Introduction and motivation
A key objective for the operation of biomass boilers is to achieve the highest possible efficiency while emitting the lowest possible pollutant emissions. In order to automate this task, CO-lambda optimization methods have been proposed in literature that ensure that the biomass boiler is operated at the lowest excess air ratio at which no relevant pollutant emissions occur, maximizing efficiency as a result. Since this optimal excess air ratio depends on various external factors, such as fuel properties, CO-lambda optimization methods continuously incorporate new measurements of the excess air ratio and the carbon monoxide content of the flue gas and estimate a new optimal excess air ratio during operation.
While achieving promising results in lab-scale tests, none of the CO-lambda optimization methods presented in literature has yet been able to gain practical acceptance. Either they are not robust enough and provide inaccurate estimates of the optimal excess air ratio or they are too slow and do not allow the optimal excess air ratio to be tracked sufficiently quickly. With the goal of providing a method that is fit for practical application, this publication presents a new modular approach for CO-lambda optimization that determines the optimal excess air ratio robustly and quickly, i.e. in real time.

Method
The new approach for CO-lambda optimization approximates the correlation between the excess air ratio and the carbon monoxide content of the flue gas, the CO-lambda characteristic, with a continuous, algebraic, non-linear model function. For this purpose, it uses a recursive-least-squares algorithm to continuously identify the model function’s parameters that lead to the optimal fit with the measured data, which are the excess air ratio and carbon monoxide content of the flue gas. From these model parameters, the optimal excess air ratio is calculated and defined as a desired value for the biomass boiler’s existing controller. This existing controller then ensures, that the biomass boiler is operated with this desired optimal excess air ratio, thus, maximizing efficiency and decreasing pollutant emissions. As a result, this new approach for CO-lambda optimization is entirely modular and can be applied to any biomass boiler with an existing control strategy capable of accurately adjusting the excess air ratio. For the measurement of the carbon monoxide content of the flue gas, a separate sensor has to be used. For this study the commercially available and proven in-situ exhaust gas sensor “KS1D” provided by the company LAMTEC has been used.

Long-term verification
The new approach for CO-lambda optimization was tested and validated at a biomass boiler with a nominal capacity of 2.5 MW that supplies a local heating network and combusts wood chips with a water content ranging from 30 w.t.% to 50 w.t.%. The long-term validation took place over an entire heating period, i.e. 5 months from November to March, during which the biomass boiler was operated alternately with the new approach for CO-lambda optimization and the standard control strategy, which means a constant desired residual oxygen content. In total the new approach for CO-lambda optimization was active for 1155 operating hours while the standard control strategy was active for 1310 operating hours. Compared to the standard control strategy, the new approach for CO-lambda optimization increased the biomass boiler’s efficiency by 3.8%, decreased total dust emissions by 19.5% and reduced carbon monoxide emissions on average (median) by 200 mg/m³. This demonstrates that the new approach for CO-lambda optimization is not only robust enough to run over a long period of time, it also leads to significant improvements in the biomass boiler’s operation. In addition, following these results, this new approach for CO-lambda optimization has also successfully been implemented and demonstrated at another biomass boiler with a nominal capacity of 1 MW where it has already been active for several months. This contribution presents the new approach to CO-lambda optimization in detail and discusses its technological and economic impact.

Fault-Detection (FD) is essential to ensure the performance of solar thermal systems. However, manually analyzing the system can be time-consuming, error-prone, and requires extensive domain knowledge. On the other hand, existing FD algorithms are often too complicated to set up, limited to specific system layouts, or have only limited fault coverage. Hence, a new FD algorithm called Fault-Detective is presented in this paper, which is purely data-driven and can be applied to a wide range of system layouts with minimal configuration effort. It automatically identifies correlated sensors and models their behavior using Random-Forest-Regression. Faults are then detected by comparing predicted and measured values.

The algorithm is tested using data from three large-scale solar thermal systems to evaluate its applicability and performance. The results are compared to manual fault detection performed by a domain expert. The evaluation shows that Fault-Detective can successfully identify correlated sensors and model their behavior well, resulting in coefficient-of-determination scores between R²=0.91 and R²=1.00. In addition, all faults detected by the domain experts were correctly spotted by Fault-Detective. The algorithm even identified some faults that the experts missed. However, the use of Fault-Detective is limited by the low precision score of 30% when monitoring temperature sensors. The reason for this is a high number of false alarms raised due to anomalies (e.g., consecutive days with bad weather) instead of faults. Nevertheless, the algorithm shows promising results for monitoring the thermal power of the systems, with an average precision score of 91%.

A sustainable energy supply can only be achieved by a flexible, cross-sectoral energy system utilizing the specific advantages of the various renewable technologies. In this workshop possible roles of different technologies will be discussed based on a previous discussion of the users’ needs among the different sectors. In this a special focus should be given on the flexibility provision via the heating sector. By bringing together different users, representing municipal and industrial energy supply, and technological experts from different IEA Technology Collaboration Programmes (TCP) the workshop should support a holistic discussion.

List of presentations: 

• Wien Energie‘s vision of a sustainable energy and ressource supply of Vienna, Teresa Schubert, Wien Energie, Austria
• Digitalization of energy management systems – optimization of internal energy use as an industrial company, Maria Lechner, INNIO Jenbacher, Austria
• Flexible Bioenergy and System Integration, Elina Mäki, VTT Technical Research Centre of Finland, Finland Task Leader – IEA Bioenergy Task 44 Flexible Bioenergy and System Integration
• Use Case: Syngas platform Vienna for utilization of biogenic residues, Matthias Kuba, BEST – Bioenergy and Sustainable Technologies, Austria
• Transformation of District Heating and Cooling Systems towards high share of renewables, Ingo Leusbrock, AEE INTEC, Austria – Lead of Austrian delegation – IEA DHC Annex TS5 Integration of Renewable Energy Sources into existing District Heating and Cooling Systems
• Opportunities offered by long-term heat storages and large-scale solar thermal systems, Viktor Unterberger, BEST – Bioenergy and Sustainable Technologies, Austria Task Manager – IEA SHC Task 68 Efficient Solar District Heating Systems
• Possibilities through digitalization on the example of District Heating and Cooling, Dietrich Schmidt, Fraunhofer Institute for Energy Economics and Energy System Technology IEE, Germany – Operating Agent – IEA DHC Annex TS4 Digitalisation of District Heating and Cooling

List of contributing IEA Tasks:

• IEA Bioenergy Task 44 Flexible Bioenergy and System Integration
• EA DHC Annex TS5 Integration of Renewable Energy Sources into existing District Heating and Cooling Systems
• IEA SHC Task 68 Efficient Solar District Heating Systems
• IEA DHC Annex TS4 Digitalisation of District Heating and Cooling

One of the main tasks for operators of medium- and large-scale biomass boilers is the continuous operational monitoring of these plants in order to assess their performance, detect errors and identify possibilities for operational optimization. However, due to the high complexity of this task, errors are frequently detected too late or not at all, which can lead to even more costly secondary errors. In addition, possibilities for optimization remain unused in many existing plants, resulting in unnecessary pollutant emissions and low efficiencies.
To assist operators in performing this task and to achieve a high level of automation, methods for the automated, model-based monitoring of such plants have been focus of recent research activities. In this contribution, we will discuss the numerous possibilities provided by the application of such methods in a practical context. For this purpose, we present selected results from previous activities, demonstrating how methods for model-based monitoring were applied at combustion plants and used to enable automated error detection and support operational optimization.

Exemplary result 1: We developed a soft-sensor which accurately estimates the non-measurable internal state of heat exchangers and implemented it at a large-scale combustion plant with a nominal capacity of 38.2 MW. This soft-sensor uses a dynamic mathematical model of the heat exchanger in combination with measured data to determine a new estimate for the heat exchanger’s internal state every second. Based on this estimate, the soft-sensor accurately detects fouling and determines the non-measurable flue gas mass flow in real time. The estimated flue gas mass flow was used in a model-based control strategy which resulted in significant improvements of the combustion plant’s operational behaviour and load modulation capabilities. These results are discussed in this contribution.

Exemplary result 2: We developed a method for the real-time estimation of non-measurable fuel properties, i.e. chemical composition, bulk density, lower heating value, in biomass boilers. These estimates were subsequently used in a model-based control strategy and enabled the improvement of the biomass boiler’s fuel flexibility. Results of this estimator achieved for different biomass fuels, e.g. poplar wood chips, corncob grits and standard wood pellets, are discussed in this contribution.
On the basis of these selected results, it will be examined which possibilities arise from the use of methods for model-based monitoring in biomass boilers and also how these results can be extended to other technologies such as biomass gasifiers.

Biomassefeuerungen spielen eine Schlüsselrolle in der Energiewende hin zu einem vollständig erneuerbaren Energiesystem. Allerdings müssen sie sich zukünftigen Herausforderungen stellen, um weiterhin relevant zu bleiben. Einerseits müssen Biomassefeuerungen mit dem höchstmöglichen Wirkungsgrad arbeiten, um wirtschaftlich rentabel zu bleiben während sie gleichzeitig eine hohe Lastmodulationsfähigkeit aufweisen müssen, um für eine breitere Palette von Anwendungen eingesetzt werden zu können. Andererseits müssen Biomassefeuerungen immer strengere Grenzwerte für Schadstoffemissionen einhalten und gleichzeitig in der Lage sein, neue und alternative Biomassebrennstoffe mit geringerer Qualität zu verbrennen.

In dieser Arbeit wird eine modellbasierte Regelungsstrategie entwickelt, die es Biomassefeuerungen ermöglicht, all diese Herausforderungen zu meistern. Diese Regelungsstrategie besteht aus drei Teilen, einer Verbrennungsregelung, einem Zustands- und Parameterschätzer und einer Methode zur CO-lambda-Optimierung. Alle drei Teile werden in dieser Arbeit hergeleitet und im Detail diskutiert, insbesondere im Hinblick auf ihre Implementierung an realen Biomassefeuerungen. Darüber hinaus werden alle drei Teile der modellbasierten Regelungsstrategie durch Simulationsstudien sowie durch eine Implementierung in realen Biomassefeuerungen verifiziert.

Als Grundlage für die modellbasierte Regelungsstrategie wird ein mathematisches Modell abgeleitet, welches das dynamische Verhalten der Prozesse in der Biomassefeuerungen einschließlich des Einflusses der Brennstoffeigenschaften beschreibt. Die berücksichtigten Brennstoffeigenschaften sind die Schüttdichte und die chemische Zusammensetzung einschließlich des Wasser- und Aschegehalts sowie der untere Heizwert.

Die Verbrennungsregelung nutz die Stellglieder der Biomassefeuerung um dessen stabilen Betrieb zu gewährleisten und schnelle Laständerungen zu ermöglichen. Diese modellbasierte Regelstrategie berücksichtigt durch ihre Formulierung, die auf dem oben genannten mathematischen Modell basiert, explizit alle relevanten Brennstoffeigenschaften. Dadurch reagiert sie gezielt auf Änderungen dieser Brennstoffeigenschaften und kompensiert direkt deren Einfluss auf den Betrieb der Biomassefeuerung. Gleichzeitig weist sie eine einfache Struktur auf und ist daher leicht zu implementieren und zu warten. Diese modellbasierte Verbrennungsregelung wird sowohl in Simulationsstudien als auch durch Experimente nach einer Implementierung an einer realen Biomassefeuerung verifiziert.

Es wird ein kombinierter Zustands- und Parameterschätzer entwickelt, der gleichzeitig die Brennstoffeigenschaften, die anschließend von der Verbrennungsregelung verwendet werden, und die Zustandsgrößen der Biomassefeuerungen in Echtzeit schätzt. Er basiert auf einem erweiterten Kalman-Filter, der das in dieser Arbeit vorgestellte mathematische Modell verwendet. Diese Methode wird für verschiedene Brennstoffeigenschaften sowohl in Simulationsstudien als auch durch Messdaten aus realen Biomassefeuerungen verifiziert. Die Ergebnisse dieser Verifikation zeigen, dass diese Methode in der Lage ist, die Brennstoffeigenschaften und Zustandsgrößen auch bei Last- oder Brennstoffwechseln genau zu bestimmen.

Um einen Betrieb der Biomassefeuerung mit möglichst hohem Wirkungsgrad und möglichst geringen Schadstoffemissionen zu gewährleisten, wird eine Methode zur CO-lambda-Optimierung entwickelt. Diese Methode verwendet einen erweiterten Kalman-Filter in Kombination mit Messdaten des Sauerstoffgehalts und des CO-Gehalts des Rauchgases zur Bestimmung eines optimalen Luftüberschussverhältnisses für den aktuellen Zustand der Biomassefeuerung. Diese Methode wird an einer realen Biomassefeuerung in einer Langzeitvalidierung über mehrere Monate verifiziert und validiert. Während dieser Langzeitvalidierung führte die Anwendung dieser Methode zur CO-lambda-Optimierung zu einer Wirkungsgradsteigerung von 3,8 %, einer Reduktion der CO-Emissionen um durchschnittlich 200 mg/m³ sowie einer Verringerung der Gesamtstaubemissionen um durchschnittlich 19 %.

Zusammenfassend ermöglicht die in dieser Arbeit vorgestellte modellbasierte Regelungsstrategie es, Biomassefeuerungen mit den geringstmöglichen Schadstoffemissionen und dem höchstmöglichen Wirkungsgrad zu betreiben und dabei ein hohes Maß an Brennstoffflexibilität und Lastmodulationsfähigkeit zu erreichen. Darüber hinaus weist die Regelungsstrategie eine geringe Komplexität auf und ist leicht in realen Biomassefeuerungen zu implementieren und zu warten. Dies ermöglicht den breiten Einsatz dieser Regelungsstrategie an bestehenden und zukünftigen Biomassefeuerungen. Dies unterstützt die weitere Verbreitung von Biomassefeuerungen im Energiesystem, was zur Reduzierung der CO2e-Emissionen beiträgt und auch die verstärkte Nutzung anderer, volatiler erneuerbarer Technologien, wie z.B. solarthermischer Anlagen, ermöglicht.

Both the planning and operation of complex, multi-energy systems increasingly rely on optimization. This optimization requires the use of mathematical models of the system components. The model most often used to describe thermal storage, and especially in the common mixed-integer linear program (MILP) formulation, is a simple integrator model with a linear loss term. This simple model has multiple inherent drawbacks since it cannot be applied to represent the temperature distribution inside of the storage unit. In this article, we present a novel approach based on multiple layers of variable size but fixed temperature. The model is still linear, but can be used to describe the most relevant physical phenomena: heat losses, axial heat transport, and, at least qualitatively, axial heat conduction. As an additional benefit, this model makes it possible to clearly distinguish between heat available at different temperatures and thus suitable for different applications, e.g., space heating or domestic hot water. This comes at the cost of additional binary decision variables used to model the resulting hybrid linear dynamics, requiring the use of state-of-the-art MILP solvers to solve the resulting optimization problems. The advantages of the more detailed model are demonstrated by validating it against a standard model based on partial differential equations and by showing more realistic results for a simple energy optimization problem.

Modern buildings with floor heating or thermally activated building structures (TABS) offer a significant potential for shifting the thermal load and thus reduce peak demand for heating or cooling. This potential can be realized with the help of model predictive control (MPC) methods, provided that sufficiently descriptive mathematical models describing the thermal characteristics of the individual thermal zones exist. Creating these by hand or from more detailed simulation models is infeasible for large numbers of zones; instead, they must be identified automatically based on measurement data. We present an approach using only open source tools based on the programming language Julia that allows to robustly identify simple thermal models for heating and cooling usable in MPC optimization. The resulting models are used in a distributed optimization scheme that co-ordinates the individual zones and buildings of a city quarter in order to best support an energy hub.

Modern buildings with floor heating or thermally activated building structures (TABS) offer a significant
potential for shifting the thermal load and thus reduce peak demand for heating or cooling. This potential can be realized with the help of model predictive control (MPC) methods, provided that sufficiently descriptive mathematical models describing the thermal characteristics of the individual thermal zones exist. Creating these by hand or from more detailed simulation models is infeasible for large numbers of zones; instead, they must be identified automatically based on measurement data. We present an approach using only open source tools based on the programming language Julia that allows to robustly identify simple thermal models for heating and cooling usable in MPC optimization. The resulting models are used in a distributed optimization scheme that co-ordinates the individual zones and buildings of a city quarter in order to best support an energy hub.

Sector coupling is expected to play a key role in the decarbonization of the energy system by enabling the integration of decentralized renewable energy sources and unlocking hitherto unused synergies between generation, storage and consumption. Within this context, a transition towards hybrid energy networks (HENs), which couple power, heating/cooling and gas grids, is a necessary requirement to implement sector coupling on a large scale. However, this transition poses practical challenges, because the traditional domain-specific approaches struggle to cover all aspects of HENs. Methods and tools for conceptualization, system planning and design as well as system operation support exist for all involved domains, but their adaption or extension beyond the domain they were originally intended for is still a matter of research and development. Therefore, this work presents innovative tools for modeling and simulating HENs. A categorization of these tools is performed based on a clustering of their most relevant features. It is shown that this categorization has a strong correlation with the results of an independently carried out expert review of potential application areas. This good agreement is a strong indicator that the proposed classification categories can successfully capture and characterize the most important features of tools for HENs. Furthermore, it allows to provide a guideline for early adopters to understand which tools and methods best fit the requirements of their specific applications.

The task of fault detection is very important for plant operators to react to faults early and to keep the
system running at optimal performance. As this task is quite complex and time-consuming, automatic
approaches can greatly support the monitoring personnel during their work. However, not many
algorithms for fault detection at solar thermal systems are available yet. Existing approaches typically
need too much configuration time, cannot be applied to complex system layouts, or only cover a small
portion of potential faults. Hence, this work introduces FaultDetective, a fault detection algorithm with
minimal needs for configuration. For each desired sensor, FaultDetective automatically identifies the
most important input features for modelling its behaviour and trains an accurate machine learning
model. Faults can then be detected by comparing measurements with predicted values. The main
contributions of this work thus are (1) a detailed description of FaultDetective and (2) its evaluation
focusing on execution-time, prediction accuracy, and the quality of detected faults.

Heat is the largest energy end-use, accounting for 50% of global final energy consumption in 2019 and
contributing to 40% of global carbon dioxide (CO2) emissions. Regarding the heat supply of buildings,
district 15 heating plays an important role and is well-established in many countries. However, most of
the district heating networks worldwide are still operated with supply temperatures of 70-120°C
(medium-high temperature) typically produced by caloric power plants. Currently operated solar district
heating (SDH) systems are mostly installed with flat-plate collectors providing either heat at lower
temperatures or with lower efficiency in case of higher temperatures. In order to increase the efficiency
of SDH systems and to support their dissemination, a new task of the International Energy Agency
(IEA) from the technology cooperation program – solar heating and cooling (SHC) is in preparation.
This contribution presents the new task, its goals, structure and preliminary outputs.

Most industrial fixed-bed biomass gasification systems usually operate at steady-state to produce the maximum amount of energy possible although they can principally modulate their loads to compensate for the fluctuations of other volatile renewable energy systems. To unleash their full load modulation capability, their typically traditional control strategies should be improved, their gas residence times affected by typically basic char removal strategies adjusted and any required manual interaction of an operator avoided. In this respect, a new controller for the char handling (accumulation and removal) of the reduction zone in a fixed-bed biomass gasifier of a representative industrial small-scale gasification system is developed and experimentally verified. This new controller consists of a recursive least-squares estimator for the flow resistance of the gasifier representing the amount of char inside and a switching controller for rotating a grate located at its bottom. The experimental verification reveals that only the traditional (pressure-based) controller requires manual adjustment of the thresholds. Moreover, the new controller (flow resistance based) significantly reduces the fluctuation range during partial load and stabilizes the temperature and pressure downstream the gasifier. This provides the basis for enhancing its fuel flexibility too and is an important feature for flexible operation in future.

Absorption heat pumping systems (AHPSs, comprising absorption heat pumps and chillers) are devices that mainly use thermal energy instead of electricity to generate heating and cooling. This thermal energy can be provided by, e.g., waste heat or renewable energy sources such as solar energy, which allow AHPSs to contribute to ressource-efficient heating and cooling systems. Despite this benefit, AHPSs are still not a widespread technology. One reason for this is unsatisfactory controllability under varying operating conditions, which results in poor modulation and partial load capability. Emloying model-based control is a promising approach to address this issue, which will be the focus of this  contribution.
First, a viable control-oriented model for AHPSs is developed. It is based on physical correlations to facilitate systematic adaptions to different scales and operating conditions and considers only the most relevant mass and energy stores to keep the model order at a minimum. The resulting model is mathematically simple but still has the structure of a nonlinear differential-algebraic system of equations. This is typical for models of thermo-chemical
processes, but is unfortunately not suitable for many control design methods. Therefore, linearization at an operating point is discussed to derive a model in linear state space representation. Experimental validation results show that the linearized model does have slightly worse steady-state accuracy than the nonlinear model, but that the dynamic accuracy seems to be almost unaffected by the linearization and is considered sufficiently good to be used in control design.
As a next step, the linearized model is used to design model-based control strategies for AHPSs. A special focus is put on redundantly-actuated configurations, i.e. configurations with more manipulated variables than controlled variables, which allows using additional degrees of freedom to extend the operating range of AHPS and hence improve their partial load capability. Two model-based control approaches are discussed: First, a linear model predictive control (MPC) approach is presented - a well-established and generally easy-to-parameterize approach, which, however, often results in high computational effort prohibitive to its implementation on a conventional PLC. Therefore, a second control approach based on state feedback is presented which is mathematically simple enough for implementation on a conventional PLC. It consists of an observer for state variables and unknown disturbances, a state feedback controller and, in case of redundantly-actuated configurations, a dynamic control allocation algorithm. Both approaches are experimentally validated and compared to a state-of-the art control approach based on SISO PI control, showing that the model-based MIMO control approaches allow for a wider operating range and hence better modulation and partial load capability compared to the SISO PI approach. This, in turn, reduces ON/OFF operation of AHPSs and also facilitates their integration into complex energy systems to generate heating and cooling in a ressource-efficient manner.

In the context of SHIP (solar heat for industrial processes) installations, solar preheating appli-cations are becoming increasingly popular for reducing fuel consumption and consequently the carbon dioxide emissions. The poster shows an algorithm, which automatically adapts the
temperature setpoint. 

A process demonstration unit for a novel aqueous phase reforming (APR) process was built and scaled up by factor 666. The set-up of this demonstration unit was supported by virtual commissioning using a virtual test bed. By using virtual commissioning, it was possible to speed-up the commissioning and to support stable, reliable and continuous plant operation for 100h.

In various applications in the field of control engineering, the estimation of the state variables of dynamic systems in the presence of unknown inputs plays an important role. Existing methods require the so-called observer matching condition to be satisfied, rely on the boundedness e variables or exhibit an increased observer order of at least twice the plant order. In this article, a novel observer normal form for strongly observable linear time-invariant multivariable systems is proposed. In contrast to classical normal forms, the proposed approach also takes the unknown inputs into account. The proposed observer normal form allows for the straightforward construction of a higher-order sliding mode observer, which ensures global convergence of the estimation error within finite time even in the presence of unknown bounded inputs. Its application is not restricted to systems which satisfy the aforementioned limitations of already existing unknown input observers. The proposed approach can be exploited for the reconstruction of unknown inputs with bounded derivative and robust state-feedback control, which is shown by means of a tutorial example. Numerical simulations confirm the effectiveness of the presented work.

This article presents a new observer design approach for linear time invariant multivariable systems subject to unknown inputs. The design is based on a transformation to the so-called special coordinate basis (SCB). This form reveals important system properties like invertability or the finite and infinite zero structure. Depending on the system's strong observability properties, the SCB allows for a straightforward unknown input observer design utilizing linear or nonlinear observers design techniques. The chosen observer design technique does not only depend on the system properties, but also on the desired convergence behavior of the observer. Hence, the proposed design procedure can be seen as a unifying framework for unknown input observer design.