Blog

Four months in, SA Tesla battery is showing mixed results in energy arbitrage

Tesla’s big battery in South Australia was officially switched on in late November 2017. At 100 MW/129 MWh, it is the largest lithium ion battery in the world and signals the beginning of a new era in how we manage the electricity system. In this article, we take a look at how the battery has operated in its first four months.

Before we delve any further, it is important to note that the Tesla battery is really two systems:

1. 70 MW/10 MWh is contracted to the South Australian state government for the purpose of providing grid stability services;
2. The remaining 30 MW/119 MWh can be used by the Hornsdale operator to trade in and arbitrage the energy market.

The economic viability of large scale batteries is largely dependent on how much revenue they are able to extract from the wholesale energy market and the various ancillary service markets. In this article, we will focus on the economic performance of the battery in the energy market only.

We use a software called NemSight by Creative Analytics (part of the Energy One group) to analyse the operation and bidding behaviour of the Tesla battery.

Tesla battery is being heavily utilised

The Tesla battery is certainly not sitting idle in the market. Figure 1 shows the operation of the Tesla battery on a five minute basis from 1 Dec 2017 to 31 Mar 2018. As can be seen, the battery is being utilised very frequently. In fact, in 63% of the dispatch intervals, the battery was either being charged or discharged. Only 37% of the time was the battery not being used in any way. Furthermore, nearly 40% of these zero utilisation periods occurred in December while the battery was still being tested.

From Dec 2017 to Mar 2018, the Tesla battery consumed an average of 116 MWh per day for charging. In contrast, it delivered an average of 94 MWh per day back into the grid. From this we can work out that the average efficiency of the battery has been 82%. As expected, the efficiency of the battery is lower under real world conditions than the spec sheet efficiency of 88%, which is calculated at 25°C.

Tesla battery made $1.4 million in the energy market, but is losing money 47% of the time

Figure 2 shows our estimate of the value that the Tesla battery received from selling electricity into the energy market versus the cost of buying electricity to charge the battery. We estimate the total net revenue from the energy market to be just under $1.4 million. The overwhelming majority of this came in January 2018 when the energy market experienced the highest volatility. In contrast, the Tesla battery barely made any money in December and March.

When arbitraging the spot market, the aim of the game is the same as trading stocks: buy low, sell high. A crucial difference is that you not only have to sell at a price which is higher than your buy price, but you also have to cover the cost of the extra energy that is needed to charge the battery (because energy efficiency is less than 100%). Therefore, the days with the most volatile pricing offer the biggest opportunities for arbitrage. Figure 3 shows that the Tesla battery made 95% of its net revenue in just five (very volatile) days.

If we exclude these five days, the average net revenue for the battery is a measly $530 per day. In fact, on 57 days (47%)  the Tesla battery actually lost money in the energy market. The total losses over the 57 days add up to about $135,000.  As mentioned earlier, this analysis looks at the revenue from the energy market only. It does not include any of the revenue received from providing ancillary services. Nonetheless, our numbers do suggest that the operators of the battery will need to be careful to avoid needlessly cycling the battery for little financial gain. This is an important consideration because the lifetime of a battery is strongly related to how many times it is cycled.

Furthermore, “buy low, sell high” is not as easy as it sounds. The energy market is incredibly complex. In our experience, a successful bidding strategy needs to be underpinned by advanced predictive analytics and co-optimisation. Failure to do so can result in the asset significantly undershooting revenue expectations. Some food for thought as the race to build big batteries begins.

Author: Marija Petkovic, Founder & Managing Director of Energy Synapse
Follow Marija on LinkedIn | Twitter

QLD smashes electricity demand record during heatwave but narrowly avoids high pricing

Midway through February 2018, Queenslanders sweltered through a severe heatwave, which saw statewide average temperatures exceed 40°C for several days. These extreme conditions resulted in electricity demand in QLD exceeding the 9,400 MW threshold for the first time ever. Not just once, but four days in a row starting from Monday the 12^th of February (see Figure 1).

The previous record for the highest ever electricity demand was 9,369 MW and was set on 12 February 2017. This figure was eclipsed during the 5pm trading interval on 14 February 2018 when total demand reached 9,840 MW. During this time, the Australian Energy Market Operator (AEMO) had also declared an actual lack of reserve (LOR 1) event for QLD. Despite these record breaking levels of demand, the 30 minute spot price remained at relatively modest levels below $350/MWh over the four days (see Figure 2).

Some might look at this modest pricing and conclude that QLD was a long way away from reaching $10,000/MWh. However, this could not be further from the truth. In order to understand just how close the market was to a price spike, we need to examine the generator bidstack.

Generator bidding data shows QLD narrowly escaped extreme pricing

The dispatch process is complicated but to put it in very simple terms: generators submit bids to AEMO for each five minute dispatch interval, signalling how much electricity they are willing to provide and at what price. AEMO’s central dispatch process then orders these bids from least to most expensive. The least cost generators (taking into account constraints) are dispatched to serve the demand in the market. The marginal bid (i.e. the last/highest cost generator that is selected) sets the price for everyone in that dispatch interval. The bidstack changes every five minutes depending on market conditions and generator bidding behaviour.

The data sets that contain the bidding information for the National Electricity Market (NEM) are notoriously difficult to access and work with. For this purpose, we use a very nifty software package from Creative Analytics (part of the Energy One group) called NemSight. NemSight allows us to extract relevant bidding data with ease and in an intuitive, easy to understand format.

For our exercise, we have extracted the bidding data for two days: 11 October 2017 and 14 February 2018. These two days were very similar in terms of price outcomes. The 30 minute spot price peaked at $259/MWh on 11 October 2017 compared with $338/MWh on 14 February 2018. Furthermore, both days had a similar average daily price ($115/MWh on 11 October compared with $105/MWh on 14 February). Despite these similarities, the bidding data tells a very different story in terms of how close each of these days came to a price spike.

Figure 3 shows a comparison of the generator bidstack for each day. We have shown the dispatch interval where pricing was the closest to spiking on each respective day. The regional bidstacks (in teal) have been adjusted to account for constraints and imports via the interconnectors. In red, we have the corresponding electricity demand for that particular dispatch interval. The data shows that 11 October 2017 was more than 600 MW away from reaching price levels greater than $10,000/MWh. In contrast, 14 February 2018 was a knife’s edge away from very high pricing with very little room for things to go wrong in the market. QLD energy consumers were very lucky to walk away from the heatwave relatively unscathed.

Coal and gas generators were less available when needed the most

In addition to the static snapshot in Figure 3, we also examined how the bidding behaviour of QLD generators evolved on 14 February 2018. The video below is an animation from NemSight. It shows how the volume offered, and the price at which it was offered, changed throughout the day. The interesting thing to note is the dropping availability in the afternoon, which pushes prices closer to the brown zone (>$10,000/MWh).

This reduction in availability was predominately due to Gladstone (black coal), Millmerran (black coal), and Braemar 1 (gas) as shown in Figure 4. These fossil fuel generators lost availability when needed the most in the late afternoon. This raises several questions for policy makers. In particular, what portion of fossil fuel capacity can reliably meet peak demand?

If there is one thing that our analysis here has shown, it is that when it comes to the energy market, there is almost always more than meets the eye. This highlights the need to understand underlying market dynamics rather than relying on conventional wisdom or incomplete data to make decisions. This is more important now than ever as we enter an unprecedented transformation of the electricity sector.

Author: Marija Petkovic, Founder & Managing Director of Energy Synapse
Follow Marija on LinkedIn | Twitter