efficiency continue to be essential for industry competitiveness. Extensive use of modeling and simulation - Model-Based Systems Engineering tools - throughout the value chain and system life-cycle is one of the most important ways to effectively target these challenges. Simultaneously there is an increased interest in open source tools that allow more control of tool features and support, and increased cooperation and shared access to knowledge and innovations between organizations.

Modelica is a modern, strongly typed, declarative, equation-based, and object-oriented (EOO) language for model-based systems engineering including modeling and simulation of complex cyber-physical systems Major features are: ease of use, visual design of models with combination of lego-like predefined model building blocks, ability to define model libraries with reusable components, support for modeling and simulation of complex applications involving parts from several application domains, and many more useful facilities. The Modelica language is ideally suited for cyber-physical modeling tasks since it allows integrated modeling of discrete-time (embedded control software) and continuous-time (process dynamics, often for physical hardware). Modelica 3.3 extended the language with clocked synchronous constructs, which are especially well suited to model and integrate physical and digital hardware with model-based software.

This talk gives an overview and outlook of the OpenModelica environment – the most complete Modelica open-source tool for modeling, engineering, simulation, and development of systems applications (www.openmodelica.org), and its usage for cyber-physical system development. Special features are MetaModeling for efficient model transformations, debugging support for equation-based models, support (via OMSimulator) for the Functional Mockup Interface for general tool integration and model export/import between tools, model-based optimization, as well as generation of parallel code for multi-core architectures.

Moreover, also mentioned is recent work to make an OpenModelica based tool chain for developing digital controller software for embedded systems, and in generating embedded controller code for very small target platforms like Arduino Boards with down to 2kbyte memory. This work is extended in the ongoing EMPHYSIS project where the FMI standard is extended into the eFMI standard for embedded systems.

ABSTRACT. The increasing amount of variable electricity generation has brought world to investigate various flexibility sources to provide power network balancing through demand side management. Therefore, it is important to create new, more thorough models that allow using smart functions to control the various electricity loads. In this paper a model to simulate a fully mixed domestic hot water tank’s behavior in 60, 30 and 15 min time resolution, and its control mechanisms were created. The model will be integrated to another smart house model to enable studying more combined smart controls and functions. Additionally, the flexibility of the hot water storage tank was investigated with the help of 4 different heating scenarios, showing its suitability for Demand Side Management, and the operation of the model was confirmed with lower time resolutions.

ABSTRACT. Energy management of small-scale renewable energy systems (microgrid) requires control of the energy consumption. Commercial buildings are sized using energy consumption criteria from the building standards (NS3031). The energy consumption of a building depends on a number of dynamic factors, including thermal loss, building climate control, and building utilization. This paper presents a simple model for predicting the thermal loss of a building. The basic simulation module is a single room model where the outer walls and windows are exposed to the ambient conditions without sun influx. We model the thermal loss based on building design data and validate the model using operational data for an actual low-energy building. The model prediction accuracy is within +/-1°C for up to 7 days when predicting thermal losses in the building construction.

ABSTRACT. Floor heating has been used for thousands of years, and essentially consists of a heat generation system and a heat distribution system with heat transfer through the floor. Modern studies for low energy buildings focus on taking advantage of water with low thermal value (lukewarm, 30-35 C), which necessitates reducing heat transfer coefficients in the system. Modern control systems allow for reducing the temperature when (part of) the building is unused, but require heating system with low heat capacity to be efficient. This implies using above-floor systems, i.e., inserting pipes in the underlayment between the subfloor (e.g., chipboard) and the floor covering (parquet, etc.).

In this paper, a heated water tank model as in Johansen et al. (2019) is modified/improved, and is extended with a water distribution system for floor heating. In comparison with the model in Lie et al. (2014), the solar heating element is excluded and the heated tank model is improved. Also, the model of the distribution system and the floor heating is improved. The developed model is compared with experimental data from Johansen et al. (2019), and the effect of the water recycling is analyzed wrt. its effect on the dynamics.

References:

Johansen, C. A., Lie, B., and Skeie, N.-O. (2019), “Models for control of thermal energy in buildings”, in E. Dahlquist, E. Juuso, B. Lie & L. Eriksson, eds, Proceedings of the 60th Conference on Simulation and Modelling (SIMS 59), no. 170 in Linköping Electronic Conference Proceedings, Linköping University Electronic Press.

Lie, B., Pfeiffer, C., and Beyer, H.-G. (2014). “Using history-based irradiance forecasts for supporting the predictive control of solar thermal systems”. Proceedings, EuroSun 2014, September 16– 19, Aix-les-Bains, France.

Sattari, S., and Farhanieh, B. (2006). “A parametric study on radiant floor heating system performance”, Renewable Energy, 31, no. 10, pp. 1617-1626.

Ho, S. Y., Hayes, R. E., and Wood, R. K. (1995). “Simulation of the dynamic behaviour of a hydronic floor heating system”, Heat Recovery Systems and CHP, 15, no. 6, pp. 505-519.

ABSTRACT. Floor heating has been used for thousands of years, and essentially consists of a heat generation system and a heat distribution system with heat transfer through the floor. Modern systems utilize “low quality” thermal energy in lukewarm water (30-35 C). To reduce the overall energy consumption, it is necessary to allow for rapid temperature changes when occupants leave or arrive, hence a low heat capacity in the distribution system is required. Maximizing the temperature in the buildings requires low heat transfer coefficients. Building energy management systems (BEMS) are used to handle the heating and water distribution.

Lie et al. (2014) discussed the use of solar heating assisted by electric heating for floor heating, and studied the use of Model Predictive Control (MPC). In Johansen et al. (2019), an improved model of an electric heater was considered, and compared with experimental data. In this paper, solar heating is excluded, but an improved model is considered: the water tank with electric heating from Johansen et al. (2019) is improved, and a distribution system with a more detailed model than in Lie et al. (2014) is used.

The emphasis of the paper is on how to use modern computer languages to analyze the model, with the ultimate purpose of control design. The focus is not on control design, but on analysis tools. Examples of analysis include output sensitivity to model parameters, model linearization, and system zeros. Parameter sensitivity is important in identifiability studies (Sarmiento Ferrero et al., 2006), linearization is relevant for control design, and system zeros are relevant for control architecture studies and the choice of sensors and actuators (Lie, 1995). Because the original stratification mechanism in the heated water tank makes it difficult to linearize the model, model modification is considered. For illustration, the model is implemented in both OpenModelica (Lie et al., 2019) and Julia (Rackauckas and Nie, 2017).

References:

Johansen, C. A., Lie, B, and Skeie, N.-O. (2019), “Models for control of thermal energy in buildings”, in E. Dahlquist, E. Juuso, B. Lie, and L. Eriksson, eds, Proceedings of the 60th Conference on Simulation and Modelling (SIMS 59), no. 170 in Linköping Electronic Conference Proceedings, Linköping University Electronic Press.

Lie, B., Pfeiffer, C., Beyer, and H.-G. (2014). “Using history-based irradiance forecasts for supporting the predictive control of solar thermal systems”. Proceedings, EuroSun 2014, September 16– 19, Aix-les-Bains, France. Lie, B. (1995). “Attainable Performance in LQG control.” Pages 263–295 in R. Berber (1995), editor: Methods of Model Based Process Control, Kluwer Academic Publishers. ISBN 0-7923-3524-4.

Sarmiento Ferrero, C., Chai, Q., Dueñas Díez, M., Amrani, S.H., and Lie, B. (2006). “Systematic Analysis of Parameter Identifiability for Improved Fitting of a Biological Wastewater Model to Experimental Data”, Modeling, Identification and Control, Vol. 27, no. 4, pp. 219–238.

Lie, Bernt, Palanisamy, Arunkumar, Mengist, Alachew, Buffoni, Lena, Sjölund, Martin, Asghar, Adeel, Pop, Adrian, and Fritzson, Peter (2019). “An OpenModelica API for Julia-Modelica Interaction”. Proceedings of the 13th International Modelica Conference, Regensburg, Germany, March 4–6, 2019, pp. 699–708. Published by Linköping University Electronic Press, ISBN: 978-91-7685-417-4, ISSN (on-line): 1650-3740, doi: http://doi.org/10.3384/ecp19157.

Christopher Rackauckas, and Qing Nie (2017). “DifferentialEquations.jl – A Performant and Feature-Rich Ecosystem for Solving Differential Equations in Julia”. Journal of Open Research Software, Vol. 5, no. 15, DOI: https://doi.org/10.5334/jors.151.

ABSTRACT. Boiling flows are widely encountered in several engineering and industrial processes. They have a special interest in nuclear industry, where a Computational Fluid Dynamic (CFD) thermohydraulic investigation becomes very popular for design and safety. Many attempts to model numerically subcooled nucleate boiling flows can be found in the literature, where several interfacial forces acting on bubbles which are interacting on the bulk fluid were neglected, due to the hard convergence of the calculations, or to the bad accuracy of the obtained results. In this paper, a sensitivity analysis is carried out for the interfacial forces acting on bubbles during subcooled nucleate boiling flows. For this purpose, 2D CFD axisymmetric simulations based on an Eulerian approach are performed. The developed models aim to mimic the subcooled nucleate boiling flows in concentric pipes, operating at high pressure. The predicted spatial fields of boiling quantities of interest are presented and commented. The numerical results are compared against the available experimental data, where it is shown that neglecting some interfacial forces like the lift or the wall lubrication forces will yield to good predictions for some quantities but will fail the prediction for others.  The models leading to the best predictions are highlighted and proposed as recommendations for future CFD simulations of subcooled nucleate boiling flows.

ABSTRACT. Gasification using a fluidized bed is a promising technology to convert agricultural residues into product gases. In this work, the syngas production potential from agricultural waste (grass pellets) is studied using a computational particle fluid dynamic (CPFD) model. The CPFD model is developed using the simulation software Barracuda and validated against experimental data. Experiments are carried out in a 20 kW bubbling fluidized bed gasification reactor that operates with air as the fluidizing gas. Grass pellets of size 5mm-30 mm in length and 5mm in diameter are used as the feed. The CPFD model considers the hydrodynamics of the gas-solid phase and the reaction kinetics involved. Influence of the static bed height, bed temperature, and air to fuel ratio on the product gases (〖CH〗_4,CO,〖〖CO〗_2,and H〗_2) and char conversion efficiency are investigated. Initial bed heights of 200 mm and 300 mm are used for the analysis. Biomass is fed at 2.46 kg/hr while the air supplied is varied to obtain the air to fuel ratio of 0.4, 0.6 and 0.8. The result shows that the increase in bed height has a significant effect on reactor temperature but very small effect on the product gas compositions and char conversion rate. An increase in bed temperature from 600℃ to 800℃ improves the gasifier performance in terms of maximum product gas quality and enhanced char conversion rate. The increase in the air to fuel ratio from 0.4 to 0.8 reduces 〖CH〗_4,CO,〖 and H〗_2 fractions in the product gas and increases 〖CO〗_2 concentration. The results obtained from the CPFD model are in good agreement with the experimental results and literature data. The CPFD model developed in this work can be utilized to optimize gasification reactors.

ABSTRACT. The chemical and thermal processes associated with the decarbonation and fuel combustion in the cement kiln process produce a large amount of carbon dioxide (CO2). Utilizing green electricity instead of fossil fuels to decarbonate the raw meal in the calciner can eliminate the CO2 emissions produced through fuel combustion and also provide a basis for simple capture of the CO2 generated through calcination, as CO2 is the only gaseous product exiting from the electrified calciner. In the current work, an electrically heated fluidized bed (FB) reactor is being developed to calcine the raw meal. The FB may replace the traditional entrainment calciner used in many plants. The purpose is to enable efficient indirect heat transfer in the bubbling bed and hence obtain pure CO2 as the gaseous product from the calciner. The minimum fluidization velocity and pressure drop of the particle bed are important characteristics in the design of a bubbling fluidized bed, and these have been measured in a cold-flow lab-scale fluidized bed unit with a bed height of 0.21 m and a circular cross-sectional area of 55 cm². The particle size distribution of the meal ranged from 0.2 – 180 µm, with a median particle size of 21 µm. The experimental results revealed that the regular cement raw meal is difficult to fluidize due to the large fraction of Geldart C particles in the meal (approximately 60%). This may be explained by inter-particular electrostatic forces forming particle clusters. The fluidization process has also been simulated with the commercial computational particle and fluid dynamics (CPFD) software Barracuda® (version 17.4.1). The purpose of using CPFD was to be able to simulate the process at cold-flow conditions and then, based on this, simulate the process at large-scale hot-flow conditions. The simulation results complied quite well with the lab-scale experiments and confirmed the difficult fluidization of the meal.

ABSTRACT. Abstract Fluidized bed reactors have several advantages, such as high heat and mass transfer rate, good mixing properties, and close to isothermal conditions. Because of these advantages, the fluidization concept is widely used in different engineering applications. In the current work, computational particle and fluid dynamics (CPFD) simulations are used to study an electrically heated bubbling fluidized bed (BFB) used as a calciner in a cement manufacturing process, applying a binary-particle fluidization system. Owing to the fine particle size (0.2 – 180 µm) of the limestone used as a raw meal in the cement kiln process, a conventional bubbling fluidized bed may be difficult to apply due to particle cohesion causing poor fluidizability of the particles smaller than 30 µm. The raw meal particles can be characterized mainly as C particles under the Geldart powder classification, and those are known to have poor fluidization properties. In the current study, to enhance the fluidization of the raw meal particles by mixing them with coarse, inert Geldart B particles was proposed. The inert particles may provide a homogeneous distribution of the raw meal particles and help to fluidize them. The aggregation and clustering of the fine particles will decrease due to collisions with inert coarse particles. The inert particles will also provide a thermal energy reservoir through their heat capacity and thereby contribute to a very stable bed temperature, which is advantageous in the control of the process. After the raw meal particles have been calcined, they have to be separated from the coarse, inert particles. This can be done by increasing the velocity of the CO2 used for fluidization to a value sufficiently high to entrain the raw meal particles, but still sufficiently low that the coarse, inert particles are not entrained. This means that the fluidized bed calciner process is designed as a semi-batch process operating in two modes: the calcination mode (with a low gas velocity) and the entrainment mode (with a higher velocity). The commercial CPFD software Barracuda® (version 17.4.1) was used for simulations to investigate suitable operational conditions at 1173 K, such as the particle size distribution of the inert particles and the fluidization gas velocity. Inert coarse particles with a diameter in the range 550 – 800 µm, an average minimum fluidization velocity of 0.129 m/s and a terminal settling velocity of 3.24 m/s appeared as suitable for the calciner operation. The impact of gas velocity variation on the fluidization of the particle mixture was studied, and an appropriate range of velocities for the calcination and entrainment modes could be determined. The simulations revealed that mixing raw meal particles with inert coarse particles could enhance the flowability in the FB reactor indicating that it is possible to apply the concept in a full-scale calcination process.

ABSTRACT. Crucial elements in offshore production units are caissons. Caissons are vented tubular structures ranging over several floors designed to receive excess process water streams. The caisson can handle different flows, either continuous or intermittently, with the purpose to diffusing the high momentum of the water flows. To reduce the forces on the structure the flows often enter tangentially into the caisson. Nevertheless, these flows can cause detrimental effects such as exceeding the design pressure of one of the connecting supports or generating mechanical vibrations, which can in turn cause fatigue. Current study focuses on a seawater caisson with typical dimensions of one placed on an oil and gas platform. Stressman Engineering AS and other supply companies to oil and gas sector are concerned about minimizing costs and weight of equipment maintaining safe operation and low maintenance. To quantify the demands of a caisson, the internal flow field have to be investigated. One of the crucial parameter affecting the design is the internal pressure distributions on the walls of the caisson. This study uses OpenFOAM as a Computational Fluid Dynamics (CFD) tool to quantify the distributed internal pressures with a multiphase formulation to handle the interactions between air and water.

ABSTRACT. Early kick/loss detection is a crucial part of safe well control, and it plays a major role in the reduction of risk and non-productive time in drilling. In conventional drilling, topside sensing is used for early kick/loss detection. Recently, a Venturi flow meter based online return flow estimation has been introduced for this purpose by the authors. In managed pressure drilling, both topside sensing and bottomside sensing can be used for kick/loss detection. Therefore, a topside return flow estimator with a bottomside well pressure and flow estimator is combined to provide a complete kick/loss detection and estimation scheme for managed pressure drilling systems. This allows improved kick/loss detection. In addition, a closed-loop kick/loss attenuation controller is used to illustrate the estimation scheme.

ABSTRACT. Integrated Autodriller Controller (IADC) is a software product by National Oilwell Varco to allow an operator to drill with a constant weight on bit (WOB). The control loop has a PI controller and PI gains are dependent on drill string compliance and response time constant (RT). The compliance is calculated based on the drill string length, but RT is difficult to predict. The existing IADC uses an empirical regression model to estimate RT based on the string length, which fails during a formation change. This work presents an Adaptive Response Time (ART), a machine learning model to predict RT based on the rate of penetration (ROP). ART can adapt to formation changes and gives smoother regulation of block speed and WOB. Simulation results also show a better torque & speed regulation of drill string using ART.

ABSTRACT. One of the main principles of improving oil recovery is maximizing the contact area between the well and the reservoir. To achieve this purpose especially in reservoirs with a thin oil column, long horizontal wells are widely used today. However, there are some challenges related to horizontal wells like water coning towards the heel due to the heel-toe effect as well as early water breakthrough owing to heterogeneity along the well. In order to tackle these issues, passive inflow control devices (ICDs) and autonomous inflow control devices (AICDs) can be used. ICDs are able to balance the drawdown pressure along the horizontal well and as a result, postpone the early water breakthrough. By applying AICDs, in addition to postponing the early water breakthrough, water can be partially choked back autonomously, and the negative impacts of early water breakthrough will be attenuated. The Johan Sverdrup field (JSF) is a giant oil field located in the North Sea and production from this field has been started recently. Since there is a plan for developing this oil field in the near future, and a few studies have been done on this field so far, further studies are needed to obtain more cost-effective oil recovery in this field. The main objective of this paper is near-well simulation of oil production from the well 16/2-D-12 in the JSF by considering ICD and AICD completions. The simulation has been conducted based on the characteristics of the reservoir near this well for 750 days of oil production. OLGA in combination with ROCX has been used as a simulation tool. The simulation results showed that by applying both ICDs and AICDs the heel-toe effect, and heterogeneity along the well can be effectively handled and the water breakthrough time can be delayed by 255 days. Moreover, it was observed that by completion of the well 16/2-D-12 with AICDs, the accumulated water production can be reduced by 11.9% compared to using ICDs. In the same way, by using AICDs the flow rate of water production is reduced by 13.4% after 750 days. Furthermore, the results showed that using AICDs has a negligible impact on both the accumulated and the flow rate of oil production compared to using ICDs. Therefore, by completion of the well 16/2-D-12 with AICDs more cost-effective oil production can be achieved.

ABSTRACT. This paper aims to describe the steady-state and dynamic heat pump models developed to study their abilities in ancillary services as well as the inter­connection between these, electrical boilers and thermal storage with the aim to balance power and heat production for a given case. An hourly steady-state system model was developed to understand the overall operational characteristics of the system for a given heat demand case in the area Kolt-Hasselager-Ormslev near Aarhus in Denmark. The model showed an annual average COP above 3,5 for a serial connected heat pump system. Detailed thermal and dynamical models of the heat pump system were developed. The models show that it will be possible to use heat pumps successfully in ancillary services. The turn-up is unproblematic but the turn down of the heat pump will be limited in a non-liquid overfed system due to risk of liquid formation in the evaporator, requiring additional heating.

ABSTRACT. Understanding the dynamics of building energy consumption and anticipating their demand is crucial to enhance demand response measures or anticipate the electricity demand on future environment. Generating electric load profiles is the first step towards a better integration and modelling of the building sector. Challenges arise at different scales, some being daily or hourly problems, other being in the matter of minutes, seconds or even smaller steps [1]. There is therefore a need to integrate better these variable optimisation steps while keeping good simulation performance and being flexible to weather, location, building design, technology development and integration, occupancy, and individual behaviour characteristics.

ABSTRACT. Compressed Air Energy Storage (CAES) is a promising alternative for energy storage. An Adiabatic Compressed Air Energy Storage (A-CAES) system has been analysed in this paper, to store excess energy production from a wind turbine generator for up to one week. Compressed air is stored in a cavern of constant volume. The heat produced by the compression of the air during the charge process is stored in several packed beds and used later during discharge to reheat the air prior to energy production.

This paper presents a preliminary thermodynamic analysis estimating the size of the system for a given quantity of energy storage, a dynamic model including the packed beds for energy storage, and a simulation made in MATLAB to analyse the efficiency of the system. The A-CAES roundtrip efficiency is 53.4%. The physical limitations of the actual compressors and expanders were taken into account when defining the operational pressures and temperatures in the model. A preliminary capital cost estimation of the system was conducted for Norway, resulting in an estimated investing cost of approximately 2 700 NOK/kWh.

ABSTRACT. This paper investigates the sensitivity on choice of heat pump concept to uncertainties and variations in boundary conditions for large scale heat pumps with a danish district heating system as case example. Large scale heat pumps have recently received increased interest due to tax-reductions, green initiatives and potential for ancillary services. A load-dependent performance map based heat pump model is used to evaluate performance of different heat pump concepts. Since large scale heat pumps have great potential for providing ancillary services, this as well as performance under normal operation should be considered in the choice of heat pump concept. A performance map based heat pump model is implemented in the multi-domain modelling language, Modelica and a micro genetic algorithm is proposed as a nonlinear optimizer for finding the optimised concept design. A sensitivity analysis is additionally carried out to identify whether and which variations in boundary conditions and cost function weighting affects the optimisation and how this consequently affects the choice of heat pump concept.

ABSTRACT. Injection of supercritical carbon dioxide (CO₂) for enhanced oil recovery (EOR), plays a vital role to minimize the impact of CO₂ emissions. CO₂-EOR refers to the oil recovery technique where supercritical CO₂ is injected in the reservoir to stimulate oil production from depleted oil fields. CO₂-EOR can be used in combination with CO₂ storage to mitigate the emissions levels to the atmosphere. The objective of this paper is to perform a computational study of CO₂-EOR and storage at the Johan Sverdrup field. The study includes simulations of oil production using the commercial software Rocx in combination with OLGA. Production with inflow control devices (ICD) and autonomous inflow control valves (AICV) shows that AICVs have an oil-to-water ratio of 0.92 compared to 0.39 for ICDs. CO₂-EOR in combination with well completion with AICVs, shows improved oil recovery, low water production, and low CO₂ reproduction. The simulations and calculations performed in this paper indicate that the Johan Sverdrup field is highly capable for CO₂-EOR and storage.

ABSTRACT. The industrial deployment of amine-based CO₂ capture technology requires large investments as well as extensive energy supply for desorption. Therefore, the need for efficient cost and economic analysis aimed at CO₂ capture investment and operating costs is imperative. Aspen HYSYS simulations of an 85% CO₂ absorption and desorption process for flue gas from cement industry, followed by cost estimation have been performed. This is to study the cost implications of different plants options. Each plant option has a different lean/rich heat exchanger type. Cost optimisation of the different heat exchangers is also done in this work. Three different shell and tube and two plate and frame heat exchangers have been examined. The minimum CO₂ capture cost of €57.9/ton CO₂ is obtained for a capture plant option having a gasketed-plate heat exchanger with ∆Tmin  of 5 ℃  as the lean/rich heat exchanger. The use of plate and frame heat exchangers will result in considerable CO₂ capture cost reduction.

ABSTRACT. In order to reduce CO2 emissions from exhaust gases, a standard method is absorption in monoethanolamine (MEA) followed by desorption. Several projects have been performed at Telemark University College and University of South-Eastern Norway about process simulation and cost estimation. An aim has been to find the process parameters which give the lowest combined investment and operating cost. A topic in these projects is economic optimization based on a spreadsheet integrated with the Aspen HYSYS simulation. The aim in this work is to calculate cost optimum process parameters and evaluate whether it is possible to perform automated cost estimation and optimization. Aspen HYSYS simulations of a standard amine based process for CO2 capture using an equilibrium based model have been performed in Aspen HYSYS version 10.0 using flue gas data from a cement plant. The capital cost of CO2 capture was estimated using equipment cost taken from Aspen In-plant and then using a detailed factor method. The cost analysis was limited to the absorption and circulation system, and compression or liquefaction was not included. Operational cost was estimated from calculated electricity and heat consumption and maintenance based on estimated capital cost. Parameters varied were the minimum temperature difference in the main heat exchanger, the number of absorption stages and % CO2 removed in the process. Optimum temperature difference in the main heat exchanger was calculated to be 10-15 °C. This was found after one simulation for each temperature. Optimum column height was calculated with 12 stages (equivalent to 12 meter of structured packing) based on one simulation for each stage number. The cost of CO2 removal was 180 NOK (19 Euro in 2019)/ton CO2 for 85 % removal and 190 NOK (20 Euro)/ton CO2 for 90 % removal. With a penalty for CO2 emissions higher than these values, 90 % removal can be regarded as most optimum. Compared to more detailed cost estimates on CO2 capture cost, the calculated cost is probably underestimated. The scope of the cost calculation is limited to the absorption and circulation system which is most important for parameter optimization. To obtain really automated calculations it is recommended to improve the robustness of the simulations. This may be achieved by making the material balances more accurate. In the parameter variation calculation, the cost of each equipment unit was based on a scaling from the base case cost. Then it should be in principle possible to optimize e.g. the temperature difference in only one automated calculation. To optimize the height (number of stages) in the absorption column automatically, a way to update the number of stages during the simulations has to be found.

ABSTRACT. Developing robust and predictable process simulation tools for CO2 absorption is important for improving carbon capture technology and reduce man made CO2 emissions. The main purpose of this work has been to fit simulated results with performance data from Test Centre Mongstad (TCM), and evaluate whether fitted parameters for one scenario (a set of experimental data at specified conditions) give reasonable predictions at other conditions. Five different scenarios from the amine based CO2 capture process at TCM have been simulated in a rate-based model in Aspen Plus and an equilibrium-based model in Aspen HYSYS and Aspen Plus. In the rate-based model, the performance data was fitted by only changing the interfacial area factor to obtain the experimental CO2 removal efficiency. The simulated temperature profile from top to bottom of the absorption column was fairly close to the measured temperature profile. In the equilibrium based model, a Murphree efficiency (EM) was specified for each of 24 stages (meter of packing) to fit both the CO2 removal efficiency and the temperature profile from performance data. In case of equal EM values for every stage, the EM is the only parameter. In this work different EM-profiles (different EM values on each stage) were examined to fit the temperature profile for a given scenario. Some of the specified EM-profiles were then used to fit performance data for other scenarios by adjusting only an EM-factor which multiplies all the EM values in an EM-profile. The performance (CO2 removal and temperature profile) was reasonably simulated for each given scenario for all the models. It was evaluated whether a fitted interfacial area (for the rate-based model) or an EM-profile (for the equilibrium based model) gave a good prediction for other conditions. The rate-based model fitted for a certain scenario was not able to predict performance well for other conditions. An indication of this was that the interfacial area fitted to the different scenarios varied between 0.29 and 1.0. Using a specific EM-profile was able to predict performance for all the scenarios fairly well. By multiplying the specified profile with an EM- factor (only one parameter), the fit at a new scenario was quite accurate. The fitted EM-factor from one scenario was fitted to values between 0.59 and 1.0 for all the scenarios. None of the models are expected to predict accurate performance for conditions far from the conditions in a specified scenario.

ABSTRACT. The transportation of CO2 is important to all Carbon Capture and Storage (CCS) projects. Both the infrastructure costs – compressors, pipelines, tanker ships, etc. – and the energy consumed in the compression or liquefaction of CO2 are significant. Understanding how the size, capacity and energy consumption of transportation alternatives varies between projects is therefore important. Modelling provides a useful insight into the performance of transportation alternatives, but the results are only useful when the basis for comparison is consistent and the impact of model input parameters is well understood. This article presents the results of sensitivity studies made using a transportation model that was developed in earlier/related work. Several important model parameters are studied using three planned/ operating CCS project cases. The results show that while the characteristics of the storage site are most important in determining the transportation system operating pressure, ambient temperature is the key parameter determining energy consumption.

ABSTRACT. Today there is an increasing focus on the possibility of developing data-driven dynamic models from “Big data”. Data-driven models are not new: the first attempts of models were based on measuring time and position of falling objects when clocks became reasonably accurate. The focus changed when more general laws were introduced by Newton and others. Goldstein (1980) gives a modern account of development of mechanistic models in classical mechanics. In parallel, data driven models made progress, e.g., by Gauss’ least squares work, and development of more formal regression models.

Forward neural networks constitute a modern example of nonlinear regression models, i.e., a static mapping from an input signal/feature vector to an output signal which can be fitted to experimental data via weight/bias parameters. Traditional use of neural networks for dynamic models are based on forward neural networks with autoregressive/delayed outputs and moving horizon/delayed inputs. Alternatively, recurrent neural networks include internal feedback paths in layers. Both approaches are based on discrete time data/models.

An alternative approach is the following. Realizing that forward neural networks essentially describe a nonlinear mapping, suppose now that both inputs and states are available over a time horizon, together with the derivative of the states. Then one can treat the inputs and states as inputs to the neural network, and the derivatives of the states as outputs from the neural network. By tuning neural network parameters, this enables fitting the neural network to the vector field of the differential equation, leading to Neural Differential Equations (Chen et al., 2018). If all states and state derivatives are known, fitting a neural network to the vector field is straightforward (e.g., Lie, 2019 for an academic example). States and state derivatives may be available in simulation studies for model reduction. However, when considering this idea for real systems with experimental data from a few sensors, two problems are faced: (i) derivatives are not available, and can at best be found by some smoothing/spline fitting, and (ii) derivatives can not be found for all states – only for a reduced order transformation given by measurements. This makes the problem more challenging.

Based on the ideas in Chen et al. (2018), a set of packages for computer language Julia (Bezanson et al., 2017) are combined to fit neural differential equation models (Rackauckas et al., 2019; Rackauckas et al., 2020) to experimental data, leading to differential equations that can be solved by standard differential equation solvers (Rackauckas and Nie, 2017). However, the packages aim higher: they allow for a very general mixing of mechanistic models and neural differential equation models in the same framework, with possibilities for the user to choose whether only parameters in the neural network model are tuned, or parameters in both the mechanistic model and the neural differential equation.

The advantage of this approach is significant: it allows for a decent mechanistic model to be used when data are scarce, e.g., at the start-up of a system, with gradual improvement of the model as more data (“Big data”) become available. Without a mechanistic model, there essentially is no model when data are scarce. In this paper, we consider a mechanistic model of a laboratory helicopter at University of South-Eastern Norway; the model is used in a course on Model Predictive Control. A mechanistic dynamic model of the system can be developed as in Gäfvert (2001) using Lagrangian mechanics. Due to the way the laboratory helicopter is designed, the helicopter model is decent when it comes to explaining responses to the pitch motor, but less good wrt. responses to the yaw motor.

Using sets of experimental data, a hybrid model consisting of the mechanistic model extended with a neural differential equation is fitted using Julia package DiffEqFlux.jl, and the model is validated. This allows for developing an improved model for MPC, and will also serve to illustrate the possibilities of such hybrid models.

References

Jeff Bezanson, Alan Edelman, Stefan Karpinski, and Viral B. Shah (2017). “Julia: A Fresh Approach to Numerical Computing”, SIAM REVIEW, Vol. 59, No. 1, pp. 65–98.

Christopher Rackauckas, and Qing Nie (2017). “DifferentialEquations.jl – A Performant and Feature-Rich Ecosystem for Solving Differential Equations in Julia”. Journal of Open Research Software, Vol. 5, no. 15, DOI: https://doi.org/10.5334/jors.151.

Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, and David Duvenaud (2018). “Neural Ordinary Differential Equations”. arXiv:1806.07366 [cs, stat], https://arxiv.org/abs/1806.07366.

Christopher Rackauckas, Mike Innes, Yingbo Ma, Jesse Bettencourt, Lyndon White, and Vaibhav Dixit (2019). DiffEqFlux.jl - A Julia Library for Neural Differential Equations. CoRR, abs/1902.02376, http://arxiv.org/abs/1902.02376.

Christopher Rackauckas, Yingbo Ma, Julius Martensen, Collin Warner, Kirill Zubov, Rohit Supekar, Dominic Skinner, and Ali Ramadhan (2020). “Universal Differential Equations for Scientific Machine Learning”. https://arxiv.org/abs/2001.04385. Herbert Goldstein (1980). Classical Mechanics, 2nd edition. Addison-Wesley series in physics, ISBN-13: 978-0201029185.

GM. Gäfvert (2001). “Modelling of the ETH Helicopter Laboratory Process”. Technical Reports TFRT-7596, Department of Automatic Control, Lund Institute of Technology (LTH).

Bernt Lie (2019). “Surrogate and Hybrid Models for Control”. Proceedings of the 60th Conference on Simulation and Modelling (SIMS 59), University of Västerås, Västerås, Sweden, August 13th–15th, 2019, pp. 1–8. Published by Linköping University Electronic Press, Series: Linköping Electronic Conference Proceedings No. 170, ISBN: 978-91-7929-897-5, eISSN: 1650-3740, doi: https://doi.org/10.3384/ecp201701.

ABSTRACT. Adaptive cruise control (ACC) for vehicles has been available for several years now. It was created on the top of cruise control that originates roughly in the 1950’s. The ACC is to enhance the automatic driving by keeping a safety distance set by a driver to another vehicle in front. The ACC is basically about balancing between the vehicle speed and the distance between two vehicles.

For control engineering education purposes, the ACC is a good and a clear example of control engineering. Basically, anyone can understand the scope of the ACC requiring little effort on motivating the need for the ACC. Second, the ACC engineering task can be divided into well-defined, separate tasks of dynamic modelling, control design and control tuning which all can be verified by applying dynamic, time-domain simulation.

The literature of the ACC recognizes several differential equation models to be worked upon. This paper introduces one of them, which despite its simplicity captures the essence of a vehicle behavior allowing the design of the ACC. Through simulations, also the control design and tuning are explained and treated with simulation results.

ABSTRACT. Drones is being slowly introduced, but are gaining more and more acceptance, in the mining industries. This study has been performing a series of tests showing that for example inspections can be supported by utilizing drones in the mining industry when planning maintenance activities and more than halve the time when performing the inspection. Stopping the production for performing maintenance is a costly event but necessary to avoid unnecessary stops and disasters. Which can be an even more expensive deal. Equipping drones with different types of sensors and in addition with thermal camera, data can be gathered and transmitted to an overall safety system. With image recognition pictures can be compared between the inspections and correct maintenance activities scheduled. The pictures from the inspections can be incorporate with machine learning and with a defined decision tree an artificial intelligence application is being taught so it can suggest maintenance activities well in advance. From the conducted tests it has been proven that only inspection with drones will not only save time and money for the mining business but also enhance the speed of performing the inspection. In addition, it has been verified that drone is a safer way to use during inspection activities comparing to sending an employee into an insecure and hazard area. For example, after blasting it is a demand, from the Swedish government, that one should not work in that area until the gas level is below a certain level. Such measurement is today being gathered manually by employees. Some employees, after a long period of being exposed to explosive gases, have developed a disease that in some cases has led to a fatal outcome. Sensors play a key role today and have been developed to be used in many applications that can be life critical as with e.g. fire alarms. Incorporating them on a drone will support todays manually task as the one above. Drones for gas measurement after blasting proved not only to be faster and safer but was also able to cover the whole blasting area when conducting the measurement comparing

to the once being manually performed. With automated drone routes used for different inspections can enhance the mining business as for increase their productivity. When mines now start investing in information systems and information technology infrastructure, they have taken one step closer to digitization and thus concede for changed working methods.

ABSTRACT. In hydro power systems, it is important to have knowledge about how changes in inflowand outflow influence the reservoir capacity, in order to optimise the power production. For some reservoirs it can be more challenging to have this knowledge due to their shape and/or implemented restrictions. Aurdalsfjord is one example. It is the intake reservoir to Bagnpower plant (operated by Skagerak Kraft AS). The paper will present a reservoir model that was developed in order to investigate the water flow and water level conditions in Aurdalsfjord. It takes the inflows and outflows as inputs and gives water levels and flows throughout the reservoir as output. The model isused to find time constants between in- and outflow locations for different inflow-values.The reservoir is discretised, where the dynamics are described with a set of mass and momentum balances as algebraic and differential equations. Modelica is used as modelling language which allows the developed models to be used in connection with models of a wide variety of other components that are present in an hydro power system(e.g., electrical, mechanical, control). Modelica is specifically aimed at implementing dynamic models with differential and algebraic equations and is available as open-source standard. The developed reservoir model can be used with an optimisation and automation infrastructure using the python application interface OMPython of OpenModelica. The University of South-Eastern Norway is developing an open-source hydropower libraryOpenHPL for teaching and research where the developed reservoir models will be integrated.

ABSTRACT. A high-head reaction-turbine hydro power system basically consists of an intake tunnel via a high-pressure steep penstock tunnel to the reaction turbines (eg., Francis turbine). A surge tank is usually placed between the intake pressure tunnel and the penstock. In case of a load rejection, the turbine valve is rapidly positioned for a required volumetric flow (discharge) of water through the turbine. During the rapid closing of the turbine valve, the water masses flowing in the intake tunnel and in the penstock are suddenly decelerated. A high-pressure region is created at the lower end of the penstock because of the obstructed water-inertia which causes pressure waves to travel in the upward direction. The magnitude of the traveling pressure wave after sudden closure of the turbine valve is termed as a water hammer. The energy of the pressure wave is released at the nearest low-pressure free water surface, i.e, at the surge tank placed between the intake tunnel and penstock (Mosonyi, 1991). In this regard, it is of interest to see the effect of the water hammer at different discharges through the turbine during the load rejection. The water inside the surge tank oscillates after the energy from the pressure wave is released at the free water surface inside the surge tank. The oscillation of water mass lasts until the pressure wave energy is fully dissipated. The design height and length of the surge tank should thus depend on the amplitude of the pressure wave, i.e., the water hammer. The amplitude of water mass oscillation inside the surge tank can be decreased using water flow-obstruction in the inlet of the surge tank, eg., in case of throttle valve surge and sharp orifice surge tank (Aronovich et al., 1970). Similarly, energy from the pressure wave can be dissipated using pressurized air inside a closed surge tank, usually referred to as an air-cushion surge tank (Vereide et al., 2014). This paper will mainly focus on the simulated response at different discharges for manifold pressure, velocity, mass flow rate and water mass oscillation inside the different kinds of surge tanks.

References (Aronovich et al., 1970) Aronovich, Grigorii Vladimirovich, Kartvelishvili, Nikolai Archilovich, & Lyubimtsev, Ya K. 1970. Water hammer and surge tanks. Tech. rept. Israel program for scientific translations. (Mosonyi, 1991) Mosonyi, Emil. 1991. Water power development. Vol. 2A. Publishing house of the Hungarian Academy of Sciences. (Vereide et al., 2014) Vereide, K, Lia, L, Nielsen, T, et al. 2014. Physical modelling of hydropower waterway with air cushion surge chamber. Page 134 of: 11th National Conference on Hydraulics in Civil Engineering & 5th International Symposium on Hydraulic Structures: Hydraulic Structures and Society-Engineering Challenges and Extremes. Engineers Australia.

ABSTRACT. A Modelica based open-source hydropower library OpenHPL 1 is under development at the University of South-Eastern Norway. OpenHPL was initiated in a Ph.D. study (Vytvytskyi, 2019). Currently, OpenHPL has units for the flow of water in filled pipes (inelastic and elastic walls, incompressible and compressible water) (Vytvytsky & Lie, 2017), a mechanistic model of a Francis turbine (including design of turbine parameters), friction models, etc (Vytvytskyi & Lie, 2018). The library also has draft models for a Pelton turbine, Francis turbine friction model, surge shaft, open channel flow, and a hydrology model. In addition, some accompanying work on analysis tools has been developed in scripting languages (Python, Julia) related to state estimation, structural analysis, etc (Vytvytskyi & Lie, 2019a). The library has been tested on real power plant data (Vytvytskyi & Lie, 2019b). The library is designed to interface to other Modelica libraries, e.g., libraries with generator models, electric grid, etc (Pandey & Lie, 2020). The models developed in OpenHPL are based on mass and momentum balance as in (Splavska et al., 2017). This paper focuses on the extension of OpenHPL with a mechanistic model of di?erent types of surge tanks and draft tubes, and their simulated response for a layout of a real hydropower plant. It is of interest to extend the simple surge tank with air-cushion type, throttle valve type, and sharp orifice type surge tank. In addition, this paper will also focus on mechanistic modeling of draft tubes, conical diffusers, and moody spreading pipes, and see their simulated response.

References (Pandey & Lie, 2020) Pandey, Madhusudhan, & Lie, Bernt. 2020 (Mar.). The Role of Hydropower Simulation in Smart Energy Systems, Submitted IEEE ESS 2020. (Splavska et al., 2017) Splavska, Valentyna, Vytvytskyi, Liubomyr, & Lie, Bernt. 2017. Hydropower systems: comparison of mechanistic and table look-up turbine models. Pages 368-373 of: Proceedings of the 58th Conference on Simulation and Modelling (SIMS 58) Reykjavik, Iceland, September 25th-27th, 2017. Linköping University Electronic Press. (Vytvytsky & Lie, 2017) Vytvytsky, Liubomyr, & Lie, Bernt. 2017. Comparison of elastic vs. inelastic penstock model using OpenModelica. Pages 20-28 of: Proceedings of the 58th Conference on Simulation and Modelling (SIMS 58) Reykjavik, Iceland, September 25th-27th, 2017. Linköping University Electronic Press. (Vytvytskyi, 2019) Vytvytskyi, Liubomyr. 2019. Dynamics and model analysis of hydropower systems. Ph.D. thesis, University of South-Eastern Norway. (Vytvytskyi & Lie, 2018) Vytvytskyi, Liubomyr, & Lie, Bernt. 2018. Mechanistic model for Francis turbines in OpenModelica. IFAC-PapersOnLine, 51(2), 103-108. (Vytvytskyi & Lie, 2019a) Vytvytskyi, Liubomyr, & Lie, Bernt. 2019a. Combining measurements with models for superior information in hydropower plants. Flow Measurement and Instrumentation, 69, 101582. (Vytvytskyi & Lie, 2019b) Vytvytskyi, Liubomyr, & Lie, Bernt. 2019b. OpenHPL for Modelling the Trollheim Hydropower Plant. Energies, 12(12), 2303.

ABSTRACT. Particle agglomeration is one of the obstacles for successful application and commercial breakthrough of fluidized bed biomass gasification. The problem is generally associated with molten ash components that interact with the bed particles, forming agglomerates that interfere with the flow behavior. In this work experimental and computational study are combined in order to gain more insight into the fluid dynamics in a bubbling fluidized bed gasifier. The goal is to develop a Computational Particle Fluid Dynamic (CPFD) model that can be used in further investigations of the correlation between flow behavior and bed agglomeration during biomass gasification in fluidized beds. The experimental part was performed in a 20 kW laboratory scale bubbling fluidized bed system. The commercial CPFD software Barracuda Virtual Reactor (VR) version 17.4.1 was used for the computational study. Simulation results were compared to the experimental data in order to validate the CPFD model. Pressure drops predicted by the simulations were in good agreement with the experimental measurements, which indicate that the model is well capable of studying the fluid dynamics in a fluidized bed system.

ABSTRACT. Fluidized bed reactors can be used for biomass gasification. The product from biomass gasification is syngas, which can be used for production of bio oil. The main challenge when using fluidized bed for gasification is ash melting and agglomeration of the bed material. Agglomeration of the bed material influences on the flow behavior in the fluidized bed reactor and thus affects the gasification efficiency. A Computational Particle Fluid Dynamic (CPFD) model is developed to predict the flow behavior in a fluidized bed gasifier. The CPFD model was validated against experimental data from a cold fluidized bed. The model was then tested against the results from a biomass gasifier, and a few modifications were needed. Glickman’s scaling parameters were used to scale up from a lab-scale to a full-scale gasifier. Simulations using the modified model were performed to study the flow behavior in a full-scale gasifier with agglomerates. It was found that the CPFD model is capable of predicting the effect of agglomerates on flow behavior in a fluidized bed gasifier.

ABSTRACT. An Entrained flow gasification reactor is studied in this work. Entrained flow reactors are best suited for a feed with small particles at large capacity, at high temperatures, pressures and with short residence time. This gives cleaner syngas as compared to fluidized and fixed bed reactors with very low amounts of tar. A kinetics-based simulation model was developed to study the entrained flow gasification process in Barracuda®. The software is based on the MultiPhase Particle-In-Cell (MP-PIC) approach. The reactor was modeled as an open cylinder with a conical shaped outlet at the bottom. The model was used to study fluid dynamics inside the reactor and the product gas composition. The product gas composition, gas temperature and gas flow rates were monitored during the simulation. The average molar composition of the produced gas on nitrogen-free dry basis is 0.467 of CO, 0.275 of H2, 0.226 of CO2 and 0.032 of CH4. The reactor temperature at different cross-sections shows that the temperature distribution becomes uniform with an increase in the reactor depth.

ABSTRACT. In subcooled flow boiling, heat transfer mechanism involves phase change between liquid phase to the vapour phase. During this phase change, a large amount of energy is transferred, and it is one of the most effective heat transfer methods. Subcooled boiling heat transfer is an attractive trend for industrial applications such as cooling electronic components, supercomputers, nuclear industry, etc. Due to its wide variety of applications for thermal management, there is an increasing demand for a faster and more accurate way of modelling.

In this work, a supervised deep neural network has been implemented to study the boiling heat transfer in subcooled flow boiling heat transfer. The proposed method considers the near local flow behaviour to predict wall temperature and void fraction of a sub-cooled mini-channel. The input of the network consists of pressure gradients, momentum convection, energy convection, turbulent viscosity, liquid and gas velocities, and surface information. The output of the model is based on the quantities of interest in a boiling system i.e. wall temperature and void fraction. The network is trained from the results obtained from numerical simulations, and the model is used to reproduce the quantities of interest for interpolation and extrapolation datasets. To create an agile and robust deep neural network model, state-of-the-art methods have been implemented in the network to avoid the overfitting issue of the model. The results obtained from the deep neural network model shows a good agreement with the numerical data, the model has a maximum relative error of 0.5 % while predicting the temperature field, and for void fraction, it has approximately 5 % relative error in interpolation data and a maximum 10 % relative error for the extrapolation datasets.