3 ms·
Solar can be regulated pretty well. If you can get good clocks to the inverters, you could actually lock the output to the fixed grid frequency, using slower, m
by namibj 7y ago
Solar can be regulated pretty well. If you can get good clocks to the inverters, you could actually lock the output to the fixed grid frequency, using slower, minute phase shifts coordinated centrally to shift load across the grid by artificially creating the speed-of-light phase difference you get over long transmission lines.
Without that central coordination you'd rely on the inverters being staggered to shut off at slightly different overfrequencies. This tends to make some inverters run less overall though, creating game theoretic issues you could fix by rotating that shut off slowly over the whole range, creating time averaging.
Alternatively you can use inverters that regulate output power down with increasing frequency. A simple P (of the PID) controller to keep the frequency in check. You'd probably want a centralized I component to be applied though, for reason of preventing long-term phase drift. Just tell the inverters once a day/week what I their control loop should target, and it's solved. This low-bandwidth signal isn't hard to transmit either.
Just try to force them to gradually change from the old to the new I, because otherwise you get ugly spiking/oscillations when they all apply the change.
It should be slow enough that synchronous machinery on the grid sufficiently dampens any such oscillations.
- drcross 7y agoIt's interesting to read your comment. Do you work in frequently regulation? Do you think any parallel concepts from the ethernet world could be deployed for decentralised frequency regulation where Carrier-sense multiple access with collision detection (CSMA/CD) is used?
- namibj 7y agoIn hope of you still reading this: No, I don't work in frequency regulation. If you want to hire me into it however, I might be interested. I don't know what sort of concepts you mean, and as I said earlier, the main issue is how you get the grid as a whole to do large-distance load-balancing/load-shifting. These decisions can't be made local for lack of insight into how loaded long-distance transmission lines are and how throttled different regions are, and therefore which regions you want the power to come from. There are certain industrial loads that can be time-shifted, and some that are only worthwhile to operate at lower prices, where reduced utilization (and thus increased capex) has to be hedged against higher energy prices (and thus increased opex). If you get fine-grained surge pricing for consumers combined with some forecasting and automatic decision making to e.g. shut devices off during surges or incentivize people to use the electric shower/kettle at a different point in time, you should get by without any fossil fuels at all. Think like adjusting clock speeds/power targets of servers depending on momentary prices, to best balance capex and opex (batch jobs only), or adjust the energy you expend on video encoding for live streams to balance perceived quality and costs. Obviously adjust EV charging. With e.g. minimal range targets for different points in time, or user-friendly ways of expressing the trade-off between higher range for the next day (and thus more freedom) and higher costs (kWh required during the EV's stay at the charger). It could even go to distribute charging stops based on forecasts for energy prices/availability, so you use the flexibility in when exactly to stop on a long trip for load-shaping the power grid.