Electricity Maps Blog
May 29, 2024
-
10 minutes
In current decarbonization strategies, the electrification of all energy usage is an important aspect of becoming independent from fossil fuels. At the same time, the amount of renewable energy installed on the grid is exploding. This increase in renewables means greater variability of emissions and price but also an explosion of negative wholesale electricity prices. It opens massive opportunities for optimization and reducing costs or carbon emissions.
Fortunately, most of the new electricity demand is flexible, meaning that usage can be shifted to times when electricity is cheaper and cleaner (see our previous work with heat pumps, electric vehicles, data centers, and home appliances).
As times of cheap electricity do not always coincide with times of clean electricity, how can we best optimize to save money AND emissions from electricity consumption?
Today, renewable energy has a very low operating cost (close to zero), much lower than fossil-fuel alternatives. On the other hand, the most polluting electricity sources (gas, oil, and coal) are also the ones with the greater operating cost. This correlation between operating costs and operating emissions often leads us to think that electricity price and electricity emissions are perfectly correlated, but is that really the case? Let’s look at an example in Germany where renewables penetration is high.
.png)
In the first week of January 2023 in Germany, electricity prices and carbon emissions peaks coincide on Wednesday morning and evening, Friday evening, and Saturday evening (blue circles). However, the correlation is not perfect (red circles). During the night between Tuesday and Wednesday, prices increased while a drop in carbon emissions was observed. Inversely, on Thursday, carbon emissions remained stable and low while prices increased significantly in the middle of the day.
Optimizing the charge of an electric vehicle based on electricity prices during this week in Germany would not necessarily have contributed to a decrease in carbon emissions compared to a baseline scenario. Why is that?
Thinking about how electricity market dispatch works gives a first answer.
The above graph shows that cleaner sources of electricity are also the ones with the lowest operating cost (also called marginal cost). Generators on the grid are turned on based on the ascending order of operating cost until total generation matches the current demand. The market price is set at the marginal cost of the last generator called (a coal-generating unit in the example of the graph).
This means two things: First, the amount of renewables on the grid will not impact the market price as long as the marginal generator (the last called generator) does not change. Second, electricity demand also plays an important role in setting the market price besides generation (volume and cost).
This impact can be best understood when looking at average price variations throughout the day in 2023 in France, Spain, and Switzerland (three countries well representative of European grids’ diversity). We observe two electricity price peaks during the morning and evening demand peaks in these three grids as in other European grids.
However, electricity carbon emissions follow very different variations between these three countries:
In France, prices and emissions follow a similar trend, but that’s less the case in Spain and trends become the opposite in Switzerland.
We took the analysis further at Electricity Maps by computing the correlation between prices and emissions on the ten European grids with the highest demand between 2020 and 2023 (excluding Great Britain). A correlation of 1 means that price and carbon emissions follow the exact same trend. A correlation of -1 means they follow the exact opposite trend. If equal to zero, it means that trends are not the same nor opposite but unrelated.
Correlation between electricity market prices and electricity carbon intensity for the top 10 European grids in 2023
A first observation is that the correlation is not stable between 2020 and 2023 as it was impacted by the COVID crisis first but most importantly the energy crisis and high prices following the Russian invasion of Ukraine.
A second observation is that there isn’t a single truth for all European grids. Some grids have good correlations, such as Germany, Spain, and Belgium, but others have much lower or even negative correlations, such as the Netherlands, Sweden, and Switzerland.
Instead of looking at a specific week and grid as in the initial example, let’s look at the bigger picture. We now evaluate the carbon impact of optimizing electricity usage based on price for several European countries throughout 2023.
In the rest of the blog post, we will investigate two specific optimization scenarios:
These scenarios consider two different periods: the day in scenario 1 and the night in scenario 2. In each scenario, baseline carbon emissions and baseline costs are calculated by assuming electricity consumption happens during the first three hours. Several optimization strategies are investigated and related emissions and costs are compared to the previously calculated baseline.
These scenarios are illustrated with an electric vehicle but also apply to any other flexible electricity usage, e.g. a dishwasher or a washing machine at home or a 3-hour flexible workload for a computing job in a data center.
First, a price-optimized strategy is applied to these two scenarios on all days of 2023 for the ten largest European grids. The impact on carbon emissions compared to the baseline scenario is discussed below.
In the first scenario where consumption is optimized over the day, the price strategy leads to carbon savings (compared to the baseline scenario) for all grids but South Central Sweden. These carbon savings however remain below the savings that would be obtained in a strategy where consumption would be optimized based on carbon. The price optimization strategy leads to 6.5% of median carbon savings compared to 10.5% for the carbon optimization strategy.
In the second scenario where consumption is optimized overnight, the results obtained by the price optimization strategy on carbon emissions are getting worse. It leads to increased emissions in Switzerland, Spain, and the Netherlands. The gap to savings achieved with a carbon optimization strategy is widening in most zones.
It can also be noted that some grids showing high carbon savings with the price optimization strategy in one scenario, yield very different results in another. Spain is the grid with the most carbon savings for an optimization over the day, yet a grid where emissions would increase for an optimization overnight.
The previous results only focused on the average impact over a year but there can be variations from one day to another. Another useful perspective to analyze the impact of a price optimization strategy on carbon emissions is to investigate the number of days where carbon emissions increase compared to the baseline without optimization.
Again, the second scenario with optimization happening overnight is the one that leads to the poorest results in terms of increased carbon emissions. Optimizing consumption based on price would increase emissions compared to the baseline for 75 days in the year for all grids. It exceeds 100 days for all grids except France and Belgium, and peaks at more than 200 days for Switzerland and Finland.
In the first scenario, a negative outcome is observed for over 100 days in the year in Switzerland, Sweden, Finland, and Italy.
As initially introduced with the example of Germany, the correlation between price and carbon emissions for most of the largest European grids is good but not perfect thus an optimization based on price can increase carbon emissions. Could we optimize electricity usage to ensure carbon and price savings at the same time?
A co-optimization algorithm is used for the following calculations. It enables simultaneous optimization of price and carbon. The details of how this optimization strategy is designed are presented in Appendix A.
To assess the outcome of this combined optimization, the two previously introduced scenarios are again considered for the ten largest European grids. Results show that carbon emissions and costs are simultaneously reduced this way.
Let’s evaluate the three different optimization strategies: based on price, based on carbon, and combined optimization. We first look at the price savings for both scenarios and all three strategies. The carbon-optimized strategy leads to price savings compared to the baseline in both scenarios for all ten grids. However, these savings significantly increase for a combined optimization and almost match savings achieved with a price-optimized strategy. Switching from price optimization to a combined optimization strategy has close to no impact on cost savings.
The benefits of the combined optimization are not only visible in cost but also in carbon savings. Shifting from a price optimization strategy to a combined optimization strategy will greatly improve carbon reductions. Moreover, the combined strategy reduces carbon emissions in all grids compared to the baseline.
These improvements are also visible in the number of days through the year with an increase in carbon emissions compared to the baseline. These are significantly reduced compared to the price optimization strategy. In the case of overnight optimization in Spain, the number of days with increased emissions compared to the baseline was reduced by more than 50% between the price optimization and the combined optimization.
This section estimates potential savings achieved in real life with a co-optimization strategy. The cost savings potential is calculated through wholesale market price which is a simplification compared to reality. An actual EV owner would be subject to a retail price which might be fixed. If subject to a variable retail price, the variation of this retail price would only in part be affected by wholesale price variations, together with other factors such as distribution costs or taxes.
Let’s look at carbon and cost savings unlocked by co-optimization in the case of Scenario 2 (optimization ran overnight for a 3-hour charging session) introduced above considering the car is charged with a 7.5kW charger (common output for a home charger).
Running a co-optimization strategy would lead to the most emissions savings in Poland with 1kg saved per charge session on average (compared to 1.3kg with carbon-optimized charging). Savings also exceed 750g per charge session in Germany and Belgium. In Belgium, this is very close to savings obtained with a carbon optimization strategy. On the other hand, optimizing based on price in Switzerland or the Netherlands would on average increase emissions by 500g and 100g CO2 per charge session respectively.
In terms of price savings, a carbon-optimized charging will yield at least 0.4€ of savings per charge session in all 10 grids. These savings increase in a co-optimization strategy. In both a co-optimization strategy and a price-optimized charging strategy, savings exceed 1€ per charging session in Italy, Poland, Germany, France, Belgium, and the Netherlands.
Considering our EV owner drives a medium EV, the three-hour charge would charge half the car. If we now consider they need to perform this charge every three days, the savings obtained over a year for the co-optimization and price-optimized strategies can be calculated.
On carbon savings, as highlighted above, Poland, Germany, and Belgium are the countries where savings would have been the greatest with around 105kgs saved over the year. This represents 1.3% of the emissions per capita in Europe in 2021.
The price optimization strategy would have enabled our owner to save at least 75€ over the year in all grids but would have also increased carbon emissions in several countries. With a co-optimization strategy, these minimum savings only slightly decrease from 75€ to 70€. Italy, Poland, Germany, France, Belgium, and the Netherlands are the grids where cost savings would have been the greatest with an average of 145€ saved over the year with price-optimized charging and 140€ in a co-optimization strategy.
Price and carbon variations are more complex than we generally believe. The correlation between the market electricity prices and carbon emissions varies across years and European grids and is also most often not perfect despite a strong correlation between operating costs and emissions.
Even in countries with a good correlation between price and carbon, optimizing based on price may lead to an increase in carbon emissions. On the contrary, optimizing solely on emissions leads to price savings on most European grids. This highlights the importance of having access to granular carbon emissions data and forecasts. Relying solely on price signals without considering electricity emissions could lead to adverse consequences for companies striving to minimize their environmental impact.
Running a combined optimization on both price and carbon greatly improves the savings in all the largest European grids. It unlocks cost and emissions savings of a similar magnitude as those respectively obtained on price and emissions optimization alone. Optimizing electricity consumption on price and emissions simultaneously is the best way to leverage the increasing potential of costs and carbon savings offered by flexible electricity usage.
Striving to empower as many organizations and individuals to reduce their emissions, we at Electricity Maps work towards offering the most comprehensive solutions that enable reductions of both carbon emissions as well as price. Already today, we provide 24-hour carbon intensity forecasts for 150 zones worldwide. We are also actively working toward expanding our forecast offering for renewables and prices across the globe.
For a real-world example of how customers leverage our forecasts for co-optimization, explore how Monta uses Electricity Maps’ API to enable EV charging optimization for their users.
To run a co-optimization on price and carbon, we compute a new metric as the sum of the normalized electricity price and the normalized carbon intensity. It is expressed as follows:
Where:
In our co-optimization, we decide to use the full year as the period to compute the mean values. Using the full year allows us to focus optimization on price during times when electricity prices are significantly higher than the rest of the year and focus optimization on carbon intensity during times when carbon intensity is significantly higher than the rest of the year.
Instead of running an optimization based on the variables pt to minimize cost, or cit to minimize emissions, running an optimization based on the variable Mt will seek to minimize both simultaneously.
Explore our free app to understand electricity grids and their CO2 emissions around the world.
