Dynamic modeling of municipal solid waste incineration

Journal Pre-proof Dynamic modeling of municipal solid waste incineration Elisa Magnanelli, Olaf Lehn Tranås, Per Carlsson, Jostein Mosby, Michael Becidan PII: S0360-5442(20)31534-6 DOI: https://doi.org/10.1016/j.energy.2020.118426 Reference: EGY 118426 To appear in: Energy Received Date: 29 January 2020 Revised Date: 8 July 2020 Accepted Date: 17 July 2020 Please cite this article as: Magnanelli E, Tranås OL, Carlsson P, Mosby J, Becidan M, Dynamic modeling of municipal solid waste incineration, Energy (2020), doi: https://doi.org/10.1016/ j.energy.2020.118426. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier Ltd. Elisa Magnanelli: Conceptualization, Investigation, Methodology, Writing - Original Draft, Writing Review & Editing Olaf Lehn Tranås: Writing - Original Draft, Writing - Review & Editing Per Carlsson: Conceptualization, Funding acquisition, Writing - Review & Editing Jostein Mosby: Data Curation, Resources Michael Becidan: Conceptualization, Funding acquisition, Project administration, Writing - Review & Editing Dynamic modeling of municipal solid waste incineration Elisa Magnanellia,∗, Olaf Lehn Tranåsa , Per Carlssona , Jostein Mosbyb , Michael Becidana a SINTEF Energy Research, Sem Sælands vei 11, 7465 Trondheim, Norway Setesdalsveien 205, 4618 Kristiansand, Norway b Returkraft, Abstract In this work, a comprehensive dynamic model of a moving grate Waste-to-Energy plant is developed using MATLAB Simulink. The objective is to develop a reliable and flexible model which can reproduce the dynamic behavior of combustion chamber and boiler. For this purpose, an extensive number of process data is used both in model development and for validation. Contrary to previous works in literature, fluctuations in both waste properties and operational set points are taken into account. The validated model is then used to study the dynamic response of the plant to changes in important process parameters. As expected, the dynamic response of the plant is faster for changes in primary and secondary air than for changes in grate speed and waste flow. The steam production response is from 1 to 4 minutes slower than the flue gas oxygen concentration response. Moreover, the response time depends to a large extent on the properties of the waste; as an example, an increase of humidity in the waste from 25% to 35% results in a 21 minute increase in the steam production response time. Such characterization of the dynamic response of the plant is fundamental to develop improved control strategies. Keywords: Municipal solid waste, Waste-to-Energy, Dynamic modeling, Process stability 1. Introduction Municipal solid waste (MSW) amounted to 2 billion metric tons in 2016 and is expected to increase to 3.4 billion tons by 2050 as a result of increased population, prosperity and urbanization [1]. Today, the average world citizen generates 0.74 kg of waste per day; in Norway, this number is 1.15 kg [2]. In the EU, waste management is heavily regulated, with the Waste Hierarchy at its centre, in order to ensure minimal environmental impact [3]. In the last years, the focus towards a circular economy has increased, to ensure a more sustainable society. In this context, a regulatory framework as well as a new citizen mindset are essential, to prevent the production of waste and to promote the direct re-use of products (for example, second-hand textile products). Furthermore, material recovery and recycling (e.g. for paper, plastic, metal, glass) is a key aspect of this effort. Finally, Waste-to-Energy (WtE) should be used to valorize the energy present in waste fractions that cannot be recycled. WtE is therefore ensuring a two-prone mission: (1) reducing the volume and weight of waste while destroying contaminants and (2) producing largely renewable heat and/or power. Indeed, the waste incinerated by WtE plants is about 50% biogenic in the EU, i.e. originating from biomass. This reduces the carbon footprint of WtE plants compared to conventional power plants [4]. Combined with Carbon Capture and Sequestration (CCS), WtE would therefore enable negative CO2 emissions that will mitigate climate change. In 2015, there were around 1200 MSWI plants worldwide, with an accumulated capacity of around 700,000 metric tons per ∗ Corresponding author Email address: [email protected] (Elisa Magnanelli) Preprint submitted to Energy day [5]. In Europe, 87% of today’s plants use a moving grate technology [6], which has proven suitable for large scale incineration of wet, bulk, mixed waste. The main advantage of this technology is that it does not require significant pre-treatment like shredding and size reduction [7]. However, maintenance costs might be higher than those of other incineration technologies such as rotary kiln and fluidized bed incinerators [8]. Although MSW incineration plants have improved significantly over the last years, challenges remain. Most notably, MSW is a heterogeneous fuel with characteristics that change continuously in an unpredictable way. Keeping a steady energy output and constant temperature in the combustion chamber is therefore challenging. Too high temperature in the furnace can cause problems like corrosion, increased downtime and reduced lifetime of key plant components. To avoid temperatures above critical levels, plants are often operated with large safety margins. If process fluctuation could be mitigated, one could reduce these safety margins. This would increase plant throughput and energy production, as well as it could reduce emissions. In order to achieve this, further knowledge on the plant dynamics is necessary. Such knowledge can be gather by dynamic modeling and simulation of the plant response to typical plant disturbances. Dynamic modeling and dynamic response of the most common thermal power plants have been extensively studied in literature [9]. However, dynamic modeling of waste-to-energy plants is scarce. Rovaglio et al. simulated the slow response during start-up of a rotary-kiln incineration plant, where constant fuel characteristics can be assumed [10]. Mahlia et al. developed a state-space dynamic model for investigation of control system strategies for a palm waste boiler [11]. However, the July 18, 2020 Nomenclature EVA Evaporator SH Superheater A Area [m2 ] C Heat capacity [J/K] cp Specific heat capacity [J/(kgK)] Subscripts F Mass flow [kg/s] aI Primary air H Heat transfer coefficient aII Secondary air h specific enthalpy [J/kg] amb Ambient conditions I Moment of inertia [kg/m2 ] att Attemperator k Fan constant [kg/kg] c Cold M Mass [kg] char Char in the waste bed N Number of control volumes cond Conduction P Production rate [kg/s] D Drag p Pressure [Pa] eva Evaporation Q Heat flow [J/s] fb Freeboard R Consumption rate [kg/s] fg Flue gas T Temperature [K] g Gas phase (flue gas, air) t Time [s] h Hot V Volume [m3 ] i i-th component in the waste in Inlet j j-th gas species in the gas phase Greek letters αg Flue gas absorptivity [-] liq Liquid water αvap Mass fraction of vapor in the drum [kg/kg] loss Thermal losses ǫg Flue gas emissivity [-] M Motor λ Thermal conductivity [W/K] m Metal ω Fan angular velocity [1/s] n n-th control volume ρ Mass density [kg/m3 ] out Output σ Stefan-Boltzmann constant [kg/(s3 K4 )] pyro Pyrolysis τ Torque [Nm] rad Radiation reac Reaction Abbreviations vap Water vapor ECO w Solid waste Economizer 2 model mainly focused on the description of the boiler. Alobaid et al. developed a comprehensive dynamical model for a fullscale moving grate MSWI plant; however, validation was only carried out against design data for steady-state nominal load operation, while the model performance for dynamic operation was not assessed [12]. A different approach to the problem is necessary. More comprehensive models and studies are needed to adequately simulate and understand the responses to change in fuel properties in a full-scale power plant. In the present work, we develop a comprehensive dynamic model of a full-scale moving grate MSWI plant, that describes both the dynamics of the combustion chamber and the dynamics of the boiler. Both dynamics play a fundamental role in the dynamic response of the plant. The model is developed using MATLAB Simulink [13]. The objective is to develop a reliable model that can describe the process dynamics at different time scales. Indeed, the process is subject to different kinds of disturbances, with different time scales. On one hand, fluctuations in fuel properties (e.g. moisture content, heating value and composition) cause fast and high frequency fluctuations of process outputs. On the other hand, changes in operational set-points cause slow, low frequency variations of process variables. Both disturbances need to be taken into consideration when studying the dynamic behavior of a WtE plant. The model is first validated by comparison with full-scale plant data. Contrary to previous studies in the literature, we take into account both fluctuations in fuel properties and changes in operational set-points. The validated model is then used to study how key process parameters influence important process outputs. With such procedure, the dynamic response of the plant can be characterized. Such knowledge on the plant dynamics is important in the development of strategies that can improve plant control. Process data from Returkraft’s plant in Kristiansand, Norway, are used for the development and validation of the model. Data from plant start-up had previously been used to calibrate the thermal inertia of the plant components [14]. After presenting the studied system in Section 2, model equations and assumptions are described in Section 3. In Section 4, the process data used for this work are presented, before results are illustrated and discussed in Section 5. Finally, conclusions and recommendations for future works are summarized in Section 6. 2.1. Process overview Waste is transported by trucks to the plant, where it is collected in a large bunker. Cranes are used for mixing the content of the bunker, as well as for conveying waste into a chute that leads to the combustion chamber. Once the waste reaches the bottom of the chute, it is pushed by a ram on a reciprocating moving grate. There, the thermal conversion process takes place. On the grate, the waste receives radiation heat from the freeboard. Moreover, the waste exchanges convective heat with the primary air that enters the combustion chamber from underneath the grate. Primary air is generally taken from the bunker. The air relative humidity in the bunker is high (90-95%) due to evaporation of water from the humid waste. To enhance combustion [15], primary air is preheated to a temperature of circa 110◦ C before it is introduced into the combustion chamber. The preheating step takes place through a heat exchanger that uses steam extracted from the boiler. The moving grate is divided into a number of zones (five, in the present case). To regulate the combustion process, grate velocity and primary air can be controlled independently in each zone. As the waste warms up on the grate, it undergoes several conversion processes. First, the moisture in the waste evaporates (drying stage). When the waste reaches a sufficiently high temperature, pyrolysis begins, and different pyrolysis gases are released (pyrolysis stage). Finally, the fixed carbon remaining in solid phase oxidizes while releasing energy (char burning stage). Inert materials and bottom ash leave the chamber as they reach the end of the moving grate, where they are collected for final disposal in landfill. With the addition of secondary air in the freeboard, the flue gas is further oxidized. Differently from the primary air, the secondary air is not preheated and enters the freeboard at ambient temperature. Secondary air is taken partly from the bunker and partly from the boiler house. The high temperature flue gas from the freeboard enters the boiler, where most of its thermal energy is recovered through a series of heat exchangers. The heat recovered from the flue gas is collected in a drum that contains saturated water and vapor. The heat exchangers making up the boiler can be categorized into evaporators (EVA), superheaters (SH), and economizers (ECO). First, the flue gas goes through the evaporators, where a saturated water stream coming from the drum evaporates as it absorbs heat from the flue gas. The saturated steam that leaves the evaporators returns to the drum. Water/steam circulation 2. System in evaporators is natural, i.e. driven by density gradients [16]. Figure 1 shows a schematic overview of the considered MSWI Then, the flue gas exchanges heat with a steam stream extracted from the boiler drum (superheater section). The superheated plant. The plant incinerates household and industrial solid waste steam is then sent to the steam turbine. Evaporators extend into (in proportion circa 60-40%) to produce heat and power. The the superheater section of the boiler, where evaporator pipes generated heat is mostly utilized by a district heating network, run into the external walls of the flue gas duct. Finally, in the while the remaining heat is used to cover the internal needs of economizers, the flue gas is used to pre-heat the water from the the plant. Similarly, the largest part of the generated electricity turbine and condenser before it returns to the boiler drum. is sold to the power grid, while a small percentage is used for Two attemperators are present in the superheater section to internal purposes. control the final temperature of the steam. Indeed, excessively 3 Figure 1: A schematic illustration of Returkraft WtE plant. Steam streams are light blue, while dark blue indicates water streams. high steam temperature should be avoided, to protect the turbine. In the specific case, the steam temperature is maintained below 425◦ C. The steam that leaves the superheater is sent to a steam turbine, where it expand to produce electricity. The low pressure steam that leaves the turbine enters a condenser where the cold water from the district heating network is heated up. The condensed stream from the turbine and condenser is then sent to the economizer. After heat recovery, the flue gas undergoes a series of cleaning treatments. These treatments are necessary to bring pollutant concentrations below national and international regulation limits. inert follows the flue gas as fly ash, while the largest fraction is collected at the end of the moving grate; – Pyrolysis products are a mixture of gases and char; – The gas species considered are limited to H2 O, N2 , CO2 , O2 , CH4 , H2 , CO, SO2 , HCl, HF, NO, NH3 and HCN and the gas phase is model as a mixture of ideal gases; – The thermodynamic properties of water ans steam are modeled according to the standard IAPWS IF-97 [17]. – Due to the high relative humidity of the incoming air, the primary and secondary air is considered to be humid air. Temperature and relative humidity determine the absolute water content of air. The dry fraction of air is considered to be 79% nitrogen, N2 , and 21% oxygen, O2 , in volume fractions. Other gases normally present in air constitute less than 1%, and are neglected [18]; 2.2. Assumptions A MSWI plant is a complex system. To reduce the complexity to a level compatible with dynamic calculations, we make the following assumptions: – Primary air distribution within the same grate zone is considered uniform; – The waste entering the combustion chamber is considered to be made up of three main fractions; a combustible fraction, an inert fraction, and moisture. The combustible fraction comprises all the waste components that can be burned. To reduce the computation level, the elements contained in the combustible fraction of waste are assumed to be C, H, S, Cl, N, O, and F. The inert fraction transits through the combustion chamber without undergoing any chemical transformation. A small fraction of – Process components are modeled as block diagrams, where block dynamics is described through a set of differential and/or algebraic equations; – The combustion chamber is modeled as two subsystems; the waste bed, and the freeboard. The waste bed is modeled as a one-dimensional system, where state variables 4 evolve along the direction of motion of the grate. The freeboard is modeled as a block diagram; n where Ctot and T wn are the total heat capacity and temperP n n n ature of the control volume, Ftot,in = i Fi,in + Fchar,in is the n total mass flow entering the control volume, FaI and T aI are the mass flow and temperature of the primary air, c p,w and c p,g are the specific heat capacities of waste and air, Qnreac is the total heat released by the waste conversion processes, Qnrad is the heat exchanged by radiation with the freeboard, and Qncond is the conductive heat. The total heat capacity of a control volume is due to its waste and char content, but also to the mass of the grate that is warmed up during the process:   X  n n n  n  Ctot =  c p,i Mi + c p,char Mchar  + c p,m Mgrate (4) – The grate is modeled as moving in a continuous way (traveling grate), even though the considered grate is a reciprocating grate [19]; – Convective heat between the waste bed and the gas phase in the freeboard has been neglected. Indeed, due to the high temperature in the freeboard and the high volume of primary air through the waste bed, its contribution is small in comparison to the other heat transfer mechanisms [20, 21]; – Heat exchangers are modeled as one-dimensional systems, where state variables evolve along the direction of flue gas flow; i where is the grate, and c p,m is its specific heat capacity. The primary air is preheated before beeing introduced in the combustion chamber. The heat necessary for primary air preheating, QaI , can be calculated as:
(5) QaI = FaI c p,g T aI − T amb n Mgrate – The description of the cleaning system, turbine and condenser is not included in the present work. 3. Model where T amb is the ambient temperature. The heat necessary for primary air preheating is extracted from the boiler drum. The heat released or absorbed during waste conversion has several contributions: In this section, the model developed to describe the different components of the plant is presented. 3.1. Combustion chamber Qnreac = Qnpyro + Qneva + Qnchar The combustion chamber is modeled as two subsystems; the waste bed, and the freeboard (Fig. 2). where Qnpyro is the heat released by pyrolysis, Qneva is the heat absorbed by moisture evaporation, and Qnchar is the heat released by char oxidation. The heat exchanged through radiation with the freeboard can be expressed as [21]: 
 Qnrad = An σ ǫg T g4 − αg T wn 4 (7) 3.1.1. Waste bed The waste bed model is considered to be one dimensional and is discretized in the direction of the moving grate as a series of control volumes. State variables within each control volume are homogeneous. The mass balance of the i-th component of the waste in the control volume n can be written as: dMin n n = Fi,in − Fi,out − Rni dt where An is the radiative heat exchange surface, σ is the StefanBoltzmann constant, T g is the temperature of the flue gas in the freeboard, and αg and ǫg are constants related to the emissivity and absorptivity of the flue gas. The radiative term has been derived following the model proposed by Hottel and Sarofim [22] for radiation in a perfectly stirred combustion chamber. Such model assumes both gas and sink surface to be grey and walls to be speckled-walls. The heat exchanged by conduction with neighboring control volumes is:     n n−1 n−1 + λc T wn+1 − T wn (8) Qncond = Qcond − Qn+1 cond = −λc T w − T w (1) where the subscript i ∈ [moisture, C, H, S , Cl, N, O, F, inert] indicates the different components in the waste. Moreover, Min , n n Fi,in , and Fi,out are the mass, the inlet mass flow and the outlet mass flow of the i-th waste component. Finally, Rni is the consumption rate of the component i. Char is produced as pyrolysis takes place. Therefore, a char mass balance needs also to be established: n dMchar n n = Fchar,in − Fchar,out + Pnchar − Rnchar dt n+1 where Qn−1 cond and Qcond are the conductive heat exchanged with the control volume to the left and to the right of the considered control volume (see Fig. 2), and λc is the overall conductive heat transport coefficient. The gases released during waste conversion and the unreacted primary air leave the waste bed and enter the freeboard. The total mass flow of the j-th gas component entering the freeboard, F j,in , can be calculated as: (2) where the subscript char indicates the char in the waste bed. Char is both produced as a product of pyrolysis, Pnchar , and consumed by char combustion, Rnchar . The energy balance for the waste bed is modeled as: n Ctot dT wn dt =  
n Ftot,in c p,w T wn−1 − T wn + FanI c p,g T aI − T wn +Qnreac + Qnrad + Qncond (6) F j,in = (3) N  X n 5 F nj,aI + Pnj  (9) Figure 2: Scheme of the combustion chamber subdivision in the model. where Pnj is production rate of the gas species j during waste conversion, and N is the number of control volumes. We use the subscript j to indicate the different gas species. gas (hot side) to the metal of the heat exchanger, and then from the metal to the water/vapor (cold side). The heat exchanged between the hot flue gas and the heat exchanger wall in the p-th control volume, Qhp , can be calculated as [23]:  p  p − T wall (13) Qhp = Hhp Fg T h,in 3.1.2. Freeboard The freeboard is modelled as a zero-dymensional system. The mass balance for the gas species j is: dM j = F j,in − F j,out + F j,aII + R j dt where Hhp is the global heat transfer coefficient, Fg is the flue p p gas mass flow in the heat exchanger, T h,in and T wall are the temperature of the flue gas entering the heat exchanger element and of the heat exchanger wall element, respectively. The subscript h is used to indicate variables on the hot side of the heat exchanger. Similarly, the heat exchanged between the wall of the heat exchanger element and the vapor/water, Qc , can be written as:  p  p Qcp = Hcp Fc T wall − T c,in (14) (10) where M j is the mass of the j-th gas species in the freeboard, F j,out is the component mass flow leaving the freeboard, F j,aII is the mass flow of species j in the secondary air, and R j is the species production/consumption rate due to reactions in the freeboard. The energy balance is: C f b,tot dT g dt = where the subscript c is used to refer to variables on the cold side of the heat exchanger. In heat exchangers, the wall separating the two fluid might constitute a relatively large thermal inertia. Therefore, a dynamic energy balance was established for the walls of the heat exchanger:     Fg,in c p,g T w − T g + FaII c p,g T aII − T g +Qreac,g − Qrad − Qloss, f b (11) where C f b,tot and T g are the total heat capacity and temperature P in the freeboard, Fg,in = j F j,in is the total flow from the waste bed, FaII and T aII are the secondary air mass flow and temperature, Qreac,g is the heat released by the reactions in the gas phase, and Qloss, f b is the heat lost from the walls of the chamber to the environment. The total heat capacity of the freeboard is: X C f b,tot = c p,g M j + c p,m M f b (12) p Mwall c p,wall p dT wall = Qhp − Qcp dt (15) To reduce the level of complexity of the model, no accumulation term is considered on the flue gas side and water/vapor side. By establishing a energy balance, the enthalpy of the fluid leaving the p-th heat exchanger element can be calculated as: j where M f b is the total mass of the metal making up the freeboard. p hout = hinp − 3.2. Evaporator, superheater, and economizer Heat exchangers are assumed to be one-dimensional systems and discretized in the direction of the flue gas flow as a series of control volumes. Heat is first transferred from the flue Qp F (16) where hin is the enthalpy of the fluid entering the control volume. 6 Table 1: Main plant parameters and nominal data. 3.3. Boiler drum The mass balance of water in the drum and evaporator can be written as [16]: Parameter Grate length Grate width Height to the top of 1 st pass Water in drum and EVA Total heat exchanger surface in boiler Waste throughput capacity Thermal power capacity Turbine nominal power dMH2 O,drum = F ECO − FS H (17) dt where MH2 O,drum is the total mass of water and steam in the drum and evaporator pipes, F ECO is the water flow that enters the drum from the economizer, and FS H is the saturated steam flow leaving the drum to the superheater. The energy balance for the water and steam in the drum and evaporator is [16]: dHdrum = QEV A + F ECO hECO − FS H hvap − QaI + Qloss,drum (18) dt where QEV A is the heat exchanged between the water/vapor mixture and the walls of the evaporator as defined by Eq. 14, hECO is the specific enthalpy of the water from the economizer, hvap is the specific enthalpy of the saturated steam in the drum, and Qloss,drum = Hdrum (T drum − T amb ) is the heat lost to the environment. As the system warms up, energy is stored both in the boiler water and steam, but also in its materials [16]. Thus, the enthalpy of the system can be written as:   Hdrum = αvap MH2 O,drum hvap + 1 − αvap MH2 O,drum hliq (19) +Mdrum c p,m (T drum − T amb ) − pdrum Vdrum where ω is the fan angular velocity, I is the moment of inertia, and τ M and τD are the motor and drag torque. The relation between volumetric flow rate and fan angular speed was found by linearizing the specific fan curve around the working point.The mass flow, F f g , can be then calculated from the fan angular velocity as: F f g = ρ f g kω 4. Plant data Full scale plant data are provided by Returkraft WtE plant in Kristiansand. The data can be categorized in plant physical data and process data. Physical data, such as components’ dimensions and volumes, were derived from plant drawings and data sheets. Some physical data were previously derived through start-up data analysis in a former work [14], where the dynamic model for the boiler was calibrated. Physical data are plant constants and do not change during operation. Process data are variable over time, and were gathered every minute, during several hours of plant operation. Table 1 reports some of the main plant parameters. A WtE plant is mainly subject to two kinds of disturbances. The fist type of disturbances has a high frequency and it is due to the high heterogeneity of the waste. The second type of disturbances has a low frequency and it is due to changes in operational set-points. The data collected during plant operation provided information not only on the disturbances the plant was subject to, but also on the overall plant response to such disturbances. As an example, Fig. 3 shows how the heating value of the incoming waste changes over one day (solid line). Since MSW is a highly heterogeneous mixture, its heating value can vary up to 11% of its average value (dashed line). As a consequence of the heating value fluctuations, the control system of the plant manipulates several plant parameters to 3.4. Attemperators In the attemperators, injectors are used to spray a flow of liquid water into the superheated steam. This is done to maintain the steam temperature below some limit. The mass balance across the attemperators is calculated as: (21) where Fatt,in is the steam flow entering the attemperator, Fatt,liq is the liquid water flow sprayed into the steam stream, and Fatt,out is the resulting mass flow. The enthalpy of the resulting flow, hatt,out , can be calculated as: Fatti n hatt,in + Fatt,liq hatt,liq Fatt,out (24) where k is a fan constant extrapolated from the fan data sheet. where the nominator represents the net heat recovered in the boiler, and the denominator represents the overall heat released by waste conversion. hatt,out = Unit m m m t m2 t/h MW MW 3.5. Flue gas fan The flue gas fan is controlled to maintain under pressure in the combustion chamber. This is done to avoid leaks of flue gas out of the plant. The dynamic behavior of the fan can be describe as: dω I = τ M − τD (23) dt where αvap is the mass fraction of water vapor in the drum and evaporator, hliq is the specific enthalpy of the saturated water, Mdrum is the total mass of drum and evaporator pipes, T drum and pdrum are the drum temperature and pressure, and Vdrum is the total inner volume of the drum and evaporator. The heat recovery efficiency of the boiler, ǫ, can be written as: FS H hvap − F ECO hECO (20) ǫ= Qreac + Qreac,g Fatt,out = Fatt,in + Fatt,liq Value 10.2 6.3 19.15 82 7660 18 55 14 (22) where hatt,in and hatt,liq are the specific enthalpies of the steam entering the attemperator and of the water sprayed in the steam stream, respectively. 7 Figure 3: Evolution of the heating value of the incoming waste during one day (solid line) and its average value in the selected time period (dashed line) on 17.05.18. The heating value is not directly measured, but calculated by the plant control system from process data and energy balance on the plant. Figure 5: Set-point for the steam mass flow to the superheater, from 01.01.18 to 14.01.18. By comparing Fig. 3 and Fig. 5, we can see how the fluctuations in waste heating value have much higher frequency (order of one over minutes) than those in steam production (order of one over days). 5. Results 5.1. Model validation To validate the model, the numerical results obtained with the dynamic model were compared against process data. During plant operation, many of the process parameters are manipulated by the plant control system. In order to validate the physical model of the plant independently from the control structure associated to it, these process parameters were assumed as inputs to the model. All model inputs were derived from dynamic process data, from the Returkraft WtE plant. They are mainly mass flows input to the process (e.g. mass flows of primary and secondary air, input waste, feed water, steam from the boiler, water to the attemperators) and their respective properties (e.g. temperature and pressure, and heating value of the waste). Figure 6 shows a comparison of process data (solid lines) and calculation results (dashed lines) for the flue gas temperature and the temperature in the boiler (black lines). Measurements of flue gas temperature are taken after the evaporator section (light grey lines) and after the superheater section (dark grey lines). Model predictions agree well with process data, and the relative error between experimental and simulated data lies within 5 % of the output absolute value. The error between experimental and simulation data is larger for flue gas temperature than for the drum temperature (experimental and simulation data for the drum temperature appear to coincide at the figure’s scale). In particular, the error is largest for the gas temperature after the evaporator (dark grey lines). This suggests that the description of the dynamic heat transfer of this component could be improved to obtain a better agreement with process data. However, it should also be mentioned Figure 4: Incoming waste flow over time (solid line) and its average value in the selected time period (dashed line) on 17.05.18. maintain important process outputs as close as possible to their set-point values. These parameters are typically primary and secondary air, input waste flow and grate velocity. As an example, Fig. 4 shows how the inlet waste flow is manipulated over time by the control system (solid line) to maintain a constant steam production in the boiler. A comparison between Fig. 3 and Fig. 4 shows that the opposite trends of waste flow and waste heating value. Indeed, waste flow is manipulated to maintain the overall heat production of the plant as constant as possible. Over longer periods of time, set-points of important process outputs are changed to optimize the performance of the process, or to satisfy some type of constraints (e.g. specific requirements on the heat delivered to the district heating network). As an example, Fig. 5 shows how the set-point for the produced steam changes over two weeks. 8 Figure 6: Temperature measurements (thick lines) and model response (thin lines) of the flue gas after the evaporator section (light gray) and after the superheater section (dark grey) and in the boiler drum (black), from 01.01.18 to 04.01.18. that it is challenging to carry out accurate measurements in the gas phase under such high temperatures, in presence of dust particles. Indeed, this specific process parameter is extracted as the average of the measurements registered by three different temperature sensors. At times, the temperature measurements reported by the three sensors differ up to 40◦ C. Thus, the deviation of the model response from process data might be partially due to inaccurate sensor measurements. On the other hand, measurements in the boiler drum are less subject to uncertainties, which can explain the better agreement between experimental and simulation data. Thus, we can conclude that the model can reproduce well the dynamic response of the plant to both high frequency (fluctuation in waste properties) and low frequency (set-point changes) disturbances. 5.2. Process dynamics A fundamental step toward the identification of operation strategies that can improve plant stability and performances is to gain a better understanding of the dynamic response of the process. In the considered WtE plant, oxygen concentration in the flue gas and steam production are two important process outputs that are monitored and controlled to ensure stable operation of the plant. The control system of the plant makes sure they deviate as little as possible from their assigned set-points. This is primarily done by manipulating waste inlet flow, grate speed, and primary and secondary air (also called manipulated variables). It is, therefore, interesting to gain insight on how these parameters influence the process outputs. Figure 7 shows the dynamic response of the plant to 10% step-wise increase in these parameters. In these calculations, all the other process parameters were kept equal to their nominal value and the composition of the incoming waste was assumed constant. In this case the steam flow from the boiler was not assigned as an input, but it 9 was determined as an output of the model. To do this, a control structure similar to the one present in the plant was added to the model. The control structure controlled the steam flow from the drum to keep the pressure in the drum as close as possible to an assigned set-point. The responses are reported as relative variation with respect to the working point (i.e. steam production of 14.5 kg/s and 6% oxygen content in the flue gas). . Figure 7a shows the plant response in steam production (solid lines) and oxygen content in the flue gas (dashed lines) to variations in primary air. As the primary air increased, the combustion rate of the waste on the grate increased rapidly, causing the steam production to increase. However, since the amount of waste entering the system stayed the same, the steam production started decreasing after few minutes. The final steady-state steam production was slightly lower than initial value. This is due to the larger amount of primary air that causes a larger velocity of flue gas, that reduces its residence time in the boiler. An increase in primary air had a positive net effect on the oxygen concentration in the flue gas. Indeed, increasing the primary air while maintaining the waste input constant, increases the air-to-fuel ratio and, thus, the excess oxygen. The effect of secondary air is similar but smaller in magnitude (Fig. 7b). Figure 7c shows the effect of increasing the grate speed. While the grate speed in different grate zones is different, the ratio between the different velocities is usually kept constant in the considered plant. Since there was no increase in any of the mass flows entering the plant, grate speed has no permanent effect on either steam production and flue gas oxygen concentration. Increasing the grate speed had the effect of temporary increasing combustion on the grate, provoking both a peak in steam production and a reduction of the flue gas oxygen concentration. It should be noticed that the grate was modeled as a traveling grate, while the considered WtE plant makes use of a reciprocating grate. Thus, the actual plant response to variations in grate speed might somehow differ from that of the model, even though the effect of changing the grate speed should show the same trend. Figure 7d shows the effect of increasing the input waste flow. In this case, there was a permanent positive effect in steam production, as more fuel was introduced in the process. However, since the amount of combustion air did not change, the airto-fuel ratio decreased, causing a permanent decrease of oxygen in the flue gas. The plant response to an increase of waste flow is much slower than in the other considered cases. Indeed, 66% of the maximum value in steam production response (i.e. characteristic response time) is reached after 23 minutes. The step response calculations showed that the oxygen response is always faster than the corresponding steam production response. Indeed, 66% of the maximum value in the oxygen response was reached between 1 and 4 minutes earlier than in the steam production response. The difference between the responses is mainly due to the thermal inertia of the boiler. The responses to variations in primary and secondary air were faster than those to variations in grate speed and waste flow. This is due to the fact that the gas phase dynamics is much faster than the dynamics of the solid phase. This is in line with the logic of the control system used by the plant, where, (a) 10% increase in primary air. (c) 10% increase in grate speed. (b) 10% (d) 10% increase in secondary air. increase in inlet waste flow. Figure 7: Dynamic response to 10% step-wise increase in primary air, secondary air, grate speed and waste flow, on steam production (solid lines) and oxygen concentration in the flue gas (dashed lines). The responses are reported as relative variation with respect to the working points of steam production (14.5 kg/s) and oxygen content in the flue gas (6% vol.). as an example, fluctuations in steam production due to high frequency disturbances (i.e. variation in waste properties) are controlled by manipulating primary air, while manipulation of waste flow is mostly used to counteract low frequency changes. A change in the considered process parameters had also an effect on the heat recovery efficiency of the boiler (see Eq. 20. Figure 8 shows the dynamic response of the boiler efficiency for 10% step-wise increase of the considered parameters. An increase in waste flow caused a higher heat recovery efficiency, while an increase in primary or secondary air had the opposite effect. Indeed, the main heat loss in the process is due to the energy that leaves the process with the exhaust flue gas. This energy depends on the temperature of the flue gas after the economizer, which is almost constant, and on the flue gas mass flow. When primary and secondary air are increased, the flue gas mass flow increases causing a larger energy loss. On the other hand, when the waste flow is increased, the energy loss through the exhaust flue gas stays approximately the same, while the energy released by the waste increases. In conclusion, the heat recovery in the boiler is affected by the fuel-to-air ratio. 5.2.1. Comparison with biomass grate firing To the best knowledge of the authors, step change experiments on real MSW moving grate plants cannot be found in literature. However, experimental results on this kind of experiments can be found for biomass grate firing. A biomass moving grate incinerator is very similar to a MSW one [20]. However, biomass is typically a easier fuel to handle than waste. Biomass (woody) is relatively homogeneous, has higher energy content (except if very wet), very low ash content and can have a high volatility/reactivity. In general, this causes less disturbances due to fluctuations in waste properties and, therefore, a easier control. Nonetheless, the dynamic response of MSW and biomass incinerators should be qualitatively the same. A comparison of the dynamic response of the WtE plant obtained in this work with experiments on a biomass grate incinerator [19], shows that the dynamic response is qualitatively the same. The velocity and intensity of the response (quantitative behavior) is different, as it depends on the plant size, configuration and fuel [19]. The dynamic behavior of the plant can be verified from a quantitatively point of view only through step change experiments on the considered plant. 10 Figure 8: Dynamic response of the heat recovery efficiency to 10% step-wise increase in primary air (solid line), secondary air (dashed line), grate speed (dash-dot line) and waste flow (dotted line). Figure 9: Dynamic response to 10% step-wise increase in waste flow, on steam production for input waste with 25% humidity (black line), 30% humidity (dark gray line) and 35% humidity (light gray line). The responses are reported as relative variation with respect to the working point (14.5 kg/s). The horizontal line indicates 66% of maximum steam production response. 5.2.2. Effect of waste properties The properties of the waste play a fundamental role in the dynamic response of the plant. As an example, Fig. 9 shows the steam production response for a 10% increase in waste flow for input waste with 25% humidity (black line), 30% humidity (dark grey line), and 35% humidity (light grey line). The composition and heating value of the dry waste was kept the same. The Initial waste flow was determined to have an initial steam production of 14.5 kg/s. Since the initial waste flow was adjusted to give the same nominal steam production, the increase in the final value of steam production was the same for the different cases. However, the moisture content of the waste determined the speed of the response of the system. Indeed, the vertical lines shows that 66% of the maximum response is reached circa 21 minutes earlier for the waste with 25% moisture than for the waste with 35% moisture. Figure 10 shows that an higher ash content also determines a slower response of the system. However, increasing ash content slows down the plant response to a smaller extent than moisture content. Higher ash content has the effect of increasing the thermal capacity of the waste. On the other hand, an higher moisture content causes a longer drying time of the waste. During real plant operation, the composition of the incoming waste is not known. Knowledge on the moisture, ash content and other waste parameters would give further information on the actual response time of the process. To retrieve this information, for example with the use of sensors, would be useful in the development of improved operation strategies of the plant. The results computed with the model showed good agreement with process data, showing that the model can reproduce well the dynamic response of the plant to fluctuations both in waste properties and in operational set-point changes. The validated model was used to explore the dynamic response of the plant to variation of important process parameters. The dynamic response of the plant is faster for changes in primary and secondary air than for changes in grate speed and waste flow. The steam production response is up to 13 minutes slower than the response in flue gas oxygen concentration. The obtained results agreed with results on biomass grate incinerators from the literature. It was also found that the response time depends to a large extent on the properties of the waste. As an example, we found that the humidity content of the waste had a great influence on the plant dynamic response. Further information on the properties of the incoming waste is therefore needed to allow for a more accurate description of the solid phase dynamics. Future works will first focus on the further validation of the dynamic response of the model by comparing it with open loop step responses of the plant. While comparison with biomass incinerator experiments showed that the qualitative dynamic response of the model is correct, quantitative aspects need to be verified through experiments on the considered plant. A fundamental challenge when conducting such experiments is the impossibility of maintaining the properties of the input waste constant. The final goal is to exploit the validated model to identify and evaluate improved operation strategies. In particular, the aim is to propose improved strategies that allow for stable and reliable operation while satisfying regulations and plant limitations. 6. Conclusions and future works In this work, a newly developed dynamic model of a WtE plant was presented. The model was developed and validated using physical and process data from a grate-fired WtE plant located in Kristiansand, Norway. 11 [12] F. Alobaid, W. A. K. Al-Maliki, T. Lanz, M. Haaf, A. Brachthäuser, B. Epple, I. Zorbach, Dynamic simulation of a municipal solid waste incinerator, Energy 149 (2018) 230–249. [13] MATLAB, version 7.10.0 (R2019b), The MathWorks Inc., Natick, Massachusetts, 2019. [14] E. Magnanelli, C. Birgen, J. Mosby, P. Carlsson, M. Becidan, Characterizing the dynamic behaviour of a WtE plant, in: Proceedings of the 30th European Symposium on Computer Aided Process Engineering, Elsevier, 2020. [15] L. Van Kessel, A. Arendsen, P. de Boer-Meulman, G. Brem, The effect of air preheating on the combustion of solid fuels on a grate, Fuel 83 (9) (2004) 1123–1131. [16] K. J. Åström, R. D. Bell, Drum-boiler dynamics, Automatica 36 (3) (2000) 363–378. [17] M. Holmgren, Xsteam: Water and steam properties according to IAPWS IF-97, www. x-eng. com (2007). [18] P. Brimblecombe, Air composition and chemistry, Cambridge University Press, New York City, New York, 1996. [19] S. Van Loo, J. Koppejan, The handbook of biomass combustion and cofiring, Earthscan, 2012. [20] C. Yin, L. A. Rosendahl, S. K. Kær, Grate-firing of biomass for heat and power production, Progress in Energy and combustion Science 34 (6) (2008) 725–754. [21] L. B. M. Van Kessel, Stochastic disturbances and dynamics of thermal processes: with application to municipal solid waste combustion, Ph.D. thesis, Technische Universiteit Eindhoven (2003). [22] H. Hottel, A. Sarofim, Radiative Transfer, McGraw-Hill, New York, New York, 1967. [23] B. . W. Company, Steam: its generation and use, Babcock & Wilcox Company, Ohio, 2005. Figure 10: Dynamic response to 10% step-wise increase in waste flow, on steam production for input waste with 25% ash content (black line), and 35% ash content (gray line). The responses are reported as relative variation with respect to the working point (14.5 kg/s). The horizontal line indicates 66% of maximum steam production response. Acknowledgments This work is part of the Waste-to-Energy 2030 project cofunded by industry and public partners and the Research Council of Norway under the EnergiX program (WtE 2030, 280949). References [1] S. Kaza, L. Yao, P. Bhada-Tata, F. Van Woerden, What a waste 2.0: a global snapshot of solid waste management to 2050, World Bank Publications, 2018. [2] Statistics norway: Avfall frå hushalda, https://www.ssb.no/ avfkomm, accessed: 2019-12-18. [3] E. Directive, Directive 2008/98/ec of the european parliament and of the council of 19 november 2008 on waste and repealing certain directives, Official Journal of the European Union L 312 (3) (2008). [4] H. K. Jeswani, R. W. Smith, A. Azapagic, Energy from waste: carbon footprint of incineration and landfill biogas in the UK, The International Journal of Life Cycle Assessment 18 (1) (2013) 218–229. [5] J.-W. Lu, S. Zhang, J. Hai, M. Lei, Status and perspectives of municipal solid waste incineration in China: A comparison with developed regions, Waste Management 69 (2017) 170–186. [6] J. Malinauskaite, H. Jouhara, D. Czajczyńska, P. Stanchev, E. Katsou, P. Rostkowski, R. J. Thorne, J. Colon, S. Ponsá, F. Al-Mansour, et al., Municipal solid waste management and waste-to-energy in the context of a circular economy and energy recycling in europe, Energy 141 (2017) 2013–2044. [7] G. Fitzgerald, Pre-processing and treatment of municipal solid waste (MSW) prior to incineration, in: Waste to Energy Conversion Technology, Elsevier, 2013, pp. 55–71. [8] L. Makarichi, W. Jutidamrongphan, K.-A. Techato, The evolution of waste-to-energy incineration: A review, Renewable and Sustainable Energy Reviews 91 (2018) 812–821. [9] F. Alobaid, N. Mertens, R. Starkloff, T. Lanz, C. Heinze, B. Epple, Progress in dynamic simulation of thermal power plants, Progress in Energy and Combustion Science 59 (2017) 79–162. [10] M. Rovaglio, D. Manca, G. Biardi, Dynamic modeling of waste incineration plants with rotary kilns: Comparisons between experimental and simulation data, Chemical Engineering Science 53 (15) (1998) 2727–2742. [11] T. Mahlia, M. Abdulmuin, T. Alamsyah, D. Mukhlishien, Dynamic modeling and simulation of a palm wastes boiler, Renewable Energy 28 (8) (2003) 1235–1256. 12 - A dynamic model of a moving grate Waste-to-Energy plant has been developed. The effect of waste composition on the dynamic plant response is analysed. Plant response was slower for higher moisture and ash content in the waste. Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: