Skip to main content

Balancing a Renewables Grid across multiple Timescales

For a hypothetical future power grid where production is dominated by variable renewable sources like PV and wind, the challenge of balancing supply and demand at all times is well know. Using again the ENTSO-E open-source data and Python linear programming solver, we extrapolate the grid level solar and wind forecast data to cover the entire yearly electricity demand while minimizing the sum of the absolute errors between supply and demand for all hours in a year.

In order to turn the problem of minimizing the sum of absolute values into a linear programming problem, we use the bounding technique presented in this tutorial.

Using for example the data for the German power grid from Spring 2022 to Spring 2023, the optimal generation mix according to the above criteria is about 75% wind and 25% PV (blue line).  With this configuration about 80% of the annual supply & demand of 475 TWh would be matched at the hourly level, leaving a discrepancy of 20% or 107TWh to be equalized somehow.

For different time-scales, the diagram shows how much imbalance is left after if by some way we could equalize the imbalance up to some time-scale - for example 4 or 12 hours, days, weeks months etc. We can see for example that for the optimal mix, the greatest drop in unmatched imbalance is in between the order of days to weeks, which matches the cycles of the general weather situation in Europe which includes periods with more or less wind and more or less clouds and precipitations. The residual imbalance beyond the month timescale is less than 5% or 20TWh, which means that there is hardly a seasonal imbalance.

Compared to the optimal mix, a 100% PV generation would be much more imbalanced: 30% intra-day plus another 30% seasonal, which require equalization across more than a 3 months. A 100% Wind (optimal onshore & offshore mix) is much less volatile across hours of the day and generally follows the seasonality of demand (more demand & generation in winter).

The optimal mix and resulting imbalance is surprisingly similar across various European countries with different climatic conditions: from the north-sea cost of Netherland to the southern mediterranean Greece to to the semi-arid plains of Spain.


The good news even without doing anything special an optimized combination of PV & wind can have a an effective load carrying capacity of around 80%, the bad news is that closing the remaining 20% will be much harder and much more expensive.

The available choices to correct the imbalance are either
  • Over-capacity with overproduction and curtailment
  • Demand flexibility (shifting demand in time to match available supply)
  • Storage
  • Long-distance HVDC interconnects between regions with different weather pattern.
  • Other intermittent energy sources which have low correlation with wind or solar.
  • Dispatchable renewable energy sources, e.g. hydro-electric.
Demand flexibility can be a very effective option to match imbalances across a few hours or days, in the case of large-scale industrial heat storage maybe days to weeks. Storage is increasingly cost competitive in the short duration of hours to days. 4 hour duration grid scale battery storage plants are currently being constructed all over the world in significant numbers to provide lucrative grid management services (frequency response, synthetic inertia, black start capabilities etc.) as well as starting to balance intraday supply and demand in PV heavy grids (e.g. Hawaii). At this point, the cost per kWh of storage capacity is roughly at par or lower with closed-loop pumped hydro, the previously only viable large scale option for true electricity storage.

The economics of storage are become very quickly much more challenging with increasing timescales: any given unit of storage capacity can be amortized more quickly the more it can be cycled - from daily (365 times per year) to weekly (52 times per year), monthly (12 times per year) or even seasonly with only one cycle per year. To be profitable, seasonal storage capacity would have to be 350 times cheaper than intra-day storage capacity. The cheapest long-term storage is hydrogen storage in underground caverns ($2-$3 capital cost per KWh of storage capacity) or as liquid storage tanks for green Methanol, not including the high cost and low efficiency of the conversion equipment.

For the longer time-scales - months to quarters, overbuilding generation capacity might end up being more cost effective. The chart below shows the same optimization for the same load and PV/wind potentials for the German grid assuming an annual overproduction between 0 to 20%:
Around 5-10% overcapacity there is no more unmatched imbalance beyond the quarterly (3 month) horizon.

For a previous analysis, we have optimized for minimal system cost using PV & wind generation with current LCOE production cost estimates 12-24h demand flexibility (for free) plus a storage cost estimate modelled after a grid scale battery storage plant or a (at this point hypothetical) H2 storage plant consisting of electrolyzers, underground cavern storage and combined-cycle gas turbines.

The resulting optimal mixes are closes to 40% PV and 60% wind (due to lower LCOE for utility scale PV) and a 20% over-production, mostly to compensate for storage losses.

At the current cost of carbon emissions, no existing storage technology can be cost competitive with traditional fossil fuel peaker plants for compensating imbalanced beyond the timescale of days. But the charts above also show how increasing deployment renewable generation and short-term battery storage are squeezing out traditional peaker plants out of the lucrative segments into being a provider of capacity of last resort. For special cases like Switzerland flexibly dispatchable hydro-power plants with large dams and season storage reservoirs could be used to provide sufficient on demand capacity to balance demand and supply across a full yearly range of time-scales.