I did my Master thesis on overvoltages in distribution grids, caused by photovoltaic generators (and sometimes also wind turbines installed in the MV grid). With increasing PV penetration this becomes more and more of an issue and therefore a number of possible remedies exist. I will describe the problem more in detail below for the people who are not familiar with the topic. The thing is, all the proposed solutions so far have different drawbacks and neither seems perfectly suited. Throughout my Master thesis (that is finished since a few months) I had an idea on a different way to solve the problem and I would like to create a discussion to hear what input you have. Perhaps we can gather some information to figure out if my solution is technically possible and if any studies have been carried out on that issue. Or who knows, maybe somebody can use this as a research topic for a thesis or a PhD?
Anyway, let's step right in. It will take some time to explain the problem and the way how to solve it but if you are familiar with it you can maybe skip over it or skim through it. I hope I will find some interested people! :-)
Load in the distribution grid lowers the voltage, while generation increases it. The allowable voltage range is 90%-110% of nominal value (often 400V in most European countries). With ever increasing PV penetration the generation during peak time (noon, summer, clear sky) is already in some grids much higher than the peak demand, e.g. in my thesis often about 2x as high.
The two most important issues that arise with that are:
- overvoltage, as the grid must be operated below 110% nominal U
- the thermal limit of cables and/or transformer is reached
The existing solutions:
1. Both problems can be fixed by reinforcing the grid. This comes however with high costs.
2. Another possibility is to reduce the power flow in the network in one of the following ways:
- Demand Side Management (increasing the load at critical times)
- Curtailment of PV generation
Drawbacks of these are: DSM is not always available and limited while storage is (so far) more expensive. The curtailment of PV generation can be a better option as it only requires curtailment during very few hours in the year (yearly energy loss is usually maximally 2-3% in most studies). But if the PV penetration shall be increased beyond that curtailment doesn't offer a solution. Also suitable controls have to be developed and care needs to be taken about fairness (one PV plant might always be curtailed as it often crosses the 110% limit, while another one is at a more favorable grid point and never faces overvoltage).
3. Furthermore, on-load tap changers (OLTC) can be installed. Effective, involves however also additional costs.
4. Reactive power consumption by PV inverters. This is the solution I also look at in my thesis. The advantages are: Low extra investment costs (inverters just need to be updated with the proper controls) and low extra operational costs (only grid losses are increased). Here the challenge is to design a control that is on the one hand effective and on the other hand easy to implement.
Two major reactive power controls exist:
- Active power dependent VAr output [cosphi(P)] => During low PV generation (e.g. <50%) PV inverter operates at cosphi=1. Then, with increasing P, cosphi is gradually lowered up to a maximum of 0.95 or 0.9. Effective and easy to implement, but it also operates during times, when there is no overvoltage problem.
- Voltage dependent VAr output [cosphi(U) or Q(U)] => During low voltage (e.g. <105%) PV inverter operates at cosphi=1. If voltage is higher than that, cosphi is gradually lowered up to a maximum of 0.95 or 0.9. This second option has the drawback that not all PV inverters within one network experience the same voltage. So the PV plants at the most critical network nodes may fully contribute, while PV plants at other, non-critical network nodes do not see a high voltage and do not or only marginally contribute. Hence, the control is not so effective, although is decreases substantially grid losses. Studies show that with this control the maximum hosting capacity (how much PV can be installed) can be increased by 20-40%.
However, one drawback with reactive power control is that it increases the power flows in the system (besides the active power there is now an additional reactive current). Therefore, it compromises the second problem in distribution grids: the thermal limit is easier reached. Therefore, this option is only applicable for distribution grids that face overvoltage and have still some free room until thermal limits are reached.
To summarize: Some of the above mentioned solutions can tackle both issues (overvoltage + thermal limit) but either involve extra expenses or are limited, whilst other solutions are not so effective or can only tackle the overvoltage problem while worsening the thermal limit problem.
OK, let's hope I have not yet turned everybody away with this boring explanation.
Now I will explain you my idea and I hope it can be understood
- Firstly, the cosphi(U) (or Q(U)) control can also work in the other direction: During times of low voltage, the inverter can inject reactive power in order to increase voltage. This control is also called Q@night (Google shows some results).
- Secondly, The transformer for an LV network has usually a no-load tap changer, so it can't automatically change the voltage like an OLTC. However, it can be set to different voltage ratios.
The idea is the following:
When a high PV penetration in an LV network is reached the tap changer ratio is permanently set to a lower value, e.g. 20kV/390V. This means the voltage on the LV side is lowered compared to the case when the tap changer is in its neutral position (20kV/400V).
This creates however possible undervoltages during times of high load and low generation (e.g. winter evening). Here comes the Q@night control into play: During these (rare?) times the voltage is increased by injecting reactive power. This requires however a PV inverter close to the lowest voltage (= highest load) in the LV network. Therefore, this can only be applied when a high PV penetration is reached and most lines have at least one PV plant installed.
- PV penetration substantially increased (a larger voltage band for PV integration is made accessible)
- No additional currents during high PV infeed (compared to reactive power control) => thermal limit is as much as possible avoided (I assume that PV infeed is much higher than the load, so during high load + extra reactive currents, the thermal limit should still be in safe margins)
- Easy to implement
- Lower voltage = higher grid losses
- Enough PV inverters need to be close to the maximum load
- Different PV inverters experience different voltages => different degree of reactive power injection => fairness?
- What is the maximum reactive power output of the inverter for Q@night? Limited only by inverter capacity? Or maybe only 10% of that? (Check Q@night in Google and you will at least find that SMA has this option for their inverters)
- How much will the lifetime of the inverter be decreased if he operates often in the Q@night control?
- How to guarantee that the PV penetration in a grid is somewhat evenly distributed? Currently, whoever wants can install a PV plant. Should there be restrictions? How to figure out after which point it is not allowed anymore to install PV capacity at certain points in the LV network?
- How much higher are the grid losses due to the lower voltage?
- How much can the PV penetration be increased by setting the tap changer down to 390V (0.025 p.u.)?
- Could reactive power supply during evening/night be desired from a system perspective for reactive power compensation?
OK, I will stop here. I am not even sure if anybody will read this, least of all join the discussion!
The topic is quite technical and as you see I still have a lot of open questions myself. Maybe you have other concerns with this idea or you know some answers? Then please be so kind and click on "Add a reply"