With the increase in the scale, size and number of solar and wind power installations and focus on cost efficiency, and effective operations and maintenance (O&M) has taken centre stage. The constant monitoring of equipment performance; measurement of current, voltage, power and energy generation; and analysis of resource data are imperative to ensure that the plant is performing at the required output level.
While data collection is the starting point for effective O&M, it is far more critical to properly analyse this data and identify anomalies. However, it is impossible to keep track and carry out a manual analysis of the huge volumes of data, with each developer and operator responsible for gigawatts of assets. This is where digitalisation and automation play a role. Automatic data acquisition is integrated with advanced monitoring and analysis systems based on artificial intelligence (AI) and machine learning (ML) software to optimise and enhance system performance.
Digitalisation is not a new concept for the renewable power sector, which has moved from computer-based systems, databases, and communication networks and servers to IoT- and cloud-based platforms, advanced analytics, predictive data analytics, asset performance management software, smart sensors and intelligent forecasting solutions. Such digital systems can perform multiple functions in the solar and wind O&M space such as improvement of power plant efficiency, reduction of operational expenses, and mitigation of unplanned outages. Digitalisation can also help in improving plant efficiencies on a real-time basis, leading to enhanced equipment life.
Transition from reactive to predictive maintenance
The advancements in the digital O&M space are driving the transition from reactive to predictive maintenance. The focus is now on predicting likely anomalies and addressing the causes beforehand, instead of diagnosing and rectifying faults after they have occurred. Predictive maintenance can be defined as an O&M practice that uses data analysis techniques for the identification of possible defects in the system components, so as to prevent any downtime. In the context of power plants, which have a large number of electrical and mechanical components prone to faults, predictive maintenance plays an important role as it helps avoid energy loss and improve revenues.
Most solar and wind project operators use supervisory control and data acquisition (SCADA) data for remote monitoring and management, which can be used for effective O&M. However, with SCADA, only a limited amount of sensor data can be monitored by operators while most failures frequently originate from previous unknown and undetected anomalies. Thus, it often becomes difficult to find the actual raise of the issue. Moreover, basic SCADA systems can only sound an alarm after a certain programmed value level is crossed, and by then it is too late to take any corrective action.
Due to this incomplete knowledge about possible faults that can impact generation, many operators have now started moving towards intelligent predictive maintenance based on advanced sensors and IoT. Defects can be recognised and rectified before any large damage occurs, which can lead to significant cost overruns, thus minimising sudden breakdowns. Predictive maintenance ultimately helps in reducing the long-term operating costs of a project as it helps to prevent system faults and product damages. Moreover, this practice enables proper planning for ordering spares and scheduling essential maintenance work. If predictive maintenance is carried out properly, O&M teams will have to make less frequent visits to the project site as the manpower requirement will be reduced for fault detection and repair activities, since minor repair work or issues can be handled remotely.
Through advanced AI techniques, equipment can be programmed to learn and react to the various operational processes and issues of a solar or wind power plant. With ML programs, these machines can be made smarter through self-learning, thereby enhancing the diagnosis and rectification of faults. Various programs are being developed in the ML domain to improve plant performance through advanced analytics. In fact, research is on to develop self-monitoring and operating machines that can keep learning and evolving to eventually become self-reliant to such an extent that they can take care of any present and future problems arising in solar or wind power plants.
Such advanced digital applications can reduce the time and effort required for due diligence, planning and analysis. In fact, a replica, or a “digital twin”, of physical assets, processes, systems and devices can be created on the virtual screen through AI. This can then be used as a benchmark for identifying faults through data anomalies. This visualisation of assets provides insights that are otherwise hard to achieve, making the O&M process highly reliable. Moreover, a digital twin helps in understanding the health of the generating asset and planning O&M activities in advance. Such digital twins are also helpful in estimating the equipment life even before it is put to use and understanding how a particular component will behave over its lifetime. This concept has immense scope for further advancements as it can be used by manufacturers to design products that have high efficiency in any given set of conditions. Further, operators can predict the performance of their assets by running various simulations and prepare maintenance strategies beforehand. This can help prevent any sudden breakages and loss in generation revenue.
Even though the application of predictive maintenance in solar and wind power plants is at a nascent stage, there are some examples of its use in forecasting, monitoring and maintenance activities in the O&M space. For instance, GE’s Predix software analyses sensor data and combines asset modelling with real-time monitoring and big data analytics. A digital doppelganger is created as a baseline to identify faults through data anomalies. Further, the asset performance management software assists in condition-based maintenance, which uses AI to execute repairs before an asset breaks down. Likewise, Siemens Gamesa uses the Pythia platform, which uses big data analytics and intelligent algorithms for running its diagnostic models. It enables early prediction of potential damages (up to three years in advance), digital twins, daily health checks, optimised spare part forecasts and risk-based planning.
Similarly, Sembcorp Energy India Limited has introduced an advanced digital tool, Virtual Brain Renewables, for its renewable operations. According to the company, this analytics-based digital asset management platform helps in monitoring more than 30 wind energy sites and provides real-time data and indicators on wind. It provides information that helps in performance monitoring, forecasting, real-time condition monitoring and anomaly detection. This, in turn, provides the operations team with insights for predictive maintenance.
Another example is GreenPowerMonitor (GPM), which is a part of DNV GL. The company has developed the GPM Horizon software, which is an integrated tool for wind, solar and energy storage. GPM Horizon assists in easy and quick visualisation of asset data brought together from a diverse range of technologies and locations to a single location so that all the assets can be monitored and operated remotely. The company has one such control centre at Bengaluru.
In another case, NEXTracker has a self-adjusting tracker control system called True Capture, which uses ML software to improve the tracking algorithms of each module row as per present weather conditions. Virtual 3D models of the plant site are created using real-time shading information, and tracking instructions for each row are updated. Software giant IBM also has a self-learning technology that integrates multiple forecasting models with weather, solar generation and power grid operation data. It predicts solar generation in advance for every 15 minutes, and has shown up to 30 per cent improvement in solar forecasting.
While AI- and ML driven O&M was initially assumed to be a distant future in India owing to the availability of cheap labour, the Covid-19 pandemic has certainly highlighted its urgency. With lockdown measures imposed, and shortage of spares and manpower experienced, developers and operators have realised the merit in transitioning to advanced digital O&M practices and predictive maintenance techniques. Thus, with rapid technology advancements in automation and digitalisation, and the resultant improvements in O&M efficiency and cost, it is expected that the majority of the O&M activities will be carried out through digitally enabled techniques in the near future in India as well as the rest of the world.
By Khushboo Goyal