Bad or non-existent data is causing expensive problems for how you work out profits at your renewable energy assets. In this post, we explain why this little-known problem is costing you money and what you can do to fix it.
The world of data is truly mind-boggling. Around 95% of businesses see managing unstructured data as a key business problem, according to Forbes.
Companies in the renewable energy sector are no exception. Wind and solar owners should be excited by the number of countries and corporates that have committed to back a ‘green recovery’. But too often, these operators are being undermined by two problems with their project performance data: either it’s missing or it’s of low quality.
In this article, we’re going to spell out why the explosion of data in renewables can be a problem for operators; we’re going to look at why machine learning and artificial intelligence can play a role in filling those data gaps; and we’re going to explain one solution that we believe could help fill those gaps with 99.9% accuracy. That equates to hundreds of thousands of euros or dollars for a sizeable portfolio in a given year.
Let’s start by looking at how we got here.
What are data gaps at renewable energy projects?
A ‘data gap’ is exactly what it sounds like. You will have a statistical record of energy production and performance of renewable energy assets, and a data gap is part of that record that is missing. These can occur when a piece of technology fails, such as a meter or your server, and means you cannot log precise performance data.
These data gaps can last hours, days and sometimes even weeks. It is impossible to recapture the data later on, and means that operators are then forced to fill in gaps in the record when they are producing reports for their shareholders.
There is also a second type of ‘data gap’, where you are receiving data on how your assets are performing but it is not accurate enough. You may not have a total gap in the production and performance log, but there is an absence of high-quality data.
This low-quality data could relate to the actual power production of assets, or power prices on the open market. This is a problem because using inaccurate data means that reports about your wind and solar assets are inaccurate too. This could cost you long-term if you have to change them later and undermine trust in your business.
The world of data is growing and evolving so fast that it can be tough to keep track of whether the information you are using is good enough. But it can cause big issues.
How bad are data gaps for renewable energy companies?
The 2020s is set to be a decade of huge expansion in the wind and solar sectors. This growth also raises commercial pressures that need quality data.
One of the most pressing of these is merchant risk. How can owners thrive in an era where governments want more renewables but are less willing to pay direct feed-in tariffs (FITs) to support operators? If FITs aren’t available, developers need to sign off-take deals that help them to take projects to financial close. The upshot is they are taking more financial risk, and need certainty on production and power prices.
That’s where data comes in.
If you’re a wind or solar operator, then you need reliable data on current and future financial performance of assets. This shows you how much energy your projects are producing and what the energy is worth, which helps you manage power price risks and agree deals with power purchase agreement (PPA) partners at the right level to take your project to financial close. Reliable data is crucial.
Unfortunately, operators too often have data gaps where those reliable statistics are meant to be. This means they are forced to fill in the gaps with estimates and, in our experience, operators tend to be happy with a margin of error of 2%-3%. In a market where operators increasingly have to take larger risks on merchant power prices and thus lower revenues, those 2%-3% equate to huge amounts over project life cycles.
But new digital approaches could make that a thing of the past.
Why are bad estimates so bad for business?
How accurate are your estimates? If you’re like operators we speak to, the chances are you think your guesswork is accurate enough to get the job done and you will be able to clear up any issues later. But why take these risks in the first place?
Here are the ways bad estimates following data gaps are impacting your renewables projects’ performance. As you’ll soon see, they can come with costly consequences.
Here are three impacts that bad estimates could have on your business.
First, being paid too little in government feed-in tariffs. These tariffs are usually worked out based on how much electricity you supplied during the year. If you under-estimate the number of hours of electricity you produced, or the value of those hours, then it can result in you being under-paid. Good luck getting that back later.
On top of that, governments work out their payments based on their own sources of data, and they may not pick up on their own data gaps. Therefore, you require data that you can rely on so that you can successfully challenge them.
Second, inaccurate invoices can damage relationships with buyers. The buyers in your PPAs want to make sure you supplied what you agreed, and so any invoices you send need to accurately reflect what you produced and supplied during the year. Failing to give that information can disrupt your relationships and cost you financially.
Third, producing bad estimates is a waste of time and money. From a business perspective, you don’t want to invest in processes that are labour-intensive and will not produce the results you need. Current approaches to producing estimates tend to rely on unlucky individuals or teams to fill in gaps at the end of a reporting period.
For example, in one case we’ve seen an employee that has to look at statistics about the performance of 80 assets at the end of each and every month, and then fill in the gaps themselves. This costs that operator tens of thousands of euros every year. It makes little sense to pay someone to do the job in a longer and less accurate way.
In addition, companies should ensure that their estimates are based on a consistent set of data and approaches that are transparent if they are challenged later. This can give full visibility over the process you used and put stakeholders’ minds at rest.
Thankfully, technology may have the answer.
How can machine learning fill renewable energy data gaps?
We know there is a great deal of hype around the potential of ‘machine learning’ and ‘artificial intelligence’ to help fix problems in data. It is an industry buzzword, and you may well be sceptical. But please give us a few paragraphs to win you round!
In broad terms, ‘machine learning’ is a process of using computer algorithms to find patterns in vast amount of data, and to improve those algorithms over time. ‘Artificial intelligence’ is the branch of computer science focused on developing machines that can think and work like humans. Both can help operators produce reliable estimates.
When it comes to working our project revenue, there are three main inputs:
- Amount of electricity produced
- Price of that electricity
- Subsidy regime or PPA that governs project returns
That is an over-simplification. In reality, we see large amounts of data coming in from the huge number of sensors in renewables projects, all of which affect their financial performance. These include hundreds of sensors in a wind or solar farm to the intra-day energy trading data. We track all of the key ones in our Monitor system.
Therefore, it makes sense to use machine learning and artificial intelligence to cope with this sheer volume of data, and identify trends in it. But how can you do it?
Which are the best data sources to produce accurate estimates?
There are five main sources of data that you should be looking at when you want to produce accurate estimates of production and performance at your assets.
Here are those five in descending order of accuracy:
- Grid meter: The second most accurate source is the meter than shows how much electricity has been exported to the grid. Not all of this may get to the end customer, but we still believe that it offers a high degree of accuracy.
- Off-taker portal: The most accurate source for working out actual production is the off-taker portal that records the power that the off-taker receives. If this information is available when producing estimates, then we’ll use it first.
- Turbine / inverter power reading: The third most accurate source we see is the performance data readings from the turbines or solar panels. These are usually accurate for production purposes, but not invoicing because of losses you get in transmission. It is a common area of dispute between operators and original equipment manufacturers.
- Power forecasts (meteorological): One of the less accurate data sources we see for estimating project performance is meteorological forecasts. These give an idea of how much a particular scheme should produce if it is operating well, but it doesn’t take into account technical issues on any given scheme.
- Long-term forecasts (energy yield assessments): The least accurate data source for working out estimates is the longer-term forecasts. We would go to these last when trying to work out estimates of project performance, but they can be useful if there are no other sources to draw on. Better than nothing.
This shows the vast number of inputs to produce good estimates, but we’re not just all talk. We’ve built an add-on for our Monitor software that does exactly that.
How can you produce estimates with 99.9% accuracy?
Our Best Estimate feature uses historic and current data to give accurate forecasts of what your project should be producing now. This can help you to fill in data gaps and identify potential problems in the quality of your current data.
This has helped us reduce the margin of error in estimating data gaps to 0.1%-0.3% in our Best Estimate product. This is a tenfold improvement, which equates to tens of thousands of euros on your bottom line over the year, but how?
Our machine learning algorithm can pull in all of these energy sources and work out how accurate they are individually, and how accurate they are related to each other. It is able to draw on 12 months’ data for each of these at every asset in a portfolio.
As a result, Best Estimate can work out the most accurate estimate for a project’s power and financial performance based on whichever sources are available. It does this in real-time, so that operators can produce the best estimate for the revenue at a project while the MWh is being produced, and so reduces uncertainty.
In practice, this helps solve three challenges we dealt with up front.
First, it can improve the accuracy of estimates and help fill in gaps in the data record, which has a direct benefit on invoices and reports, and do it consistently.
Second, it shows wind turbines or solar panels that are not performing as well as they should, because you can compare performance with other machines in your portfolio, and so helps you to find low-quality data. Small improvements here can lead to big returns.
Finally, it achieves estimates with a 0.1% margin of error and without the labour-intensive process that companies currently have to go through. It’s just one click.
There are other benefits in terms of long-term production and revenue forecasting at your assets – but we’ll deal with the future in another post soon.
How can I sum this up to win over our finance director?
Simple. The benefits of improving data quality go far beyond better data. A tenfold boost in estimate accuracy can give a major financial edge at the end of the year – equivalent to hundreds of thousands of euros or dollars in your portfolio. This gives you far more control over your financial performance.
Do you have any burning questions? Would you like to see how this could work for your current project? We’d love to chat over a coffee (or Zoom, Microsoft Teams or any other reputable videoconferencing platform!).