Asset risk is one of the biggest challenges utilities faces. Due to an aging infrastructure, utilities run the risk of catastrophic events like fires and unexpected outages if they don’t focus on asset management. These lead to a slew of challenges from risk management and operational inefficiencies, to creating a less-than-positive experience for the customer. Therefore, utilities must innovate and find new ways to monitor and maintain their critical assets from cradle to grave.
With the availability of drones, sensors, smart meters and other intelligent devices, utilities collect petabytes of data every day. Predictive analytics is the path utilities should take to tackle asset risk. Leveraging historical and current data, utilities can build machine learning algorithms and models to predict the health of critical assets and be proactive about their maintenance.
Predictive analytics allow us to understand reasons for failure events and make appropriate adjustments to avoid these occurrences in the future. They also help minimize total maintenance costs, revenue loss and fines for non-compliance. Predictive analytics helps organizations move from reactive to proactive maintenance.
A completely proactive and predictive maintenance goal in the immediate term is a utopian one for a utility. Assets will continue to have unexpected failures and will require reactive maintenance. Predictive analytics should primarily be used to identify critical assets that should be prioritized (based on risk of failure factors) for maintenance during planned outages. Over a period, the need for reactive maintenance will decrease and allow utilities to adjust workforce requirements and improve operational efficiencies.
What questions do Predictive algorithms answer?
Predictive algorithms and models are the building blocks of all analytic solutions. Identifying the specific models and techniques that best suit an organization’s needs is paramount to ensure they get the most impact from their enterprise data and analytics solutions
Predictive algorithms can answer several questions related to asset maintenance, including:
- What elements/characteristics play a role in the failure of the assets?
- What is the correlation of failures between assets?
- What trends and root causes do we see for asset failure?
- How can we optimize to prevent asset failures in the future?
- What is the impact of preventive maintenance on the life of an asset?
What are some of the challenges to using predictive models?
Building predictive models can be challenging, and utilities must be willing to make significant investments to glean the most value from it. Some common challenges include:
- Predictive algorithms need a lot of historical data points to identify patterns and learn from them. Utilities may not have sufficient historical failure data for results to be meaningful.
- Advanced skills required for building reliable predictive algorithms, both business and technical, are hard to source
- Building the trust and acceptance of these predictive models can be challenging. Overlaying human experience and intuition on top of model-generated results will help overcome this skepticism over time.
So, how do we make a Preventive Analytics deployment successful?
Data, data everywhere; but how do we make sense of it all?
Without a clear strategy and plan to manage and use this large amount of data to generate insights, collecting all this data would be a wasted effort. It is also important to recognize large sets of data cannot automatically be used as is.
Here are a few factors utilities should consider when initiating a move to predictive analytics:
- Which sets of data are critical to building algorithms and decision making?
- How will all the collected data be stored and integrated from where these models can be built?
- How will the data be screened, cleansed, validated and prepared to be useful in analysis?
- Keeping in mind the end goal of preventive asset management, what actionable practices can be generated from the analysis?
Predictive analytics is a powerful tool utilities can leverage for making significant improvements to the way they maintain assets today. However, predictive analytics is not without its challenges, and success is dependent on having an effective data strategy, the right people involved and a focused scope. Unify has a successful track record of supporting complex E&U organizations that need to build models to support predictive use-cases.