System Stability and Virtual Synchronous Machines. Why we need grid forming converters, and the metrological challenges of incentivising them. Dr Andrew Roscoe. 8 th January 2018
HPoPEIPS High Penetration of Power Electronic Interfaced Power Sources
The situation today : sufficient synchronous generation that the network appears to users as an AC voltage source. Aggregated generation PCC [Point of Common Coupling] Non-linear load Voltage source I 1 Load demands non-sinusoidal current. Voltages V V V V 2 Voltage at PCC is slightly distorted, but still roughly sinusoidal.
The situation if the generation is dominated by conventional converters using vector current control : the network appears to users as an AC current source. PCC [Point of Common Coupling] Aggregated generation Non-linear load Current source Injected current (converter goal) I I 1 Converters attempt to inject Sinusoidal current. V V 2 But the load is non-linear. The resulting voltages (and currents) are highly non-sinusoidal and very hard to define.
In a real network, three-phase loads (and faults) may be unbalanced, non-linear, or both. In the extreme case below, the 3 phase loads consist of a non-linear load on phase A, a linear load on phase B, and no load (open circuit) on phase C. The upper power system is viable, with loads being provided with acceptable power quality. The voltage sources need to provide power with ripple at twice the nominal frequency to deal with the unbalanced fundamental plus other frequencies to deal with unbalanced non-linear loads. Balanced, sinusoidal voltage sources Balanced, sinusoidal current sources The lower power system is unviable. It results in unbalanced voltages and poor power quality, with potentially dangerous over-voltages on the un-loaded phase. If the two sources were paralleled, the required unbalance and most of the harmonic load current components will be supplied by the 3-phase voltage source.
VSM0H converter (Virtual Synchronous Machine with Zero Inertia) 6
Proposed VSM algorithm (Virtual Synchronous Machine) Virtual Rotor Drive E αβ I abc ~ X ' Filter inductor is virtual X V αβ Virtual Stator/PCC Frequency droop D f Power set-point P set Frequency set-point f set All values in per-unit except f 0 (Hz) and angles (radians) f pu P e Governor model. Could be updated in real time Prime Mover model. Could be updated in real time I abc, V αβ Fault current limiting. DC current managemen t E E Field,Q,V controller P e Measured output power E V Rotor dynamics Estimated / Inferred X ' 1 X ' d e s 2 f 0 d m s 2 f 0 P δe P de P dm P m + + E x P e V + - - H could be optimised and adjusted in real time 1 2Hs f pu Possible direct adjustments during the most violent transients to avoid pole slip or excessive δ angles. 2 f 0 s Rotor
1) You cannot stabilise the network with a very small capacity of voltage sources, even if these provide huge per-unit values of inertia. 2) The network requires ENOUGH capacity of Voltage Sources. 3) This is fundamentally more important than inertia. 4) A stable network can be produced with zero inertia. Case A 4 GVA network, Viable. Havg=4 H=8 H=8 Loads Case B 4GVA network, Unviable. Havg=4 H=80 0.2 GVA 1.8 GVA Loads Case C 4GVA network, Viable. Havg=0 Loads Case A represents the situation today. A reasonable average inertia H is spread across the voltage sources (synchronous machines). Both inertia and synchronising torque are sufficient. Case B shows why inertia is not everything. This network has the same average H, delivered by a fraction of the generation capacity which is equipped with oversized flywheels. It would not work. The network is dominated by current sources which rely on a stiff voltage source for correct operation, and the only voltage source has a low rating (and hence high coupling impedance) compared to the rating of the network. The network has sufficient inertia, but there is not enough synchronising torque. Case C is a functional stable network, with. ROCOF may be transiently high, but both frequency and power quality can be kept within acceptable bounds. There is no inertia, but there is enough synchronising torque. Voltage source Current source
The metrological challenges : Assessing (and paying for) Inertia, Fast Frequency Response, and Power Quality improvement. Metering assessment, in real time: Energy (kw-h) every n minutes Fast Frequency Response How is the response tallied against real-time ROCOF? What is the definition of ROCOF? The signature of Inertia. How do you discriminate Inertia from Fast Frequency Response, using current and voltage measurements? What are the units for metering? A payment for availability in (MVA.s)-hours? Or a payment for actual delivered response in useful MW-s where the MW excursion is a genuine response to an event. 9
The metrological challenges : Assessing (and paying for) Inertia, Fast Frequency Response, and Power Quality improvement. Metering assessment, in real time: How do you recognise and reward the signature of a device which provides any or all of: Voltage source Syncronising Torque Grid forming (and black start contribution) Power Quality Improvement? What are the units of measurement? For PQ improvement, requires careful real-time assessment of voltage and current harmonics, in both amplitude AND phase, to determine positive of negative impacts. 10
The metrological challenges : Assessing (and paying for) Inertia, Fast Frequency Response, and Power Quality improvement. ~0.001-0.2Hz (5s or longer dynamics) ~0.2-5Hz (0.2s to 5s dynamics) >2Hz (Dynamics less than 0.5s) ~0.001-0.2Hz (5s or longer dynamics) ~0.2-5Hz (0.2s to 5s dynamics) >2Hz (Dynamics less than 0.5s) 11