Identification of ORC unit operation in biomass-fired cogeneration system

Renewable Energy 142 (2019) 400e414

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

Identification of ORC unit operation in biomass-fired cogeneration system
Jacek Kalina*, Mateusz Swierzewski Silesian University of Technology, Institute of Thermal Technology, Konarskiego 22, 44-100 Gliwice, Poland

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 November 2018 Received in revised form 29 March 2019 Accepted 17 April 2019 Available online 25 April 2019

This paper presents an analysis of operational parameters of the commercial Organic Rankine Cycle (ORC) cogeneration unit integrated with biomass-fired boiler and municipal heating network. The analysis is based on field measurements in real operational conditions using standard industrial sensors installed within the system. Regression based mathematical modelling have been applied to develop robust predictive models of the ORC system for its diagnostics and production planning. Historical data collected within the Supervisory Control and Data Acquisition System of the plant have been used to establish correlations between key thermodynamic parameters. Results reveal off-design performance characteristics of the ORC unit and its individual components such as turbine, evaporator and condenser. There are also demonstrated results of application of the model for technical condition and performance monitoring, which can support decisions on maintenance activities. © 2019 Elsevier Ltd. All rights reserved.

Keywords: Biomass Organic rankine cycle Cogeneration Process identification Diagnostics

1. Introduction Organic Rankine Cycle (ORC) technology has been considered for many years as an interesting option for utilisation of locally available biomass resources. The technology have received a lot of attention [1,2] and nowadays, there is significant number of plants re et al. [3,4] the in operation worldwide. According to Tartie cumulated installed electric power of the biomass-fired ORC plants in 2016 was around 301 MW and the estimated number of the ORC units was 332. Most of the plants were built within the years 2004e2016 in Germany, Austria and Italy as the result of effective system of incentives and financial support for investment projects. The ORC technology has been also considered as the key component of complex structures of renewable energy plants [5,7]. However, although the technology is mature and commercially available, there are still some issues preventing successful market uptake. The biggest barrier is modest financial attractiveness of investment projects. Special conditions are required to achieve reasonable level of profitability and payback period [2,6]. Moreover, under current legal and market conditions the financial performance of biomass cogeneration projects varies within operational

* Corresponding author. E-mail addresses: [email protected] (J. Kalina), mateusz.swierzewski@polsl.
pl (M. Swierzewski). https://doi.org/10.1016/j.renene.2019.04.080 0960-1481/© 2019 Elsevier Ltd. All rights reserved.

phase [7]. Therefore, there are required correction actions [8e10] and continuous monitoring of plant’s performance [11]. A certain possibility to improve the cost-effectiveness of projects comes from the possibility to participate in the reserve and balancing electricity markets [12,13]. However, making profits from volatile electricity prices requires a plant-specific control strategy. The best operational effects can be achieved if there is knowledge available on technical condition of equipment, off-design characteristics and limitations as well as flexibility and possible control settings. Therefore plant owners and operators should be supported by effective software tools that would provide credible information for on-line diagnostics and optimal plant control under variable load conditions, biomass properties, electricity and fuel prices. Such tools require system-specific mathematical models that can generate relevant data for verification of measurement devices, performance assessment by comparing measured parameters against model calculations (on-line diagnostics) and optimal production planning [13,15,16]. Nowadays, effective semi-empirical models can be created using existing Supervisory Control and Data Acquisition (SCADA) systems and collection of long-term historical measurement data which would allow identification of the energy conversion processes. A comprehensive review of modelling approaches and tools for the off-design simulation of ORC systems for performance prediction and design of control strategy has been carried out by Liu et al. [17]. Their findings show a variety of models of different

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

applicability, accuracy, computational complexity, data requirement and physical robustness. In several works presented in the literature there have been demonstrated application of mathematical models for performance improvement of industrial-scale plants. Salogni et al. [9] presented experiences from operation and maintenance of the combined heat and power plant (CHP) in Cremona (Italy), with Turboden 14 CHP module of 1.0 MW of gross electric power and 5.5 MW of heating capacity. In the paper, the actual performance was compared against the reference performance. Key mathematical relationships of working fluid evaporation and condensation pressures, electric power and estimated fluid flow were discussed as they allowed detection of turbine faults and supported decision on an early overhaul. Prando et al. [10] presented operational experiences from the plant in Renon (Bolzano, Italy). Basing on the results from a calibrated Matlab/ Simulink thermodynamic model they proposed measures to improve performance of the ORC module. The improvements consisted mainly on different management strategies such as decrease of the district heating network temperature. Erhart et al. presented several studies on equipment performance of biomass fired cogeneration plant in “Scharnhauser Park”, a quarter of the city of Ostfildern, Germany [18e20]. There were examined correlations between various parameters of the ORC module. In Ref. [18] electric power of the system was correlated with heat flux delivered to the cycle and thermal oil temperature at the evaporator inlet. Variability of parameters such as turbine isentropic efficiency, evaporation and condensation pressures was also discussed and respective correlations were proposed. A study dedicated to the impact of condenser operation on the entire cycle has been performed in Ref. [20]. The authors claim that transient sink conditions is the main cause of unsteady operation for the entire system. They conclude that by changing cooling water mass flow rate through the condenser the electric efficiency can be improved by 1e1.5% point. This corresponds to a difference in the electric power in the range of 10%. In addition, it was demonstrated in Ref. [20] that the best performance requires low degree of working fluid superheating in the evaporator and high evaporation to condensation pressure ratio. Sami et al. [21] demonstrated the use of simplified dynamic model of biomass fired ORC cogeneration system as an optimization and design tool. The results revealed that at the given biomass inlet mass flow rate biomass energy conversion efficiency decreases with increasing heating value of the feedstock. . Similar work was presented by Calise at al. [22], who developed a zerodimensional dynamic simulation model of solar-geothermal hybrid cogeneration plant based on an ORC unit. The model was developed in Engineering Equation Solver using energy and mass balances as well as empirical off-design performance correlations for turbine and pump. Dickens et al. [23] demonstrated the use of a validated offdesign model for assessment of the highest achievable net power of a test rig ORC system over an extensive range of operating conditions. They claim that the off-design performance optimization is computationally-costly and it is unlikely to be performed in real time. Therefore, it would be very useful to predict performance parameters using simple formulas. They demonstrated that twoparameter correlations between normalised (dimensionless) system variables can be developed to effectively map its performance. In this paper the ORC unit installed at the biomass-fired cogeneration plant of the city of Krosno (Poland) has been examined using long term measurement data collected within existing SCADA system. These data have been used to develop a semiempirical regression model of the ORC module. Operational experiences of this plant under real load conditions have been presented in our previous work [7]. Since the year 2017 the plant has been participating in the electricity balancing market. Currently there

401

are on-going design works on a heat storage system, which would increase operational flexibility of the plant. Therefore, in the near future the CHP plant will run under variable heating load focusing on the best economic performance. However, the results will depend on off-design characteristics of the key components of the system. In this work there is studied off-design characteristics of the Turboden 14 CHP Split ORC module [24], which is one of them. As the mathematical models for thermal diagnostics of energy conversion plants operation should have simple structure and short computing time [13], the approach adopted in this work is based on process identification. In general, this is a procedure of finding a simplified mathematical representation of a physical system that would enable fast predictions of its parameters of state and output quantities with an acceptable accuracy. Process identification is typically used for estimation of output signals of dynamic systems due to the change of input signal or disturbances. However, in static processes this technique is useful for obtaining accurate correlations for unknown and difficult to estimate quantities, such as turbine isentropic efficiency, friction losses and heat transfer coefficients. In cases where large collections of measurement data are available process identification techniques can be used instead of or together with physical modelling. In such cases empirical or semiempirical models of a process can be developed [13]. All the computations have been performed using Matlab software. Thermophysical properties of fluids were determined using the Open Source CoolProp library [25]. Empirical correlations between operational parameters of the ORC unit have been developed using historical data acquired and collected over the period from 2013 to 2017. The paper shows preliminary results of modelling and compares them against measurement data. It also discusses possible functionality of the model and potential benefits from its implementation in the plant’s on-line monitoring system. 2. ORC cogeneration unit operation Since the CHP module was commissioned in 2013, the technological system of the Krosno plant has consisted of two main blocks namely heating subsystem with coal-fired heat-only boilers and biomass-fired cogeneration subsystem. The CHP unit has been sized for the base heating load of the heating network and it is the only source of heat out the heating season. Within the heating season cogeneration runs in series with coal fired boilers. Indicative energy flows within the integrated system in the two respective modes of operation have been depicted in Figs. 1 and 2. The ORC unit delivers heat to the heating network from condenser ðQ_ Þ. out;C

However, the total heating output of the cogeneration system also includes heat inputs from exhaust gas latent heat recovery unit (HER) and from biomass boiler grate cooling (GC). As demonstrated by Erhart et al. [19], condenser operation is strongly influenced by the heating network input and output flows of hot water. Therefore in summer season the HER unit is by-passed and exhaust gas heat is not being recovered in order to increase electric power output of the ORC generator Pel;g. The ORC unit receives heat input from combustion system heat exchanges TOX, ECO1 and ECO2 by two thermal oil circuits: HTTOC and LTTOC. The main energy input is the heat flux to evaporator and working fluid heater Q_ . Heat in;EV

from low temperature oil circuit LTTOC is used in additional working fluid preheater (Split unit), which is a specific solution of the studied type of ORC unit In this work the main effort is payed to the modelling of performance of the ORC module. In regard to the biomass combustion boiler only the main operational features have been addressed. The reason for this is twofold. Firstly, the ORC unit is the prime component of the system that follows the heating load and sends

402

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

Fig. 1. Indicative energy flows in winter mode of operation (Legend: BC e biomass combustion chamber; GC e grate cooling; TOHX e high temperature thermal oil heat exchanger; ECO1, ECO2 e thermal oil economisers; CAH e combustion air heater; CAV e combustion air ventilators, EF e electro filter; EGV e exhaust gas ventilator; HER e exhaust gas latent heat recovery; HTTOC e high temperature thermal oil circuit; LTTOC e low temperature thermal oil circuit; ORC e ORC cogeneration unit).

Fig. 2. Indicative energy flows in summer mode of operation.

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

control signals to the combustion system. This is typical way of operation of such systems [18,19]. Secondly, there are many unknown and strongly variable parameters of biomass boiler operation such as biomass composition and mass flow [26], and therefore modelling of the process requires significant efforts that will be undertaken in a future work. Schematic diagram of the Turboden 14 CHP SPLIT module is presented in Fig. 3 and its technical specification is given in Table 1. This is module with the Split heat exchanger that allows better recovery of heat from biomass combustion gasses. The ORC working fluid is octamethyltrisiloxane C8H24Si3O2 (short name: MDM, CAS No. 107-51-7). The heat between biomass combustion system and the ORC unit is transferred by the thermal oil Therminol 66. The empirical correlations of the model have been developed using standard industrial measurement devices, which had been delivered together with the system. The ORC unit is controlled by the Siemens programmable logic controller Simatic OP 77A. The control system communicates with the Proficy HMI/SCADA iFIX delivered by the GE Digital [27] using the Ethernet TCP/IP protocol. Key measured thermodynamic parameters are: thermal oil temperatures TO,1 - TO,5, MDM temperatures T1 - T5, network water temperatures Tw,1 - Tw,2, evaporation pressure p1, condensation pressure p2. The flow of working fluid in not measured. The measured flows are thermal oil flows m_ O,1 and m_ O,2 in high and low temperature circuits respectively as well as the flow of condenser cooling water m_ w,1. The results of measurements are stored within the GE Proficy Historian ver. 4.5 database server.

403

Table 1 Technical specification of the ORC unit TURBODEN 14-CHP SPLIT [24]. Quantity Thermal oil loop Nominal temperature HT oil loop (in/out) Thermal power input HT loop Nominal temperature LT oil loop (in/out) Thermal power input LT loop Overall thermal input Cooling water circuit Heating network water temperature (in/out) Thermal power to the water circuit Performance Gross active electric power Gross electric energy efficiency Captive power consumption Net active electric power Net electric efficiency Indicative turbine isentropic efficiency Indicative biomass consumption

Unit

Value

 C kW  C kW kW

310/250 6130 250/130 585 6715

 C kW

60/80 5341

kW % kW kW % % kg/h

1317 19.6 62 1255 18.7 Up to 90 2935

3. Modelling of the Organic Rankine Cycle In recent years many papers have been published on theoretical and experimental performance of the ORC technology. However, there is lack of available information regarding real ORC units on industrial level [17]. Moreover, according to Park et al. [28] there is a large gap between research and development for source and sink temperature differences above 150  C, and the majority of published experimental works were performed for micro- and mini-

Fig. 3. Schematic diagram of ORC system studied in this work (Legend: 1- Split preheater; 2 - MDM heater; 3 e MDM evaporator; 4 e turbine by-pass valve; 5 e turbine admission valve; 6 e turbine; 7 e electric generator; 8 - regenerative heat exchanger; 9 e condenser; 10 e filter; 11 e MDM circulation pump).

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

404

scale ORC systems. Examples are works presented by Dickes at al. [23,29] and Lecompte et al. [30]. In general there are different approaches to modelling of power generation equipment performance. There are being developed physical models (“white box”), semi-empirical models (“grey box”) and empirical input-output models based on soft computing approach (“black box”). In a “white box” model no approximations of system parameters are made and such models are usually less accurate in predicting behaviour of real equipment. For diagnostics, dispatch and control purposes there are typically “grey box” and “black box” models used as presented for example in Refs. [31e33]. In this work such regression based models have been developed and tested. Overall, it is not an easy task to establish an effective mathematical model for on-line performance evaluation of an ORC system. The main reasons are the lack of a proper instrumentation, the inaccuracy of the available measurements and the uncertainty on fluid thermodynamic properties [32] as well as the effects transient operation due to variability of noise signals and the nature of the control process. The problem is usually tackled using sophisticated strategies, redundant measurements and calculation algorithms tailored for the purpose, as presented in Refs. [29,31,32]. On the other hand it is usually not economically justified in the case of small-scale distributed plants and in the case of existing plant only an existing SCADA system can be used for data acquisition. It this work these data are used for modelling and condition monitoring. Operation of the ORC system can be described theoretically by several fundamental equations of the thermodynamic processes constituting the Rankine cycle. In normal operation mode, with the turbine by-pass valve closed and the turbine inlet valve fully open the power output of the generator is:


 P_ el;g ¼ m_ wf ;T h1 jT1 ;p1  h2 jT2 ;p2 þDpRH;h hm;T hg

(1)

Where the turbine outlet enthalpy is:






 h2 jT2 ;p2 þDpRH;h ¼ h1 jT1 ;p1  h1 jT1 ;p1  h2;s s

1 ;p2 þ

DpRH;h

hi;T

(2)

Energy balance of the evaporator is:

  Q_ in;EV ¼ m_ O;1 hO;1  hO;2 ¼

vUEV þ m_ wf ;T h1 jT1 ;p1  m_ wf ;P h8 jT8 ;p1 þDpEV þ DQ_ EV vt

(3)

is:

(4)

Heat flux delivered to the cycle in Split heat exchanger is:

  Q_ in;SP ¼ m_ O;2 hO;4  hO;5
 ¼ m_ wf ;SP h6 jT6 ;p1 þDpEV þDpH  hP;out þ DQ_ SP

(5)

Heat flux transferred to the network water in condenser is:

  Q_ out;C ¼ m_ w;1 hw;2  hw;1 ¼

vUC þ m_ wf ;T h3 jT3 ;p2  m_ wf ;P h4 jT4 ;p2 DpC  DQ_ C vt

Electric power delivered to the pump is:

P_ el;motor ¼ m_ wf ;P 4h4 jT4 ;p2 DpC DpF


3 vP;in pP;out  pP;in 5 þ

hi;P

1

hm;P (7)

Where pump outlet and inlet pressures are:

pP;out ¼ p1 þ DpEV þ DpH þ DpRH;c

(8)

pP;in ¼ p2  DpC  DpF

(9)

Heat transferred from the hot side to the cold side in the regenerative heat exchanger is:



m_ wf ;P * h2 jT2 ;p2 þDpRH;h  h3 jT3 ;p2 ¼ m_ wf ;RH;c h5 jT5 ;pP;out DpRH;c  þ DQ_ þh P;out

RH

(10) Where the mass balance of flow splitter is:

m_ wf ;P ¼ m_ wf ;SP þ m_ wf ;RH;c

(11)

In addition, heat flux transferred from hot side to cold side at particular heat exchangers is given by the formula:

Q_ hc ¼ UADTm

(12)

The momentary gross efficiency of electricity generation, which is one of the indicators of system performance, can be presented as:

P_

el;g hel ¼ _ Q in;EV þ Q_ in;H þ Q_ in;SP

(13)

Although the presented equations explain principles of the system operation and allow theoretical analysis of its characteristics, they have very limited ability to predict operating parameters of a real unit under real load conditions. As the result of measurement system design the variables m_ wf ;T , m_ wf ;P , m_ wf ;SP , m_ wf ;RH;c , hm;T , hg , hi;T , hP;out , hi;P , hm;P as well as pressure drops Dpi and heat losses DQ_ are unknowns that can’t be precisely determined. The i

Heat flux delivered to the cycle by the thermal oil in the heater

  Q_ in;H ¼ m_ O;1 hO;2  hO;3
 ¼ m_ wf ;P h8 jT8 ;p1 þDpEV  h7 jT7 ;p1 þDpEV þDpH þ DQ_ H

2

(6)

variables that can be manipulated by the operator are m_ w;1 , hw;2 (through Tw;2 ), hO;1 (through TO;1 ). The parameters that should be monitored for evaluation of system’s condition are P_ , T , p , p as el;g

1

1

2

well as U, hi;T and hel , that typically are not taken into consideration by standard SCADA system of the plant. Useful information, such as for example temperature differences in heat exchangers, can be also drawn from the visual analysis of the cycle. Example of the T-s diagram of the thermodynamic cycle created using on-line reading of measurement data is presented in Fig. 4. Within the system being studied the heat flux to the ORC evaporator is controlled qualitatively, i.e. the flow of thermal oil is constant whereas the forward TO,1 and return TO,2 temperature varies with the load. The Split heat exchanger is supplied with heat by changing both the thermal oil inlet mass flow and its temperature. In the winter mode the CHP unit works at full load whereas in the summer mode it is set to the part load operation. In both cases operational parameters of the ORC unit vary significantly due to variable biomass quality and heating network water flow and temperature [7]. The system is relatively small and therefore its thermal inertia is low. It is never in steady state conditions as the amount of working fluid in low and high pressure sections varies in time. Additionally the cycle is influenced by vacuum pump operation for separation of noncondensing gasses. Fig. 5 depicts the

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

405

Fig. 4. Sample T-s diagram of real ORC cycle of the T14 CHP SPLIT unit at full load.

Fig. 5. Winter (a) and summer (b) dynamics of the ORC unit.

dynamics of the ORC unit within sample hours of operation in winter and summer day respectively. The key parameters are presented in the dimensionless normalised form, i.e. all the values are have been divided by the value at the beginning of an hour (see Fig. 6). 4. Measurement errors and accuracy of empirical correlations An important issue of regression modelling of industrial equipment is accuracy of data in field conditions. In laboratory conditions an extensive analysis of measurement errors can be performed as presented in Ref. [30]. However, in field operating conditions of industrial-scale equipment both input signals and noise can’t be fully controlled. Therefore, the accuracy of measurements can’t be fully determined from statistical analysis and information on precision of measuring instruments. In complex dynamic energy conversion systems there are several error components such as: sensor error, sensor mounting error, sensor

cleanliness and wear error, calculator error, property function error, signal transfer error, variability of operating conditions, system dynamics and thermal inertia, time delays and other. In such conditions overall measurement errors can be only roughly assessed and classified to type B uncertainty according to Ref. [34]. In this case the accuracy is estimated from any information other than statistics. This could be information from past experience of the measurements, from calibration certificates, manufacturer’s specifications, from calculations, from published information, and from common sense [34]. For example, Butler et al. [35] presented results of research carried out on heat metering accuracy, with a specific focus on installed heat meters rather than laboratory testing. The findings depicted that the error of majority of tests was within the range of 4e7%. In Ref. [7] raw measurement data were presented indicating variability quality of the obtained values. It was for example depicted that impossible states were suggested by some measurements such as liquid state of working fluid at evaporator outlet as well as the state of superheated vapour at condenser outlet. According to Ref. [34] the coverage factor, which is used as a

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

406 Table 2 Identification model variables. Parameter

Symbol

Method of determination

Measurement device

Estimated expanded uncertainty

Electric generator power output

Calculated

Power meter EQM 3*58/100V, 5A, P-0,5S, Q-1

±1.0%

Evaporator pressure Condenser pressure Condenser heating output

P_ el;g p1 p2 Q_

out;C

Direct measurement Direct measurement Calculated

Pressure transmitter SITRANS P, DSIII 7MF4233 Pressure transmitter SITRANS P, DSIII 7MF4233 Heat meter Supercal 531

±0.5% ±0.3% ±5.0%

Cooling water flow Water temperature at condenser outlet Condenser mean temperature difference Thermal oil evaporator inlet temperature Degree of superheating Evaporator heating output

m_ w Tw;2 DTm TO;1 DTsup Q_

Direct measurement Direct measurement Calculated Direct measurement Calculated Calculated

Flow meter SONIX 10D Teletrans TS200 Pt500 e PT100 sensor e Fischer DE38

±1.5% ±0.3% ±1.0% ±0.5% ±3.0% ±5.0%

Power generation efficiency Turbine isentropic efficiency

hel hi;T

Calculated Calculated

e e

±10.0% ±10.0%

in;EV

multiplier of the combined standard uncertainty in order to obtain an expanded (expected) uncertainty is typically in the range 2e3. The task, which has been undertaken in this paper is not to determine the true values of the measured parameters, but to find the relationship between the measured parameters, which are described by the values measured in the industrial SCADA system. Therefore, in many works in this field the issue of measurement errors has not been addressed [20,21,23] or only expectations are presented [19] what is in line with the guidelines. In cases where redundant measurements are available data reconciliation techniques can be applied to increase accuracy of the model [14,31]. The key system parameters, which has been identified in this work using regression models are given in Table 2. In the first stage of the analysis historical data stored in the iFix Historian database were carefully examined in order to define the approach to the identification process. One minute average values calculated for the entire year period were used for the regression analysis. In order to select adequate set of data system stability checks and data filtration were performed applying the linear regression of measurements within a specified moving time

window. The results have been presented in Ref. [7]. After data filtration the Matlab software was used to establish correlation between key parameters of the ORC system. As the model is going to be used within an online diagnostics tool comparing current performance against the reference performance all the correlations were established using filtered measurement data. Accuracy, or goodness of fit, of correlations have been analysed by means of standard measures, i.e. coefficient of determination R2 and Root Mean Squared Error (RMSE):

Pi¼N 2 ðxi  b xiÞ R2 ¼ 1  Pi¼1 i¼N 2 i¼1 ðxi  xÞ

(14)

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u i¼N u1 X 2 ðx  b xiÞ RMSE ¼ t N i¼1 i

(15)

The highest electric power of this unit was achieved in the year 2015 after initial years of operation and calibration of the control system [7]. Therefore it was decided to use data from this year as

Table 3 Coefficients for condenser correlations. Equation

R2

RMSE

a

b

c

d

(17) (18) (19) (20)

0.9811 0.979 0.9542 0.9718

2.966*104 MPa 3.13*104 MPa 4.619*104 MPa 0.03348

0.3458 0.2239 0.2074 0.4122

4.314*106 4.031*104 10.86*104 0.6539

10.45*103 7.613*103 3.524*103 e

9.035*105 7.43*105 e e

Fig. 6. Condensation pressure according to Eq. (17).

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

407

Table 4 Coefficients for evaporator correlations. Equation

R2

RMSE

a

b

c

d

e

(21) (22)

0.9992 0.8425

7.325*103 MPa 0.01005

11.93 1.789

4.43*104 5.613

0.06444 1.557

e 4.503

e 0.3961

Table 5 Coefficients for degree of superheat correlations. Eq.

R2

RMSE

a

b

c

d

e

f

(23) (24) (24.1)

0.9642 0.9754 0.9689

0.5654  C 0.4692  C 0.4751  C

34.06 24.22 32.52

0.001417 0.08875 0.08045

0.3579 0.2286 0.2892

9.709*104 7.27*104 7.456*104

e 8.636*104 8.481*104

e 4.507*104 5.583*104

the reference data to establish model correlations for reference system performance.

p2 ¼ a þ bQ_ out;C þ cTw;2 þ dT 2w;2

(17)

p2 ¼ a þ bm_ w þ cTw;2 þ dT 2w;2

(18)

p2 ¼ a þ bDTm þ cTw;2

(19)

5. Identification of key model parameters According to Ref. [17] in regression models typically secondorder multivariate polynomials are applied for developing the component’s performance correlations. The same approach has been adopted in this work. Examination of historical data revealed that although there is the thermal oil admission valve installed at the ORC inlet its position is fully open nearly throughout the entire period of operation. Therefore, firstly a correlation between the Therminol 66 inlet temperature and the temperature of water at the condenser outlet as well as the heating output was examined:

TO;1 ¼ 155:7 þ 0:8288Tw;2 þ 0:023Q_ out;C   2

 8:91105 Tw;2 Q_ out;C þ 1:954107 Q_ out;C

(16)

The coefficient of determination of the correlation (15) is R2 ¼ 0.9825, RMSE ¼ 3.83  C. Several correlations have been tested for assessment of condensation pressure and better goodness of fit was obtained for two-parameter formulas. Best fitted correlations are Eq. (16)e(19). Coefficients of equations are given in Table 3. Fig. 6 presents condensation pressure as a function of heating output and water outlet temperature that results in the best goodness of fit factor R2.

Fig. 8. Evaporation pressure as function of the degree of superheat for a given range of thermal oil inlet temperature.

Fig. 7. Degree of superheat according to Eq. (23).

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

408

U Umax

 ¼

a

Tw;m Tw;m;max

Q_ out;C

TO,1. The correlation is as follows:

!b (20)

Q_ out;C;max

In the case of evaporator, correlations took different form. Evaporation pressure appeared to be best fitted by linear formula (21) whereas heat transfer coefficient was estimated using formula (22):

p1 ¼ a þ bQ_ in;EV þ cTO;1 U Umax

 ¼aþb þe

TO;m



TO;m;max Q_ in;EV

(21)

þc !2

Q_ in;EV

Q_ in;EV;max

!

2  TO;m þd TO;m;max

Q_ in;EV;max (22)

An important parameter of the evaporator is the degree of superheat that shows how much the working fluid is superheated above the saturation temperature for a given pressure. This parameter for the overall range of working conditions was correlated with evaporator heat input and thermal oil inlet temperature

DTsup ¼ a þ bQ_ in;EV þ cTO;1 þ dT 2O;1

(23)

Another correlation for the degree of superheat tested dependency of this quantity on pressure ratio:

DTsup ¼ a þ b

  2  p1 p p þ e 1 TO;1 þ fT 2O;1 þ cTO;1 þ d 1 p2 p2 p2 (24)

Coefficient for correlations (21) to (24) are given in Tables 4 and 5. Correlation (24) is depicted in Fig. 7. It has to be noticed that due to measurement errors also negative values of the degree of superheat occur. In Ref. [7] that at part load operation in some points measurements suggest the state of subcooled liquid at evaporator outlet. After filtering out these data the coefficients of correlation (24) change as presented in Table 5. for (24.1). It should be also emphasised that for a given inlet temperature of thermal oil higher values of the degree of superheat result in lower evaporation pressure what is presented in Fig. 8. Therefore the degree of superheat should be as low as possible in order to maximise electric power output. Generator power output and isentropic efficiency were

Fig. 9. Electric generator output power according to Eq. (25).

Fig. 10. ORC turbine isentropic efficiency according to Eq. (26).

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

409

Table 6 Coefficients for turbine correlations. Eq.

R2

RMSE

a

b

c

d

e

f

g

(25) (26) (27)

0.999 0.8426 0.9954

11.24 kW 0.02013 1.409  C

2.103e-07 33.12 11.47

0.04223 0.01467 0.885

3.981 1.507e-05 0.03935

e 0.3875 e

e 1.478e-03 e

e 1.882e-06 e

e 4.592e-05 e

Table 7 Coefficients for overall system correlations. Eq.

R2

RMSE

a

b

c

d

e

f

(28) (29) (30) (31)

0.9812 0.9820 0.7695 0.9980

255.1 kW 47.34 kW 0.00325 0.8452 kW

1452 183.3 0.02492 3.945

1.054 87.49 0.0839 0.01623

e 0.3291 0.01069 2.092e-05

e 14.21 2.88e-06 0.07888

e 0.03988 e e

e 4.998e-06 e e

Table 8 Sample results of power output estimation. Load case

Source of value

Measured

Grey box

Black box

High

Gross electric power P_ el;g , kW Relative error, % Gross electric power P_ el;g , kW Relative error, % Gross electric power P_ , kW

1341.8

1364.5

1342.7

e 686.2

1.69 663.4

0.07 785.8

221.6

3.32 237.8

14.51 298.3

Relative error, %

e 270.4

7.31 280.8

34.61 298.5

e

3.87

10.40

Medium Low (1) Low (2)

el;g

Gross electric power P_ el;g , kW Relative error, %

correlated with turbine pressure ratio and turbine inlet temperature. For isentropic efficiency evaluation usually either mass flow rate [19] or pressure ratio [15,20] is used as the only independent variable. In our case, as the mass flow rate of the working fluid is not measured, pressure ratio and turbine inlet temperature were used. The following correlations have been established:

 b p P_ el;g ¼ a 1 T c1 p2 

hi;T ¼ a þ b

p1 p2



(25)  2  p p þc 1 þ dT1 þ eT 21 þ fT 31 þ g 1 T1 p2 p2 (26)

 p T2 ¼ a þ bT1 þ c 1 p2

(27)

Measured data for correlations (25) and (26) are shown in Figs. 9 and 10. Additional correlations were developed for the overall system performance parameters. They are simple input-output correlations that do not consider internal processes and can be regarded as the black box model of the ORC. Such correlations have been developed for heat input, generator power output and electricity generation efficiency. These parameters were correlated with condenser heat output, thermal oil and water temperatures:

Fig. 11. Results of grey box model validation.

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

410

Q_ in;EV ¼ a þ bQ_ out;C

(28)

 TO;1 TO;1 2 TO;1 _ þ cQ_ out;C þ d þe Q P_ el;g ¼ a þ b Tw;2 Tw;2 Tw;2 out;C 2
þ f Q_

(29)

out;C

hel ¼ a þ b



TO;1 Tw;2



 TO;1 2 þc þ dQ_ out;C Tw;2

 2
p P_ el;c ¼ a þ bP_ el;g þ c P_ el;g þ d 1 p2

(30)

(31)

Coefficients for correlations (25)e(31) are given in Tables 6 and 7. 6. Model validation and use The grey box model of the ORC unit has been established using correlations (17), (21), (24) and (25). Condenser heat output, water outlet temperature and Therminol 66 oil inlet temperature have

been selected as independent parameters as they are manipulated by the operator. Additionally correlation (16) can be used as a recommendation of the thermal oil temperature set value. In the first step condenser and evaporator pressures are being determined from (17) and (21). Then (24) is used to determine evaporator outlet temperature. Finally power output is calculated using correlation (25). Fundamental thermodynamic equations (1)e(13) can be used to estimate other internal parameters if necessary. The inputoutput black box model consists of equations (28)e(31). Sample results of model calculations are presented in Table 8 in respect to electric generator power output. At full load operation, that in fact takes place for around 4500e5000 h per year, calculations using both tested approaches result in a relatively good agreement with measured values. At part load conditions results of model calculations are of slightly worse quality. For filtered historical data collected in 2015 (362695 samples) the average relative errors of the models in case of power output were 1.1% in the case of grey box approach and 3.4% in the case of black box approach. Validation of the grey box model of power output is presented in Fig. 11. Most of the predicted values are within the tolerance of ±5% from the measured data. Better quality of prediction has been reached at medium load. Figs. 12 and 13 depict results of electric

a)

b)

Fig. 12. Generator power output within sample hour of operation in winter (a) and summer.

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

a)

b)

Fig. 13. Condenser pressure within sample hour of operation in winter (a) and summer (b).

Fig. 14. Condenser pressure within sample hour of operation in December 2017.

411

412

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

power and condenser pressure predictions in winter and summer modes of operation respectively. The first functionality of the model is comparison of current performance parameters against the reference values, for which the model was calibrated. As it was presented in Ref. [7], a slight deterioration of the ORC unit performance has been noticed in the year 2017. The achievable electric power has decreased from the one observed in 2015 at the same external conditions. It appeared after implementation of the regression model that not only electric power but also other thermodynamic parameters have differed from the values predicted by the model. Examples are presented in Figs. 14e16. Modelling results suggest maintenance activities such as examination of working fluid condition and check of the amount of working fluid in the system. Parameters, which change independently from the operator are ambient conditions, network water return temperature and biomass properties. In addition, market prices of electricity vary in time. These parameters can be to a certain extend forecasted and then an optimal production plan can be decided. In this context the model was used for sizing hot water storage tank and optimization of operational strategy with sales of electricity on the power

balancing market [36]. The developed correlations were used in the Ebsilon Professional [37] simulation model of the cogeneration plant. On this basis, there was performed simulation of annual operation of the facility under real load conditions with a time step of 1 h. The simulation was repeated for different trial sizes of hot water storage tank. In this way the optimal size of the tank as well as optimal operational strategy were decided. 7. Conclusions Knowledge on the operating characteristics of equipment can be useful for planning changes in system structure, new control strategies as well as for decisions on maintenance and repairs. However, field examination of power machinery is not a trivial task. In this paper we propose using available measurements and historical data for condition monitoring of equipment. The paper presents the concept of a tool that could help monitoring of operational parameters and fast diagnostics of the commercial biomass-fired ORC cogeneration unit installed in municipal heating plant. The tool uses mathematical model of the plant, which has been calibrated using data acquired at the assumed reference state of the

Fig. 15. Degree of superheat within sample hour of operation in December 2017.

Fig. 16. Generator power output within sample hour of operation in December 2017.

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

machinery. The concept assumes comparison of current state parameters against the results of modelling. Together with predictions of load, biomass quality and electricity prices such model can be also an effective tool for improvement of economic performance of the system. Static empirical correlations between key thermodynamic parameters as well as grey box and black box regression models of the ORC unit have been developed using available historical measurement data. The model correlations revealed satisfactory goodness of fit and relatively good ability to predict state parameters of the ORC unit under different external conditions (input signals), for which the model was calibrated. In addition, due to relatively fast dynamics and small time delays of the processes within the ORC cycle the static correlations were able to reproduce time series of process parameters. However, for input data from a later period of operation higher differences were revealed, which indicated deterioration of system performance. In addition, it was found out that the grey box approach, in which internal state parameters of the system were calculated, resulted in a better accuracy than an inputoutput black box model. Acknowledgements This work was carried out in the project IntBioCHP titled: System integration of biomass fired cogeneration plants. The project is financed by German Federal Ministry of Education and Research and Polish Ministry of Science and Higher Education within the framework of the Polish - German Sustainability Research Programme STAIR. Nomenclature A h L_ m_ s p P Q_ U U T

DTm

heat transfer area, m2 specific enthalpy, J/kg heat loss flux, W mass flow, kg/s specific entropy, J/kgK pressure, Pa power, W heat flux, W internal energy, J heat transfer coefficient, W/m2K temperature,  C mean logarithmic temperature difference

Greek symbols h efficiency D difference v specific volume, m3/kg Subscripts c c C el EV F g h H i in/1 m m

captive cold side condenser electric evaporator filter electric generator hot side heater isentropic inlet mechanical mean

O out/2 P RH s SP T w wf

413

Therminol 66 thermal oil outlet pump regenerative heat exchanger after isentropic process Split preheater turbine water working fluid

References [1] G. Taljan, G. Verbi c, M. Pantos, M. Sakulin, L. Fickert, Optimal sizing of biomass-fired Organic Rankine Cycle CHP system with heat storage, Renew. Energy 41 (2012) 29e38. [2] M. Uris, J.I. Linares, E. Arenas, Techno-economic feasibility assessment of a biomass cogeneration plant based on an Organic Rankine Cycle, Renew. Energy 66 (2014) 707e713. re, ORC market: A world overview, last access: 04.05.2018, http://orc[3] T. Tartie world-map.org/analysis.html. re, M. Astolfi, A world overview of the organic rankine cycle market, [4] T. Tartie in: Proceedings of the IV International Seminar on ORC Power Systems, ORC2017, 13-15 September 2017, vol. 129, Energy Procedia, Milano, Italy, 2017, pp. 2e9. [5] A.M. Pantaleo, S.M. Camporeale, A. Sorrentino, A. Miliozzi, N. Shah, Ch. N. Markides, Hybrid solar-biomass combined Brayton/organic Rankine-cycle plants integrated with thermal storage: techno-economic feasibility in selected Mediterranean areas, Renew. Energy (2018) (Article in press), https:// doi.org/10.1016/j.renene.2018.08.022. (Accessed 28 March 2019). [6] A. Duvia, A. Guercio, C. Rossi, Technical and economic aspects of biomass fuelled CHP plants based on ORC turbogenerators feeding existing district heating networks, in: Proceedings of the 17th European Biomass Conference, Hamburg, Germany, June. 2009, pp. 2030e2037.
[7] J. Kalina, M. Swierzewski, R. Strzałka, Operational experiences of municipal heating plants with biomass-fired ORC cogeneration units, Energy Convers. Manag. 181 (2019) 544e561. [8] A. Rettig, U.Ch. Müller, L. Gasser, J. Hurter, Dynamic simulations supporting the design process of a real combined heat and power application in Switzerland, in: Proceedings of the IV International Seminar on ORC Power Systems, ORC2017, 13-15 September 2017, vol. 129, Energy Procedia, Milano, Italy, 2017, pp. 644e651. [9] A. Salogni, D. Alberti, M. Matelli, R. Bertnanzi, Operation and maintenance of a biomass fired e organic Rankine Cycle e CHP plant: the experience of Cremona, in: Proceedings of the IV International Seminar on ORC Power Systems, ORC2017, 13-15 September 2017, vol. 129, Energy Procedia, Milano, Italy, 2017, pp. 668e675.  [10] D. Lon car, Z. Guzovi
c, N. Serman, Performance improvement of biomass cogeneration plant by dual loop organic rankine cycle, Chem. Eng. Trans. 39 (2014), https://doi.org/10.3303/CET1439062. [11] D. Prando, M. Renzi, A. Gasparella, M. Baratieri, Monitoring of the energy performance of a district heating CHP plant based on biomass boiler and ORC generator, Appl. Therm. Eng. 79 (2015) 98e107. €ge, O. Renner, R. Pohl, Profitability of participation in control [12] T. Muche, Ch. Ho reserve market for biomass-fueled combined heat and power plants, Renew. Energy 90 (2016) 62e76. [13] N. Kumbartzky, M. Schacht, K. Schulz, B. Werners, Optimal operation of a CHP plant participating in the German electricity balancing and day-ahead spot market, Eur. J. Oper. Res. 261 (2017) 390e404. [14] M. Plis, H. Rusinowski, A mathematical model of an existing gas-steam combined heat and power plant for thermal diagnostic systems, Energy 156 (2018) 606e619. [15] G. Szapajko, H. Rusinowski, Theoretical-empirical model of the steam-water cycle of the power unit, Acta Montan. Slovaca 15 (1) (2010) 24e27.
[16] J. Kalina, M. Swierzewski, M. Szega, Simulation based performance evaluation of biomass fired cogeneration plant with ORC, Energy Procedia 129 (September) (2017) 660e667. [17] L. Liu, T. Zhu, N. Gao, Z. Gan, A review of modeling approaches and tools for the off-design simulation of organic rankine cycle, J. Therm. Sci. 27 (4) (2018) 305e320. [18] T. Erhart, U. Eicker, D. Infield, Part-load characteristics of organic-rankinecycles, in: Proceedings of the 2nd European Conference on Polygeneration e 30th March-1st April, 2011, pp. 1e11. Tarragona, Spain. [19] T.G. Erhart, U. Eicker, D. Infield, Influence of condenser conditions on organic rankine cycle load characteristics, J. Eng. Gas Turbines Power 135 (April 2013), 042301-1 to 9. [20] T. Erhart, M. Van den Broek, U. Eicker, Dynamic models for a heat-led organic rankine cycle, in: Proceedings of the International ASME-ORC Power Systems Seminar 2013, October 2013. Rotterdam, Netherlands, October, 2013, pp. 7e8. [21] S. Sami, E. Marin, A numerical model for predicting dynamic performance of biomass-integrated organic rankine cycle, ORC, system for electricity generation, Am. J. Energy Eng. 4 (3) (2016) 26e33.

414

J. Kalina, M.
Swierzewski / Renewable Energy 142 (2019) 400e414

[22] F. Calise, D. Capuano, L. Vanoli, Dynamic simulation and exergo-economic optimization of a hybrid solaregeothermal cogeneration plant, Energies 8 (2015) 2606e2646. [23] R. Dickes, O. Dumont, S. Quoilin, V. Lemort, Performance correlations for characterizing the optimal off-design operation of an ORC power system, in: Proceedings of the IV International Seminar on ORC Power Systems, ORC2017, 13-15 September 2017, vol. 129, Energy Procedia, Milano, Italy, 2017, pp. 907e914. [24] https://www.turboden.com/applications/1051/biomass. (Accessed 28 March 2019). [25] I.H. Bell, J. Wronski, S. Quoilin, V. Lemort, Pure and pseudo-pure fluid thermophysical property evaluation and the open-source thermophysical property library CoolProp, Ind. Eng. Chem. Res. 53 (6) (2014) 2498e2508.

˛ ka, Operational [26] J. Kalina, A. Sachajdak, R. Strzałka, M. Swierzewski, J. Cwie Experience and Modelling of Biomass Combustion Process in Cogeneration Systems with ORC Units, E3S Web Conf. 82 (2019). https://doi.org/10.1051/ e3sconf/20198201008. (Accessed 28 March 2019). [27] https://www.ge.com/digital/products/ifix. (Accessed 28 March 2019). [28] B. Park, M. Usman, M. Imran, A. Pesyridis, Review of organic rankine cycle experimental data trends, Energy Convers. Manag. 173 (2018) 679e691. [29] R. Dickes, O. Dumont, R. Daccord, S. Quoilin, V. Lemort, Modelling of organic Rankine cycle power systems in off-design conditions: an experimentallyvalidated comparative study, Energy 123 (2017) 710e727. [30] S. Lecompte, S. Gusev, B. Vanslambrouck, M. De Paepe, Experimental results of

[31]

[32]

[33]

[34]

[35] [36]

[37]

a small-scale organic Rankine cycle: steady state identification and application to off-design model validation, Appl. Energy 226 (2018) 82e106. L. Zanellato, M. Astolfi, A. Serafino, D. Rizzi, E. Macchi, Field performance evaluation of geothermal ORC power plants with a focus on radial outflow turbines, Renew. Energy (2018) 1e9. Article in press, https://doi.org/10.1016/ j.renene.2018.08.068. (Accessed 28 March 2019). Z. Li, R. Outbib, S. Giurgea, D. Hissel, A. Giraud, P. Couderc, Fault diagnosis for fuel cell systems: a data-driven approach using high-precise voltage sensors, Renew. Energy 135 (2019) 1435e1444. G. Cavazzini, P. Dal Toso, Techno-economic feasibility study of the integration of a commercial small-scale ORC in a real case study, Energy Convers. Manag. 99 (2015) 161e175. Joint Committee for Guides in Metrology: Evaluation of Measurement Data d Guide to the Expression of Uncertainty in Measurement, JCGM, 2008. https:// www.bipm.org/en/publications/guides/gum.html. (Accessed 28 March 2019). D. Butler, A. Abela, C. Martin, Heat Meter Accuracy Testing, Final report, Department of Energy & Climate Change, London, UK, November 2016.
M. Swierzewski, J. Kalina, Implementation of heat storage and network water cooler for improvement of energy and economic performance of municipal heating plant with biomass fired cogeneration module, in: Proceedings of the 5th International Conference on Contemporary Problems of Thermal Engineering CPOTE 2018, 18-21, September 2018 (Gliwice, Poland). STEAG energy services: Ebsilon professional. https://www.ebsilon.com/en/. (Accessed 28 March 2019).