Design and Operation Optimization of a Hybrid Railway Power ...

Design and Operation Optimization of a Hybrid Railway Power Substation

PANKOVITS Petronela

Design and Operation Optimization of a Hybrid Railway Power Substation Petronela Pankovits1,3, Maxime Ployard2, Julien Pouget1, Stephane Brisset2, Dhaker Abbes3, Benoit Robyns3 1. SNCF Innovation & Research, 40 avenue des Terroirs de France, 75611 Paris, France E-Mail: [email protected], [email protected] 2. University Lille Nord de France, Laboratory of Electrical Engineering and Power Electronics of Lille (L2EP), Ecole Centrale de Lille, Cité Scientifique – BP 48, 59651 Villeneuve d’Ascq, France E-Mail: [email protected] 3. University Lille Nord de France, Laboratory of Electrical Engineering and Power Electronics of Lille (L2EP), Ecole des Hautes Etudes d’Ingenieur (HEI) 13, rue de Toul, F-59046 Lille, France E-Mail: [email protected], [email protected]

Acknowledgements The presented topic is performed under CONIFER project, sponsored by the French National Research Agency (ANR). The final purpose of this project is to develop new tools and methods for the design and energy management of future hybrid railway network based on smartgrid technology, for the French National Railway Corporation.

Keywords «Railway substation», «Hybrid power integration», «Energy storage», «Design», «Optimization», «Energy efficiency».

Abstract Railway traffic increases and electricity market liberalization constrain the railway actors to consider new solutions to handle the energy consumption. Hence, a technology change in the railway electrical systems is considered through the integration of renewable energy sources and storage units. In this context, a relevant methodology is proposed here for optimal design and operation analysis of railway hybrid power substations. This method is useful for the analysis and improvement of future railway power network efficiency.

Introduction Nowadays, railway traffic is rising in all over the countries, implying an increase in electric power consumption. To overcome this, it is important to ruminate on new efficient railway substation power systems with different architectures and more power electronic devices (power converters) without disturbing the traffic or the energy quality of railway lines [2]. In addition, with electricity market liberalization, railway actors are determined to search for techno-economic solutions to face future energy demand increase. One solution is the integration of renewable energy sources and storage units in the railway electrical power systems. This solution may contribute to a partial independence from energy producers and improve power quality through storage units (power smoothing). Before adapting this solution and developing the future Hybrid Railway Power Substations (HRPS), it is essential to consider the new network architecture and to presage its optimal mode of operation before any implementation. This paper deals with this last issue. First, a new network architecture of the electrical railway traction power system has been established, considering constraints related to renewable energy sources integration in the power sector [13] and possible configurations for the future energy system [8], [10], [12].

EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.1


PANKOVITS Petronela

Then, for the retained architecture, a pre-design process has been developed through an optimization stage and analysis of the operation mode. In this context, many research studies are focusing on optimal design and operation of isolated or grid-connected stationary hybrid renewable power systems [11]. In [5] and [14] authors proposed the sizing and techno-economic optimization of an autonomous hybrid power system considering wind turbine swept area, photovoltaic generator installed surface and batteries capacity. However, they haven’t optimized batteries charge and discharge profiles. In [3] a method for optimal design of a hybrid stand-alone system including aero-generators, a photovoltaic generator and a battery is implemented minimizing the cost over its 20 years of operation. The optimal size of components (units’ number of wind generators, PV arrays and battery storage) is calculated using Particle Swarm Optimization (PSO) and three states of operating strategy as constraints (i.e. charging the battery when renewable power generation is greater than load demand and use it in the contrary case). Furthermore, three scenarios are set for choosing the minimum cost in meeting the power demand. In [9] an optimal design analysis of a stand-alone photovoltaic and fuel cell hybrid system is carried out using genetic algorithm (GA). Optimal power of renewable sources, battery capacity and state of charge was determined by considering the cost as the objective function and analyzing the influence of each variable on the total cost. In this case, the battery state of charge enables to start and stop the fuel cell, thus setting the charging/discharging control of the storage system during the operation strategy. Ashok, et al. [4] explicit a method for optimal sizing of a micro-hydro-wind and PV hybrid system by finding the trade-off between reliability and life cycle cost. Besides of units sizing, an optimal scheduling of hourly load and renewable energy during a day is accomplished using Quasi-Newton algorithm. Hence, the operation strategy for the storage system is a priori defined in the optimization method. It is in [6] that the authors explicit a particularly approach dedicated to optimal planning of the operation mode in the energetic sketch up for a household PV multi-source system. They applied linear mix integer optimization to show that PV production is advantageous despite its intermittent nature. Our contribution is consistent with all these researches since we are interested in the techno-economic analysis of the new HRPS. Nevertheless, another goal is the optimal sizing of the storage system charging/discharging profile considering power balance and financial criteria. Thus, an analysis of the optimal operation strategy of the renewable hybrid power system is led. So our paper is organized as follows: First generic architecture for future HRPS is presented. Then, design phase is exposed. It is based on an iterative optimization process using two different algorithms (SQP and GA) and is carried out with a quasi-static power flow model simulated on Matlab/Simulink. Finally, proposed method is applied successfully to a case study. Optimal planning of electrical sources and loads for a typical day is obtained and results and conclusions are carried out.

Generic architecture for future HRPS The French National Rail network is electrified in two main systems: 1.5kV DC and 25kV 50Hz AC [2]. According to specific characteristics of each system (i.e. parallel connection between two railway substations in DC system unlike AC system), we are determined to treat each system separately. In addition, since the railway power supply for electrical traction is a very particular system, it is better if the hybrid system composed of renewable sources and energy storage will be attached to the substation without changing its natural structure. Hence, it will just improve it by allowing isolation mode if needed. Therefore, the proposed generic configurations for the new HRPS, for both DC and AC system are presented in Fig. 1 and Fig. 2. A multisource system (PV, wind turbine, foreseeable source and associate storage units) can be connected to the railway substation through DC or AC bus with appropriate power electronics converters to obtain the catenary voltage level (1500V DC and 25 kV 50 Hz AC). This connection will enhance new energy flow between the electrical grid (Smartgrid 1) and a neighboring railway substation (HRPS b). This is possible directly by catenary (Smartgrid 3), either by dedicated feeder (Smartgrid 2). It’s interesting to mention that the train is seen as a consumer (C) in traction mode and as a source (S) in case of braking recovery. EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.2


Fig. 1: Generic architecture for DC HRPS system

PANKOVITS Petronela

Fig. 2: Generic architecture for AC HRPS system

System modeling For the design phase, system modeling is based on power and financial balances for the global system. In fact, power flow models for PV, wind turbine and storage system in addition of economic models for cost function are introduced:

Wind turbine model The hourly output of wind power, as expressed in [4], is: ·

· 0.5 ·

·

·

·

(1)

with the turbine efficiency, the generator efficiency (both obtained from the manufacturer data), the power coefficient of wind turbine (depends on the sails), the air density, the wind turbine rotor swept area and the wind velocity.

PV model For photovoltaic power output ·

, PV surface is considered as decision variable [1], so :

·

(2)

where is the solar radiance, surface area of PV panels.

is the power conversion efficiency of the PV module and

the

The wind profile considered for the optimization process is given in Fig. 3. The hourly data wind speed corresponds to the railway substation area. The global irradiation profile for the PV system is also given in Fig. 4. According to the graphs, photovoltaic potential is prevailing on wind potential for this location and a priori the site is not favorable for a wind farm installation. 1200

4 3.5

1000

Irradiation (W/m²)

wind speed (m/s)

3 2.5 2 1.5

800

600

400

1 200

0.5 0 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 time (hour)

Fig. 3: Wind speed profile

EPE'13 ECCE Europe

0

0

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 time (hour)

Fig. 4: Irradiation profile

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.3


PANKOVITS Petronela

Power flow model of battery storage The storage system considered in the design phase is a battery suitable for railway applications. It is mathematically represented with a power flow model as explained in [7]. We search to obtain not only the unit sizing (i.e. battery numbers), but also the optimal reference power of the storage system for 24 hours’ time interval, in order to analyze the optimal operation strategy. Hence, we create 24 variables . Various losses have been considered by introducing the energy efficiency for storage power if 0 and 1/ if 0). Negative power values mean that the battery is discharging whereas positive power values correspond to battery charging. Stored energy, the day as follows:

, depends on the initial state of charge and the energy stored every hour during 3600 · ∑

(3)

_

So, the energy amount stored in the end of the day will be: 24 · 3600

(4)

Particular problem formulation for our typical energy management problem (5) the network power, , storage power, is the power balance of the global system, with and wind generators the substation power, and the renewable power coming from PV panels

, .

Cost function (6) represents the total cost for the grid, is the global cost, where the wind production cost and the storage units cost. system,

is the total cost for the PV

Based on the forecast information such as meteorological condition, load demand and electricity prices (buying and purchasing price), we search to anticipate the management operation plan for the HRPS system. The total energy cost for a day is considered with 400 €/m² for the PV system cost, 70 €/MWh for the wind energy production cost and 220 €/kWh for the storage cost (battery system suitable in railway applications). The storage cost strongly depends on the density of mass energy and power density characteristics. This is why we will consider and as decision variables in the _ _ optimization model.

Optimal pre-sizing and operation mode analysis In the first step of the design process, the aim is to analyze the new network architecture of the hybrid renewable power system and to anticipate its optimal operation mode. The techno-economic analysis of the system is based on power and financial balances. An iterative optimization process is implemented with the following characteristics: • • •

system modeling; problem optimization specifications : decision variables (wind turbine rotor swept area, PV generator surface, batteries number and batteries hourly state of charge), feasibility constraints and objectives to minimize; adequate optimization algorithm.

Optimization is carried out using a GUI that adapts the dynamic model of the system under Matlab/Simulink to optimization procedure. This optimization tool enables to use overall optimization EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.4


PANKOVITS Petronela

approach based on existing algorithms in the Matlab Toolboxes reducing the convergence problems or the risk of hitting local optima. Also, it allows the use of distributed computing and the export of results for better analysis. The flowchart of the implemented procedure is presented in Fig. 5:

Fig. 5: Flowchart of the optimization procedure Optimization has been made with two different algorithms: SQP (determinist method) and GA (stochastic method). SQP algorithm is preferred due to constraints and specifications of the optimization model as described in Table I. SQP is well adapted to threat dynamic modeling problems including nonlinear and complex system. Precise and rather fast, SQP allows using a number of initial points highly superior to variable numbers. In the same time its convergence depends on the initial points and the solution found can easily hit a local optima. To avoid local optimum, GA (global research method) has been used. However, GA works better for an optimization model without constraints. So, the specifications used in the SQP optimization model have been changed in order to apply the GA. To facilitate the convergence, we propose a new problem formulation when applying GA optimization by eliminating constraints (including equality constraints). This formulation implicitly allows reducing optimization variables.

Table I. Specification of the optimization model for SQP algorithm Variables , with 0 , with 0 , with _ _ , _

_

Constraints _

Objective function Minimizing total cost Min(C)

_

1. .24

0

_

_

_

Table II shows, the specifications of the optimization model when using GA algorithm. In this case we express the energy constraints (Table I) in terms of storage power per hour ( _ ) and we assume that initial state of charge ( ) is zero. This implies having fewer variables in the optimization model.

Table II. Specification of the optimization model for GA algorithm

_

, and

EPE'13 ECCE Europe

Variables , with 0 , with 0 1. .23 with _ _

_

Objective function _ _

Minimizing total cost Min(C)

∑

_ _

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.5


PANKOVITS Petronela

When applying optimization techniques to a dynamic model, the convergence time is increasing with variables number. However, optimizing per hour is permitting to size the storage capacity and also to analyze its hourly management. This explains the 24 decision variables associated to the state of charge distribution in time, every hour, leading to an optimal predictive profile for the storage system.

Application study case An application study case for a particular railway substation is proposed. The nominal power of the substation is 20 . The feasibility when integrating renewable sources and energy storage is studied. Proposed sizing and design method is applied to a typical day according to the load profile of the railway substation. Considered sources profiles (hourly data of wind speed, global irradiation) , ) as correspond to the railway substation site. First, the areas of renewable energy generators ( well as the reference of the storage energy ( ) are determined. Then, analysis of the operation _ strategy proposed by the solution found with the specific algorithm is accomplished through two different scenarios.

Scenario 1 The first solutions found by the SQP algorithm are strongly determined by the cost values given for system components. With actual cost, one of the solutions is to install 2000m² of PV and a very week utilization of storage system. As expected, a very week utilization of wind turbines is suggested due to the low wind speed profile. This however allows validating the efficiency of the optimization algorithm.

Scenario 2 In order to show the interest of the optimization strategy in the design stage of a hybrid power system, in particular for its management operation analysis, a second case scenario is required. This is based on economical speculations concerning renewable energy technologies. The optimization results are obtained using SQP and GA and supposing lower cost for renewable sources (e.g. half price for the PV production) and storage units (e.g. almost zero cost) in order to validate optimization approach and associated developed tool. So, for the load demand of the HRPS system, the power distribution of system components is presented in Fig. 6. This time, using storage power ( in Fig. 6) is presented as a solution in the optimization process and its distribution per hour permits to analyze the operation management during the day. The storage system appears to charge when PV production is maximum and discharges when load demand is high (meaning high electricity price). Numerical values from Table III help to a better comparison between SQP and GA algorithm. Even if this values are extremely high, they actually correspond to the power demand by the consumer, 20 . The best solution is found by SQP algorithm. The PV surface is very large and no power is taken or sent to the grid. Solution found with SQP

Solution found with GA 2

2

P

P

PV

PV

P

P

sto

1.5

sto

1.5

P

P

grid

grid

1 Power (p.u.)

Power (p.u.)

1

0.5

0.5

0

0

-0.5

-0.5

-1

-1

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hour)

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hour)

Fig. 6: Optimization results for the considered load profile of the railway substation

EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.6


PANKOVITS Petronela

Table III. Numerical results of the optimization procedure SQP GA

Optimization variable (Spv) 200000m² 221500m²

Cost function (C) 39736k€ 42342k€

Evaluations 5351 26100

Conclusions In this study, new network architecture for the future hybrid railway substations (HRPS) and its preliminary analysis and design optimization are presented. A dynamic model based on quasi-static and power flow equations including financial aspects has been presented in the first part. The sizes of the energetic systems as well as a first analysis concerning the optimal energy management strategy based on two optimization procedures was developed in the second part. The obtained results for an application study case were discussed based on two different scenarios. For the second case, the results are obtained supposing low cost renewable sources and storage units in order to validate optimization approach and associated developed tool. It appears that SQP algorithm is faster and more accurate than GA to solve this power flow optimization problem. In future work, an energy management strategy will be applied to HRPS system. This will require experimental verification of the obtained results in a reduced scale.

References [1]

Abbes, D., A. Martinez, G. Champenois, and J. P. Gaubert. "Multi-objective Design Optimization of a Hybrid PV-wind-battery System." ELECTRIMACS. Cergy-Pontoise, France, 2011.

[2]

Aeberhard, M., C. Courtois, and P. Ladoux. "Railway Traction Power Supply from the state of the art to future trends." Power Electronics Electrical Drives Automation and Motion (SPEEDAM), International Symposium on. 2010.

[3]

Ardakani, J. F., and G. H. Riahy. Renewable Energy - Trends and Applications, Chapter11 - Optimum Design of a Hybrid Renewable Energy System. Croatia (Hrvatska), 2011.

[4]

Ashok, S. "Optimised model for community-based hybrid energy system." Renewable Energy 32, no. 7 (2007): 1155-1164.

[5]

Celik, Ali Naci. "Techno-economic analysis of autonomous PV-wind hybrid energy systems using different sizing methods." Energy Conversion and Management 44 (2003): 1951-1968.

[6]

Clastres, C., T.T. Ha Pham, F. Wurtz, and S. Bacha. "Ancillary services and optimal household energy management with photovoltaic production." Energy 35 (2010): 55-64.

[7]

Jaafar, A., C. R. Akli, B. Sareni, X. Roboam, et A. Jeunesse. «Sizing and Energy Management of a Hybrid Locomotive Based on Flywheel and Accumulators.» IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 58, n° 8 (2009): 3947-3958.

[8]

Keyhani, A., M. N. Marwali, and M. Dai. Integration of Green and Renewable Energy in Electric Power Systems. Edited by Wiley & Sons. 2010.

[9]

Lagorse, J., S. Giurgea, D. Paire, M. Cirrincione, M. G. Simoes, and A. Miraoui. "Optimal Design Analysis of a Stand-Alone Photovoltaic Hybrid System." Industry Applications Society Annual Meeting IAS '08, IEEE, 2008.

[10]

Liserre, M., T. Sauter, and J.Y. Hung. "Future Energy Systems Integrating Renewable Energy Sources into the Smart Power Grid Through Industrial Electronics." Industrial Electronics Magazine, IEEE 4 (2010): 18-37.

EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.7


PANKOVITS Petronela

[11]

Luna-Rubio, R., M. Trejo-Perea, D. Vargas-Vazquez, and G.J. Rios-Moreno. "Optimal sizing of renewable hybrids energy systems: A review of methodologies." Solar Energy 86 (2012): 1077-1088.

[12]

Nehrir, M.H., et al. "A Review of Hybrid Renewable/Alternative Energy Systems for Electric Power Generation: Configurations, Control, and Applications." IEEE Transactions on Sustainable Energy 2 (2011): 392-403.

[13]

Robyns, B., A. Davigny, B. François, A. Henetton, and J. Sprooten. Electricity Production from Renewables Energies. Edited by Wiley & Sons. 2012.

[14]

Wang, L., and C. Singh. "Multicriteria Design of Hybrid Power Generation Systems Based on a Modified Particle Swarm Optimization Algorithm." IEEE TRANSACTIONS ON ENERGY CONVERSION 24, no. 1 (2009): 163-172.

EPE'13 ECCE Europe

ISBN: 978-90-75815-17-7 and 978-1-4799-0114-2

P.8