- Mingyang has over 15,000 wind turbines each with hundreds of sensors generating data every second. That adds up to a massive scale of hundreds of millions of data records every day — which a six-node TDengine deployment can ingest and store with no performance deterioration.
- Their databases now contain over 4 billion records with an average of 700 columns each, but their data only occupies 24 TB on disk, coming in at a compression ratio of 10%.
- Even on this massive scale, the complex aggregate queries that Mingyang uses to monitor turbine status information return in a fraction of a second.
Wind farms are an increasingly important component of power generation today — according to the Global Wind Energy Council’s latest report, a total of 117 gigawatts of new wind power was installed in 2023, representing 50% growth year over year. These wind farms must run at peak efficiency in order to provide the most benefit to our society in terms of clean energy production. But wind turbines generate more than just electricity; they also generate vast amounts of time-series data that must be stored and processed in a scalable, affordable manner.
Operators rely on these data to make informed decisions that have a direct impact on the efficiency of their farms, and as such the selection of a suitable data platform is extremely important. Earlier this year, Mingyang Group, one of the world’s top wind operators, chose TDengine as the foundation for their smart wind energy system. TDengine’s high performance and scalability now enable Mingyang’s real-time monitoring and prediction applications for their wind turbines, delivering a wealth of data-driven possibilities.
Company Overview
Mingyang Group is a global leader in wind energy. For the past 30 years, Mingyang has continued to develop and innovate large-capacity, low-wind-speed onshore and offshore wind turbines, overcoming key technical industry challenges and greatly improving the utilization rate of wind resources in various regions.
Case Background
At present, over 15,000 Mingyang wind turbines have been deployed in more than 800 projects worldwide, and each of these turbines has hundreds to thousands of data collection points that each generate one record per second — adding up to hundreds of millions of data points every day. These data points are stored and used in essential applications such as centralized monitoring and analytics.
For Mingyang, it’s essential to monitor the latest status information of each data collection point on each turbine. Due to the vast number of metrics collected on wind turbines, creating the data schema for this dataset is challenging. The data is stored in wide tables with as many as 884 columns, which can require special configuration to ensure performance.
Product Selection
Mingyang was building an energy big data application system and required a way to improve time-series data processing. With this in mind, they began testing TDengine in 2022, though at that time there were concerns regarding its performance in wide table scenarios. Since the release of TDengine 3.0, however, they have been encouraged by a number of optimizations in this area. In fact, within two or three versions, the performance issues that they experienced during testing had been resolved.
The LAST
function, which returns from cache the last non-null value in a column, is a good example. In scenarios with wide tables having many columns mainly consisting of null values, the LAST
function took minutes to produce results in earlier versions of TDengine. At present, thanks to adjustments and optimizations, the function returns in a matter of milliseconds.
Project Results
The company deployed TDengine in single-replica mode on six servers each having 4 cores, 24 GB of RAM, and 10 TB hard disks. Over 400 databases have been created on this TDengine cluster to store the data from more than 10,000 wind turbines. The databases contain more than 4 billion records with an average of over 700 columns. This occupies 24 TB of disk space, indicating a compression ratio of around 10%.
Mingyang collects data in two ways: by transmitting real-time data over a network to their data center and writing it into TDengine, and by preparing data files covering certain time periods and then copying them into TDengine. TDengine displays excellent performance with Mingyang’s data: they use complex LAST
/LAST_ROW
queries across multiple tables to monitor the latest status of their turbines, and these queries return results in as little as 0.2 seconds.
TDengine is now a key component of Mingyang’s energy big data system, responsible for the efficient storage and processing of data. Through data sharding and partitioning, TDengine flexibly scales to avoid bottlenecks. Thanks to the fast query performance that TDengine delivers, Mingyang has been able to implement real-time monitoring and prediction of wind turbines, allowing the company to detect anomalies and failures quickly and perform maintenance or adjustments as appropriate. Other business applications are more convenient now that they have access to the data in TDengine.
Over the past few years, Mingyang has witnessed TDengine’s commitment to providing a time-series database optimized for industrial scenarios like their own. Going forward, Mingyang intends to continue uncovering TDengine’s potential to create more possibilities in the field of wind energy.