Meridia Insight Tech for Good Frontiers

A converter can steady a DC grid without controlling its voltage

A controller that never directly regulates DC voltage can still make a meshed network settle—because synchronized AC terminals quietly create a network-wide

4 terminals, no direct DC-voltage regulator: synchronization itself closes the droop that makes the mesh solvable.

A four-terminal DC grid may not need any converter to regulate its voltage directly. In the controller studied by Hisham Mahmood and Marcelo A. Elizondo, each converter instead adjusts its AC frequency in response to two local DC measurements. Once the AC terminals synchronize, that frequency response has a less obvious consequence: it creates an effective voltage droop at every terminal, and the meshed DC network settles to a unique operating point (Mahmood and Elizondo, 2026).

That is a potentially useful answer to a problem facing power systems built around converters. Their connections are not always backed by a strong, synchronous AC grid that naturally provides a stable reference. Meanwhile, multi-terminal DC networks are expected to move power among distant resources and loads. The paper’s result is not that a new controller has solved every grid-stability challenge. It is that a control strategy designed to form the AC side need not leave the DC side untethered.

The Science

A converter links two electrical worlds: alternating current, or AC, on one side and direct current, or DC, on the other. A meshed DC network connects multiple converter terminals through branches, creating several paths for power to travel. That is unlike a simple point-to-point link: a change at one terminal can affect voltages and currents throughout the network.

Many converters use grid-following control. They measure the AC grid’s phase and frequency, often through a phase-locked loop (PLL), then align their output to it. That approach can become less robust when the surrounding grid is weak. Grid-forming controls take a different role: they establish a voltage waveform and its frequency rather than merely following one. The distinction matters increasingly as converters connect large loads and renewable generation to electrically weak parts of the grid.

The proposed dual-measurement frequency-droop controller, or DMFD, gives each terminal a frequency command based on two DC-side signals: deviation of the bus voltage from its nominal value, and deviation of DC-port current from its scheduled value. The current signal is filtered to avoid responding directly to fast measurement fluctuations. In simplified form, the frequency change is proportional to a weighted combination of those two deviations.

“Droop” means that a control variable changes gradually as the system departs from its target, encouraging multiple devices to share the response. Conventional DC droop usually adjusts a converter’s power or current directly in response to DC voltage. DMFD instead adjusts AC frequency. The surprising part of the paper is the connection between the two: if an AC terminal synchronizes to a fixed frequency, the frequency command and its two measurements imply a relationship between DC current and DC voltage. That relation acts like a DC droop even though the controller never commands DC voltage or current directly.

The authors represent the DC grid as a network of nodes and resistive branches. Kirchhoff’s current law says that the currents entering and leaving each node must balance; each branch’s current follows from the voltage difference between its ends and its resistance. Incorporating the DMFD relationship produces a linear system for the node-voltage deviations. The authors show that, when the network is connected and the droop gains are positive, this system has one solution. In practical terms, the model predicts one steady DC-voltage and current pattern for a given set of injections and frequency conditions.

This is a theoretical result under stated assumptions, not a claim that the real hardware will always reach that point. It assumes unsaturated operation for the core steady-state analysis. The paper separately includes current and power limiters, because real converters cannot deliver unlimited power. Those protection and limiting functions matter especially during faults and large disturbances.

Baseline DMFD gains shown in the damping study

The paper marks this gain combination as its baseline point in the damping-ratio plot; the gain coefficients have different definitions and should not be compared as a common physical quantity.

Baseline DMFD gains shown in the damping study
LabelValue
Voltage gain m_dc50 puP/puV
Current gain m_i1.5 puP/puV
Frequency gain m_f0.01 puP/puV

What They Found

First, the model turns disturbances that might seem to require separate analyses into equivalent current offsets. Suppose one AC grid runs at a steady frequency different from nominal. The affected converter’s frequency response changes its DC current; the mesh then distributes the resulting voltage and branch-current changes. The calculation is a single linear network problem. The paper therefore gives a way to trace a frequency bias at one terminal through the rest of the DC system, rather than treating it as a local AC issue alone.

The same framework applies when one terminal becomes islanded—separated from the larger AC grid and responsible for serving a local power balance. In that case, its frequency is no longer imposed by an external grid. The paper estimates the frequency shift from the power required by the island and the converter’s frequency-to-power sensitivity, then uses that frequency as an equivalent input to the DC network equations. This gives operators and designers a way to ask whether the other terminals can absorb the imbalance, and what DC voltages and branch currents would result.

Second, the controller’s two measurements play different roles. The voltage term helps shape the steady sharing of power across the DC mesh. But in the small-signal model—the model that describes small, rapid movements around an operating point—voltage-only droop leaves an oscillatory mode undamped. The authors find that adding the current channel damps that mode. So the second measurement is not merely a refinement to improve sharing: in the studied model, it is important to stability.

Fig. 7: Damping ratio ζdom of the dominant oscillatory mode over the (mdc,mi) gain plane at mf=0.01. The baseline point (mdc,mi)=(50,1.5) is marked by a square, and the contour denotes ζdom=0.7.
Fig. 7: Damping ratio ζdom of the dominant oscillatory mode over the (mdc,mi) gain plane at mf=0.01. The baseline point (mdc,mi)=(50,1.5) is marked by a square, and the contour denotes ζdom=0.7. Source: Hisham Mahmood, Marcelo A. Elizondo

Figure 7 maps the dominant mode’s damping ratio over the voltage- and current-gain plane. Damping ratio describes how quickly oscillations die away relative to how strongly they swing; a higher value generally means more rapid decay. The figure marks a baseline gain setting of m_dc = 50, m_i = 1.5, and m_f = 0.01, and plots a contour at a dominant-mode damping ratio of 0.7. These are design references, not a universal prescription: the suitable gains depend on the network and converter dynamics.

Damping-ratio design contour

Reference damping-ratio contour in Fig. 7. The paper describes this as a contour on a gain plane, not as a single measured operating result.

Damping-ratio design contour
LabelValue
ζ_dom contour0.7 damping ratio

Third, the researchers test the ideas in averaged electromagnetic-transient simulations of a four-terminal ring. The scenarios include uneven loading, the islanding of a terminal, and a three-phase fault on a weak AC connection with a short-circuit ratio near 2. Short-circuit ratio (SCR) is a rough measure of grid strength relative to the converter; a value around 2 signals a weak connection. In the reported comparisons, DMFD maintains stable operation in conditions where a PLL-based grid-following baseline with the same DC-droop gains loses stability. The simulations also follow the steady-state characteristics predicted by the network model.

Those results are encouraging, but they have a defined scope. They come from averaged simulations, not a deployed multi-terminal DC network or a hardware demonstration. The excerpted paper material does not report a broad quantitative comparison of settling time, losses, or performance across many network designs. The strongest evidence here is the combination of a solvable steady-state model, a small-signal explanation of damping, and time-domain simulations of a specific four-terminal topology.

Why This Changes Things

The central contribution is architectural. Grid-forming control and DC-grid coordination are often treated as separate jobs: one establishes a reliable AC waveform, while another directly regulates DC voltage or power. DMFD proposes that a frequency-based grid-forming action can also induce the DC sharing behavior needed for the mesh to settle. If this works across diverse terminals, it could make a coordinated control design less dependent on assigning one converter the special job of holding DC voltage.

That matters in networks where terminals connect to different AC systems. A shared frequency signal cannot be assumed when those systems have separate operating conditions. The paper explicitly handles a steady frequency deviation at one terminal as a disturbance that propagates through the DC network. Likewise, if one terminal islands while others remain connected to their own grids, the rest of the mesh can contribute to balancing the change. That is a more flexible picture than a system in which all terminals either remain grid-connected together or island together.

The method also offers a planning advantage. Because the steady-state equations are linear in the modeled network quantities, a designer can estimate how droop ratios and net injections affect voltage deviations and terminal currents before running detailed time-domain simulations. These “design envelopes” can help screen gain choices against operating bands and current limits. The damping maps add a second filter: a setting can be checked not only for steady voltage and current consequences, but also for whether oscillatory modes are adequately damped.

The paper’s claim should not be confused with a claim that DC voltage no longer matters. Voltage remains a vital measured signal and an operating constraint. Nor does a unique mathematical equilibrium guarantee that the system can safely reach it under every fault or saturation event. The study’s explicit power-limit loop and fast current limiter acknowledge the physical limits that droop equations alone cannot erase.

The results also speak to weak-grid concerns without implying that DMFD is a universal replacement for grid-following controls. The simulations indicate an advantage in the particular weak-tie fault case examined, including an SCR near 2. Yet real networks bring protection systems, delays, vendor differences, line losses, unmodeled interactions, and operating schedules that can change the outcome. A promising controller must be evaluated against those conditions, not only against the elegance of its equations.

What’s Next

The next test is breadth. The four-terminal ring is a useful case because it exposes interactions among nodes, but practical systems will vary in topology, line resistance and inductance, converter ratings, and AC-grid dynamics. Testing the method across those variations would show how widely its effective-droop and uniqueness results translate into acceptable operating margins. Hardware-in-the-loop studies and laboratory demonstrations could then test implementation details that averaged models cannot fully capture.

A second question is how to coordinate limits. During a severe disturbance, a converter may reach its current or power limit, changing the controller’s effective response. The paper includes supplementary power limiting and current limiting, but the steady-state and small-signal analysis focuses on unsaturated operation. Quantifying the transition between normal droop behavior and constrained operation will be important for protection design and for deciding how much reserve each terminal must carry.

There is also a design trade-off embedded in the controller. Stronger droop can improve sharing or reshape voltage deviations, while gain choices also influence dynamics. The damping map is a start toward making that trade-off explicit, rather than relying on trial and error. A broader set of validated gain maps could help system planners choose settings that are robust to uncertain loading and network conditions.

The wider promise is not that frequency control can make power networks simple. It is that one local control law may help converters contribute to both sides of a difficult interface: forming a stable AC waveform and supporting a shared DC network. If future tests confirm the result under real limits and diverse grids, DMFD could widen the options for building converter-rich systems that remain stable even when no single strong grid can do all the work.

Comments (0)

No comments yet. Be the first to share your thoughts.