Frequency and Voltage Control in Microgrids: Modeling and Simulations in Islanded Mode Mohsen Nemati Stefan Tenbohlen University of Stuttgart-IEH Stuttgart, Germany mohsen.nemati@ieh.uni-stuttgart.de stefan.tenbohlen@ieh.uni-stuttgart.de 2 Muhammad Imran 2 Holger Mueller 2 Siemens AG Erlangen, Germany imran.muhammad@siemens.com hmmueller@siemens.com 3,4 Martin Braun 3 Fraunhofer IWES, 4 University of Kassel Kassel, Germany martin.braun@uni-kassel.de Abstract Given the intermittent nature of renewable generation, in order to successfully accommodate it into electrical grid and to reduce the control burden on utility networks, the idea of smartly managing and controlling smaller subsystems either continuously operating in isolation from utility networks or as an extension to existing power supply structures has been a focus of extensive research for a decade and is lately receiving considerable commercial attention. The design, development and analysis of such microgrids demand the availability of dynamic models which are capable of reflecting the newly introduced control capabilities and which offer the flexibility of simulating the operation scenarios which have not been taken into account in traditional practices of power system analysis and design. This paper focuses on an implementation of secondary frequency control and the so called grid forming capability of converter based renewable generators. Commercial simulation software PSS NETOMAC and PSS SINCAL have been employed because of their widespread use among the network development professionals. These models can be used in a network development process, in stability studies, and in online dynamic security assessment systems, in order to assess the capability of microgrids in providing a reliable power supply under different operating conditions, and to evaluate their dynamic response and the effectiveness of proposed control strategies. Keywords Distributed Generators, Dynamic Models, Secondary Control, Microgrids, PSS NETOMAC and PSS SINCAL I. INTRODUCTION In this paper, a microgrid hybrid system combining battery storage, photovoltaic and diesel generators has been taken into account and the focus has been on modeling of the control features pertaining to primary and secondary control, and on analysis of their interactions in the system. The concept of a microgrid assumes a cluster of controllable loads, microsources and storage devices operating as a single controllable system that provides power to the customers [,2] and facilitates the integration of distributed renewable generation while avoiding the stability related problems faced by the utility networks. [3] In case of a fault in the upstream network, or for planning and economic reasons, a microgrid can be automatically transferred to island mode by disconnecting it from the distribution network. During the island mode the subsystem must be capable of maintaining a specified degree of power supply reliability. [4] Thus the stability of these systems is an important aspect of power supply quality and reliability and therefore it should be analyzed in appropriate detail [5]. The assessment of the dynamic performance of networks in an islanded mode temporary or permanent demands the availability of proper accurate dynamic models of the grid components, which model the behavior of generators for various dynamic phenomena. Among these components are distributed generators (DG), which are dynamic systems and when connected to the power system will affect its dynamic behavior. The relevant control aspects are primary and secondary control as well as the so called grid forming capability of power electronic converters. [6] These models are expected to serve as a basis for dynamic performance studies of microgrids. They combine the achievements of research activities in modeling the physical processes of DGs together with the extensive experience of the authors in modeling and analyzing power system dynamics. Several control modes have been taken into consideration, namely, primary and secondary frequency control, voltage control as well as active and reactive power control. Depending upon the capabilities and needs of a specific hardware, the user will be able to choose from various operating modes. The control parameters of the dynamic models are adjustable, so that they can be tuned to the settings and characteristics of a specific hardware system. The models are valid for the so called RMS (root mean square) simulations with a focus on the symmetrical representation of the network. In the time frame of their validity, the current control of the converter system is modeled in a simplified way, as explained in the following sections.
Moreover, the models can be coupled with a SCADA type system which might then employ them in online dynamic security assessment of the microgrid under different control modes and foreseen operation scenarios, thereby enabling the implementation of a wide range of intelligent and optimized control strategies. II. DESCRIPTION OF MODELS These models represent the recently available technology of power electronic converters (e.g., Sustainable Automation s GRIDFORM, SMA s Sunny Island) which offer the capability of imposing a desired frequency on the electrical system, thereby forming an electrical grid, in addition to the traditional grid following capability. When only the grid following capability is available the converters are able to operate only in parallel with synchronous generators which serve to provide the frequency reference required for the control of a converter. [7] The droop concept of grid forming converter is that the converter acts as a voltage source behind an impedance (representing filter elements) and the frequency of synthesized voltage waveform is varied according to the power output of the converter thereby enabling a power sharing among parallel operating converters according to their respective droop settings. [8] In order to represent the above mentioned features in simulation models the ac side of the converter is modeled as a voltage source to represent a grid forming element as well as a current source to represent grid following element. In both representations the converter models implement primary (droop) and secondary frequency control, voltage control, as well as active and reactive power control. Large power systems are operated in a very tight margin of frequency, therefore constant frequency has been traditionally assumed while performing the dynamic analysis in the so called RMS domain. In island networks, this situation will change and more frequent activation of secondary frequency control will be experienced, owing to high share of intermittent power generation and implementation of a smart regulation of generation as well as load. Hence, the dynamic models should offer the capability of testing the performance of such control features. In the development these simulation models this need has been addressed as well. [9] The overlaying frequency command signal and power command signals are supplied to implement the so called secondary frequency control. These signals can be supplied from a central energy management system when the models are simulated online for dynamic security assessment. The schematic control structure is given in the Fig.. A. Island System: The purpose of this investigation is to simulate primary and secondary control functions of the aforementioned models in a realistic island grid consisting of primary controlled small (2-3kW) diesel generators. The electrical network consists of a medium voltage central substation which is supplied via step up transformers from diesel generators, batteries (kw) and photovoltaic sources (2kWp) operating at low voltage level. Fig.. Pset Qset Fset Vset Filters Magnitude Primary control Secondary control Voltage vector transformation Schematic control structure Connection Point Phase P Q Filter Elements v(t) The loads (total kwp) are supplied via medium voltage feeders consisting of overhead lines emanating from the central substation. The diesel generators are equipped with appropriate governor and exciter models, the overhead lines are represented by a pi model, and the loads act as constant impedances. Due to privacy restrictions of the island system s data not all the details can be published. The network structure is shown in the Fig.2 Fig. 2. Schematic network structure III. TEST SIMULATIONS Traditionally, it has been recommendable to operate all primary controlled units at equal or very close percentage droop settings in order to ensure a proportional-to-size sharing of primary reserve and to avoid dynamic interactions among individual generator units. In the context of emerging (smarter) control strategies, this concept might change, and it might be desirable (for example during peak PV production hours, when batteries are fully charged and very few synchronous generators are online) for the grid forming battery storage to considerably larger share of incoming load, resulting into a smaller droop setting.
Moreover, over an operational period, it might even be desirable to implement several different droop settings and switch from one setting to another under different operation scenarios. When different power sources have different droop settings, the steady state frequency deviation will be dictated by an equivalent droop calculated as: _ i n Ri Sb sys System Droop () Where i _ Sr i R, S are respectively the droop setting and the i r _ i MVA rating of ith unit; n is total number of units participating in droop based frequency regulation; and S is the chosen b _ sys system base which might be equal to the sum of individual ratings of the participating units. The primary control performance has been verified by employing this formula. A. Simulation Scenario This scenario demonstrates the performance of battery in primary frequency control mode and furnishes the methodology employed for validating the simulation results. The battery is modeled as a grid forming voltage source. In an actual island system, the same calculations can be performed by the central energy management system to assess the dynamic state of the system in advance and thereby smartly adjust the individual droop settings of the equipment in order to ensure frequency stability. The load flow state has been set up in a way that, initially the battery is charging (from diesel; the purpose is to run the diesel at high utilization and hence high efficiency in low demand hours), and when a load change occurs the battery starts releasing the stored energy, thereby exhibiting primary frequency control. A lower droop setting of a given generator results into a larger active power share from that generator. In order for the battery to contribute a larger power share compared to that of diesel, the battery droop has been set to 3% whereas the droop setting of diesel is %. A summary of the relevant data for equivalent droop calculation and major calculation steps are given in Table. In order to calculate the equivalent system droop the equivalent system droop is determined by: f System Droop P.44 pu.365pu.6 pu.36hz This frequency deviation can be confirmed from the simulation results depicted in Fig. 3. Table : Data for system droop calculation. Equipment identifier G G4 Bat Bat2 S r_i [MVA].37.425.75.25 R i [pu]...3.3 Calculation of frequency deviation resulting from the load step Selected System Base S b_sys [MVA] S r_i =.795 Load Step [MW].245 Load Step [pu] at S b_sys.245/.795 =.365 System DROOP [pu] at S b_sys.44 Frequency Deviation [pu] at 6Hz.6 Frequency Deviation.36 Final Frequency 6-.36 = 59.64 S r_i : Size [MVA] of ith unit R i : DROOP [pu] of ith unit based upon S r_i 6 6.5 6 59.5 59..2.3.4.5.6.7.8.9..2.3.4.5 3 25 2 59.64 Hz 5..2.3.4.5.6.7.8.9..2.3.4.5 Diesel Gen - Active Power _ G(37kVA): Red, G4(425kVA): Blue Batteries - Active Power _ Bat(75kVA):Red, Bat2(25KVA):Blue 5..795.37 System Droop.3.795.75.44 pu 4.4% at S b _ sys..795.425.3.795.25 Thus a load step of.245mw (i.e..365 pu at S b_sys ) would cause a frequency deviation of.36 Hz according to:..2.3.4.5.6.7.8.9..2.3.4.5-5 - -5 Fig. 3. The simulation results in Scenario system frequency, battery and diesel generation active power B. Simulation Scenario 2 The steady state aspect of this operation scenario represents a low load situation. Thus fewer units are dispatched, which results into a system with low primary reserve and low inertia.
In such operation scenarios a grid forming battery source is expected to provide the so called inertial response to load changes as well as the primary reserve for the load following. In the scenario under investigation the dispatched diesel units are assumed to be running close to their rating in order to operate them at maximum efficiency while any surplus (diesel generation minus load) power is being used to charge the batteries. The negative sign of active power represents battery charging. Diesel generators are operating in primary control mode with 2% droop. Batteries are in grid forming mode and take part in primary control with 3% droop. At t=.5 s, an additional load (approx. 5% of total generation) is switched in, which is subsequently shared among all the sources taking part into droop control. This results into a steady state frequency deviation which should be removed by activating secondary frequency control. To this end, at t=7 s a secondary frequency control command is given to the larger battery unit (Bat_75kVA). As a result, the battery supplies the deficit power by bringing the system frequency back to nominal value. The simulation results appear in Fig. 4. 6 6 whereas Bat2_25kVA is operating in PQ control mode. Both batteries are modeled as grid-following sources. The diesel units have 2% droop setting. The droop setting of Bat_75kVA is set at 5% in order to demonstrate a scenario where significantly different droop settings result into a stable control performance. When an additional load (approx. 5% of total generation) is switched in at t= s, the Bat_75kVA responds due to its being in droop control mode, whereas the second battery unit keeps its P and Q at pre-disturbance values. At t=6s a P command is issued to battery unit, representing a dispatch control signal from a supervisory control system. The absence of inertial response by having the grid following (instead of grid forming mode) results into a larger dynamic frequency deviation as compared with the previous scenario where the battery was modeled in grid forming mode. This scenario therefore demonstrates the performance of combined primary control with significantly different droop settings on individual participating generators, along with the impact and performance an overlying P control. Again, by employing these models, several combinations of control modes, control parameters and control signals a stability assessment of a system can be performed at planning and operation stages. The simulation results are illustrated in Fig. 5. 59 2 3 4 5 6 7 8 9 2 3 4 62 2 75 5 25 Diesel Gen - Active Power _ G(37kVA): Red, G4(425kVA): Blue 6 58 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 3 Diesel Gen - Active Power _ G(37kVA): Red, G4(425kVA): Blue 75 25 75 25-25 2 3 4 5 6 7 8 9 2 3 4-75 -25 Batteries - Active Power _ Bat(75kVA):Red, Bat2(25KVA):Blue Fig. 4. The simulation results in Scenario 2 system frequency, battery and diesel generation active power C. Simulation Scenario 3 The steady state significance of this scenario is that it represents a high load situation where the load is being supplied by diesel generators as well as by the batteries. From dynamics point of view, as enough primary reserve is available therefore one battery unit can be operated in PQ control mode thereby reducing the control burden on the unit. Thus Bat_75kVA is operating in droop control mode 2 2 3 4 5 6 7 8 9 Batteries - Active Power _ Bat(75kVA):Red, Bat2(25KVA):Blue 2 5 5 2 3 4 5 6 7 8 9 Fig. 5. Simulation results in Scenario 3 system frequency, battery and diesel generation active power
D. Simulation Scenario 4 This scenario investigates the interactions of photovoltaic sources operating in PQ control mode with the grid forming battery sources operating in droop control mode. The grid forming capability of the battery converters is essential in this case as no diesel (synchronous) units are present in the system. The batteries are operating with 3% droop setting. At t=.5s an additional load (about 5% of total generation) is switched in, the batteries follow the load which results into a steady state frequency deviation. This deviation is then removed by issuing a secondary frequency control command to the larger battery unit (Bat_75kVA) at t=2s. This action brings the frequency back to the nominal. Thus, the simulation scenario demonstrates the successful emulation of a system operation with % renewable generation. It is also noticeable that the absence of diesel generators (which impart longer response times) results into faster dynamic transitions from one state to the other. Such a behavior is expected from the sources which rely on power electronics and hence offer faster control response and hence releasing the primary reserve very quickly. Fig. 6 shows the simulation results of this scenario. 6 included in the load model in order to reflect the increased reactive load demand during reacceleration of the motor which results into a delayed voltage recovery. The motor is connected to the medium voltage substation via a transformer. The following simulation results show the voltage on the MV substation as well as on the motor terminals. With the given sizing of the generation equipment and the models with typical parameters, the slowest voltage recovery is observed in the case where only diesel generation is present. This can be attributed to limited size of available diesel generation and slow response of exciter controls. Referring to Fig. 7, the voltage recovers to 4% soon after the fault is cleared, and then slowly restores to the nominal value. Such behavior can result into a stalling of the motors and eventually to a system blackout. [pu] Voltage - MV Substation: Red, Motor Terminal: Blue.8.6.4.2.2.4.6.8.2.4 6 59.2.4.6.8.2.4.6.8 2 2.2 2.4 2.6 2.8 3 Photovoltaic - Active Power _ PV-/2/3/4 _ each rated 25kVA 8 75 7 65 6.2.4.6.8.2.4.6.8 2 2.2 2.4 2.6 2.8 3 Fig. 7. Simulation results in Scenario 5 voltage recovery performance with diesel generation On the contrary, the voltage recovery is faster in the cases where the load is being supplied by converter based (e.g., battery) generation. Such improved voltage recovery performance can be attributed to relatively large size of available converter infeed and a faster response of voltage (or reactive current) controls. Referring to Fig. 8 and Fig. 9, the voltage recovers to about 6% soon after the fault is cleared, and thereafter it reaches the nominal value quite faster as compared with the previous case. 3 2 Batteries - Active Power _ Bat(75kVA):Red, Bat2(25KVA):Blue.2.4.6.8.2.4.6.8 2 2.2 2.4 2.6 2.8 3 - Fig. 6. Simulation results in Scenario 4 system frequency, battery and photovoltaic active power E. Simulation Scenario 5 This simulation scenario demonstrates the impact of a three phase short circuit fault of ms duration on the voltage recovery performance of the island system under consideration. A kw low voltage induction motor has been [pu] Voltage - MV Substation: Red, Motor Terminal: Blue.8.6.4.2.2.4.6.8.2.4 Fig. 8. Simulation results in Scenario 5 voltage recovery performance with current source based converter infeed Thus, the availability of converters can be utilized in improving the voltage stability of island network dominated with photovoltaic and battery based generation whereby their capacity provides a voltage support function resembling to that of a STATCOM.
[pu] Voltage - MV Substation: Red, Motor Terminal: Blue.8.6.4.2.2.4.6.8.2.4 Fig. 9. Simulation results in Scenario 5 voltage recovery performance with voltage source based converter infeed IV. CONCLUSION AND FURTHER WORK A demand for the availability of dynamic models in the commercial software packages which have a widespread use in professional circles has increased with the advent of commercial popularity of smart grids. The focus of this contribution has therefore been on developing generic dynamic models which emulate the behavior of newly introduced control features in DG s as accurately as necessary for dynamic stability investigations. The models combine the graphical user interface of PSS SINCAL with the computing power of PSS NETOMAC. They are available as macros, which are linked to a network defined in PSS SINCAL. The models have been made suitable for large scale network development studies involving microgrids and island electrical systems and as such they are expected to help engineers plan, expand, and assess the performance of microgrids as an extension to their existing network models. Thereby an efficient use of already available power system databases and widely popular software can be realized. The structure of the models has been made flexible and modular so that more elaborate control and compensation loops can be easily added to the models in order to experiment novel control strategies. On the other hand the future advancements in the hardware of renewable sources can be easily reflected into the existing model structures In order to test the models performance, The simulation scenarios have been designed to demonstrate the positive impact of fast primary regulation provided by grid forming converter; the successful implementation of significantly variant droop settings; the confirmation of resulting equivalent droop; and an emulation of a system operation with % renewable generation. From the perspective of RMS dynamics on a seconds timescale, the simulation results show a reasonable behavior and therefore the models can be employed in the planning and analysis of microgrids. The simulation results demonstrate how a grid forming converter can be included into dynamic stability investigations; their behavior can be analyzed; and the control performance can be tested with different control parameters (e.g., the impact of droop settings, the impact of filter parameters on oscillations and their damping, etc) by employing the developed models. Such investigations can be performed while planning an island network as well as during actual network operation for the realization of relevant online security assessment procedures by employing the power system simulation software PSS Sincal and PSS NETOMAC. REFERENCES [] Yubing, Duan, et al. "Modelling and simulation of the microsources within a microgrid." Electrical Machines and Systems, 28. ICEMS 28. International Conference on. IEEE, 28. [2] Nemati, Mohsen, et al. "Development of generic dynamic models for distributed generators in microgrids." PowerTech (POWERTECH), 23 IEEE Grenoble. IEEE, 23. [3] Dondi, Peter, et al. "Network integration of distributed power generation." Journal of Power Sources 6. (22): -9. [4] Kroposki, Benjamin, et al. "Making microgrids work." Power and Energy Magazine, IEEE 6.3 (28): 4-53. [5] Mohamed, YA-RI, et al. "Adaptive decentralized droop controller to preserve power sharing stability of paralleled inverters in distributed generation microgrids." Power Electronics, IEEE Transactions on 23.6 (28): 286-286. [6] Piagi, Paolo, et al. "Autonomous control of microgrids." Power Engineering Society General Meeting, 26. IEEE. IEEE, 26. [7] PhotoVoltaic Power Systems & Micro-Grids, AAU http://www.et.aau.dk/digitalassets/3/389_photovoltaic-powersystems-and-micro-grids.pdf [8] Bollmann, Andrew Mark. An experimental study of frequency droop control in a low-inertia microgrid. Diss. University of Illinois, 29 [9] Feng, Da, et al. "System control of power electronics interfaced distribution generation units." Power Electronics and Motion Control Conference, 26. IPEMC 26. CES/IEEE 5th International. Vol.. IEEE, 26.