Alperia improves safety by modeling penstock stress
Alperia is a 100% sustainable green tech company that produces energy from renewable sources. It owns and operates 34 hydro-electric plants in South Tyrol; 6 district heating plants; and manages an electricity grid spanning 8,735 km. It is the third largest provider of renewable energy in Italy, providing service to nearly 341,000 customers in the country.
As one of the biggest producers of energy in the region, safety is a top priority for Alperia. One area of particular concern for the company is monitoring and maintaining its penstocks–groups of pipes that transport pressurized water from reservoirs or dams to turbines in their hydro-electric power plants.
Penstocks are particularly vulnerable to rupture. In addition to the incredible strain of carrying thousands of gallons of pressurized water a day, they are subject to environmental stressors like weather and temperature fluctuations. Even seemingly minor seismic movements could potentially compromise their integrity–a risk exacerbated by the fact that Alperia’s penstocks are located in a mountainous region of Italy prone to landslides. A pipe rupture has the potential to trigger a catastrophe of epic proportions, so ensuring that their penstocks remain in optimal condition is more than just an operational concern for Alperia–it’s a matter of life and death.
A Healthy Distance
In addition to improving safety, Alperia wanted to improve its remote monitoring system to save on hardware, maintenance and operational costs. Its penstocks were located high up in the mountains, a good 6-8 hour drive away from civilization, so the more they could monitor from a distance, the more the company would save in time, maintenance and manpower. But the gauge sensors typically used to measure penstock strain came with a hefty price tag; buying enough of them to install on all the company’s penstocks would be exorbitantly expensive.
An AI Partner with Industry Experience
Alperia wanted to see if machine learning technology could help the company improve its current process, so it opened its business case up to pitches at the 2019 WhatAVenture Innovation Summit. Fieldbox won the pitch contest with a proposal on leak detection.
Fieldbox’s team worked closely with Alperia’s engineering experts, examining the on-site challenges from every angle, reviewing historical data, and assessing the company’s budget and available resources to decide on the best plan of attack. They put together a project schedule, then opened up the Fieldbox platform to Alperia’s experts so the team could work together collaboratively using common tools.
Saving on Sensors with Behavioral Modeling
The project team decided to install strain gauges and 3D GPS sensors on a select sample of penstocks to monitor pipeline strain and behavioral response to weather, temperature, energy production and other environmental stressors. The strain gauges captured traditional data, while the GPS sensors captured micromovements in the penstocks and the concrete blocks containing them.
The team then used the data collected and AI technology to create a behavioral model that could accurately pick up on danger signs and accurately predict when a penstock was likely to rupture using only GPS data, since GPS sensors are much cheaper and easier to install than strain gauges.
Fieldbox also built a dashboard that allowed Alperia’s operations team to monitor sensor data from a remote, centralized location, saving them unnecessary trips to and from the mountains to check on penstocks in person. In case of trouble, the dashboard provided alerts with precise GPS data pinpointing the problem area, so that Alperia’s team could travel directly to the right location without wasting any time.
Improve Hydroelectric Infrastructure Safety
by using AI to better understand how penstocks and galleries react to surrounding environmental factors.
Reduce Hardware Costs
by implementing a virtual monitoring system that measures penstock strain with GPS sensors that are cheaper and easier to install than strain gauges.
Reduce the Cost of Penstock Safety Operations
with better reporting (real-time, centralized, remote) and system alerts.
Assess the Benefits of a Centralized Data Platform
to develop, train and deploy other AI models for hydraulic engineering.
The Hydrobox Solution
By providing Alperia with real-time control of their penstocks, the new digital monitoring system, dubbed Hydrobox, has not only improved hydroelectric infrastructure and worker safety, it allows the company to gather data that helps it dig deeper into the root causes of penstock rupture, allowing Alperia to maintain its position as a leader in the field of safety innovation.
Having an accurate behavioral model and ongoing access to real-time data also helps Alperia plan maintenance in advance and design a schedule that reduces both inspection and maintenance downtime. And because the solution optimizes GPS sensor use, the company saves money because it needs to purchase fewer sensors that are also less expensive.
A Promising Future
With the success of the Hydrobox project, Alperia is optimistic about the potential for AI in the future of energy production. “We think AI solutions can assist us in optimizing maintenance operations, for instance by monitoring acoustics in plant components to detect anomalies early,” says Theiner. “And if we look at production planning, combining an AI analysis of weather data with predictive market analysis, could potentially help us optimize production in order to obtain the best spot prices in the electricity market.”
As for Fieldbox, the insights gleaned from developing Hydrobox will help the AI operator develop future behavioral models for large infrastructure projects, such as dam or railway surveillance.
The road ahead looks promising for AI innovation, energy and infrastructure safety.