In Croatia, wind output swung from 958.83 MWh in its strongest hour in January 2026 to just 3.79 MWh in its weakest, illustrating how little annual capacity factors reveal about a large wind portfolio. The fleet’s hourly variability has operational and commercial implications for the electricity system and investors. February brought a similar pattern, with output falling to zero in the weakest hour.
In January, Croatia had 30 wind farms in regular operation with total installed capacity of 1,237.15 MW. Two more projects, totalling 77 MW, were under construction or in trial operation. The operating fleet generated 250,003 MWh that month, giving it a capacity factor of 25.57%.
Daily production in January ranged from 1.79 GWh to 19.12 GWh. In February, the strongest hour recorded 1,016.42 MWh, while the weakest had no wind output. The month’s capacity factor was 26.2%.
Hourly variability and balancing costs
These swings pose a balancing challenge for the electricity system. A portfolio capable of producing around 1 GW in one hour can behave almost like a different asset just hours later. The financial effects start with forecasting and the management of deviations from expected output.
Wind farms nominate their expected production before delivery, and traders can adjust their positions on the intraday market as forecasts improve closer to real time. Some difference between nominated and actual output nevertheless remains. The balancing system must absorb it, while the market and contractual structure determine who ultimately pays.
A wind farm may carry imbalance exposure directly, or a trader may take it on in exchange for a fee or a less attractive PPA price. An aggregator can manage several assets together. In each case, variability has an economic cost even though the wind resource itself is free.
Correlation within wind portfolios
Aggregation does not automatically remove that cost. Projects exposed to the same regional wind pattern can see their output rise and fall together, making the diversification benefit smaller than the number of assets might suggest. The relevant measure is the correlation between their generation profiles.
Wind farms hundreds of kilometres apart can still have strongly correlated output if they are exposed to the same large-scale weather system. By contrast, assets in different wind regimes may offset one another even within a smaller geographic area. This distinction is increasingly important for wind portfolios across Southeast Europe.
Developers assembling multiple assets should assess not only each project’s P50 production but also the covariance between their hourly production profiles. Lenders financing a portfolio should determine whether the projects genuinely reduce volatility or simply combine generation that moves in tandem. Both assessments affect the value of diversification.
PPAs and grid concentration
The same issue applies to PPAs. A corporate buyer purchasing output from several wind farms may expect diversification to produce a relatively smooth delivery profile, but that assumption can fail when the plants respond similarly to regional weather. The portfolio may then overproduce collectively during strong-wind periods and underproduce together when winds are weak.
Exposure to replacement power can increase precisely when market prices are high. Transmission geography adds another form of correlation: many Croatian wind farms are concentrated in coastal and southern areas with attractive wind resources. HOPS reported that the operating fleet in early 2026 was particularly concentrated in Zadar, Šibenik-Knin, Split-Dalmatia, Dubrovnik-Neretva and Lika-Senj counties.
Clustering can create economies of scale in project development, but it can also concentrate injections into common parts of the network. The combination of correlated weather and shared grid infrastructure therefore warrants greater attention in asset valuation. Geographic spread alone does not establish that a portfolio’s output will be diversified.
Storage and balancing services
Storage could eventually mitigate some variability, but it is not a one-size-fits-all solution. A battery designed to absorb brief forecasting errors needs a different operating strategy from one intended for two-hour wholesale arbitrage. A battery firming a PPA may also require different power and energy characteristics from one used mainly for balancing services.
Croatia’s balancing framework is becoming more open. HOPS allows technically prequalified individual network users, aggregators, independent aggregators and renewable-energy communities to provide balancing services, subject to qualification and contractual requirements. This creates an opportunity to turn variability into flexibility, but it requires understanding the asset’s physical behaviour.
Annual production smooths out the extremes, while debt, PPAs and electricity systems face them hour by hour. For investors across Southeast Europe, the issue is not only how much energy a portfolio produces over a year. It is also how often its wind farms rise and fall together, and what that correlation costs when the system has to respond.
