Intelligent Control of the Energy Storage System for Reliable Operation of Gas-Fired Reciprocating Engine Plants in Systems of Power Supply to Industrial Facilities

Abstract

Gas-fired reciprocating engine plants (GREPs) are widely used in power supply systems of industrial facilities, which allows for ensuring the operation of electrical loads in case of accidents in the power system. Operating experience attests to the fact that during islanded operations, GREPs are shut down by process protections or protective relays in the event of severe disturbances. This leads to complete load shedding, which is accompanied by losses and damage to industrial facilities. Severe disturbances include the following ones: large load surges on GREPs due to one of them being switched off, the group starting of electric motors, and load shedding (more than 50%) during short circuits or disconnection of process lines. Energy storage systems (ESS) have the ability to compensate for instantaneous power imbalances to prevent GREPs from switching off. The authors of this study have developed methods for intelligent control of the ESS that allow one to solve two problems: prevention of GREPs shutdowns under short-term frequency and voltage deviations as well as preservation of the calendar and cycling lifetime of battery storage (BS) of the GREP. The first method does not require performing the calculation of adjustments of control actions for active and reactive power on the ESS online but rather determines them by the value of frequency deviations and the voltage sag configuration, which greatly simplifies the system of automatic control of the ESS. The second method, which consists in dividing the steady-state power/frequency characteristic into sections with different droops that are chosen depending on the current load of the ESS and the battery state of charge, and offsetting it according to a specified pattern, allows for preventing the premature loss of power capacity of the ESS BS.

1. Introduction

Many countries are phasing in long-term action plans to achieve carbon neutrality by 2050. This is possible either by reducing carbon dioxide emissions to zero or by maintaining a balance between emissions and removals, including through offsetting [1].

Therefore, in the electricity sector, the use of non-renewable energy sources is decreasing, and the construction of generating facilities based on renewable energy sources (RES) is increasing [2,3]. The great bulk of new generation capacity additions come from wind and solar energy, which in recent years have exceeded the volume of capacity additions from conventional thermal, nuclear, and hydropower plants worldwide [4].

A distributed, low-carbon energy infrastructure, equipped with digital devices and systems, allows the most efficient use of all available types of primary energy resources [5,6]. An important precondition is to meet the needs of society for electricity, heating, and cooling in the necessary amounts and at affordable prices, as well as to ensure the reliability, availability, and safety of energy supply.

According to estimates by the International Energy Agency, decentralized energy can provide up to 75% of new consumer connections, as well as cover all of the needs to increase the capacity of electricity consumption by existing consumers in the future until 2030.

In order to minimize the negative impact on the environment and reduce greenhouse gas emissions, industrial facilities implement measures to utilize secondary energy resources [7]. Secondary energy resources include blast furnace gas, converter gas, mine gas, biogas from wastewater treatment plants, etc. Gas-fired reciprocating engine plants (GREP) of various manufacturers of a wide range of capacities have become widely used in this sector of the electric power industry [8,9].

Another factor encouraging industrial facilities to commission GREPs is the need to ensure reliable power supply and power quality at the buses of critical electrical loads involved in the continuous technological process [10,11].

Industrial facilities are usually powered by distribution grids, but it is possible for them to operate in an isolated state [12,13]. Additionally, the power supply system of an industrial facility may, for various reasons, be designed for islanded operation, with the load supplied by one or more GREPs. In these cases, the maintenance of active and reactive power balances must be provided entirely by GREPs [14].

Operating experience shows that under isolated or islanded operation, GREPs are shut down by protective relays (PR) during sudden load surges/dumps, which are caused either by technological process features or emergency disturbances. GREP shutdowns are related to their design features, the algorithms of automatic control systems (ACS) set, and the parameters of PR settings [15,16].

GREP shutdowns in most cases lead to disconnections of all electrical loads with the corresponding damage from emergency shutdowns and losses from lower manufacturing output by industrial facilities [17].

To solve a wide range of problems in the electric power industry, energy storage systems (ESS) have been widely used in recent years [18,19,20]. Given the high speed of the inverters, ESSs can effectively compensate for instantaneous imbalances of active and reactive power when they occur, preventing GREPs from shutting down.

An important issue in ensuring the reliable operation of the ESS together with the GREP is to maintain its state of charge (SoC) at an optimal level. Without an appropriate control algorithm, the ESS cannot perform this function to ensure the regulation of parameters of operating conditions within a given range, because the energy capacity of the battery storage (BS) of the ESS is chosen relatively small to minimize its cost [21,22].

The purpose of this article is to present the method developed by the authors for independent control of the active and reactive power of the ESS to prevent shutdowns of GREPs during short-term frequency and voltage deviations in isolated or islanded operation of the power supply system of an industrial facility. Furthermore, we present the method developed by us for maintaining the SoC of the ESS, which consists in segmenting the steady-state power frequency characteristic into individual sections with different droops and then offsetting it according to a specified pattern. This is necessary to prevent GREP shutdowns and ensure reliable power supply to critical electrical loads involved in the continuous technological process.

2. Overview of Causes of GREP Shutdowns and Ways to Prevent Them

As a rule, the maximum permissible short-circuit (SC) clearing time in power supply systems should not exceed 0.15 s. In the technical documentation that is shipped with GREPs, one of the manufacturers specified the following parameters for settings of protective relays whose action shuts down GREPs: voltage deviations in three phases above 110% or below 90% of U_nom with a time delay of 0.2 s. Other manufacturers have the time delays of protective relays at a voltage drop ranging from 0.5 to 5 s, however, at the operation of backup protections, it is impossible to prevent shutdowns of GREPs [9].

In short-circuit/automatic reclosing (AR) cycles as well as when the automatic transfer switches (ATS) are triggered, which are accompanied by the self-starting of electric motors, short-term voltage drops are observed [23]. In the event of the above voltage dips and interruptions, GREPs are normally shut down by PRs.

As per the technical documentation supplied with GREPs by one manufacturer, parameters of adjustment of the GREP PR by frequency such that they lead to GREP shutdowns are as follows: frequency deviations above 51.5 Hz or below 49 Hz with a time delay of 0.2 s. Other manufacturers have GREPs shut down when the frequency rises above 55 Hz with a time delay of 4 s or when it falls below 47.5 Hz with a time delay of 2 s.

Short-term frequency and/or voltage deviations in the isolated or island operation of the power supply system are observed quite often. They can occur as a result of short circuits. Additionally, they can occur in certain circuit diagram configurations and operating conditions. For example, during the direct start of large electric motors or groups of electric motors of a single process line, as well as during disconnection of large electrical loads or their groups [24].

A design feature of the GREP is that it requires a greater air supply for normal operation: the more power it builds up the greater the air supply is required. When the shaft speed is reduced below the minimum allowable value the cylinders fail to provide the pressure necessary to ignite the air-and-fuel mixture in the compression phase.

Air is supplied to the GREP through a compressor, which is rotated by a compressor turbine, and its working medium is the exhaust gases from the GREP. At the load surge, the GREP ACS, due to a decrease in the shaft speed, increases the fuel supply, the GREP power goes up, and the amount of exhaust gases increases. In addition, the compressor and compressor turbine speed increases smoothly, ensuring the supply of larger amounts of air to the GREP [25].

The high moment of inertia of the compressor turbine/compressor system prevents rapid changes in the amount of supplied air, so the transition of the GREP to a higher power output state is delayed (about 1 s) relative to the electrical load surge. Delayed air supply during load surges leads to overheating of the GREP, so to prevent damage to the GREP, they are shut down by PRs [26].

It is important to note that the setpoints of GREP PRs set by their manufacturers are not subject to change without approval during the entire warranty period. If the owner of the GREP unilaterally changes the setpoints, the warranty obligations will be unilaterally canceled by the manufacturer.

The above approaches of GREP manufacturers to the choice of operation parameters of PRs lead to a narrowing down of the range of permissible operating conditions and GREP shutdowns, given the correct actions of PRs in the power supply system of an industrial facility. This creates emergency conditions for continuous technological processes.

Therefore, it is necessary to find such engineering solutions, the implementation of which would allow one to significantly reduce the number of GREP shutdowns under disturbances with short-term frequency and/or voltage deviations.

Let us list some examples of auxiliary engineering measures that should be implemented in the power supply system of an industrial facility:

• Maintaining the good working condition of electrical equipment to minimize the number of short circuits;
• Changing the network topology opening of sectional and bus coupling circuit breakers to minimize short circuit currents and the depth of voltage dips during short circuits [27,28];
• Bringing electrical equipment up to date: replacement of high-voltage circuit breakers with modern ones with a shorter intrinsic time of tripping in order to minimize the short-circuit clearance time, installation of current-limiting devices or isolation transformers to reduce the depth of voltage dips during short-circuits;
• Upgrading protective relays (use of fast-acting main and backup protections) and network automation (use of fast-acting ARs and ATSs);
• Using soft starters or variable frequency drives on large electric motors;
• Ensuring consecutive starts in groups of electric motors with voltage control on GREP buses or installing dynamic voltage compensators that provide reactive power injection to minimize the depth and duration of voltage dips.

However, the implementation of a large number of measures at an existing industrial facility can take quite a long time, and, moreover, will require significantly larger financial outlays.

The most effective solution, which can be implemented quickly, is the installation of an ESS. This will compensate for momentary imbalances of active and reactive power that cause short-term frequency/voltage fluctuations and prevent GREPs from shutting down.

3. Overview of Ways to Maintain the State of Charge of the ESS

The following operating conditions of the ESS are known, the choice of which is conditioned by the tasks to be solved: U/f = const, PQ-control, operation with steady-state power frequency, and/or power/voltage characteristic [29,30].

In the U/f = const (voltage source) mode, the ESS maintains the frequency and voltage in the power supply system regardless of the current values of active and reactive power generation by GREPs, acting as the master inverter. The ESS ACS sets the amount of active and reactive power generation necessary to maintain the frequency and voltage within the specified setpoints, thus implementing overall primary regulation.

In the PQ-control mode (current source), active and reactive power generation by the ESS is kept constant if the frequency and voltage are within the range of permissible conditions. Consequently, the ESS is not involved in the overall primary regulation. P and Q setpoints are set either by the ESS ACS or by the power system ACS, which allows the ESS to participate in the secondary frequency and voltage regulation [31].

If the ESS operates with a steady-state power frequency and/or power/voltage characteristic, it facilitates frequency/voltage regulation conditions in the isolated or islanded operation of the power system for other GREPs by participating in secondary frequency/voltage regulation [32].

Regardless of the chosen operating conditions of the ESS, an important aspect of its operation is to keep the SoC at an optimal level. At the extreme values of the SoC (0% or 100%), there are no possibilities for the ESS to participate in regulating the parameters of operating conditions. The studies [33,34] consider various ways to maintain the SoC of the ESS.

The first group of them includes methods for ESSs that implement the function of the master inverter, in which the dedicated GREP for maintaining SoC receives information about the current value of SoC through the data link. This information is received by the ACS of the GREP, and the power of the GREP is set in proportion to the deviation of SoC from the specified value. This approach makes it possible to ensure high accuracy in maintaining the mean value of SoC, with the relative simplicity of implementing the algorithm in the ESS ACS.

The disadvantages of this method include the fact that the capacity of the ESS inverter must be designed for 100% of the load in the power supply system. In addition, it is necessary to create a data link from the ESS to GREP, and its damage leads to the SoC value exceeding the permissible range, as well as requires reconfiguration of the GREP ACS for interaction with the ESS.

The second group includes the methods that do not use the data link between the ESS ACS and the GREP ACS and are based on the application of the steady-state power frequency characteristic. In [35,36,37] it is proposed to change the slope of the steady-state power frequency characteristic of the ESS according to a specified pattern. In this case, in the discharge/charge processes, the ESS will produce/consume more or less active power [38,39,40,41]. This approach is effective for aligning SoCs that operate in parallel.

In [42,43,44,45] it is proposed to offset the steady-state power frequency characteristic of the ESS along the frequency axis, depending on the SoC, while the ESS discharge/charge processes will change the intensity of the response to frequency changes. This approach has high reliability in maintaining the specified SoC value, provides the possibility of parallel operation of several ESSs with the same SoC values, and does not require reconfiguration of the GREP ACS.

The disadvantages of this method include the fact that the ESS is not involved in the overall primary frequency regulation but only in the secondary frequency regulation. In this case, the SoC value is maintained within a permissible range rather than near the specified value. At the same time, the ESS ACS has a more complex structure and algorithms.

4. Key Defining Features of Circuit Diagram Configurations and Operating Conditions of GREP Operation and the Choice of ESS Parameters

Isolated or islanded operation of the power supply system of an industrial facility with GREPs under various circuit diagram configurations and operating conditions has markedly distinct features. All circuit diagram configurations and operating conditions can in principle be divided into two groups:

• At the beginning of the transient, the voltage is undervoltage due to a short circuit or connection of a group of motors with slips much larger than normal. Self-starting or group starting of electric motors in many cases require additional reactive power generation, which should be taken into account when choosing parameters for the ESS;
• At the beginning of the transient, the voltage is close to the nominal value, so there are no conditions for braking the motors. If a shortfall of active power occurs and it is necessary to increase the frequency to the nominal value while preventing the shutting down of the GREP PRs, then under-frequency load shedding (UFLS) devices should be used and, as a supplement, the ESS should be used.

The first group includes circuit diagram configurations and operating conditions arising in the case of triggering of ATS devices with a dead time of 2–3 s. Maintaining the reactive power balance by ESS is economically not feasible. It is necessary to apply control actions (CA) to load shedding (LS) and after normalization of operating parameters by the action of special automatics, which is to ensure successive starts of small groups of electric motors.

The second group includes circuit diagram configurations and operating conditions occurring when one or more of the GREPs that are switched on are suddenly switched off. If the total available capacity of the GREPs remaining in operation, taking into account their overload capacity, is insufficient to maintain the balance of active power, it is necessary to implement LS CAs:

• With the use of UFLS devices that have a frequency setpoint higher than f_min of the GREP PRs. In this case, the LS value must not be less than the power of the largest of the GREPs switched on, if their powers are different. Experience attests to the fact that implementing LS by the action of UFLS devices leads to the need to disconnect a larger amount of load, which proves critical for industrial facilities. This is due to the fact that UFLS devices have a time delay of 150–300 ms, and with a high rate of frequency reduction during islanded operation, their speed may not be enough;
• With the use of a special automatic load shedding system (SALSS), which acts when one or more GREPs are switched off and records the actual value of the disconnected generated active power using a device for monitoring prior operating conditions. The implementation of the LS CA should be carried out when a specified frequency setpoint is reached.

Let us consider a simplified single-line diagram of the power supply system of an industrial facility with several GREPs as shown in Figure 1.

In Figure 1, the infinite power bus (IPB) shows all external sources of equivalent unchanged electromotive force (EMF) behind a resistance corresponding to the short-circuit power. Electricity consumption in the power supply system of an industrial facility—20 MW, and the total capacity of the GREP—10 MW. Power consumed by electrical loads as percentages of the total value: synchronous motors—10%, asynchronous motors (AM)—62%, static load—28%. The parameters of the equivalent AMs differ: M_stat/M_nom = 0.2 (for AMs accounting for 20% of their total consumption), 0.4 (70%), and 0.8 (10%), respectively. The total power of the operating GREPs in the transient calculations varied. The CA is applied to the ESS connected to the GREP buses at node A (Figure 1).

GREPs are equipped with standard synchronous generators. The difference lies in the use of a gas-fired reciprocating engine to drive the synchronous generator, which differs from the steam or gas turbine in several key ways. The difference is in the maneuvering characteristics of the GREP: the allowable load surge value and the time between successive load surges. Modern gas-fired reciprocating engines are equipped with a turbo-supercharging system, which provides the necessary amount of air into the combustion chambers when the load on the synchronous generator changes. Air is supplied to the gas-fired reciprocating engine through a compressor, which is rotated by a compressor turbine, and its working medium is the exhaust gases from the GREP. The high moment of inertia of the compressor turbine/compressor system prevents rapid changes in the amount of turbo-supercharging, so the transition of the GREP to a higher power output state occurs with a delay (at least 1–2 s) relative to the electrical load surge. If the load surge is greater than the allowable value (at an initial load of 60% it is no more than 10–15%) GREP will be switched off either by process protection or by synchronous generator frequency reduction protective relays. To prevent GREP from switching off during short-term frequency and voltage deviations in islanded operation of the power supply system, the ESS is used.

Figure 2 shows the results of calculations of transients when one of the four GREPs is switched off by the process protection (without short-circuit) P_{GREP nom} = 2.5 MW. Additionally, the total load in the power supply system is P_load = 9.2 MW in the islanded state, with losses factored in. Before the shutdown, this GREP was generating about 2.5 MW, while the other three GREPs were generating 2.24 MW each.

If one chooses not to apply an LS CA, then the power supply system sets f = 46.7 Hz (Figure 2a) with the frequency regulation droop close to zero (the available power of the switched-on GREPs is equal to 110%). If in the case of disconnection of one GREP and frequency decreases down to f = 48 Hz by the action of the SALSS the CA is applied in the amount of 2.5 MW step-wise within 0.1 s, the frequency is restored up to f = 50.0 Hz (Figure 2b). The action of the SALSS is a response to shutting down the GREP before the frequency reduction allows one to minimize the amount of LS, thus preserving the power supply to a greater number of electrical loads.

If the shutting down of a single GREP occurs as a result of a short-circuit, then during the transient process the remaining GREPs may be shut down by the action of the PRs because of a short-term voltage drop, if their setpoints are close to U_nom and the time delay is less than the allowable short-circuit duration.

It is not economically feasible to apply CAs to the ESS to increase the voltage at the GREP buses until the short-circuit is cleared, because significant amounts of reactive power will be required. To reduce short-circuit clearing time it is necessary to reconstruct the protective relays, with the installation of high-speed protections, both in the distribution grid and in the power supply system of an industrial facility. It is often effective to install current-limiting devices to reduce the depth of voltage dips at the GREP buses during short circuits in the electrical distribution network.

Voltage control in power systems is realized by two methods:

• Regulation proportional to deviations of the current voltage value from the specified value (with possible use of signals, including those proportional to time derivatives of operating conditions parameters);
• Discrete boost, with two levels of reactive power output: normal and boosted.

The magnitude of the control error is much smaller. However, the connection to the ESS GREP buses with the ACS, in which the control algorithm is implemented, may require them to be reconfigured in a joint fashion. This is important to avoid oscillatory instability under different circuit diagram configurations and operating conditions, as well as to ensure good damping of oscillations in case of disturbances. High accuracy is not required to solve the problem of voltage maintenance with the aid of ESS at GREP buses.

The reactive power control of the ESS as per the boosting principle is the simplest and most effective solution. In this case, the ESS does not output reactive power as normal but delivers the maximum amount of reactive power when the voltage drops below the specified setpoint. The boosting mode can be set at different voltage values by the value of reactive current I_CA or reactive power Q_CA.

If the parameters of operating conditions during a transient are close to normal values, it is necessary to reduce the CA on the ESS with respect to I_CA. This is to be completed at the end of the self-start of electric motors due to a significant decrease in the value of the current consumed by them. Otherwise, a voltage increase above U_max is possible, which will cause the GREP to shut down by the overvoltage protection. If, on the other hand, in the ESS ACS it is not I_CA but Q_CA that is set, then when the voltage rises, I_CA will decrease without additional actions. In the calculations of the transients given in this article, this is the method adopted for applying CAs [46].

One should not withdraw the CA on the ESS as long as there is a possibility of repeated braking of the electric motors with a corresponding voltage drop. On the other hand, when most motors have reached normal speed, the excessive reactive power output causes the voltage to rise. Analysis of the results of transient calculations showed that the optimum voltage of the CA withdrawal is approximately equal to the nominal voltage of the network to which the GREPs are connected. In the calculations presented in the article, the bus voltage at which the CA acting on the ESS is withdrawn is 6 kV.

The specifications of the ESS should be chosen with the following in mind:

• ESS should contribute to the normalization of parameters of operating conditions after the elimination of an emergency disturbance to prevent GREP shutdowns, as well as to ensure the successful self-starting of electric motors in the power supply system of an industrial facility;
• Considering the effect of the ESS on the operating conditions of the entire power supply system, in order to minimize its power and energy intensity it is necessary to simultaneously implement additional engineering measures, a list of which is given above;
• A check of the possibility of disconnection of GREPs by PRs must be performed for all types of emergency disturbances, including remote ones (for example, non-synchronous AR on power supply lines; short circuits at the buses of adjacent substations; load surges/dumps in the power supply system, etc.), with a high probability, and not only for the most severe disturbance (nearby three-phase short circuit);
• A check of the possibility of disconnection of the GREP by PRs should be performed under asymmetric emergency disturbances as well as under post-emergency voltage asymmetry.

Requirements for the formation and application of CA to the ESS to prevent shutdowns of the GREP:

• It is necessary to introduce the CA quickly in a way appropriate to the circuit diagram configurations and operating conditions (implementation time not more than 30 ms) that indicate the occurrence of such operating conditions where it is possible for the GREP to be shut down by the PRs;
• CA introduction until the multiphase short-circuit is cleared is of no effect, especially for nearby short-circuits;
• Taking into account the nature and parameters of the transient, it may be necessary to reapply the CA to the ESS; in such case starting up is triggered by a deviation of the monitored parameter, for example, by a voltage drop;
• The output of active/reactive power of the ESS during the transient should be performed at the current frequency value (ESS frequency control is what is being controlled) to eliminate an increase in complexity of the ESS ACS [47].

Calculations of transients were performed with the aid of the MUSTANG-90 software package (Russia) for various circuit diagram configurations and operating conditions in the power supply system of an industrial facility with GREPs. The calculations attested to the efficacy of the methods developed by the authors.

5. Method for Independent Control of Active and Reactive Power of the ESS

Preventing the shutdowns of GREPs by PRs during short-term frequency deviations is especially relevant in isolated or islanded operations of the power supply system. This is due to the small values of the mechanical constant of inertia of the GREP, the lag of air supply to the GREP, as well as setpoints of protections with respect to f_min and f_max, close to f_nom = 50 Hz.

Let us list the features of the method developed by the authors for independent control of active and reactive power of the ESS, aimed at stabilizing the frequency in the power supply system of an industrial facility with GREPs:

• When the balance of active power in the power supply system is disturbed and is accompanied by a frequency deviation from f_nom, CAs are applied to the ESS. CA adjustments are chosen so that the frequency does not go beyond a range narrower than the range of frequency values, as specified by the setpoints of GREP PRs.
• The choice of CA adjustments applied to the ESS is specified differently for the initial stage of the transient (emergency disturbance) and for the intermediate points of the transient at unsteady parameters of operating conditions.
• At the initial stage of the transient, when a rapid change in the GREP active power output (∆P_GREP) is observed, it is necessary to ensure with minimum delay the application of the CA to the ESS with respect to the active power (CA_ESS-P) in the amount of ∆P_ESS ≈ ∆P_GREP. It is difficult to perform a direct measurement of P_GREP in a transient because switching in the power supply system causes rapid changes in the subtransient and aperiodic components of currents as well as their harmonic components due to individual phase imbalance. Measuring the correct value of P_GREP can lead to performance degradation in the event of applying CA_ESS-P.
• The application of CA_ESS-P at the initial stage of the transient in the amount of ∆P_ESS ≈ ∆P_GREP should be carried out in proportion to the frequency derivative given islanded operation of the power supply system:

  ∆P_ESS = – K_df ∙ (df/dt)₀,

  (1)

  where K_df is the coefficient of proportionality. Calculation of K_df is performed when loads of different magnitudes are connected to GREPs, recording the value of load increment on the GREP (∆P₀) and the frequency derivative (df/dt)₀ (Figure 3).
  The graph of the relationship ∆P_GREP.₀ = φ(df/dt) in the power supply system with GREPs (Figure 3) is obtained for the considered power supply system of the industrial facility, the process flow diagram of which is shown in Figure 1. For some other power supply systems in islanded operation, such a graph will have to be obtained on a case-by-case basis, taking into account the GREP per-unit capacity based on contingency calculations of transients. Figure 3 corresponds to an emergency disturbance with the shutdown of a single GREP.
• In the transient with load surge at GREPs, the unambiguous correspondence between ∆P_GREP and df/dt is broken due to the influence of dynamic power components, so the CA_ESS-P should be implemented differently. In addition, the boundary values of frequency f₁, f₂ are set to implement frequency control in the power supply system, in addition to f_min, f_max, which characterize the setpoints of the GREP PRs, as follows:

  f_min < f₁ < f < f₂ < f_max.

  (2)
• In order to control the transient, the ESS ACS monitors the frequency in the power supply system at equal time intervals T. If f < f₁, it outputs a CA to increase P_ESS so as to reduce the load at the GREP and increase the frequency. If f > f₂, it outputs a CA to reduce P_ESS. Optimal values of intervals T and ESS power steps (+∆P_ESS, −∆P_ESS) are chosen based on the results of transient calculations.
• It is possible to increase the efficiency of ESS control if the values of power steps (+∆P_ESS, −∆P_ESS) are changed in proportion to the current values of the frequency derivative in the power supply system (df/dt)t.
• After completion of the controlled part of the transient, when there is no need to change ∆P_ESS within 6–10 s, i.e., f₁ < f < f₂, the ESS ACS smoothly (or in several steps) zeros out the value of CA_ESS-P. After that, the ESS SoC recovery algorithm is launched.
• Since in isolated and islanded operation during surges of active load at the GREP, the frequency in the power supply system decreases for a short time, when choosing the CA_ESS-P adjustment it is necessary to factor in the value of the power frequency characteristic to reduce the amount of active power imbalance.

Let us consider the results of calculations of transients in which the GREPs were not shut down by the PRs in the islanded operating state, due to the application of a CA to the ESS with respect to active and reactive power.

The following setpoints of the PRs with respect to the frequency of the GREP are set: f_min = 47.5 Hz with a time delay of 2 s, f_max = 55 Hz with a time delay of 4 s. To control the power of the ESS the boundaries of the range of permissible values are set as follows: f₁ = 49, f₂ = 51 Hz. It is taken into account that violation of the condition f₁ < f < f₂ is permissible for a time shorter than the time delay of the GREP PRs. T = 300 ms was assumed in the calculations (changes in the range of 200–500 ms have no effect on the calculation results), and ∆P_ESS steps are +10% and −20% of the preceding value of active power.

In the calculations of transients, it was assumed that at the initial stage of the transient the control of the active power of the ESS is introduced with a delay of 60 ms (triggering of the starting element; calculation of (df/dt)₀ and ∆P_ESS; taking over the CA by the ESS ACS), subsequent control with respect to frequency smaller than f₁ or larger than f₂ is performed with a delay of 30 ms.

Let us consider examples of CA_ESS-P,Q application, where in fundamentally different transients with severe disturbances, the method developed by the authors for ESS control allows one to prevent GREPs shutdown while having the same CA adjustment. Consequently, an important advantage of this method is the possibility to determine the CA adjustment as applied to the ESS by the magnitude of frequency deviations, without the need for online calculations.

1. Load surge at GREPs during group starting of electric motors.

Figure 4 shows the calculation of the electromechanical transient during islanded operation of the power supply system operating under minimum load conditions Under such conditions, a group of electric motors with a total capacity of 0.94 MW (37.5% of P_GREPnom) must be started from one running GREP. Taking into account the factory-set setpoints of the GREP protective relay for frequency reduction, without employing the ESS, it is possible to successfully implement the direct starting of electric motors with total power not more than 0.25 MW (10% of P_GREPnom). Since the process flow of the industrial facility requires performing simultaneous direct starting of electric motors with a total capacity of 0.94 MW (there are no soft starters and variable frequency drive for electric motors), it can be implemented only through the use of the ESS. “Sawtooth” characteristics are due solely to the active power control algorithm of the ESS. Given the speed of the ESS inverter, which implements the CA in about 10 ms, the transient graph (one-second range) shows it as “sawtooth” characteristics.

In the baseline mode, the load of the GREPs is 30% of P_GREPnom. The total nominal power of electric motors during group starting is 37.5% of P_GREPnom. The results of the calculation of the transient given a group start of electric motors by a single GREP, with the application of the CA to the ESS with respect to P and Q according to the method developed herein, are shown in Figure 4.

Analysis of Figure 4 shows that the application of the CA to the ESS with respect to P and Q makes it possible to provide successful group starting of electric motors of large total power in islanded operation without switching on additional GREPs. Such conditions can occur at minimum power consumption at an industrial facility when one of the four GREPs remains in operation.

2. Shedding 95% of the load from the GREP at nominal initial load.

Under the baseline conditions, the load of the GREP is 100% of P_GREPnom. The load shedding from the GREP is 95% of P_GREPnom. In this case, there is a transient process with a significant short-term increase in frequency, at the beginning of which the CA are applied to the ESS in the amount of active power consumption equal to the value of load shedding ∆P_ESS = –2.3 MW (Figure 5).

If large power drops occur quite often at an industrial facility, then in the ACS of the ESS under normal operating conditions an algorithm must be implemented to maintain such a value of the ESS SoC that would allow operating with the power +∆P_GREPnom. Such operating conditions can occur, for example, when a large process line is shut down.

The analysis of Figure 5 shows that the application of the CA to the ESS with respect to P, in accordance with the method developed, makes it possible to prevent the GREP shutdown in the case of shedding 95% of P_GREPnom, accompanied by a short-term increase in frequency under islanded operating conditions. This makes it possible to provide a reliable power supply to the remaining electrical loads in operation [48,49].

3. Three-phase short circuit in the power supply system with GREPs.

In the baseline mode, the load of the GREP is 2.24 MW. As a result of a three-phase short circuit in the power supply system of an industrial facility, the power supply to electrical loads with a capacity of 27% of P_GREPnom is disrupted. Next, in 0.38 s, the power supply to the disconnected electrical loads is restored by the action of the ATS device. The transient during this emergency disturbance is shown in Figure 6.

The analysis of Figure 6 shows that this transient differs from the other two (Figure 4 and Figure 5) in that the use of a higher-power ESS increases the efficacy of control. There are no general patterns in this case, because at different depths of voltage dips, short-circuit duration, and different load parameters, the optimal adjustments of the CA applied to the ESS will be different. The choice of the adjustments of CAs applied to the ESS with respect to P and Q should be made on the basis of the results of the analysis of short-circuit probabilities with the use of standard means of increasing the dynamic stability of the load.

Establishing technical specifications for the ESS should be based on the results of calculations of transients. The analysis of the transients shown in Figure 4, Figure 5 and Figure 6 made it possible to establish the requirements for the ESS, which are shown in

Table 1. Specifications to be met by the ESS in the power supply system with GREPs.

Figure	Unom, κV	Imax, A	Pmax, MW	Energy Intensity ESS, MW∙s
Figure 4	6.3	220	+2.1	+25
Figure 5	6.3	220	–2.4	–12
Figure 6	6.3	130	+1.0	+10

(p > 0 when the ESS power is delivered).

The presented method allows one to minimize the requirements for power and energy intensity of the ESS BS, and hence the cost of the ESS. This was made possible by maximizing the use of the GREP overload capacity, since the implementation of the CA applied to the ESS is chosen so as to take into account the GREP protective relay setpoints with respect to frequency.

It is possible in principle to increase the voltage at the buses of GREPs to prevent them from shutting down by applying a CA to the ESS with respect to Q. However, as long as the multiphase short-circuit is not cleared, it is possible to increase the voltage only when having the power Q_CA significantly higher than the power of the short-circuit. In addition, the output of a significant amount of reactive power from the ESS at the moment of a short circuit can significantly increase the value of the short-circuit current. This may be unacceptable due to the insufficient breaking capacity of high-voltage circuit breakers or will lead to shortening their electrical lifespan at an accelerated pace.

Consequently, the voltage control by applying CAs to the ESS with respect to Q is effective when implemented immediately after clearing the short circuit. If the short-circuit durations are such that the GREP PR used to lower voltage can trip during the short circuit, then the reactive power output from the ESS is of no effect. Therefore, the reduction of the short-circuit clearing time by replacing the PRs and high-voltage circuit breakers is the only engineering solution. If the time delay of the protection by voltage drop at the GREP is 0.5–5 s, which is triggered in the process of self-starting electric motors in the load, then applying the CA to the ESS with respect to Q is feasible.

Figure 7 shows the results of calculations of transients at a three-phase short-circuit at node B when four GREPs are in operation (Figure 1). For comparison, four calculations of transients are given: without applying the CA, as well as with applying the CA to the ESS with respect to Q after clearing the short circuit. The value of Q_CA = 10 MVar, and the duration of the CA varies. In Figure 7: ΔU—increase in the voltage at GREP buses in case of applying the CA to the ESS; Δt—the time during which the voltage at GREP buses is below U_min = 5.67 kV (90% of U_GREPnom = 6.3 kV); blue arrows indicate the time during which the voltage is below U_min.

In the method developed, the control of the ESS with respect to Q is implemented as per the principle of boosting the excitation of generators: ∆Q_ESS > 0 if U < U_min. The values of these parameters are chosen on the basis of preliminary results of transient calculations. In the transients shown in Figure 7, the CA applied to the ESS with respect to Q was introduced in 30 ms after clearing the short circuit, and the response time of the ESS was assumed to be 5 ms.

The analysis of Figure 7 shows that the duration of the CA applied to the ESS has no significant effect on Δt if it is not less than the duration of self-starting of the main electric motors in the power supply system. At the same time, the application of the CA to the ESS with respect to Q allows one to prevent GREP shutdowns.

The magnitude of the instantaneous voltage rises at the moment of applying the CA to the ESS (ΔU) depends on the extent of the CA (Q_CA) and is close to linear for different load compositions. The decrease in time Δt is more noticeable, the greater the available power Q_CA, and the greatest effect is obtained when Q_CA/P_GREP ≈ 0.4–0.5. Parameters that have a significant effect on the value of Δt are the relative power P_GREP/P_load and the short-circuit power at GREP buses.

The method developed by the authors for independent control of active and reactive power of the ESS differs from the known ones by the fact that it does not require performing calculations of CA adjustments applied to the ESS online. We propose to determine CA adjustments applied to active power by the value of frequency deviations, which greatly simplifies the implementation of the ESS ACS. The reactive power CA is proposed to be performed in line with the principle of excitation boost, which allows not to perform a general reconfiguration of the ACS of active elements in the power supply systems. This makes it possible to prevent GREPs shutdown during short-term frequency and voltage deviations that result from severe disturbances during islanded operation of the power supply system, as well as to ensure reliable power supply of process lines from GREPs, preventing possible damages and losses from the reduced output.

6. Method for Maintaining the ESS State of Charge Depending on the Current ESS Load and SoC Values

Depending on the task being solved by the ESS, the SoC value may be different under different circuit diagram configurations and operating conditions. Let us address the issue of maintaining a given level of ESS SoC as applied to solving the problem of preventing GREP shutdowns by frequency protection during its short-term deviations in transients.

In the method developed by the authors for maintaining the SoC of the ESS operating in the PQ-control mode, the required SoC value is maintained by changing the slope of the ESS steady-state power frequency characteristic and its offset along the frequency axis depending on the SoC value. In the general case, the steady-state power frequency characteristic is described by the equation:

$$f-f_{\mathrm{nom}}=-K_{\mathrm{ESS}} \left(P_{\mathrm{ESS}}-P_{\mathrm{ESS}. 0}\right),$$

(3)

where f is the current frequency value, Hz; f_nom is the nominal frequency, Hz; K_ESS is the ESS droop, Hz/MW; P_ESS is the current active power of the ESS, MW; P_ESS.₀ is the active power of the ESS at f_nom, MW.

The K_ESS factor captures the relationship between the change in frequency and the active power output of the ESS during isolated or islanded operation of the power supply system, as a response to the change in frequency.

The first of the degrees of freedom in Equation (3) is K_ESS, which determines the slope angle of the steady-state power frequency characteristic of the ESS. By adding to Equation (3) the factor that takes into account the change in SoC, it is possible to change the K_ESS and, hence, the intensity of the ESS response to the change in frequency at different values of SoC. According to [34,44], assuming P_ESS.₀ = 0, Equation (3) for the charging and discharging modes, respectively, takes the form:

$$f-f_{\mathrm{nom}}=-K_{\mathrm{ESS}}·P_{\mathrm{ESS}}=-\frac{K_{\mathrm{ESS}. 0}}{{\mathrm{SoC}}^n}· P_{\mathrm{ESS}}, P_{\mathrm{ESS}}\ge 0,$$

(4)

$$f-f_{\mathrm{nom}}=-K_{\mathrm{ESS}}·P_{\mathrm{ESS}}=-K_{\mathrm{ESS}. 0}·{\mathrm{SoC}}^n·P_{\mathrm{ESS}}, P_{\mathrm{ESS}} \le 0,$$

(5)

where K_ESS.₀ is the droop of ESS at SoC = 1, Hz/MW, n is the exponent of the power function (n > 0).

Let us obtain the current ESS power from Equations (4) and (5):

$$P_{\mathrm{ESS}}=-\frac{f-f_{\mathrm{nom}}}{K_{\mathrm{ESS}.0}}·{\mathrm{SoC}}^n, P_{\mathrm{ESS}} \ge 0 (\mathrm{discharging}, \mathrm{if} f< f_{\mathrm{nom}}; \mathrm{S}o\mathrm{C} \downarrow ),$$

(6)

$$P_{\mathrm{ESS}}=-\frac{f-f_{\mathrm{nom}}}{K_{\mathrm{ESS}.0}}·\frac{1}{{\mathrm{SoC}}^n}, P_{\mathrm{ESS}} \le 0 (\mathrm{charging}, \mathrm{if} f> f_{\mathrm{nom}}; \mathrm{S}o\mathrm{C} \uparrow ).$$

(7)

When the frequency deviates downwards from f_nom, the ESS AC generates a power setting of P_ESS ≥ 0, and the ESS begins to discharge, delivering power to the power supply system. In this case, the available power of the ESS decreases simultaneously with SoC. When the frequency deviates upwards from f_nom, the ESS ACS generates a power setting of P_ESS ≤ 0, ensuring the ESS charge. In this case, the available power of the ESS grows simultaneously with the SoC (Figure 8).

The exponent n in Equations (6) and (7) has a significant effect on the power of the ESS, depending on SoC. This way of regulating the ESS (minimum response time) can lead to dynamic instability during short-term frequency changes given islanded operation of the power supply system. This should be taken into account when performing transient calculations in the design process. Small values of n lead to an insignificant change in the slope of the steady-state power frequency characteristic, with K_ESS.₀ ≈ K_ESS, and with its large values the slope of the steady-state characteristic changes significantly when changing the SoC value.

Figure 9 shows graphs of the sensitivity of P_ESS to variations in n in the range from 0.1 to 2 at the chosen droop with respect to frequency K_ESS.₀ = 10⁻⁶ Hz/W as a function of the value of SoC. Figure 9a shows the ESS charging mode with frequency change f − f_nom = 0.5 Hz, and Figure 9b shows the ESS discharging mode with frequency change f − f_nom = –0.5 Hz. The power of the ESS is limited by the ACS at 2.5 MW.

The analysis of Figure 9 shows that as the value of n increases, the ESS response to changes in SoC increases significantly, so it is advisable to limit the range of SoC change (set the setpoints), within which the method developed by the authors will be effective.

The second degree of freedom in Equation (3) is the frequency (f); therefore, it is possible to offset the steady-state characteristic along the frequency axis by some value Δf as a function of the SoC value. In this case, Equations (6) and (7) for the ESS power take the form:

$$P_{\mathrm{ESS}}=-\frac{\left(f-f_{\mathrm{nom}}\right)-\mathsf{\Delta }f}{K_{\mathrm{ESS}.0}}, P_{\mathrm{ESS}} \ge 0 (\mathrm{discharging}, \mathrm{if} f< f_{\mathrm{nom}}; \mathrm{S}o\mathrm{C} \downarrow ),$$

(8)

$$P_{\mathrm{ESS}}=-\frac{\left(f-f_{\mathrm{nom}}\right)+\mathsf{\Delta }f}{K_{\mathrm{ESS}.0}},{ }_{\mathrm{ESS}} \le 0 (\mathrm{charging}, \mathrm{if} f> f_{\mathrm{nom}}; \mathrm{S}o\mathrm{C} \uparrow ).$$

(9)

The value of Δf can depend linearly on the SoC, as in Equation (10), be set as a fixed value depending on the current value of the SoC, or have the form of a piecewise linear function.

Δf = m_SoC (SoC − SoC_set),

(10)

where m_SoC is the coefficient of proportionality; SoC_set is the SoC setpoint value.

When the ESS is discharged, the static frequency characteristic should gradually shift downward along the frequency axis, causing the ESS discharge power to decrease linearly, and when in the case of charging it should shift upward, causing the charge power to increase, as shown in Figure 10.

An analysis of the sensitivity of the P_ESS to changes in m_SoC is not presented in the article, but it should be noted that the maximum value of Δf should not be greater than the actual frequency deviation in the islanded operation of the power supply system. Otherwise, there may be a sharp shift of the operating point of the characteristic to the area of the opposite mode of operation of the ESS. In this case, the ESS ACS will form an incorrect command (for example, the command to charge instead of discharge) and the parameters of the operating condition will go beyond the range of permissible values, which is unacceptable.

Based on the results of the analysis of the features of the steady-state characteristic offset depending on the value of SoC, the ESS power regulation was assumed to be in accordance with Equations (11) and (12), aimed at stabilizing the frequency and maintaining SoC within a given range.

$$P_{\mathrm{ESS}}=-\frac{\left(f-f_{\mathrm{nom}}\right)-\mathsf{\Delta }f}{K_{\mathrm{ESS}.0}}·{\mathrm{SoC}}^n, P_{\mathrm{ESS}} \ge 0 (\mathrm{discharging}, \mathrm{if} f< f_{\mathrm{nom}}; \mathrm{SoC} \downarrow )$$

(11)

$$P_{\mathrm{ESS}}=-\frac{\left(f-f_{\mathrm{nom}}\right)+\mathsf{\Delta }f}{K_{\mathrm{ESS}.0}}·\frac{1}{{\mathrm{SoC}}^n},{ }_{\mathrm{ESS}} \le 0 (\mathrm{charging}, \mathrm{if} f> f_{\mathrm{nom}}; \mathrm{SoC} \uparrow )$$

(12)

For effective frequency control in the case of islanded operation of the power supply system with GREPs, the range of power variation of the ESS in both directions should be the greatest possible, which is usually achievable at SoC = 0.5 (50%). Regardless of the control algorithms specified in the ESS ACS, there is an optimal SoC range for frequency control purposes, whose boundaries it is not recommended to exceed [50].

Deep cycling of the ESS battery storage as well as regular charging to 100% reduces the calendar and cycling life of the battery storage, which is expressed in a premature loss of energy capacity [51]. Therefore, the ESS should operate within the optimal SoC range with a steady-state power frequency characteristic offset depending on the SoC value, as recommended in [43]. Additionally, it is reasonable to change the slope of the steady-state characteristic depending on the SoC in the regions of the upper and lower limit states of the ESS SoC, as shown in Figure 11.

The Δf quantity in Equations (11) and (12) is determined by a piecewise linear function:

$$\mathsf{\Delta }f=\left\{\begin{array}{c}-P_{\mathrm{ESS}}·R, \mathrm{SoC} \ge {\mathrm{SoC}}_{\max 2}, \\ f_{\max }-f_{\lim .\max }, {\mathrm{SoC}}_{\max 1}\le \mathrm{SoC} \le {\mathrm{SoC}}_{\max 2},\\ 0, {\mathrm{SoC}}_{\min 1}\le \mathrm{SoC} \le {\mathrm{SoC}}_{\max 1}, \\ f_{\min }-f_{\lim .\min }, {\mathrm{SoC}}_{\min 2}\le \mathrm{SoC} \le {\mathrm{SoC}}_{\min 1},\\ -P_{\mathrm{ESS}}·R, \mathrm{SoC} \le {\mathrm{SoC}}_{\min 2}, \end{array}\right.$$

(13)

where R—droop with respect to frequency for the GREP generator; f_max/f_min—critical maximum/minimum frequency value; f_lim._max/f_lim._min—maximum/minimum frequency offset limit.

Changing the slope angle of the steady-state characteristic in the regions of upper and lower limit states of the SoC strengthens the response of the ESS in the right direction as the SoC approaches them.

Figure 12 shows a general view of the segmented steady-state characteristic of the ESS with an offset along the frequency axis and a change in the droop depending on the SoC value (black line—steady-state characteristic of the GREP; green line—steady-state characteristic of the ESS divided into sections at SoC = 0.5; vertical section on the frequency axis—dead band of the ESS frequency control algorithm).

In the inclined sections of the characteristic (Figure 12), the power is distributed between the GREP and the ESS, while in the horizontal sections, the ESS power is maintained so as to ensure constant delivery of the maximum permissible power of the GREP, preventing its emergency shutdown. When the ESS is discharged, the steady-state characteristic gets offset upwards, and when it is charged, it gets offset downwards by Δf. When the lower or upper limit state of the SoC is reached, the slope of the steady-state power frequency characteristic of the ESS changes to increase the discharge power at high SoC values or decrease it at low SoC values. The movement of the operating points of the ESS and GREP in Figure 12 can be traced by the gray dashed lines. A detailed description of the frequency regulator can be found in [43].

The method developed by the authors for maintaining the SoC of the GREP does not require the availability of a data link between the ESS and the GREP. This method consists in dividing the steady-state power/frequency characteristic into sections with different droops that are chosen depending on the current load of the ESS and the battery state of charge and offsetting it according to a specified pattern. This makes it possible during islanded operation of a power supply system of an industrial facility with GREPs to ensure reliable operation of the ESS, preserving the calendar and cycling service life of the battery storage, as well as preventing premature loss of its power capacity.

7. Conclusions

GREPs shutdowns by process protections or protective relays under severe disturbances in the islanded operation of power supply systems of industrial facilities lead to damages and losses due to load shedding. GREPs shutdowns can be prevented by using ESSs to compensate for instantaneous imbalances of active and reactive power, accompanied by short-term frequency and/or voltage deviations during transients.

The most severe disturbances during islanded operation are large surges of load on GREPs due to one of them being switched off, the group starting of electric motors, and load shedding (more than 50%) during short circuits or disconnection of process lines.

The method developed by the authors for independent control of active and reactive power output/consumption of the ESS does not require the online calculation of CA adjustments at the ESS. We propose to determine CA adjustments applied to active power by the value of frequency deviations, which greatly simplifies the implementation of the ESS ACS. Optimal values of frequency control time intervals and CA adjustments at the ESS are chosen based on the results of calculations of electromechanical transients. We propose to perform the reactive power control in line with the principle of excitation boost, which allows one not to perform a general reconfiguration of the ACS of active elements in the power supply systems, providing good damping of oscillations in case of emergency disturbances.

The method developed by the authors for maintaining the SoC of the GREP does not require the availability of a data link between the ESS and the GREP. This method consists in dividing the steady-state power/frequency characteristic into sections with different droops that are chosen depending on the current load of the ESS and the battery state of charge and offsetting it according to a specified pattern. This helps to preserve the calendar and cycling lifetime of the ESS BS, preventing premature loss of its energy capacity.

The methods developed by the authors for intelligent ESS control allow one to prevent GREPs from switching off during islanded operation, even under severe disturbances, which is proved by the results of calculations of electromechanical transients. This makes it possible to ensure a reliable power supply of process lines at industrial facilities from GREPs, preventing possible damage from their shutdowns and losses from the reduced output.