Remote monitoring of solar photovoltaic (PV) plants is now standard in almost all projects. The primary goal of remote monitoring is to collect data that can be used to develop corrective actions, if performance falls short of expectations. On the surface, remote monitoring appears to be a straightforward platform for providing data that can be used to identify and correct errors or problems. Because of this simplicity, the majority of remote monitoring websites are data storage platforms, rather than data analytics platforms. Today, customers have many options, ranging from firms offering enterprise-grade software solutions to inverter original equipment manufacturers (OEMs) that provide a free portal for remote monitoring. As a result of numerous options and assumed simplicity, remote monitoring has become ineffective.
The following are the most significant observations based on the past eight years of managing solar PV assets:
Errors, data loss and non-standard data structures
Devices from various OEMs provide data in a variety of formats and structures. This inconsistency in format and data structure leads to data handling errors. Furthermore, the internal data logging method, frequency and accuracy vary, depending on the OEM. This variation, combined with poor network connectivity, causes shambles in data handling. To deal with these data inconsistencies, remote monitoring portals frequently manipulate, resulting in the removal of vital information that may reveal an underlying pattern. We found that various parameters recorded and the frequency of data captured vary greatly between portals and 75 per cent of portals log an insufficient set of data about inverters at a higher frequency.
Rationalisation
With different devices passing up different parameters and the number of parameters recorded varying, data rationalisation is completely missed. This leads to data overload, making it difficult to find critical information about the performance of a plant.
Data without context
On the surface, solar PV data appears to be a collection of numbers. This becomes meaningful only when it is presented with a context. Different data inferences from the same dataset are needed by a developer of a solar PV project, an asset manager, a site engineer and an OEM. Giving all users the same data representation does not aid in decision-making. With most remote monitoring portals providing non-contextual data, operating a solar PV plant within financial goals is impossible.
Absence of a provision to initiate actions
While the primary goal of a remote monitoring platform is to identify and fix errors on-site, most portals do not provide the option to initiate actions based on the data. In order to take action, the user will need to rely on additional tools and software to ensure that identified performance issues are addressed on-site. Most remote monitoring platforms lack this complete cycle of identifying issues, initiating action and following up on the desired performance after correction.
AI and autonomous monitoring
Data from remote monitoring systems is traditionally handled and evaluated by an expert to assess the performance of a solar PV plant. This procedure requires a significant amount of effort and frequently results in numerous errors and high uncertainty. Recent advancements in big data acquisition, storage and analytics are aimed at streamlining the process and improving the accuracy of analysis. Machine learning engines, such as Bia (developed by Inspire Clean Energy), which are hardware independent and device agnostic, are assisting in reducing such human interventions and errors. Furthermore, artificial intelligence (AI) techniques seek to develop innovative, autonomous and intelligent condition monitoring concepts, aiming to make smart decisions to improve solar PV equipment life.
The way forward
Remote monitoring would eventually become a unified software platform, allowing asset owners, managers and field teams to harness the power of connected devices, resulting in increased productivity and efficiency. Small residential plant users or large IPPs with a fleet of assets want a system that can help with performance visualisation, problem identification, action proposal, field intervention initiation and learning to predict failure based on the intervention so that appropriate precautions can be planned all through a single log-in. Recent products in the market come with configurable dashboards that let users customise data representations to meet their unique needs. Collaboration with industry alliances such as SunSpec is helping to address data rationalisation to a large extent. Many providers still do not follow these standardisation procedures, but as time goes on, customers will start to impose them in order to make managing solar plants less of a hassle. Newer platforms tend to understand the role of different users and provide different data representation, enabling them to make faster and smarter decisions.
