Home » Blog » Application of Time-Series Database in Nuclear Power Safety

Application of Time-Series Database in Nuclear Power Safety

The Russia-Ukraine war is still the focus of global attention. As a traditional nuclear power, Ukraine possesses multiple nuclear power plants within its borders, and their safety has raised widespread concerns. Today, let’s discuss the applications of time-series databases in the realm of nuclear power safety.

Monitoring System of Nuclear Power Safety

First of all, any nuclear power plant will have its own safety monitoring system. We can generally call it a DCS monitoring system. Simply put, it is a distributed control system. The nuclear power plant will collect relevant data through the DCS system to realize the operation of the nuclear power plant. In particular, it is necessary to monitor the operation status monitoring, alarm display, failure prediction and other operations of the reactor in an all-round way. Therefore, in the monitoring process, it will generate a large amount of time series data. This monitoring system is mainly composed of three parts: equipment monitoring, variable monitoring, and algorithm monitoring. Its goal is to realize remote monitoring covering all aspects of equipment, variables, and algorithms in the reactor protection control device. The system has the functions of variable online maintenance, parameter remote update, remote control, trend line display, visual alarm, historical data recording, operation playback, etc. It breaks through the traditional on-site monitoring, on-site maintenance and control mode of the reactor. It is of great significance in operation and maintenance scenarios such as design verification, operation debugging, operation maintenance, and fault location of the monitoring system.

Operation and Maintenance Data Related to Nuclear Power Safety

The operation and maintenance data involved in the DCS online monitoring system include operating status data, process data, parameter data, diagnostic result data, and operating status data, including data representing the operating status of equipment, such as power status, CPU usage, communication occupancy, and equipment temperature. etc.; process data refers to various signal data, logic data, and feature data after feature processing or calculation that participate in reactor control; parameter data describes the information of the unit, such as site configuration, site connection, channel configuration, parameter settings, etc. Data; Diagnosis result data is the results of power supply, communication, and operating status obtained by the safety-level DCS based on operating status data and characteristic data for diagnostic analysis.

Problems of Using Traditional Databases in Nuclear Power Safety

First, it is the storage capacity issue. Traditional security-level DCS remote online monitoring systems generally use relational databases for storage. For example, MySQL or Oracle, etc., require a large-capacity disk array. Taking the storage of 10,000 floating-point analog points with time scale as an example, sampling once per second and storing them for half a year requires more than 4,400 TB of space. The sampling period of the safety-level DCS is generally at the millisecond level, and the shortest even reaches 20 ms. Faced with the exponential growth of data volume, traditional relational databases cannot meet the demand.

The other problem comes with the acess performance. Traditional relational databases usually adopt methods such as partitioning and indexing to improve retrieval and access efficiency. However, in the face of the explosion of data volume, its access efficiency will gradually decrease. The DCS system needs to provide historical data functions such as running process playback and trend display. In addition, it will also provide data access services for upper-layer applications, such as health management, intelligent operation and maintenance, and digital twins. Taking drawing a 24-hour second-level trend line as an example, 86,400 pieces of data need to be read at one time. The above functions usually need to obtain tens of thousands or even hundreds of thousands of data at one time, which puts forward high requirements for fast access to the database.

Time Series Database Greatly Improves Storage & Access Performance

To meet the needs of real-time, high-frequency, and massive data storage, this is the best place for time-series databases. In terms of storage, the time-series database has a high data compression ratio, which can ensure maximum disk space saving without destroying data integrity and authenticity. For example, when a single server processes 300,000 points and the sampling period is 1s, the time-series database stores 200 hours of data and only occupies 4 GB of space; while under the same conditions, the relational database stores 5 hours of data already occupies 4 GB space. In addition, in terms of query, time-series databases have the characteristics of low latency and high concurrency. By optimizing common query modes and using indexing techniques, query latency can be reduced; query concurrency can be improved through caching, routing, and other technologies, and data access capabilities can reach to ten thousand pieces per second.