E.ON uses artificial intelligence to warn of power failures

Artificial intelligence has advanced predictive maintenance on a north German regional grid
Published: Wed 20 Jun 2018

Artificial intelligence with machine learning is arguably the most significant of modern analytics techniques being exploited to extract value from the growing mass of modern utility data.

A key use case is around asset management and maintenance, which E.ON north German regional grid operator subsidiary Schleswig-Holstein Netz AG has implemented on its medium-voltage grid.

After nine months of use, the results, which by extension serve as a power failure warning indicator, have apparently even astonished the company.

“The probability that we can predict a defect in the power grid has increased by a factor of two to three,” says Thomas König, who is responsible for E.ON's German grid business. "And our customers benefit as well because possible sources of error that we identify in advance reduce the number of faults and make our grid more stable."

Asset predictive maintenance

E.ON’s solution uses a variety of internal and external data to detect patterns and inconsistencies, including the age and type of the power lines, maintenance and weather data as well as real-time information such as the current load behaviour.

Combining comprehensive data analysis with a self-learning algorithm enables the prediction of faults and failures before they occur. According to the company, with the information they have gained, around a dozen maintenance projects in Schleswig-Holstein Netz AG's grids were brought forward in the recent months.

Other benefits the company foresees from the technology includes improved planning for grid construction projects, with the prospect also of improved investment planning and budget allocation.

An E.ON statement says the company has continuously worked to improve the security of supply in its networks in recent years. At Schleswig-Holstein Netz AG, for example, the average downtime in 2016 was just 8.85 minutes, around a third lower than the nationwide average of 12.8 minutes.

Artificial intelligence use cases

E.ON – now in a bid to acquire RWE’s Innogy and become a networks and customer solutions business – considers itself a pioneer in artificial intelligence. Its solution was developed in house by experts from its distribution system operator subsidiary HanseWerk AG and the E.ON DataLab.

Initially its aim was to predict when medium voltage cables in windfarms are needing to be replaced. In that case inputs include existing data including geo information, asset and grid operations data, along with external data on the weather, lightning and salt content.

In addition to the power cable use case, the solution is being applied to substations and the intention is to extend the analytical model to also cover other assets, such as low voltage cables.

According to the company, research has shown that the intelligent approach provides a 30% improvement compared to the conventional approach.

Earlier this year E.ON partnered with Sight Machine, a San Francisco software start-up, to develop artificial intelligence-based solution to improve energy efficiency in the commercial and industrial (C&I) sector in Europe.

“We believe that innovative technology and applications are the future and that they will give us a big advantage over our competition,” says E.ON in a statement.

Artificial intelligence in the grid

Artificial intelligence has been identified by Navigant Research as one of the ‘enabling technologies’ of the future grid, effectively providing the ‘brain’ for what the company has termed the ‘neural grid’.

Currently, however, few advanced solutions encompassing predictive modelling and optimisation have been deployed. According to Navigant, true artificial intelligence will only become reality when ubiquitous sensing and connectivity are available, leading to the volume and quality of data necessary to make solutions powerful.

In addition to operational functions such as asset management, artificial intelligence is expected to inform all manner of tasks and prepare utilities to use their customer data, load data, distributed resource integration data, etc. to develop new service offerings.

Indicative of the potential, Navigant projects the analytics market to more than triple from now to 2025, led by developments in North America, Europe and Asia Pacific and with asset analytics and customer operations analytics showing the highest growth rates.