By: Vladimir Milovanovic - Vice President of Power Systems, Anglophone Africa at Schneider Electric 

The integration of Artificial Intelligence (AI) and Machine Learning (ML) is significantly transforming the management of utilities by providing advanced technology that delivers real-time insights into the operational conditions of facilities.

This enhances situational awareness and enables more informed decision-making in utility management, enabling utilities to monitor their operations more effectively by transitioning from routine (preventative) to predictive maintenance. 

Utilities, especially municipalities in South Africa, have varying degrees of visibility into their network infrastructure. This depends on factors such as the age of the existing infrastructure, the sophistication of their network planning and implementation of connectivity technologies on the network assets

This visibility is crucial for leveraging AI and ML, as AI is built on algorithms derived from ML, which requires access to substantial amount of high-quality, accurate data across the utility's grid. With the right data, ML can transform utility uptime by providing deeper insights into asset conditions

Leveraging this enables a shift from rigid, time-based maintenance to a more predictive, proactive approach, resulting in optimised use of (always limited) field resources, reduced outages and enhanced operational efficiency and customer satisfaction. However, it all starts with building that foundational network visibility and data infrastructure.

Preventing equipment failure

Preventative maintenance is the most common strategy used by utilities. This time-based approach has the utility define the key components of its network and analysing factors like mean time to failure, statistical data, available workforce and operational volume. The goal is to derive a maintenance schedule that prevents equipment failure and downtime.

However, this is very difficult to achieve in practice as utilities are often constrained by limited resources, and the information they have may not be fully accurate or account for the evolving nature of their network. As a result, preventative maintenance frequently devolves into reactive maintenance, where action is only taken after a component has already failed and resulted in downtime. This impacts key performance metrics and customer satisfaction. 

The better approach is predictive maintenance, which leverages data and analytics to proactively predict and prevent asset failures. However, establishing the right data infrastructure is crucial to enabling this strategy.

By transitioning to an AI-powered predictive maintenance strategy, as opposed to the more traditional time-based preventative maintenance approach, utilities can reduce asset maintenance costs by optimising workforce management and resource scheduling, maximise the impact of available field resources and improve customer satisfaction.

Additionally, utilities can also benefit from enhanced safety and compliance with regulations, including sustainability and environmental friendliness by mitigating risks like fires, oil spill, and threats to operator and public safety.

However, the success of predictive maintenance depends on the quality, quantity and duration of data collected. More data over time enables deeper and more impactful trend analysis, the first step of the predictive maintenance value chain is detecting abnormal asset behaviour, which requires extensive grid connectivity to feed rich data inputs.

Real-time asset visibility 

Thus, the more sensors and natively connected devices integrated, the better the utility's real-time asset visibility. This data-driven foundation then powers the subsequent predictive maintenance steps, but it all starts with the initial detection of abnormal conditions – making connectivity and data collection essential for an effective predictive maintenance programme.

Yet, transitioning to predictive maintenance is not an "all or nothing" approach. Rather, it requires the utility to take a long-term, strategic view of its network and future needs. The first step is equipping the utility's most critical assets with the right technologies to leverage predictive maintenance models, even if in some cases it means retrofitting This provides the data foundation to start implementing AI-based predictive maintenance.

Concurrently, the utility can then dedicate more resources to address known problematic or high-risk areas of the network that may not yet have the necessary facilities for predictive maintenance. By building out this digital roadmap and prioritising the right assets, the utility can start realising savings through optimal resource use and minimised downtime – gradually transitioning away from reactive maintenance.

To transition to a predictive maintenance approach, utilities should work with a trusted technology partner who can assist them in developing a realistic and achievable digital roadmap. The digital roadmap is key to implementing the software layer, which is crucial for harnessing data and gaining real-time visibility and insights that empower utilities to make quick and efficient operational decisions.