Dr. Xiaoqin Ma, Contributor
Digitalization has become a worldwide focus and the race to become the global leader in Artificial Intelligence (AI) is becoming increasingly competitive, with many countries, including Canada, China, and the UK releasing strategies in the last twenty months to promote the use and development of AI.
Wind energy is one of the most innovative, forward-looking and fast-paced industry sectors. With the U.S. wind energy cost continuing to fall below the cost of coal and gas, wind energy asset owners and operators are redefining their approach to operations and maintenance (O&M) through adopting technologies that introduce greater efficiencies and streamline processes. Many have invested, and are continuing to invest, in predictive maintenance solutions, which combine improved SCADA data analytics, CMS systems, and oil monitoring with AI and machine learning.
AI platforms work by identifying deviations between expected component behaviors and actual behaviors, which are then flagged allowing an inspection team to identify the fault and fix it in advance of the fault causing a catastrophic failure or significant down time. This, as a result, extends the lifetime of the asset in question.
Ideally, such a solution should serve to not only flag to an asset owner that a failure is occurring, but also the nature of the issue, the rate at which the failure is occurring and the remaining useful life of the asset. This is where real-world engineering expertise plays a key role in shaping the AI solution.
Though AI, machine learning and digital twin are current buzzwords, they have the potential to create significant confusion in our industry. Below, we will examine six key market trends in order to cut through some of the “chatter” around these terms, to determine with greater clarity how to make digitalization work for the wind energy industry, rather than the other way around.
Bigger Is Not Always Better
Small, agile companies may have the advantage over large industrials when bringing innovative digital products into the market. Independent players that are closely connected to the marketplace, thus ascertaining a clear sense of market need, have an advantage over bigger, more generalist companies. The ‘top-down’ approach employed by larger industrials can come with a significant price tag attached; often proposed solutions are abstract in practice and constrained by their legacy systems, creating a challenge for the asset owners and operators when it comes to implementing the solutions. The digital products offered by larger industrials have typically been developed around their own equipment, hence, the analytics are optimized for their equipment. Oftentimes, independent, agile companies can offer more flexibility by the way of pricing, customized solutions and equipment-agnostics analytics. It is interesting to note that it is often these bigger players that appear to struggle in the wind energy market.
Digitalization Focused on Value, Not Tools
Numerous multi-national corporations have embarked upon a digitalization journey by investing heavily in digital technologies and tools only to have these investments fail to materialize in business value. To avoid repeating these costly mistakes, a clear understanding of the business value delivered by digital tools needs to be established as the first step.
After that, an investment strategy to develop, partner or acquire the relevant digital technologies should follow. This will lead to a more “value-focused” digital transformation, which should, in time, result in a clear demonstration of the value that digital technologies bring to the industry.
Digitization Before Digitalization
Despite the importance of data and information to unleash the full power of AI and machine learning, some critical wind energy data falls into the category of not being “digitized” or stored in an “AI ready” or “IIoT” format. Furthermore, O&M teams currently spend 80 percent of their time organizing inspection data, which means they have less time to focus on utilizing their core skills to address engineering issues.
Tools that enable the O&M teams to “digitize” their inspection and maintenance data enable them to spend an increased amount of time tracking failures, responding to developing technical and safety issues and determining where the maintenance budget should be focused.
A fully digitalized and connected predictive maintenance approach is becoming increasingly important sector-wide; asset owners require the right data to be in place in order to maximize the investment of AI/machine learning. Asset and operations managers have cited a need for higher-quality data to improve the reliability of their organizations’ assets. The challenge lies in ensuring data is stored in a unified format and easily accessible. Cloud-based mobile inspection and maintenance software transforms inspection and maintenance reporting practices, creating substantial safety and efficiency gains for the organization, and generates good quality data, increasing the efficiency also of Digitalization processes.
Overcoming “False Positives” And “False Negatives”
In its enthusiasm to adopt innovative digital solutions, the industry may be in danger of undermining the progress made in predictive maintenance to date, by driving an increase in “false positives” that, in turn, will result in increased OPEX costs for operators. Clearly, there are only so many times that maintenance personnel can be sent out to respond to identified failures — which fail to materialize into an actual fault — before confidence is lost in the technology concerned.
Equally, “false negatives” are bad news to the industry too. A cautionary tale of “false negatives” in AI involves researchers training a Neural Network (NN) to detect camouflaged tanks in photography for the US Army, succeeding, only to realize the NN had learned to distinguish cloudy days from sunny days, instead of distinguishing camouflaged tanks from empty forest. These issues with AI and machine learning must be carefully managed to avoid the industry experiencing a turn from enthusiasm to disappointment whilst AI is still in its infancy stage in wind energy.
The AI technology currently available in the market is narrow AI, which means that it is able to handle just one particular task. It is still impossible for AI technology to fully complete the work of an all-round engineer solving complex engineering problems. Development of Industrial AI is even more challenging than the AI we use in our daily life through the likes of cloud-based smart home hubs, due to the lack of good-quality training datasets.
U.S. operators are finding that AI needs to be informed by real-world engineering expertise. A focus on engineering-led approaches — from agile companies which have grown up from an engineering base with a strong industry track record — is best positioned to harness all the improvements promised by AI.
It is therefore imperative that the two approaches, AI and engineering-led predictive maintenance, are not seen as opposites, and that operators’ use of emerging technologies is underpinned by deep engineering knowledge.
During this digital transformation journey, turbine owners are increasingly running into obstacles relating to limited data access from their own assets. Without full access to data, owners cannot fully understand the health of their asset and manage it in an optimal way. Ultimately, results from AI/machine learning are only as good as the data available. It is possible for asset owners to come up against a barrier in a number of different ways, from insufficient data acquisition systems or infrastructure, meaning data isn’t collected in the first place, to a contractual limit that means owners and operators aren’t able to freely access all the data collected about their turbine performance.
That said, it is possible for the industry to surmount challenges in the areas of data collection, data handling and data access, so that turbine performance can be optimized, and the U.S. wind market remains competitive. For the former, innovative digital hardware and IIoT sensors are becoming increasingly cost-effective, making it easy to justify investment in improving data acquisition and infrastructure systems. With the latter, if U.S. wind farm owners collectively start holding their suppliers to account when it comes to data access, greater volumes of data will become increasingly available, ensuring a smoother procurement process.
DIY or Do-It-For-Me?
These factors naturally lead to one final consideration in evaluating your own predictive maintenance approaches: whether to carry out all assessment and monitoring in-house or outsource it to a predictive maintenance provider. The former, DIY approach may lend itself well to organizations whose personnel possess a broad and in-depth O&M experience in industries with a long track record in applying predictive analytics, and who are well placed to advise on the implementation of predictive maintenance approaches for wind assets. Equally, you may already be using some form of condition monitoring technology, and this familiarity could serve as the basis for a more fully developed predictive maintenance solution. Over time, after working with and alongside experts from your predictive maintenance supplier, you will develop levels of expertise that require less hand-holding and make some level of DIY feasible.
Alternatively, it may prove more resource- and cost-effective for your predictive maintenance partner to continue to use their existing expertise to monitor, detect and predict failures remotely and advise on the most suitable course of action, at least in the short-term. In the long run, and as your on-site team become more familiar with predictive maintenance technology and techniques, you could consider moving from this Do-It-For-Me approach, to a semi-DIY set-up before taking full control. Of course, for this to be an option, you’ll need to be working with a provider that can deliver this level of flexibility and a tailored approach.
If an asset owner or operator decides to go with the DIY approach eventually, it is still advisable to invest in a powerful digital platform, which could increase the efficiency up to ten-fold. After all, the core competence of asset owners and operators usually falls on energy production instead of AI/machine learning and digital product development. To have a strong technology partner in this area will enable a pain-free implementation of the DIY strategy in the short term as well as the longer-term maintenance of the most up-to-date digital technology in the current competitive and fast-evolving market.
Understanding these myths, challenges and potential approaches is a key first step in taking advantage of the opportunities on offer. When combined with real-world engineering expertise, AI and machine learning have the potential to transform decision making for wind energy asset owners. With benefits on offer including the ability to assess turbine condition, optimize maintenance programs and bring down the levelized cost of energy, asset owners would do well to take the lead in adopting approaches that combine AI and engineering expertise.
Author Dr. Xiaoqin Ma is Head of Technology and Marketing at ONYX InSight. She holds a PhD in Electrical, Electronics and Communications Engineering from Xi’an Jiaotong University