The benefits of minute-by-minute local energy trading
In our recent post The Minuteman of the Half-Hour, we explained some of the benefits of local energy trading based on one minute settlement periods. The whitepaper below, produced for Swanbarton’s LEMDEx local energy market project, explores the issue in a little more depth.
A Local Energy Market (LEM) is an intuitively simple idea. A consumer is using energy, and their neighbour, who has a solar panel, say, is generating more than they can use. Rather than selling this energy back to a supplier, who then sells it on to another consumer, the generator sells direct to their neighbour at a mutually beneficial price that’s negotiated between them. The generator gets a better price, incentivising renewable generation and benefitting the environment, their neighbour gets cheaper energy, and the varying price of electricity encourages behavioural change and investment in storage, easing pressure on the distribution network.
This seductively simple description, however, masks a more complex reality. Because of the expense of purchasing and installing small scale renewable generation assets, these assets can’t compete directly with large scale generators such as power stations in terms of wholesale price. Once the network charges and levies which, at present, make up almost half of the average energy bill are factored into the price of small scale renewable energy, the price may become too high for consumers to bear. The business case for small scale renewable generation post-Feed-in-Tariff could, however, be made more attractive in the context of a LEM. This is because renewable energy that is generated and simultaneously consumed in the same local area is not only environmentally friendly, but also has little to no impact on the transmission network and a greatly reduced impact on the distribution network. It’s therefore arguably deserving of reliefs on network charges and green levies. This would reduce a consumer’s bill and allow a higher proportion of that bill to be paid to local generators producing green energy. Government and regulators have the opportunity to demonstrate their commitment to renewable energy by making that change.
However, the assumption that locally matched energy has a low impact on network infrastructure is not always a safe one. At present in Great Britain, energy is traded based on half hour, or longer, totals. But if local trading aims in part to balance the distribution network, then this simply won’t do. Although a producer may generate the same total amount of energy over a half-hour period as their neighbour consumes, it’s unlikely that this energy will have been produced or consumed at a constant rate within the half hour. Renewable generation output tends to be subject to rapid changes in the weather, and the power used by a consumer can change literally at the flick of a switch. In a worst-case scenario, it may be that a producer generates at 1 kWh in the first fifteen minutes of the half hour period, but the consumer uses 1 kWh in the second fifteen minutes. Whilst this is a match in terms of half hour energy volumes, it’s obvious that this match offers no benefit to network infrastructure; energy rushes out of the local network area and into the transmission network during the first fifteen minutes of the half hour and then rushes back in during the second fifteen minutes. Worse, flexible assets dispatched to match half hour energy totals might actually make network balancing more challenging. This is an extreme example, but our research shows that this would be a significant problem for any real-world LEM that attempted to operate on a traditional half hour basis.
Let’s take a look in more detail: The graphs below show current measurements for a solar array and a light industrial consumer over the course of 12 hours, first averaged at one second intervals (a good proxy for instantaneous power), then at one minute and half hour intervals.
These graphs show that a LEM based on half hour totals would significantly overestimate the total matched energy, that is, the total amount of time for which a producer and consumer are matched in terms of power. The ‘true’ matched energy is the area in common below both lines on the above graphs; any area between the lines is not really matched. If we look around the eight-hour mark, we can see clearly that, as the averaging periods increase, the blue line moves from above to below the orange line. In fact, if we were to take an average of the whole twelve hours, we would have two straight lines, with the blue line completely underneath the orange one. Therefore, unless the two lines of instantaneous power never cross, the longer the timestep over which average energy is calculated, the more the true matched power will be overestimated. In the course of the day above, the half hour overestimation was 6.73%, whereas the one-minute overestimation was less than 1%.
The next set of graphs show how the one second average consumption power of a light industrial consumer compares to first the one-minute average and then the half hour average. It’s immediately apparent that the one minute average is usually closer to the one second ‘‘true’ power than the half hour average. Therefore, a flexible asset dispatched to match the one minute average will be much closer to matching real power, for much more of the time, than one dispatched to match the half hour average.
It might be thought that variations within the half hour are unimportant, because, with large enough numbers in a LEM, opposite variations will occur, thus making matches based on half hour totals more or less right on average. However, this won’t always be true. On the graph below, for example, we can see that, during the period 7-7.30am, consumption jumps from around five amps below the half hour average to around 40 amps above the half hour average.
A generator considered to be matched in terms of half hour total energy with this consumer would be out in terms of power by up to 200% for a good proportion of the half hour. Moreover, this jump isn’t a random error that will be cancelled out; as timed processes, such as heating, come online, industrial shifts start, and so on, it’s likely that other consumers will also be beginning to consume more around this time of day.
The situation for generators is likely to be as bad, if not worse, as all renewable generators in a local area are likely to change their output at similar rates at similar times, due to changes in light levels and prevailing weather conditions. The graph below shows the one second and half hour average output of a solar array over the course of a day.
The output from 8-8.5 hours into the data is about 3 amps on average, but the spread in terms of real time current is around 10 amps, with a clear downwards trend that will likely be shared by all local PV generators. A consumer that tried to use or store the output of this array based on half hour total energy, i.e. by consuming at a constant 3 amps, would in reality hardly absorb any of the energy being produced to begin with, and towards the end of the half hour would actually be taking energy from the grid.
By contrast, the graph below, focusing on this half hour period, shows how much better a LEM based on one minute totals would perform.
Whilst the trend within a minute may be shared by other PV arrays in the local area, minute-by-minute matching allows a flexible asset to track the trend across the half hour. The spread across any particular minute is much smaller than across the whole half hour, meaning that a flexible asset dispatched to match the output of the PV array minute by minute would be much closer to the true match. There are still variations within a minute, but these are most likely caused by very localised weather phenomena, such as small clouds, which won’t affect every PV generator in a local area simultaneously (similarly for wind turbines and gusts of wind). These variations are therefore much less likely to be replicated in the output of other generators in the local area than is the downwards trend across the half hour. On average, therefore, a LEM based on one minute energy totals would be much closer to real time power balance than one based on half hour energy totals.
Swanbarton’s Real Time Trading Platform (RTTP) is able to arrange energy trades between generators, consumers and flexible assets on a minute by minute basis. Trials have shown that, in a LEM consisting of industrial and commercial consumers, with PV generators and battery storage, this can reduce peak power flows into and out of the local network area by 55% and total energy flows by 36%. As well as reducing the magnitude of peak power flows, RTTP can also greatly reduce the amount of time for which these peaks exist. Most curtailment events in constraint managed zones occur within three minutes of critical power thresholds being exceeded. This is too soon for most traditional balancing services to be able to step in. But by providing a price signal for the dispatch of flexible assets within a minute, RTTP can dramatically reducing the need for curtailment of precious renewable energy, impacts on productivity, and the strain on local network infrastructure. The result is less money spent on network reinforcement and the savings can be passed on to consumers and renewable energy generators.