“Systems of systems were informally developed by AC grid engineers,” explains Benchaib. “Today we’re formalising the concept in DC grids. It is an active example of ‘thinking globally and acting locally’.” This brings to mind the theory of complex systems. In a school of fish, for example, each fish keeps a certain distance and follows the fish in front. The result is that each fish acts like the whole school, while the whole school acts like an individual fish. “Systems of systems are naturally occurring forms of organisation,” states Benchaib. “They will underpin the 2030 transmission grid in which control will be distributed amongst the nodes.”
Distributed control will enable adjustments at different hierarchical levels and at different timescales to take place across the entire network. Indeed, a key component of systems of systems thinking is the notion of scale. Flexibility comes from the ability to switch from long timescales and large space scales to small ones, then back again. Again taking the example of the school of fish, should one or more individual fish quit the school, there is no effect on the school or its pattern of movement. All of which points to the tantalizing prospect of plug and play. “It should be possible to add a load to a system or to remove it without destabilising the system in the DC power network of 2030,” states Benchaib. AC networks, of course, have that plug-and-play capability, developed piecemeal over the years. However, with mass RES their control systems are unlikely to be able to cope. To enable plug-and-play behaviour in a multi-terminal DC (MTDC) system that integrates offshore wind farms, Benchaib and his research team are exploring primary, secondary and even tertiary control functions. They operate on different timescales and at different system levels to ensure grid stability. Primary control represents the variation of the power/current injection set-points caused by disturbances in the system. This is different from the local controller of the (AC/DC) converter, the role of which is to track any set-points. In the “system of systems” approach, a solution with different timescales has been adopted.
The time dynamic response of the primary control should therefore be slower (longer timescale) than the controller of the converter, which is around a few milliseconds. This primary control “time constant” will not depend only on the reactivity of the converter but also on the power injection ability of all connected nodes (production, storage, loads, etc.), which theoretically should be from a few hundred milliseconds to a few seconds, as for AC systems. Primary control is global but distributed. A primary control strategy that has proved itself in AC grids is the so-called droop control technique. This could be applied for MTDC power grids. In such networks the distributed control will then employ the droop mechanism to regulate the DC voltages by adjusting the power (or current) injections from the converters.
The converters simultaneously and individually inject the active power needed to restore balance to the MTDC grid. Thus, for the same voltage deviation two converters could inject different amounts of power. “Droop control is in each node. This is another example of ‘thinking globally, acting locally’ at work,” stresses Benchaib.
At a different and slower timescale, secondary control regulates the voltage deviation across the grid – or the next system up in the system of systems – on a timescale of several seconds. It uses master-slave strategies where one or more converters act as a slack bus (provide or absorb the power needed to achieve power balance within the DC grid), while other converters control their power exchange between the AC and DC systems. “The work we have put into formalising a system-of-systems approach for DC grids could usefully be applied to several other fields,” says Benchaib. “It’s a whole philosophy.”
Fault clearance speed: a system-wide challenge to HVDC grids
Unlike AC networks, which have been around for over a hundred years, HVDC grid technology is still in the making. Few of the principal protection systems from AC grids would work for DC grids, though the main requirements still hold. These are sensitivity (detection of all faults), selectivity (ability to isolate and trip only faulty parts), speed and reliability.
An issue that cuts system-wide across those criteria is the circuit breaker (CB) speed. CBs in AC systems typically clear a fault in 70-100 milliseconds (ms), including the fault detection time. However, DC networks have low impedance, so faults go deeper and faster. To be absolutely reliable, HVDC CBs should have completely cleared the fault current in less than 5 ms after the fault inception. Communication speed is again a challenge in current differential protection, an AC protection system that is actually applicable to DC grids. “Alstom recently doubled the speed of its differential protection’s communication channel from 64 to a very fast 128 kbps. It reduces the fault detection time to less than 17 ms. However, that’s still far short of the 3-4 ms required to protect HVDC grids,” explains Dr Oleg Bagleybter, Programme Manager (Networks).
Backup protection solutions, where there is often an additional time lag, will demand even higher CB speed, and without backup there can be no reliability. New strategies are needed to integrate fault detection, discrimination and isolation for DC grids. A possible future solution may be the Alstom proposal of autonomous HVDC breaker operation, minimising breaker duty while also reducing the time the power interruption is “seen” by the interconnected AC systems.
Multi-terminal DC Grid Primary & secondary controls