Written by: Lachlan Hensey

A koan is a riddle with no sensible answer, used in Zen Buddhism to challenge logical thinking and to learn to live with ambiguity. For example, this one is well known: “Two hands clap and there is a sound – what is the sound of one hand?” It is the impact of the two hands – their interaction – that produces the sound, which can’t be attributed to one or the other.

Do batteries reduce greenhouse gas emissions? Inadvertently, this title appears to be a koan. The deeper one looks, the less sense the question seems to make, for reasons elaborated over the course of this article. Of course, to decarbonise the energy system, batteries are essential. But identifying precisely how – and how much – they reduce emissions is vexing. Like the sound of clapping, emissions intensity is the result of interactions within the energy system – market bidding, generator dispatch, network constraints, for example. Beyond fossil fuel generators, it is therefore no easy task to attribute responsibility for emissions, or their reduction, to one element in a complex, interconnected system.

This article seeks to do a few things: first, we will have a closer look at the title question and contextualise it in the role that battery energy storage plays in the energy transition; second, we will unpack a few ways of calculating carbon abatement (a.k.a. emissions reductions) and consider which, if any, are appropriate in the case of batteries; and finally, I reflect on what we’re really trying to do by measuring emissions reductions and – when it comes to batteries – whether we should even care. The article predominantly regards “small” unscheduled batteries (i.e., neighbourhood or community batteries), however much of the content is broadly applicable to utility-scale and even household batteries.

1. Can batteries reduce emissions?

Battery energy storage systems (BESS) are net energy consumers due to roundtrip losses – that is, for a typical lithium-based BESS [1], for every 100kWh that is stored, only about 80-90kWh can be discharged because of energy losses. These occur when AC power is converted to DC power (and vice versa), and when powering parasitic loads within the BESS such as the control devices and cooling systems. While BESS datasheets might claim roundtrip efficiency of >90%, this is generally a calculation of the (theoretical) efficiency of the battery modules and inverters, and ignores parasitic loads. This is reasonable enough – parasitic loads are variable and context-dependent – but it doesn’t reflect real-world battery operation.

So, if batteries increase the total amount of energy consumed, can they reduce emissions? My perspective is similar (while being an entirely different sentiment) to this guy:

 

Batteries don’t reduce emissions, clean energy does [2]. But a renewable energy-based system needs batteries for various reasons – principally, to store surplus energy for when it’s needed (firming), and to provide the system services traditionally provided by spinning machines in fossil fuel plants.

The joke on Mr Larson’s t-shirt is about how we attribute responsibility in a collaboration. When batteries collaborate with renewable energy (RE), the system benefit is greater than either technology could achieve alone, but it’s a stretch to definitively attribute that benefit to batteries (recall the koan at the beginning).

Let’s rephrase the initial question to: how can batteries enable emissions reductions? Now we have a few reasonable answers:

• By charging when RE is abundant, batteries increase system load, reducing curtailment of RE generation caused by thermal/voltage constraints on the surrounding network, the emergency backstop, or even negative wholesale prices. Ultimately this increases the overall percentage of RE in the generation mix, reducing the emissions intensity of energy.

• By providing essential system services such as frequency control, synthetic inertia and voltage regulation, batteries enable the exit of fossil fuel generation as we no longer depend entirely on spinning machines for the stability and security of the system.

• By discharging when RE is scarce, batteries firm the grid and may displace fossil fuel generation.

Although these are genuine benefits, astute readers will notice that they remain conditional and/or somewhat indirect. When coal is eventually put out to pasture, the role of batteries in ensuring system security may go largely unheralded. Additionally, although some functions (e.g., frequency response) are ‘global’ to the system, some benefits depend very much on where the battery is installed or when it operates. For instance, to reduce solar curtailment due to voltage rise in a neighbourhood, the battery must be connected to the same LV network as homes that experience curtailment. The same goes for utility-scale curtailment on the HV network.

I tend to think of carbon abatement via batteries in two categories: operational emissions reductions, whereby a battery displaces emissions-intensive generation – for example, offsetting the use of a gas peaking plant; and long-term decarbonisation, whereby batteries facilitate a predominantly renewable energy system by providing necessary services and functions.

Operational emissions reductions are immediate and, theoretically at least, measurable. Long-term decarbonisation is indirect, highly complex, future-dependent, and therefore a battery’s contribution is only measurable in a loose, conditional, modelled sense, if at all.

 

2. Calculating carbon abatement

There are several approaches to measuring carbon abatement, each with some degree of legitimacy and usefulness. Unfortunately, measuring the absence of something re-introduces the attribution ambiguity we just tried to avoid by focusing on enabling emissions reductions rather than reducing emissions directly. If you want to claim to have reduced 1 tonne CO2-e, you have to show that whatever intervention you made is the reason why those emissions didn’t happen. If you install rooftop solar and your grid energy consumption goes down 30%, this is fairly straightforward. If you are a battery shifting energy from one part of the day to another, this is decidedly tricky. We are back trying to attribute responsibility within a complex system.

When working on community battery projects, we sometimes hear stakeholders refer to the battery as ‘charging on local solar’ and therefore reducing emissions. This is an understandable perspective, but it stands on shaky ground.

A grid-connected (front-of-meter) battery charges from the grid. There may be energy on the surrounding network that was exported by rooftop solar systems, but it makes no difference in the eyes of the energy market, or from a system-level emissions perspective, where the energy came from. Consuming energy next to a wind turbine is no cleaner than consuming energy next to a coal plant if both generators feed into the same shared pool of energy.

More to the point, the battery will not increase the overall proportion of RE in the system by storing it for later unless the solar would have otherwise been curtailed. If there is no such curtailment, the exported solar would still be consumed, either by a neighbour or, in extreme circumstances, by flowing back up into the HV distribution network to flow elsewhere (although in this case there is some likelihood of local solar curtailment from voltage constraints).

On the other hand, if a neighbourhood is awash with solar and customer voltage levels are pushing 253V, then the load of a charging battery will help to avoid derating or shutdown of solar inverters by lowering the voltage, enabling more clean generation. For instance, the Fitzroy North Community Battery typically charges at ~53kW and lowers the voltage on each phase of the network about 3V.

2.1 Using the National Greenhouse Account Factor 

One standard method for calculating carbon abatement is using the Australian National Greenhouse Account Factors, a document which includes emissions factors from a plethora of activities, from industrial processes to land use change. The 2023 scope 2 emissions factor for electricity consumption in Victoria is 0.79kg CO2-e/kWh, which means that for every kWh of energy one consumes, an estimated 0.79kg of carbon dioxide equivalent is released into the atmosphere.

A handy rule of thumb, perhaps, but as there is only a single emissions factor applicable for the whole year, it is hardly useful when measuring time-sensitive activities like a battery charging and discharging within a day. This is because the emissions intensity of energy generation fluctuates constantly depending on the generation mix. All we end up calculating is a rough guess of the emissions associated with the energy lost through charging and discharging, while the emissions impact associated with the energy stored and then discharged simply cancel out to nil.

2.2 Using the Average Emissions Factor  

A more time-sensitive approach is to use the average emissions factor (AEF) for every relevant 5-min energy market trading interval. The AEF is formulated by taking into account the emissions intensity of every generator powering the system at a given time and weighted according to how much energy it generated. The main benefit here is that the AEF reflects far more accurately the likely emissions associated with consumption at a specific time. Note, however, that AEMO’s published emissions factors do not include rooftop solar.

For example, at the time of writing (1pm Fri Dec 13th, 2024), renewables make up 72% of generation in Victoria. The grid-average emissions factor is 0.31kg CO2-e/kWh, and last night the peak in emissions intensity was 0.79kg CO2-e/kWh (incidentally, the 2023 National Greenhouse Account Factor). It is reasonable to suggest that charging a battery when the grid-average emissions intensity is low and discharging when it is high would result in a net reduction of emissions (‘emissions arbitrage’, perhaps). Better yet, this would generally be effective in generating revenue through wholesale market arbitrage, since prices are typically low when average emissions intensity is low, and vice versa.

Using the figures above, charging a 100kWh capacity battery right now would result in 31kg CO2-e of emissions, and discharging 85kWh (accounting for roundtrip losses) during last night’s peak emissions intensity – ignoring the time travel required – would offset 67.15kg CO2-e for a net reduction of 36.15kg CO2-e.

 

There are two issues with this. First, simply charging and discharging would only result in emissions reductions (according to this approach) if the difference in emissions intensity between the time of charging and discharging is significant enough to offset the roundtrip losses. Depending on the roundtrip efficiency of the battery and the RE penetration, this ‘spread’ in emissions intensity must be at least roughly 18-25% just to break even. For much of the year – as shown in the example above – this is easy. But over the course of the year, it’s possible that the battery would increase emissions.

 

 

The second issue is that it doesn’t reflect how the energy market actually works. If demand is forecast to increase by 10MW from one interval to the next, generators don’t simply scale up equally to meet demand [3]. For each market interval, generators bid their available power capacity at different prices, and the AEMO dispatch engine finds the cheapest combination of generators that meet demand and abide by the physical constraints of the network. Hence, bids are accepted largely in order of price, with the bid of the most expensive generator – the marginal generator – required to meet demand, setting the wholesale price of energy paid to all generators (regardless of what they bid). Currently Speaking provide an excellent explanation here.

This implies that the decision to charge the battery is potentially implicated in how generators are dispatched. Demand can directly influence emissions intensity by causing additional, potentially dirty, generators to come online or increase their output. Consequently, some have suggested using the Short-Run Marginal Emissions Factor (MEF), the emissions intensity of the marginal generator, to measure the emissions impact of battery operation.

2.3 Using the Short-Run Marginal Emissions Factor (SR-MEF) 

The SR-MEF reflects the immediate impact of changes in electricity demand on emissions, considering which generators ramp up or down in real-time. The rationale for using the SR-MEF [4] is that it more accurately reflects the emissions impact of consumption (and displaced generation) at a specific time by taking into account how generators are dispatched through the market. Like the AEF approach, one could multiply the kWh of charged energy by the emissions intensity of the marginal generator at the time that it was charged, and then the same for discharged energy. The net result would show the emissions impact.

It may seem like a sharper tool, but that doesn’t mean it’s useful. Following the SR-MEF approach would tell you to charge the battery when the SR-MEF is close to 0 (i.e., a wind, solar or hydro marginal generator) and discharge the battery when the SR-MEF is high (i.e. fossil fuel marginal generator). Take a look at the graph below showing the proportion of time each generation source was the marginal generator by time of day in Q2 2024 (whole of NEM; adapted from AEMO data).

 

The first thing to notice is that hydro is the most likely marginal generation source at almost every hour. While hydro is clean, leaving aside environmental flows and other hydro complexities, it’s not ideal to be charging a battery from another form of stored energy that’s often required to meet peak demand – and tends to be expensive.

A second observation is that solar and wind are the least likely marginal generation source (if we lump black and brown coal together). Because solar and wind have zero fuel costs, they are likely to bid their full capacity at very low or negative prices and are unlikely to be able to increase their output if demand increases unless they were previously being curtailed. If wind or solar are setting the price, it’s likely negative.

Coal generators, on the other hand, typically can’t operate below 25-40% of their maximum capacity, so they may also bid at negative prices to avoid having to shut down and restart, which may take over 24 hours. By virtue (*cough*) of rarely shutting down – at least, deliberately – and through spurious bidding practices enabled by market dominance, coal ends up as the marginal generator with some regularity, but very rarely when batteries would prefer to discharge: in the evening peak demand period.

Following the SR-MEF approach to its logical conclusion would mean charging most often during the expensive peak demand period (lunacy!) and discharging in the early morning (or even in the middle of the day), helping no one. You could moderate that approach to avoid doing silly things (technical terminology), but you would most likely end up with an ostensibly random dispatch approach that neither progresses the energy transition by firming renewable energy, nor generates revenue through market arbitrage.

If my opposition to the SR-MEF approach was purely about ensuring the commercial viability of energy storage, you could say ‘them’s the breaks, batteries shouldn’t increase emissions to make money’. But there are several reasons why the SR-MEF approach is unsatisfactory:

1. The SR-MEF approach assumes that increasing system load by charging a battery makes the battery responsible for the consequent increase in emissions from energy generation. In any moment in time, loads fluctuate significantly across the NEM, and it doesn’t pass the sniff test to attribute a net rise in demand to a specific battery – particularly, for example, a 100kW neighbourhood battery – simply because it is the object of analysis. You could divide responsibility equally among all loads, but that’s another approach entirely.

2. Batteries are not only reducing operational emissions by displacing fossil fuel generators. Recall that batteries are essential for firming the variable supply of renewable energy, providing system services, and addressing curtailment – all of which, I posit, result in far greater carbon abatement over time by enabling decarbonisation of the energy system rather than focusing on moment-to-moment reduction in operational emissions.

3. Decarbonising the energy system requires lots and lots and lots and lots (ad infinitum) … of energy storage, which simply won’t happen if batteries operate at a loss (even if subsidised by governments, i.e., taxpayers).

Simply put, given that the energy sector is crying out for more battery storage to address the duck curve and enable more RE generation, one should be suspicious of any method that suggests doing the opposite. Even if we were to conclude (in my view, wrongly) that the SR-MEF approach was correct for calculating emissions reduction, we shouldn’t extrapolate that it shows the best way to operate a battery to maximise decarbonisation of the energy system.

The SR-MEF approach partially addresses the dynamics of the energy market but inherently ignores what we want to achieve through the energy transition, specifically vis-à-vis load management and firming. In this context, its real-time granularity is also its key drawback. But perhaps the most significant critique of the SR-MEF approach is that AEMO’s generation dispatch is based on demand forecasts, not real-time fluctuations in demand, which are typically met by frequency regulation services.

2.4 Considering balancing services

If small batteries don’t directly influence marginal generation, what effect does their operation have on the system? As demand across the NEM is constantly fluctuating, generation must vary in alignment to maintain a stable frequency of 50Hz. If demand drops, scheduled generators must gently throttle back their output, or the frequency will rise due to the oversupply of power (and vice versa).

This is achieved primarily through two mechanisms: Primary Frequency Response (PFR), whereby scheduled generators are required to respond to minor deviations in frequency; and Regulation Frequency Control Ancillary Services (regulation FCAS), whereby AEMO sends a signal to market participants every 4 seconds to manage frequency deviations. The generators primarily responsible for these balancing services are scheduled, and predominantly fossil fuel, generators. If an unscheduled (as they typically are) community battery starts charging at 100kW, it won’t be solely responsible for increasing the output of the marginal generator; most likely, scheduled generation would increase its output an incredibly small degree.

Now, one could go about calculating the emissions intensity of the generators providing these balancing services at the time of charging and discharging in order to estimate a battery’s emissions impact, but frankly, this feels absurd and disconnected from our understanding of the role of batteries. It makes little sense to suggest that because a load is small it only affects the balancing services.

Whether you charge a battery or boil a kettle, balancing services respond to the aggregate balance of supply and demand within the entire system, not individual loads and generators, and small loads have large aggregate impacts. The interconnectedness of innumerable small parts of the energy system prevents us from isolating a single battery to understand “its impact”. Its impact depends on various other elements in the system, all of which are in a collective, dynamic and contingent relationship with each other.

This isn’t to say ‘oh, it’s all too complex, let’s just not.’ My point is rather that the question of how much a battery reduces operational emissions is, to me at least, intrinsically irresolvable. The more fine-grained an approach one takes, the more accurate one tries to be, the more counter-intuitive the approach seems and the less sense it seems to make. Just like the sound of one hand clapping, it makes no sense to measure the emissions impact of a battery when this impact is a consequence of relationships among a multitude of other factors beyond the battery.

Unlike community batteries, the operation of scheduled batteries is captured in AEMO’s forecast and dispatch, so balancing services are largely irrelevant to their emissions impact. They can directly influence generation dispatch through their bidding.  This highlights an important point: when using the above methods, the operation and rules of participation in the energy market are of greater consequence than how the battery is actually operated in determining the battery’s purported emissions impact.

3. Reflections: What are we trying to do here?

Let’s sum up some key takeaways before reflecting on some broader questions. Here’s a few:

• The role of batteries isn’t to reduce emissions per se, but to maximise the contribution of renewable energy and take on some responsibility for managing the stability and reliability of the power system.

• Using the methods discussed, the operation and rules of participation in the energy market are of greater consequence than how the battery is actually operated in determining the battery’s purported emissions impact.

• Trying to calculate the emissions impact of battery operation inescapably requires making some judgements – even if you are not aware of doing so – about how to attribute responsibility for consumption and generation dynamics, and therefore which method you deem appropriate.

• The various methods described in section 2 are all legitimate, rational and potentially useful in specific contexts, and yet they would all yield very different answers (and guidance) vis-à-vis battery operation.

First, calculating emissions reductions is often largely unscientific and more akin to economic forecasting, which is hardly a beacon of reliability. Perhaps that’s why it’s called carbon accounting. The fact that different methods can yield very different but legitimate answers is the first clue that perhaps we haven’t quite done enough of the hard work in defining what we are trying to measure and accomplish.

Second, if a renewable power system relies on batteries, what is the point of measuring their operational emissions impact? As a thought experiment, imagine an energy system with a minimum of 85% RE penetration. The battery would almost always reduce emissions if operated to do so, but to an increasingly smaller degree as RE penetration increases. This system would only be feasible by virtue of the role batteries play in firming variable renewable generator power output and providing system services, so to measure the emissions impact seems, to me, of minor relevance. The operational emissions impact of the batteries would be marginal, but the long-term decarbonisation impact is significant but difficult to measure.

Given that measurement matters, whether for policymaking, reporting, or grant applications, I’m not suggesting we do away with it entirely. But we need to reconsider what we are measuring, and whether it’s aligned with what we’re trying to achieve. Otherwise, we are evaluating potential projects or interventions according to criteria that don’t reflect our objectives.

Some initiatives might receive support even if they don’t help much, while others might be ignored despite making a larger contribution to decarbonisation in the long run. This is because their benefits take longer to show and don’t immediately reduce emissions from daily operations. This is a symptom of carbon fetishism, whereby the basis of the global response to climate change is the commodification of CO2.

The collective preoccupation with CO2 as the villain of the story blinds us to the wider systemic causes, impacts and possible responses to climate change. Measuring the operational emissions impact of battery operation will simply reflect the emissions intensity of the system it interacts with. So, what could we measure instead, and how could we do it? Ideally, we want to know whether a battery operates in a way that would help to:

• unlock additional RE generation,

• firm variable supply,

• reduce peak wholesale prices,

• reduce the risk of minimum system load,

• incentivise further investment in renewable generation,

• or reduce reliance on fossil fuel generation.

Over the long term, these factors are arguably more important to decarbonising the energy system than day-to-day shifts in operational emissions. And although difficult to measure with confidence, there are justifiable proxies we could start with. For example:

• To measure utility-scale renewable enablement (reduced curtailment), one could measure energy charged by a battery during periods of curtailment. Conceivably, any kWh charged during this period may have otherwise been curtailed.

• To measure reduced risk of minimum system load, one could measure a battery’s charging power during periods approaching or below an agreed threshold of system load. Note that renewable curtailment and minimum system load are likely to coincide.

• To measure contribution to firming, one could measure energy throughput when time-shifting energy from periods of high to low renewable penetration.

It may be that the suggestions above remain too simplistic and share some of the same challenges as measuring carbon abatement. The advantage to this approach is that we’re at least trying to measure the things that matter most in the energy transition.

Our climate response won’t fail because batteries sometimes increase operational emissions, but it certainly will if we don’t enable the exit of fossil fuel generation from the grid. If that is successful, the emissions intensity of generation becomes largely irrelevant. Let a thousand batteries bloom!

 

[1] We’re primarily considering LiFePo and Li-NMC chemistries since they dominate the battery storage market but recognise there are other lithium batteries that have superior roundtrip efficiency.

[2] As do reducing energy consumption, load management, etc., but that’s hardly as catchy.

[3] Although this is what they do for minor fluctuations in demand, which we’ll get to in section 2.4.

[4] There is also the long-run marginal emissions factor, which is forward-looking and attempts to account for structural changes in the energy system over longer time periods.