Demand Response Will Be Key If We’re To Integrate a Large Percentage of Renewable Energy
Here’s a wonderful article that deals with a subject that is of significant concern: the intermittence of solar and wind energy. It provides an interesting case study of how demand response can figure into the equation, a phenomenon that will play a huge role in a future in which we try to integrate a growing percentage of renewables.
Demand response formed a large part of my project report when I studied for my Masters degree in Renewable Energy Systems.
My report related to Western Denmark, which at the time had the highest penetration of wind power anywhere in the world. I looked at various energy storage and demand response type options for maintaining a stable grid with a large amount of intermittent power production.
The article referred to covered only one half of the demand response equation – that of coping with a sudden drop off in wind power production.
There are two facets to the intermittent nature of wind which must be taken into account.
1. Predicted variation – when looking 4 to 6 hours ahead in a country with a lot of widely dispersed wind power, it is possible to have a reasonably good idea of how much wind power will be generated. Prediction is usually within around
+/- 10% of the rated wind capacity on the system, and can be much closer for part of the time when weather systems are relatively stable.
Predicted variation can be planned for and other types of generation ramped up or down to balance the system.
2. Unpredicted variation – As above, suppose the system has 3 GW connected, and the prediction for 6 hours time is for 1.6 GW, you can expect error bars on the prediction of up to +/- 300 MW depending on how stable the weather is at that particular time.
In this prediction, you may get anywhere from 1.3 GW to 1.9 GW (with occasional outliers beyond this range)and will need to plan to handle this degree of variation from prediction.
Traditionally, this has been done almost entirely by adjusting power production, but if fossil fuel use is to be reduced as rapidly as possible, this is no longer enough.
In my report, I proposed to schedule power equal to the negative error bar to dispatchable power uses – such as running heat pumps on Denmark’s many district heating networks. Predicted variation can then be managed by moving either more or less power onto loads which are not time critical.
In this example, 1.3 GW of wind power can be cautiously regarded as firm 6 hours ahead – with any production above this going to uses where time of use is not of any significant importance.
If demand for unscheduled uses exceeds prediction, it is likely that some power can be diverted back from heating water for district heating, pumping water supplies to local distribution reservoirs and such like to meet immediate needs – the water can always be heated in a few hours time when power supplies are more readily available, and there is always a enough treated water in the local distribution reservoir to allow a temporary interruption to pumping.
This scenario can deliver extra power to the general user when the wind dies down more than predicted, and can also absorb unexpected bonus production when the wind suddenly increases – giving a full two way spinning reserve substitute.