下载PDF
Predicting Rare Failures in Hydro Turbines
技术
- 分析与建模 - 机器学习
- 分析与建模 - 预测分析
- 功能应用 - 远程监控系统
- 传感器 - 加速度计
- 传感器 - 液位传感器
- 传感器 - 温度传感器
- 传感器 - 振动传感器
适用行业
- 公用事业
适用功能
- 维护
用例
- 预测性维护
挑战
运营水轮机的公用事业公司在执行定期维护以防止意外故障方面具有既得利益。大多数维护都是在计划的基础上进行的，如果需要，资产会离线、检查和主动维修。水轮机机组非常可靠，这意味着很少有计划外停机的例子。然而，这些故障对他们的运营商来说是非常昂贵的。
鉴于运营商对计划外停机时间的敏感性，许多运营商已经为涡轮机和发电机配备了传感器和平台，以实时收集有价值的性能信息。但是由于历史上的水电故障很少可以比较，丰富的流数据和基于传统统计的分析在预测真正的故障事件方面并不是很准确。事实上，他们经常通过良性误报使监控团队超载而产生更多问题，从而导致不必要的停机时间进行评估。这就引出了一个问题：人工智能能否帮助维护团队从数据中提取更多价值？
客户
未公开
关于客户
一家领先的水电公用事业公司，评估了他们如何应用机器学习来更好地分析有问题的涡轮机行为。该公司在其现有分析平台中受到误报的困扰，最近有一个大
解决方案
对于这项合作，水电公用事业公司希望评估 SparkPredict 在两个关键领域的表现，并将结果与他们现有的基于统计的解决方案进行比较：
• SparkPredict 能否在其中一台涡轮机发生已知事件之前检测到发电机的变化？
• SparkPredict 可以为工程团队提供哪些额外的见解来帮助他们改进维护？
SparkCognition 获得了一台涡轮机的两年数据，该涡轮机在 2015 年至 2016 年期间经历了多次故障事件。
从给定的传感器集合中，SparkPredict 数据科学家确定了 40 个重要标签，这些标签测量了发电机速度、功率输出、温度读数、油位、振动和轴间隙测量等重要变量。在收到数据的六周内，SparkPredict 数据科学家研究、训练并构建了一个无监督机器学习模型。无监督学习在预测罕见故障时特别强大，因为故障信息很少，而且没有一个始终如一地记录在案。这种类型的模型不是与以前已知的指标进行比较，而是适应资产的独特行为。通过标记异常行为，SparkPredict 可以识别未知故障的指标。
使用这种方法，SparkPredict 准确识别数据集中的涡轮机故障事件，并提前一个多月发出预警，成功满足第一个项目标准。
此外，SparkPredict 通过可解释的 AI 公开了事件的关键领先指标。特别是，该公用事业公司希望了解多个传感器如何相互关联，以及与历史测量值的微小偏差如何导致初期故障模式。 SparkPredict 提升了 9 个值发生变化的传感器。
收集的数据
Oil Level, Power Output, Temperature, Vibration, Generator Speed
运营影响
数量效益
相关案例.
Case Study
IoT Solutions for Smart City | Internet of Things Case Study
There were several challenges faced: It is challenging to build an appliance that can withstand a wide range of voltage fluctuations from as low at 90v to as high as 320v. Since the device would be installed in remote locations, its resilience was of paramount importance. The device would have to deal with poor network coverage and have the ability to store and re-transmit data if networks were not available, which is often the case in rural India. The device could store up to 30 days of data.
Case Study
Automation of the Oguz-Gabala-Baku water pipeline, Azerbaijan
The Oguz-Gabala-Baku water pipeline project dates back to plans from the 1970’s. Baku’s growth was historically driven by the booming oil industry and required the import of drinking water from outside of the city. Before the construction of the pipeline, some 60 percent of the city’s households received water for only a few hours daily. After completion of the project, 75 percent of the two million Baku residents are now served around the clock with potable water, based on World Health Organization (WHO) standards. The 262-kilometer pipeline requires no pumping station, but uses the altitude differences between the Caucasian mountains and the capital to supply 432,000 m³/d to the Ceyranbatan water reservoir. To the people of Baku, the pipeline is “the most important project not only in 2010, but of the last 20 years.”
Case Study
GPRS Mobile Network for Smart Metering
Around the world, the electricity supply industry is turning to ‘smart’ meters to lower costs, reduce emissions and improve the management of customer supplies. Smart meters collect detailed consumption information and using this feedback consumers can better understand their energy usage which in turn enables them to modify their consumption to save money and help to cut carbon emissions. A smart meter can be defined in many ways, but generally includes an element of two-way communication between the household meter and the utility provider to efficiently collect detailed energy usage data. Some implementations include consumer feedback beyond the energy bill to include online web data, SMS text messages or an information display in consumers’ premises. Providing a cost-effective, reliable communications mechanism is one of the most challenging aspects of a smart meter implementation. In New Zealand, the utilities have embraced smart metering and designed cost effective ways for it to be implemented. The New Zealand government has encouraged such a move to smart metering by ensuring the energy legislation is consistent with the delivery of benefits to the consumer while allowing innovation in this area. On the ground, AMS is a leader in the deployment of smart metering and associated services. Several of New Zealand’s energy retailers were looking for smart metering services for their residential and small business customers which will eventually account for over 500,000 meters when the multi-year national deployment program is concluded. To respond to these requirements, AMS needed to put together a solution that included data communications between each meter and the central data collection point and the solution proposed by Vodafone satisfied that requirement.
Case Study
NB-IoT connected smart meters to improve gas metering in Shenzhen
Shenzhen Gas has a large fleet of existing gas meters, which are installed in a variety of hard to reach locations, such as indoors and underground, meaning that existing communications networks have struggled to maintain connectivity with all meters. The meter success rate is low, data transmissions are so far unstable and power consumption is too high. Against this background, Shenzhen Gas, China Telecom, Huawei, and Goldcard have jointly trialed NB-IoT gas meters to try and solve some of the challenges that the industry faces with today’s smart gas meters.
Case Study
OneWireless Enabled Performance Guarantee Test
Tata Power's power generation equipment OEMs (M/s BHEL) is required to provide all of the instrumentation and measurement devices for conducting performance guarantee and performance evaluation tests. M/s BHEL faced a number of specific challenges in conducting PG tests: employing high-accuracy digital communications for instrumentation, shortening setup and dismantling time, reducing hardware required, making portable instrument setup, avoiding temporary cabling work and the material waste costs
Case Study
British Gas Modernizes its Operations with Innovative Smart Metering Deployment
The UK government has mandated that smart meters are rolled out as standard across Great Britain by end of 2020, and this roll-out is estimated to create £14 billion in net benefits to the UK in consumer energy savings and lower energy generation demand, according to the Oxford Economics report, “The Value of Smart Metering to Great Britain.” While smart-metering systems have been deployed in many countries, the roll-out in Great Britain is unique because it is led by energy retailers, who have responsibility for the Electricity and Gas meters. The decision to have a retailer-led roll out was made by DECC (Department of Energy and Climate Change) to improve customer experience and drive consumer benefits. It has also led to some unique system-level requirements to support the unique local regulatory model.
