India has 41 GW of installed wind power capacity, which requires effective operations and maintenance (O&M) for maximum energy generation, reduced opex costs and enhanced equipment life. As asset size and scale are increasing, the wind power industry is transitioning towards advanced analytics and digital technologies such as machine learning (ML) and artificial intelligence (AI) for optimisation of asset management. Thus, predictive maintenance is gaining traction across the country for its cost-efficiency benefits. At a recent webinar organised by Renewable Watch on “Predictive Maintenance of Wind Turbines”, senior executives of leading wind independent power producers (IPPs) discussed the current trends, challenges and role of predictive maintenance in their O&M strategies…
Kapil Kasotia
The wind energy segment entails several challenges. Looking at wind turbine performance, one sees that over the last decade or so, different types of wind farms, machines and original equipment manufacturers (OEMs) have come up. The most challenging aspect is to actually receive the amount of energy that one expects from a turbine. Turbines in India are not performing up to the mark. This is because of a “blackbox”, in some sense. One primary yet fundamental reason for the poor output is often considered to be wind itself. Over the past few years, surface winds have decreased all over the world by up to 10-20 metres per second. This has been exacerbated by extreme weather events and El Nino. Year after year, lower wind speeds have been recorded. Furthermore, wind turbines tend to break down more often without the reasons for such breakdowns being apparent. Moreover, the accuracy of these observations is questionable. As a result, identifying the reasons for inefficiency in a particular turbine becomes a challenging mystery.
To predict all significant parameters accurately, it is important to place sensors in each part of the turbine. At present, there is a mismatch in access to data, as OEMs collect all the data regarding a multitude of parameters required for immediate and real-time calculations, and performance indication. We have tried to use AI- and ML-based solutions. Our learnings suggest that we do not possess sufficient data to support such solutions. IPPs have access to limited data, primarily for dashboarding and not for deducing logic between the observed parameters. However, we need existing data to utilise ML algorithms in order to generate predictions about, for instance, time to failure and the remaining life of the equipment. Data can enable us to improve performance, which may improve the overall value generation by wind turbines.
Going forward, we can focus on low-hanging fruits such as light detection and ranging. Analysing the total amount of energy that we are producing and losing is also crucial. We are moving towards facilitating the smooth adoption of new-age technology for predictive maintenance and data collection. I believe that the wind turbines that have been operational for the past several years are like data gold mines. Collaborating with existing analytics companies to mine data from these turbines may help us better analyse the strengths and weaknesses in their performance, based on which digital models and simulations may be created. At present, we have many wind assets that are not working to their full potential. Apart from the reasons discussed earlier, lack of accountability may also be factored in. In order to improve the output and performance of wind turbines in the coming years, it is critical to focus on aspects of data sharing and effective data utilisation, while establishing answerability for all the stakeholders involved.
Atul Pachauri
OEMs operate nearly 90 per cent of the installed wind turbine generators in India. As a result, it is essential for IPPs to collaborate with OEMs on several wind O&M issues. One challenge that we face is that of decreasing wind speeds. And when there is wind, breakdowns become a challenge. A major issue is when a turbine is in breakdown mode but wind is blowing, thereby leading to loss of revenue. Therefore, it is important to approach OEMs in such a way that their guarantees and warranties cover more than just machine availability. Currently, the key performance indicators are plant or wind turbine availability. Given today’s needs, predictive analysis is necessary. Everyone is concerned about the cost of operations and plans to improve their IRRs. Predictive maintenance can be done using the sensors or the equipment included with each wind turbine generator (WTG), since the OEM for the WTG would already have included nearly a hundred sensors. As a result, the data is already synchronised and is continuously sent to the supervisory control and data acquisition system and the central monitoring system. Predictive analysis can be done with0 this data, which the engineer has access to.
The necessity for external sensors, which is also underlined by the use of vibration analysers (installed externally over the foundations), can be met with predictive analysis. Since WTGs need to be positioned at a specific height, if lightning strikes, one of the many turbines in a farm may have problems with its blades. With the use of drone inspection, even such issues can be discovered at a preliminary stage. Wind turbine experts prefer ground-level jobs to climbing towers. Because of this, some professionals have shifted from wind to solar. This is the main cause for the shortage of specific expertise.
O2 Power intends to concentrate on both its technological and commercial objectives. Technically, the company will continue to concentrate on doing all of the tasks, such as conditional monitoring, that significantly reduce breakdowns. The company needs to have as many turbines available during the windy season as possible, especially when the wind is at maximum speed. Commercially, the key priority is to reduce the impact of the age of the nearly 10-year-old turbines and the old OEM contracts, which are expensive. Therefore, O2 Power will now concentrate on lowering this expense in order to increase its investment in other products providing O&M services.
Rajiv Babaji Samant
Most IPPs have started taking baby steps in wind predictive O&M as they have set up their central monitoring room in each of their offices. Thus, whatever is happening to the turbine, the data from the project site is reaching their head office, where they are monitoring. The monitoring is done in terms of what is happening with respect to turbine temperature and power generated by each of the turbines. If any anomaly is observed in turbines, then alarms are raised and sent to the O&M department. Hence, the rectification could be done at the level of IPPs.
A key strategic area that our company is currently working on is anemometer readings. It is an equipment that is deployed behind wind turbines, measuring wind speed and direction. We did an exercise with a glider on a few turbines and found out that there were many errors in anemometer readings. After this exercise, we tested the anemometer and discovered that its alignment was not proper. After the completion of this exercise, all errors in other turbines in lieu of these errors were rectified. The other activity that we are actively taking care of is how we can carry out all preventive maintenance. In addition, we ensure that all work pertaining to preventive maintenance is done during low wind conditions and not during high wind conditions. We are trying to ensure 99.5 per cent availability of the turbine during high wind speed conditions.
Furthermore, most wind turbines in India are designed to operate in temperatures of up to 40 degrees Celsius. However, the majority of sites where these turbines will be installed have temperatures of more than 45 degrees, which will pose a major challenge. Thus, it raises a big question about whether these manufactured wind turbines are designed to be operated in Europe or are they suitable for Indian weather conditions as well. A detailed study or analysis with regard to this needs to be undertaken. Looking at the challenges faced in carrying out O&M, the budget for operating expenditures, while considering a purchase, is very limited. When a new gadget comes into the market, it may look very attractive, but all IPPs may not have the budget for it. Even if the product is good, there could be budget constraints. A leasing model, wherein there is a lease of equipment and then the creation of a revenue sharing model, with whatever gadgets a company comes up with, can be viable in the industry. Going forward, we plan to reduce O&M costs further. Also, we do need to know more about the emerging technologies and new products coming into the market.
Niraj Shah
A key challenge for an operator of a wind farm is to make informed decisions based on various systems, which makes the task extremely vulnerable. The operation of wind turbines includes analysis based on outputs from the supervisory control and data acquisition (SCADA) system. There is an AI-ML system, which takes limited data from the SCADA system, makes some interpretation and sends back to the operator. In addition, there is a condition monitoring system, which is an independent system requiring time and effort from the O&M operator. Therefore, we need to see how all these different independent functions can be collaborated. To scale up operations, specialisation of O&M products in only one of these different technological aspects is not the way forward. Too many products may confuse the O&M operator, who may then start missing out on important issues emerging from the wind farms.
Another key challenge is with respect to the quality of manpower from the contractor’s end. Human errors need to be avoided as this space involves the analysis of a huge amount of data. I observe that from the OEMs’ end, there is a lack of transparency. Even after paying for the turbines, there is a lack of access to the software of the turbine. Besides, the operator does not have access to switch on or off the turbines. This is a precarious situation that needs to be amended. This is especially so now because various technological changes are taking place that make operations more flexible. Given these changing market dynamics, the lack of transparency among different stakeholders is not appreciated.
In terms of technological advancements, we have been using an AI-ML platform for all our wind turbines for the past two years. We have not found any catastrophic failures so far from our projects using these platforms, although many deviations are being reported. A key issue with AI-ML platforms is that it cannot be gauged what these platforms are learning themselves and changing in their algorithms. I have not come across technology providers who share the probability of the success of these AI-ML algorithms. Basically, the AI-ML platforms look at historical data and based on regression analysis, try to analyse what will happen in the future. However, as mentioned, the lack of transparency regarding what the algorithm is learning is a big handicap for an O&M operator. In addition, other important information from different platforms (for instance, condition monitoring) cannot be integrated with the AI-ML platform if the solution provider does not accept that information. Therefore, such AI-ML platforms have limitations and for gathering any new information, new platforms need to be added, which creates hurdles. Too many systems lead to many alarms, which need to be attended to by a single person, thereby impacting the productivity of operations. Also, often unnecessary alarms are raised, leading to wastage of time. Therefore, we are planning a slightly different SCADA set-up, which has the capability to handle big data.
Nguồn: https://renewablewatch.in/2022/09/22/turbine-upkeep/