Energy Mix Model

Electricity is unusual among commodities: it cannot be stored in the network itself. Generation and consumption must match continuously, not on average over a year. If they drift apart, system frequency moves away from 50 Hz, and if it moves far enough, protection equipment disconnects plant and load until balance is restored. That is what a blackout is.

This is why "we built enough solar panels to cover annual demand" is not the same as "the lights stay on". Try it: build 60 GW of solar in Central Europe with no storage. Annual generation looks impressive, but coverage sits near 88% because none of it arrives at night. The simulator counts every hour separately, which is the only way the problem becomes visible.

The gap between annual energy and hourly adequacy is the central difficulty of decarbonising electricity, and it is why systems need a mixture of energy sources, firm capacity, storage and flexibility rather than the single cheapest technology.

Variable generation follows the weather. Solar and wind produce when the resource is there, not when demand asks for it. Dispatchable generation can be scheduled: hydro with a reservoir, bioenergy, and to a lesser extent nuclear, which is technically capable of load-following but so capital-intensive that it is usually run flat out.

Three distinct things fix a variability problem, and they are not interchangeable:

Storage shifts energy in time but never creates it. A four-hour battery is excellent at moving midday solar into the evening peak — move to the Mediterranean region and watch the price fall sharply as you add the first few gigawatts. It is useless against a five-day winter wind lull, because it simply does not hold enough energy.

Dispatchable low-carbon generation covers exactly that gap. Turn on the wind drought event in the Atlantic region with no bioenergy built and you will see hundreds of hours of shortfall; add 2 GW of bioenergy and it nearly disappears. Fuel-limited plant that runs rarely is expensive per unit of energy but cheap insurance per unit of reliability.

Overbuilding and curtailment is the third option, and it is a real strategy rather than a failure. Build enough variable capacity that even a poor day meets demand, and accept throwing away surplus on good days. Watch the curtailment figure climb past 50% in high-renewable portfolios: that wasted energy is a genuine cost, and it is why the price stops improving as you keep adding panels.

Comparing generation technologies by capital cost alone is misleading, because a euro spent on nuclear and a euro spent on solar buy very different things. The price figure in this simulator annualises capital cost over each technology's lifetime at a 6% cost of capital, adds fixed and variable operating costs and fuel, and divides by the energy actually delivered to consumers.

Two consequences follow, and both are visible as you move the sliders.

First, utilisation dominates. Nuclear has by far the highest capital cost per kilowatt here, yet a nuclear-heavy portfolio produces some of the lowest prices in the tool, because those kilowatts run more than 90% of the year. The same logic works against dispatchable bioenergy, which is cheap to build but expensive per megawatt-hour when it only runs a few hundred hours.

Second, marginal value falls. The first gigawatt of solar in a sunny region displaces expensive supply at exactly the right time. The twentieth arrives when the system is already in surplus, gets curtailed, and earns nothing while still costing its full capital charge. This is why every technology in this simulator has a point beyond which adding more makes the price go up rather than down — and finding those points is the most useful thing you can do here.

Note also what is deliberately excluded: transmission reinforcement, distribution networks, system services, and the cost of land. Real system costs are higher than the figures shown.

The most important lesson in this tool is that there is no universally correct generation mix. The same portfolio performs completely differently depending on where it is built.

Build a solar-dominated system that works beautifully in Mediterranean Europe, then switch the region selector to Northern Europe without changing a single slider. Coverage collapses. The reason is not that northern solar panels are worse in some vague way — it is that the seasonal profiles are in opposition. At 62°N, winter solar output is roughly a fifteenth of summer output, and winter is precisely when heating demand pushes the load to its annual maximum. The resource and the demand are anti-correlated.

The Mediterranean has the opposite structure: demand peaks in summer because of air conditioning, which lines up reasonably well with solar output, though the daily peak lags several hours behind midday — hence the value of storage there.

The Atlantic coast shows a third pattern, where persistent wind provides high annual energy but still leaves multi-day lulls that need covering. The Alpine region can lean on reservoir hydro for seasonal balancing that no battery could provide. The Baltic region is the hardest case in the tool: northern demand seasonality without the hydro resource to balance it.

The practical conclusion is that copying another country's energy strategy is rarely sound. Optimal portfolios are a function of local resource, local demand shape, and local geography.

Every model omits things, and being clear about what is missing matters more than the numbers themselves. This simulator makes the following simplifications:

No power flow. The grid is treated as a single node. Transmission constraints, voltage, stability, and reactive power are not modelled. The transmission outage event is a crude derating, not a network study.

No system services. Real grids need inertia, frequency response, and reserves. High-inverter systems can supply these, but not for free, and none of that cost appears here.

One synthetic weather year. Results come from a single seeded year. Real planning uses decades of data, because the year that determines whether a system is adequate is usually an unusual one, not an average one.

Daily perfect foresight for hydro and bioenergy. Dispatch allocates each day’s limited energy to the highest residual-demand hours using perfect knowledge of that day. Real operators rely on forecasts and operating constraints rather than perfect information.

Simplified costs. Single representative capital costs per technology, a uniform 6% cost of capital, no learning rates, no supply chain constraints, no grid connection charges. A common cost of capital is applied to every technology deliberately, to isolate the engineering differences between them rather than let financing terms dominate the comparison — in reality each technology attracts a different risk premium and financing structure. The capital cost values are representative figures, not quotes; technology costs move quickly, so they should be updated periodically to stay current.

Fixed import price. The price of imported electricity is held constant. Real interconnector and market prices move hour to hour with neighbouring systems' conditions, and would rise exactly when a shortfall makes imports most needed — so a system leaning heavily on imports here looks cheaper and more secure than it would in practice.

Battery boundary and degradation. The annual simulation solves for a cyclic state of charge, so the battery finishes the year at approximately the same charge level at which it started and cannot supply free initial energy. Storage is otherwise modelled at fixed round-trip efficiency and capacity. Real batteries fade with cycling and age, lose usable capacity and may require augmentation or replacement — effects not captured here.

Lifecycle emissions are indicative. The bioenergy figure in particular depends enormously on feedstock and land-use assumptions, and reasonable analysts disagree by an order of magnitude.

Use this to build intuition about trade-offs. Do not use it to make a decision.