Modern Active Voltage Control in Distribution Networks, including Distributed Generation, Using the Hardware-in-the-Loop Technique

Abstract

Voltage constraints usually place restrictions on how distributed generation (DG) can be connected to weak distribution networks. As DG capacity increases, active voltage control techniques are needed. Active approaches can greatly lower connection costs while boosting the capacity of connectable DG when used in place of the passive strategy. In this article, a modified active voltage control algorithm is used on an IEEE 33 bus system to test the robustness and reliability of the control algorithm under severe conditions. The simulations are carried out using the hardware-in-the-loop (HIL) method. Real-time simulations are used to test data transfer and the reliability of the control algorithm’s execution. The analysis is based on a three-phase symmetric power system.

1. Introduction

Power systems are increasingly reliant on distributed generation (DG). External, non-electric factors such as appropriate geographic locations of wind and solar resources influence the location of DG units. As a result, whether planned or not, they are frequently attached to distribution networks at the closest location. Furthermore, all energy producers now have access to distribution networks as a result of energy market deregulation laws that include incentives for renewable energy production [1,2]. As a result, much research has been conducted to analyze issues concerning DG market access, such as locational fixed costs [3], the long-term implications of feed-in tariffs [4], and a carbon tax strategy for system cost-effectiveness [5]. Many strategies were proposed concerning the connection of DGs to the networks. A strategy proposed a fair and equitable electricity tariff for the calculation of the distribution network usage fee [6]. Another one suggested providing DGs with an incentive to locate themselves in the most needed areas of the network [7]. Optimization techniques were utilized for the optimization of the insertion of DGs into the network [8,9]. The voltage rise effect, particularly in weak distribution networks, is the most critical factor restricting the capacity of connected DGs. Currently, this is frequently handled by connecting DGs to specialized feeders or any other possible network upgrading, such that the power flow of the entire network keeps the extreme voltages (maximum generation/minimum loading) inside network boundaries, as they are traditionally connected in a “fit and forget” style [10]. Because the traditional distribution network is passive, this inefficient mode of operation is possible. Certainly, this significantly raises the cost of the DG connection because its control is solely dependent on the primary substation and ignores DG capabilities in network voltage regulation. In the passive configuration, if the increase in available low running cost DG units threatens network safety, the distribution network operator (DNO) must disconnect some DG parks, wasting some installed capacity. Instead, coordinated active control methods that make advantage of DG features can significantly increase the connected capacity [11].

Several coordinated control algorithms utilizing both the capabilities of DG units and the network were published [12,13]. It has been aimed to mitigate high ramp-rate variables in the network [14], while other articles implemented coordinated control strategies to minimize voltage deviations [15]. A hierarchical coordinated voltage correction scheme was proposed. This scheme consists of two layers of control: a regional layer based on the local measurement information and a feeder line coordination layer based on a communication network [16]. The charging and discharging of a battery-based storage system were controlled [17]. Control is divided into two layers: centralized for the entire network and decentralized in each DG [18]. An optimization problem is constructed and controlled by the DG output to make a balance between voltage regulation and reactive power transfer during islanding [19]. Moreover, these control strategies are accompanied by marginal incentives for DG owners to connect them at the optimized locations [20] and with incentives for more applications of customers’ demand response [21]. An analysis of the latest studies on local, centralized, distributed, and decentralized voltage control algorithms and a comparison between them were introduced [22]. Moreover, the frequency was added to an adaptive coordinated control scheme [23]. Model predictive control techniques [24,25,26,27] and fuzzy control techniques [28] were utilized for time-saving during control. Optimization techniques were utilized to reach the optimal decisions for coordinated control [29,30]. Distributed control algorithms were applied with no need for communication infrastructure [31,32]. The symmetry of the voltage profile over the network is considered a major concern for electric power utilities. Thus, operators apply strict constraints to keep the voltage within acceptable boundaries.

This article follows the trend of a specific rule-based algorithm that selects the quantity to be controlled according to certain conditions and priorities [10,33].

In this article, the results of real-time simulations, including the hardware-in-the-loop (HIL) technique, of an optimal improvement of a control algorithm are demonstrated. The modified algorithm is applied to the IEEE 33 bus radial distribution network with arbitrary DG units. The goals of this application are to validate the coordinated control algorithm for more general networks, as well as to inspect new conditions and scenarios.

This article is organized as follows: Section 2 shows the tools, techniques, and coordinated control algorithm, Section 3 delves into real-time simulations of the algorithm being applied to the IEEE 33 bus radial distribution network in a variety of circumstances, and Section 4 wraps up the conclusions.

2. Coordinated Control Algorithm

The DNO is equipped with a number of tools to manage the network state. They are outlined beneath, along with the benefits and drawbacks of each. They are arranged according to increasing price:

• On-load Tap Changer (OLTC): Utilizing the OLTC of the primary transformer is the simplest and cheapest way to regulate the network voltages, with no cost of high infrastructure for communication links or huge power losses through the network. However, it is inefficient in some cases, where the difference between the extreme (maximum and minimum) voltage of the network is close to the difference between the predefined limits. In these cases, when the maximum voltage exceeds the upper limit of the network and the OLTC decreases the whole network voltage, the minimum voltage of the network will decrease below the network’s lower limit, causing the OLTC to be in a state of successive ups and downs. This problem is solved by deactivating the OLTC in case the other extreme voltage is too close to its limit [34]. Another adaptive OLTC voltage control focused only on the correction of the false image of the network load that has not taken the influence of DG into consideration [35]. In some networks, DNOs prefer to utilize other control tools before the OLTC for mechanical purposes [36].
• Reactive Power Control Using DGs: DG units receive orders from the coordinated control algorithm to absorb or generate some reactive power within their limits. It is more efficient than the OLTC, although it requires a communication network among the DNO and DG units and highly increases network losses. However, it is still cost worthy, considering that normally the cost of losses is much less than the cost of curtailed energy of DGs. Voltage control loss factors were proposed as means of understanding the interactions between reactive power flows, losses, and curtailment [37].
• Curtailing the Real Power of DGs: This tool, like the one before it, depends on a system of communication between DNO and DG units. It is regarded to be the most efficient method for maintaining the network voltages within limits (as proven in [11]). However, since it is the most expensive, it ought to be employed as the last choice after all other alternatives have been exhausted.

The Hardware-in-the-loop (HIL) simulation is a technique that substitutes the actual component of a machine or system with a simulation in order to design and test control systems that handle complicated machines and systems. This technique is used in this article to verify the reliability of the algorithm upon implementation in real networks, considering all technical issues such as delay periods, the convergence of calculations, data connection, safe testing, etc.; it is not just a theoretical algorithm.

The hardware component in this article is an Arduino Mega 2560 board utilizing an ATmega 2560 microprocessor. It is programmed with the control algorithm intended to govern the network’s overall voltages. Applying MATLAB Simulink, the parameters of both the network and DGs’ are modeled. Power flow analyses are carried out employing the forward/backward sweep method [38]. The forward/backward sweep method is selected in this article for its trustworthiness for radial networks in real-time tests since it converges to an output of reasonable accuracy in a relatively short number of iterations. While Simulink transmits the variables intending to evaluate the state of the network to Arduino, the latter returns the decisions to Simulink. This closed-loop system functions with a suitable sample time. Figure 1 summarizes the cycle of operations of the HIL technique. Figure 2 shows the experimental setup for the latter cycle. The image was taken while the configuration was displaying the live results of the simulation.

The control algorithm consists of two main portions: basic and restoring.

The basic part covers the measures necessary to keep network voltages within prescribed limits in the event that one of the extreme voltages exceeds the allowed limits. The restoring part tends to return both the power components of DGs, as well as the position of the OLTC to their optimal levels, as far as the network state permits. The restoring part functions only when all network voltages are within acceptable boundaries; consequently, the basic part is not functioning. Both parts employ an approximate law [39] to estimate the voltage sensitivities of buses of maximum and minimum voltages to the real and reactive powers of all DGs. It can also give a rough estimation of how much reactive power must be absorbed or how much real power must be curtailed of a particular unit in order to achieve a certain voltage change in a certain bus. It should be highlighted that this algorithm is applicable only to radial networks. The operational principles of the whole algorithm are illustrated in Figure 3.

The following active tools must be available for the actions to be applied:

• OLTC: the OLTC is regarded as inaccessible for control if the other extreme network voltage is too near to the other boundary by less than a single tap step added to a reasonable margin. Otherwise, the desired number of taps is computed so that the overstepped voltage returns within boundaries. Consequently, it is actuated after a preset delay period of time to skip short voltage variations.
• Absorption or production of reactive powers from DG units: The voltage sensitivities of the buses of extreme voltages with regard to all of the reactive power controllable units are estimated beforehand [39]. The highest sensitivity unit (i.e., lowest reactive power that is needed from it) to the node of the overstepped voltage is picked, provided that this unit has not reached its full capacity of reactive power and will not lead other voltages to cross the other boundary simultaneously. The quantity desired to be absorbed/produced by this unit is estimated, and the requisite quantity is subsequently implemented after a predefined delay period. In case the stated conditions are not fulfilled, then this tool is deemed unavailable for voltage regulation.
• Curtailment of active power produced by DGs: It is similar to the preceding one except that it functions for the real powers of DGs, which only the DNO is capable of lowering. If the network cannot maintain all voltages within boundaries, even if no real power is injected into the network in any way, load shedding is the only alternative left in this situation.

The basic control continues to employ the accessible active tools sorted ascendingly by their costs one after another until the achievement of its target. The basic control algorithm is improved such that if the two extreme voltages of the network violate their limits at a certain instant, the basic control tends to restore the voltage violating first (regardless of the implications of its actions on the other voltage), then tends to restore the other. The restoring control is modified to check if it is capable of recovering any decision conducted by the basic control, fully or partially, to reduce the running cost. That may be possible in case of the removal of a particular unit or an escalation in loading. It functions the opposite of how the basic part does. Real power is examined first to recover any prior curtailments since it causes the most expensive cost, followed by reactive power, and lastly boosting the OLTC as high as is feasible to reduce network losses. It ought to be regarded that restoring control picks the relevant resource with the lowest sensitivity coefficient in order to recover as much power as allowed.

The delay mechanism operates as follows:

• Both the basic and restoring control parts possess their own time delay. When activating a control part and deactivating the other one, it must wait for a preset time delay for the purpose of bypassing transient and rapid voltage surges.
• The DNO must wait after each decision for its special time delay. It ought to examine its implications in the network before deciding the next action. That time delay is prescribed based on the number of taps, mechanical activation time of the OLTC, time constants of generators, and the quantity of power upgrading.
• If a decision has been proven to be insufficient after its delay time, the system performs the next step immediately and does not apply the time delays of basic and restoring parts. The purpose is to prevent the risk of an extreme voltage violation for an extended period of time. The procedure is repeated until the regulation of all voltages for the basic part, or the optimal situation for restoring the part is achieved.
• In case a voltage boundary breach occurred during the application of a restoring control, the restoring control is immediately stopped, and the basic control restarts functioning.

3. Real-Time Simulations

The operation of the algorithm under investigation is tested in the IEEE 33 bus radial distribution network. Data of its loads are tabulated in Appendix A. The standard network is modified by adding three units of DGs, each one having a rated power of 6 MW. The units are located in buses 23, 19, and 22, respectively, as shown in Figure 4. The simulation is based on the assumption of a three-phase symmetric power system, as in practice, significant imbalances do not often take place in medium voltage networks. The tap step voltage of the main transformer is assumed to reach 1.67% of the nominal voltage and a 1 s mechanical activation time for each tap. The voltage of the substation is initially set to 2 taps above 1 p.u. with no intervention from DG units. The lower and upper limits of network voltages are set to be 0.95 and 1.05 p.u. The limits of DGs’ reactive powers are 0.33 * P_rated. The delay time is 4 s for the basic part and 6 s for the restoring part.

In the simulation test applied, the DGs are connected to the network (with their full capacity) at t = 30, 50, and 70 s, and then detached (modeling the case of non-availability of real power) at t = 110, 150, and 180 s, respectively. The simulation results are illustrated in Figure 5.

The network maximum voltage oversteps the upper boundary due to the connection of the first unit at a time of 30 s. The basic control drops the OLTC set point by a tap after the delay time has lapsed, and the tap changer functions after its mechanical activation time (1 s), returning the maximum voltage inside network boundaries.

None of the increasing network voltages has violated the voltage limits after the connection of the second unit at a time of 50 s. Therefore, no actions have been implemented by the coordinated control.

At a time of 70 s, the third unit is activated, and the network maximum voltage oversteps again. Initially, the basic control cannot set a further tap changer operation, as lowering the substation voltage further leads the minimum voltage to drop below the network voltage lower boundary. Thus, the OLTC option is inaccessible at this moment. Hence, reactive power control is selected. The entire reactive power control capacity of all units is employed one after another split by appropriate delays according to the time constants of the generators, such as the delay mechanism states. They are arranged in descending order by the coefficients of voltage sensitivity with respect to the bus of the overstepped voltage. The network maximum voltage continues to be beyond its boundary, even after the reactive power control, which has become currently inaccessible. As a result, real power control is employed. The third unit is picked for its highest sensitivity. Its real power is reduced by an appropriate quantity to maintain all voltages within acceptable limits.

When the first unit is removed at a time of 110 s, the network voltages decrease, causing some of them to violate the minimum voltage limits. The basic control algorithm has been reversed to boost the network voltage. No tap changer operations can be implemented, as raising the substation voltage would raise the network maximum voltage above the network voltage upper boundary. As a result, the availability of reactive power control is utilized. The reactive power control capability of the first two units is utilized not only to end the absorption of reactive power, but also to generate their full capacity of reactive power one after another as before. The minimum voltage is not yet restored within limits. However, the maximum voltage of the network also violated the upper boundary.

Some remarks should be noted:

• The control algorithm would not tend to recover the maximum voltage until the minimum voltage is restored within limits first (the first event to occur is handled first.)
• The option of the OLTC is not unavailable anymore, as the maximum voltage has overstepped the upper limit already. When such a scenario occurs (a violation of both limits at the same time), the algorithm obviously cannot regulate both voltages at the same time without causing instability in the system. The voltage control algorithm is set to regulate the voltage violated limits, first neglecting its implications to the other extreme voltage, then regulating the latter voltage afterward.

The tap changer is raised by a single tap operation. Finally, the minimum voltage of the network is restored within limits. However, the maximum voltage is above limits, returning the network to a similar condition when all the units are connected. The reactive power of the first two units has been reversed to their full capacity of absorption. That is not adequate, leading to further curtailment of the real power of the third unit, restoring the network voltages within limits. The final steady state result is the utilization of the OLTC (which is not chosen first) and curtailing more active power, while keeping the reactive power control at its state. This result verifies the robustness of the modified algorithm for applying the optimized actions on the network under such scenarios.

The network minimum voltage decreases below the lower boundary again at time 150 s when the second unit is disconnected. The tap changer cannot be activated, as the maximum voltage is too close to the upper boundary. Consequently, the reactive power control is activated. It is adequate that the first unit gives up its reactive power absorption and operates at unity power factor to restore the network voltages within standard boundaries.

The same condition is repeated at a time of 180 s when the third unit is disconnected. The other units operate at a unity power factor as initial conditions.

Table 1. Data of special points in simulations.

Time (s)	Vmax (p.u)	Vmin (p.u)	OLTC Ref.	Vss (p.u)	Pref (MW)			Qref (MW)			Ploss (kW)
					DG1	DG2	DG3	DG1	DG2	DG3
0.00	1.033400	0.950164	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	0.00	186.27
30.01	1.033400	0.950398	1.0334	1.0334	6.00	0.00	0.00	0.00	0.00	0.00	184.45
33.89	1.050005	0.972217	1.0334	1.0334	6.00	0.00	0.00	0.00	0.00	0.00	217.78
38.00	1.050722	0.972651	1.0167	1.0334	6.00	0.00	0.00	0.00	0.00	0.00	222.59
39.00	1.034271	0.954817	1.0167	1.0167	6.00	0.00	0.00	0.00	0.00	0.00	230.23
50.01	1.034308	0.954855	1.0167	1.0167	6.00	6.00	0.00	0.00	0.00	0.00	230.34
69.97	1.037571	0.958392	1.0167	1.0167	6.00	6.00	0.00	0.00	0.00	0.00	292.77
70.01	1.037603	0.958428	1.0167	1.0167	6.00	6.00	6.00	0.00	0.00	0.00	293.60
70.34	1.050374	0.959376	1.0167	1.0167	6.00	6.00	6.00	0.00	0.00	0.00	363.32
74.43	1.111057	0.961428	1.0167	1.0167	6.00	6.00	6.00	0.00	0.00	−1.98	857.86
78.75	1.077556	0.959709	1.0167	1.0167	6.00	6.00	6.00	−1.98	0.00	−1.98	976.50
83.22	1.075382	0.956471	1.0167	1.0167	6.00	6.00	6.00	−1.98	−1.98	−1.98	1048.58
87.51	1.057947	0.955625	1.0167	1.0167	6.00	6.00	4.20	−1.98	−1.98	−1.98	878.75
89.83	1.049988	0.955362	1.0167	1.0167	6.00	6.00	4.20	−1.98	−1.98	−1.98	801.72
110.01	1.049076	0.955115	1.0167	1.0167	0.00	6.00	4.20	−1.98	−1.98	−1.98	790.69
110.29	1.048322	0.949761	1.0167	1.0167	0.00	6.00	4.20	−1.98	−1.98	−1.98	729.14
114.47	1.045909	0.932724	1.0167	1.0167	0.00	6.00	4.20	1.98	−1.98	−1.98	697.21
118.70	1.048015	0.940633	1.0167	1.0167	0.00	6.00	4.20	1.98	1.98	−1.98	591.13
119.44	1.050030	0.941211	1.0167	1.0167	0.00	6.00	4.20	1.98	1.98	−1.98	575.12
123.00	1.051768	0.941708	1.0334	1.0167	0.00	6.00	4.20	1.98	1.98	−1.98	568.03
124.00	1.068187	0.959811	1.0334	1.0334	0.00	6.00	4.20	1.98	1.98	−1.98	550.40
124.52	1.068173	0.959664	1.0334	1.0334	0.00	6.00	4.20	−1.98	1.98	−1.98	549.99
128.55	1.066865	0.952039	1.0334	1.0334	0.00	6.00	4.20	−1.98	−1.98	−1.98	624.49
132.58	1.061454	0.950706	1.0334	1.0334	0.00	6.00	3.36	−1.98	−1.98	−1.98	667.21
136.44	1.049993	0.950303	1.0334	1.0334	0.00	6.00	3.36	−1.98	−1.98	−1.98	581.01
150.01	1.049718	0.950261	1.0334	1.0334	0.00	0.00	3.36	−1.98	−1.98	−1.98	578.51
150.11	1.048874	0.949928	1.0334	1.0334	0.00	0.00	3.36	−1.98	−1.98	−1.98	567.26
154.13	1.040971	0.947281	1.0334	1.0334	0.00	0.00	3.36	0.00	−1.98	−1.98	495.19
155.68	1.041271	0.950035	1.0334	1.0334	0.00	0.00	3.36	0.00	−1.98	−1.98	444.10
180.01	1.040835	0.950745	1.0334	1.0334	0.00	0.00	0.00	0.00	−1.98	−1.98	431.49
180.58	1.033400	0.949966	1.0334	1.0334	0.00	0.00	0.00	0.00	−1.98	−1.98	329.88
184.60	1.033400	0.949168	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	−1.98	282.40
188.62	1.033400	0.949810	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	0.00	213.11
189.43	1.033400	0.950011	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	0.00	193.08
194.73	1.033400	0.950164	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	0.00	186.29
250.00	1.033400	0.950164	1.0334	1.0334	0.00	0.00	0.00	0.00	0.00	0.00	186.27

collects the data of special points in simulations to ease the follow-up process. It should be noted that:

• Time delays in both the algorithm and the connection of HIL components are apparent. For example, when the maximum voltage of the network exceeded its upper limit at t = 70.34 s., the first control action was taken at t = 74.43 s. after a period of time slightly larger than the 4 s. set for the delay of the basic control algorithm.
• Power losses may be relatively high in some instants. They increased from 186.27 kW in case of no contribution of DG units, to 801.72 kW (a steady state value at t = 89.83 s) in case of the maximum possible contribution of real powers from DG units and their full capacity of reactive power absorption. However, they are negligible compared to the real powers enabled by DG units (16.20 MW), which may be stored in energy storage systems or transported and sold to other MV networks, as the simulations emulate the case of maximum generation/minimum load (the worst case for voltage regulation algorithms).

4. Novelty/Contributions of the Article

The contributions of this article can be summarized as follows:

i.

The basic part of the control algorithm was modified such that if the two extreme voltages of the network violate their limits simultaneously at a certain instant the basic control tends to restore the voltage violating first (regardless of the implications of its actions on the other voltage), and then it tends to restore the other one. That boosted the network’s stability and prevented stalling of the system or even the occurrence of wrong actions that may cause unjustified losses. This is considered a major breakthrough since no publications, as far as the authors know, discussed such severe cases, as they often cause the traditional algorithms to get stuck and lead to instability of the network.

ii.

The restoring control is modified to check if it is possible to restore any action performed by the basic control, totally or partially, to lower the running cost (due to a disconnection of a certain source or an increase in network loads). Thus, the proposed algorithm guaranteed that more curtailed active power and absorbed reactive power would be restored.

iii.

Implementation of the HIL technique has proven the ability of the system to handle its actions at the right time, demonstrating its reliability and robustness.

5. Conclusions

In this article, a modified control strategy that is HIL-based is proposed to handle a severe situation. The modifications to the strategy proved the capability to maintain all network voltages within acceptable boundaries, even if both extreme voltages violated both limits simultaneously. Moreover, these modifications proved the reliability required to avoid instability or stalling of the system, or even wrong actions that may cause unjustified losses. This is considered a major breakthrough, as no publications, to the best of our knowledge, discussed that scenario. The modifications are performed upon restoring control within the proposed algorithm, guaranteeing more curtailed active power and consumed reactive power to be brought back, as the algorithm focuses on how much active and reactive power could be recovered, as long as the network voltages are maintained between the predefined boundaries without the need to set other stricter boundaries. Real-time domain simulations obtained from the implementation of the HIL technique verified the reliability and robustness of the modifications applied to the voltage control algorithm, as that algorithm applied its decisions at the optimal time.