• Twitter Social Icon
  • LinkedIn Social Icon
  • Facebook Social Icon

Visit the Elastacloud website 

  1. Elastacloud Channels
  2. Customer Stories
  3. Using Machine Learning to predict and respond to electricity line faults, a project with SSE
Search
rebecca.bourke
7 days ago
  ·  Edited: 7 days ago

Using Machine Learning to predict and respond to electricity line faults, a project with SSE



Using cutting edge cloud technology and innovative data solutions the Elastacloud team has delivered industry-leading business intelligence insights in a range of projects within the energy sector. Last year we worked with SSE, a large UK energy generation company to deliver industry-leading solutions to their data challenges.


SSE

SSE plc is primarily a developer, operator and owner of low-carbon energy assets with a strategic focus on regulated electricity networks and renewable energy. Their headquarters are in Scotland but have locations all over the UK, and they are listed on the London Stock Exchange whilst being a constituent of the FTSE 100 index. SSE also maintain and invest in transmission and distribution networks, generating and delivering electricity across the UK.


For SSE, faults along their networks, which are often caused by lightning strikes, are very costly to business operations with great impact on their customers. When blackouts occur, SSE receives costly fines from OFGEM, and as climate change continues to influence extreme weather, blackouts and the subsequent fines are likely to increase. SSE´s goal was to protect their customers in the areas with the highest chance of lightning strikes, and to do so they had to identify the top 100 locations that have the highest risk of faults and the greatest impact on customers when those faults occur. Additionally, with these insights they can better and more cost-effectively install surge resisters to reduce risk of faults from lightning strikes.

SSE consulted Elastacloud because they wanted to build a Machine Learning model that would target these high-risk areas using historical and current data. They wanted an interactive dashboard platform that would help them make better business and operational decisions.


The Data

SSE collect large amounts of geographical, meteorological, engineering and network data to understand the key features of their locations such as weather patterns and customer demographics. The potential of these large data sets a yet untapped and could be used to better predict and respond to faults. The Elastacloud data science team collected these data sets and explored correlations between them.


The first step was to join the data sets to each other by connecting the geological, meteorological, customer and faults data. For instance, the lightning strikes were compared to the faults data to identify if lightning was a potential cause. The team were then able to combine geographical data and frequency data to show the areas of most probable lightning strikes, which included analysing engineering and land usage data to discover which areas were most susceptible to faults.

Using this data the team were able to build machine learning models to find the lightning strike hotspots.

A Machine Learning Answer

The Elastacloud team used machine learning algorithms to build AI models that used the historic ´training data´ to make predictions or decisions for future faults involving lightning strikes on a given task. The models consider a variety of features of the lightning strikes such as distance, duration and frequency. Adjoining this to probability data, information about the cause of a fault and the likeness of it occurring can be seen with great accuracy.


With complex data sets like this, the presentation needs to be clear. Our BI Team used SQL and Power BI capabilities to structure and categorise the features of the data into multiple segments. The platform created from these tools gave SSE the ability to view the dates, equipment identification numbers and regions in a single comprehensive report with an intuitive and interactive front end. The report gave SSE insight into the following features of the faults:

· Time occurred

· Cause

· Location

· Frequency

· Consumer impact

· Costs

The models provide SSE with cutting edge insights around power blackouts and surges in SSE’s generation and identifies the top 100 hotspot areas along their networks. This allows SSE to make smarter, more targeted and cost-efficient decisions when implementing surge resistors using the machine learning models as a validation tool. Through this, SSE is better equipped to reduce the number and duration of electricity blackouts foregoing heavy fines from OFGEM and improving electricity supply efficiency for their customers.


Find out more about Elastacloud in the Energy Sector here.

Or if you want to read more stories of our projects and customer experiences click here.

0 comments0 replies
0
51 views
0 comments
0 replies
Newest PostsA Higher Education Project: Transcription Service Powered by Microsoft Azure Cognitive ServicesAn Elastacloud Energy Solution: The BSUoS Forecast.Applying Learning Analytics to Better Understand and Develop Your Students
Categories
Team Elastacloud
Community
Parquet.Net
AutoExploreR
Cloud Operating Model
See all categories
New Posts
  • george.darby
    Dec 5, 2019

    A Higher Education Project: Transcription Service Powered by Microsoft Azure Cognitive Services

    Discussion 
    Elastacloud have been working with a large, top-50 ranked university on several projects that leverage and deliver increased use of digital, data science and cloud solutions within the university. Recently, they approached us with a problem: they had over 4000 hours of interview research per annum in audio format that, when possible, is manually transcribed into text. The information consisted of valuable project research, but the process of manual transcription is very costly, time consuming and not as secure as they would like. To solve this problem, The Elastacloud team delivered a self-service web portal fully integrated with the university’s existing ecosystem to allow users from the University to simply drag, drop and submit multiple audio files for processing; the subsequent natural language transcripts being received in a few minutes. This allows much greater value to be realised from the data now in text format, for example search and analytics. Additionally, the university's existing Microsoft-powered digital landscape tied in well with our Microsoft technology expertise. Therefore, we were able to help the University build a platform that combined seamlessly with their portal, security and user experience. In the absence of a platform to transcribe audio data, universities often consult third party transcriptionists. The standard industry rate in the UK for hiring a transcriptionist is anywhere between £50-90 per hour. Whereas, the default rate for Azure cognitive services speech to text function, which we use to run the platform, is 75p per hour. We understood the potential for Universities to make significant monetary savings, and forego the time spent transcribing manually. The project successfully established an intuitive platform that has reduced the Universities costs and improved speed, efficiency and security. Currently, the university use the service to transcribe research-based audio information internally but will aim to extend the service from researchers to a wider audience. Therefore, the next step for us is to look at the wider applications of this service… Students Adopting a transcription service as a University brings a range of benefits not only for operational use, but potentially for staff and students throughout the University. With access to transcription services, students can undertake more varied and dynamic research outside the classroom such as interviews, where 1 to 1 conversation can be recorded, transcribed in minutes, and saved for later reference. It also makes reviewing and finding specific data significantly easier without missing key points. Faculty For the faculty, it is important to realise that many of your students may already be using, or thinking about using, external applications and services to transcribe audio data. With a service built into their platforms, the university has ownership of this and can overlook student engagement with the service and the entire virtual learning environment. In use, the lectures can be swiftly transcribed and posted online for students to review without missing out on key pieces of spoken information. It would also prove significantly beneficial to international students on the course who could use the transcribed copies of the lesson to study and review at their own pace.
    0 comments
    0 replies
    0
  • george.darby
    Nov 5, 2019

    An Elastacloud Energy Solution: The BSUoS Forecast.

    Discussion 
    Today, energy companies are at the forefront of cutting-edge technology, with an onus on supplying energy in the most sustainable way. The diversification of the UK energy supply mix incorporates low-carbon renewable energy sources such as wind, hydro and solar. Today more than 20 percent of energy in the United Kingdom comes from renewable sources. In summer, it’s often closer to 100 percent. Many energy companies have diversified their supply sources to incorporate these low-carbon and sustainable energy sources, but this has led to complications in balancing demand and supply. The Balancing Services Use of System Charge (BSUoS), charged by National Grid to help recoup the costs of balancing the grid from energy production, is a tariff which energy generators are liable to pay. The fee is paid to the system operator in half-hour periods for balancing the supply to meet demand, ensuring providers are generating energy efficiently. But if generation is not tightly monitored, the charge accumulates, severely damaging profitability. Difficulty predicting charges The charge is also difficult to predict. The price is volatile with external factors, such as weather conditions, which reflects the energy mix of the UK. The intermittent power generation and intermittency of weather conditions make calculating predictions complicated. With the ability to predict the charge, generation companies can better respond to the inherent intermittency of renewable energy generation and pricing. Equally, they will learn more about the drivers of the BSUoS charge and how to avoid over or under producing energy. The Elastacloud team were approached by a large UK energy generation company to discover if this could be possible with their data, and so, in a recent project we built the BSUoS Forecast. The solution: the BSUoS Forecast The forecast was built using Microsoft Azure machine learning services. Our Data Scientists used historical BSUoS, demand, and generation data to build the machine learning model for the forecast that generates automated reports for the user. These reports provide a prediction of the BSUoS charge for the customer with significantly greater accuracy than other industry standard models. The service enables energy companies to reduce costs by avoiding generation during the high price periods, and a reduced error risk. The proficiency and accuracy of the model not only provides advanced, industry-leading predictive reports, but makes it understandable and intuitive. As for the wider energy sector, the Forecast is a market-leading data science service, allowing generation companies to more profitably diversify their energy mix through improved optimisation of their own generation profile. Our unparalleled experience using Azure machine learning cloud services, combined with our energy sector expertise has enabled us to provide our clients in the industry with cutting-edge business solutions. The BSUoS forecast is just one example of how we boost our customers’ value through the optimisation of data. Get in touch today to find out more about the BSUoS forecast. The tool is available for energy companies who want to predict their BSUoS charge.
    0 comments
    0 replies
    0
  • george.darby
    Oct 18, 2019

    Applying Learning Analytics to Better Understand and Develop Your Students

    Discussion 
    In the last decade or so the Higher Education (HE) sector has become a very competitive environment. Year-on-year, applications come in from across the UK, with institutions fighting to fill their courses with students who they believe have the best chance of succeeding for themselves and for the university. In this increasingly competitive market, attracting and retaining students is a top priority for all HE institutions. The general shift toward a more student (or customer) focused business model by many institutions shows evidence of this and suggests that universities are strategizing how they can best acquire and retain students. Institutions are beginning to recognise the importance of data science practices to enhance their understanding of their students and staff alike. In many cases, the application of data science techniques and platforms to analyse large data sets has proved extremely beneficial. Better quality, more detailed, and wider-spread data allows for better decision making and accurate strategic management of business operations. These capabilities can provide a university with a competitive edge over rivals in acquisition and retainment of their students. A New Challenge Elastacloud's work in the Higher Education (HE) sector aims to bring intelligence to institutions' data, helping to inform decisions and operations. Uniquely, we combine expert data science and machine learning capabilities with a high level of industry knowledge, helping us to provide the best solutions to our customers. Within the last year, a large UK university with a recognised diverse student base consulted Elastacloud, wanting to understand how students from different demographic groupings at their university engaged with the digital practice enhancement applications they had created. They also wanted to measure the students' digital capabilities and find out the extent to which interventions enhanced their skills. The university wanted to understand these key points so they could make informed decisions about how to help improve their students' digital capabilities; for this they needed Learning Analytics. In the context of HE, Learning Analytics (LA) is the measurement, collection, analysis and reporting of data about students for the purposes of understanding and optimising learning and the environments in which it occurs. Universities implement LA for a number of reasons, chiefly to retain students. The project sought to understand what effects using LA to make educational interventions has on distinct demographic groupings. The Project We began to explore and aggregate the university's demographic, biographic, and Virtual Learning Environment (VLE) usage data. Our Data Scientists were there to guide and inform the data exploration process. Through applying LA, machine learning technology, and data science techniques, we were able to identify cohorts of students and observe which demographic groups had low engagement, low attainment, or had high drop-out rates. The importance of this part of the project was that we were able to use engagement with VLE as marker for students at risk of dropping out of university. By using LA programmes, Elastacloud provided the university's project team with key information and insight into VLE usage and student digital behaviour trends. In addition, a baseline of engagement with the VLE was determined so that when interventions were made, they could be analysed for their effectiveness. The objective was to increase student and staff involvement with VLE and deliver student behaviour measurement capabilities for the university. The Outcomes The project found that there is a relationship between a student’s engagement with the VLE and their likelihood of dropping out. As a result, the university now has actionable insights into their previously untapped data, allowing them to work towards reducing drop-out rates amongst all demographics of students. Additionally, on our journey to answering the key questions posed by this project's aims, we were able to create new opportunities with machine learning analytic tools. The university aims to deploy and monitor these tools for student and employer use, ensuring engagement goals are met and feedback capabilities are encouraged. These include self-auditing staff programmes and student skills reflection areas which enable them to find help and access personalised feedback to improve their digital capabilities. The adoption of these machine learning tools has given the university the means and understanding to identify students requiring extra support and make interventions accordingly thus deterring them from dropping out. The university was impressed by what we were able to achieve together and how quickly we were able to achieve it. As such they continue to consult Elastacloud on large scale data projects as we help to answer some of the sector's most pressing questions.
    0 comments
    0 replies
    0
  • Channels

  • Contact

  • Main Site

  • More