Pumped storage power station data development method

Abstract

The invention relates to a pumped storage power station data development method. The method is suitable for the technical field of pumped storage power station data processing and value mining. The technical problem to be solved by the invention is to provide the pumped storage power station data development method. According to the technical scheme, the pumped storage power station data development method comprises the steps that data value weights and scene weights are given to pumped storage power station data on a storage device, the data value weights are assigned based on the influence degree of the data on services, and the scene weights are assigned based on a pumped storage power station operation scene; calculating the comprehensive priority of each data by combining the data value weight and the scene weight of each data and the access frequency in the latest t time based on the current service content and the current operation scene every t time; based on the comprehensive priorities of the data, the storage areas of the data on the storage device are adjusted, and the storage areas comprise a hot area, a warm area and a cold area.

Description

TECHNICAL FIELD

[0001] The present application relates to a pumped storage power station data development method. It is suitable for pumped storage power station (hereinafter referred to as "pumped storage power station") data processing and value mining field, and can be applied to pumped storage power station unit operation monitoring, fault early warning, peak shaving efficiency optimization, equipment life prediction and other core business scenarios. BACKGROUND

[0002] Pumped storage power station is an important device for peak shaving, valley filling, frequency regulation and phase regulation in modern power system. As a core regulation resource of new power system, its operation process will generate massive multi-source heterogeneous data, mainly including three types: 1. Unit operation data: such as water pump water turbine speed, guide vane opening, volute water pressure, generator stator temperature and other time series data (sampling frequency 1-10Hz); 2. Grid interaction data: such as grid peak shaving / valley filling load instruction, grid-connected power, voltage frequency and other structured data (update frequency 1-5 minutes); 3. Equipment state and environment data: such as unit vibration waveform (unstructured data), plant temperature and humidity, equipment insulation value and other mixed data (sampling / update frequency 0.1-1Hz).

[0003] At present, the pumped storage power station data processing mainly adopts traditional ETL (Extract-Transform-Load) integration method and static layered storage method.

[0004] Among them, the ETL integration method realizes the "extraction-static conversion-loading" of multi-source data through batch processing tools (such as Hive and Spark Batch of Hadoop ecosystem): first, data is extracted from sensors, SCADA (Supervisory Control And Data Acquisition), grid dispatching system and other sources, then data conversion is completed according to pre-defined fixed rules (such as unified field format, unit conversion), and finally loaded into data warehouse (such as Greenplum). But the traditional ETL has the following problems: ① Poor real-time performance, unable to meet the needs of emergency scenarios. Traditional ETL adopts batch processing mode (such as processing a batch of data every 5 minutes), and the data integration delay is generally more than 5 minutes. While pumped storage power station unit fault early warning (such as bearing overheating), grid emergency peak shaving and other scenarios require real-time data support (time delay requirement <30 seconds), and batch processing delay will cause missed fault warning window or peak shaving response lag.

[0005] ②, low adaptability, relying on manual intervention: data conversion rules are statically predefined, when multi-source data formats or features change (such as the sampling frequency of the newly added vibration sensor is changed from 1Hz to 10Hz, and the "peak shaving priority" parameter is added in the power grid load instruction field), manual reprogramming of the conversion script is required, and the average adjustment time is more than 4 hours, which cannot adapt to data changes.

[0006] The static hierarchical storage method statically divides data into three levels of "hot, warm, and cold", and allocates storage media based on fixed periods or access frequencies. Hot data (such as near 7-day unit operation data): stored in memory (such as Redis) to meet high-frequency access; warm data (such as near 1-3 month data): stored in SSD (Solid State Drive) to balance access speed and cost; cold data (such as 3 months or more historical data): stored in HDD (Hard Disk Drive) to reduce storage cost. However, the static hierarchical storage method has the following problems: Ⅰ, serious waste of storage resources: the hierarchical rule is based on fixed periods (such as hot data fixed for 7 days), without considering the scene characteristics of pumped storage power stations - the access frequency of unit data during peak shaving period (19:00-22:00) is 3-5 times that of flat period (10:00-15:00), but static layering will cause hot data to occupy memory resources during flat period (utilization <30%), and memory resources will be insufficient during peak shaving period (warm data needs to be temporarily loaded, time delay increases by 2-3 seconds).

[0007] Ⅱ, insufficient data value mining: only "access frequency" is used as the basis for layering, without considering data value - for example, vibration data 1 hour before unit failure (low access frequency, but high value for fault tracing) will be classified as cold data by static layering, and needs to be loaded from HDD (time delay >5 seconds), resulting in low efficiency of fault root cause analysis. SUMMARY

[0008] The technical problem to be solved by the present application is to provide a pumped storage power station data development method.

[0009] The technical solution adopted by the present application is: a pumped storage power station data development method, comprising: assigning data value weights and scene weights to pumped storage power station data on the storage device, wherein the data value weights are assigned based on the influence of the data on the business, and the scene weights are assigned based on the pumped storage power station operation scene; every t time, based on the current business content and the current operation scene, combining the data value weights and scene weights of each data, and the access frequency in the last t time, calculating the comprehensive priority of each data; Adjust a storage area of each data on the storage device based on a comprehensive priority of each data, the storage area including a hot area, a warm area and a cold area.

[0010] The storage device includes Redis as the hot area, an SSD array as the warm area, and an HDD array as the cold area.

[0011] The comprehensive priority of each data is calculated based on the current business content and the current running scenario, in combination with the data value weight and the scenario weight of each data, and the access frequency in the last t time, and includes: Comprehensive priority = (data value weight × a) + (access frequency in the last t time × b) + (current running scenario weight × c) Wherein, a is the priority calculation weight of the data value weight; b is the priority calculation weight of the access frequency; c is the priority calculation weight of the scenario weight; and a+b+c=1.

[0012] Further comprising: When accessing data in real time, extracting data features of the current data, the data features including time sequence features, scenario demand features and format features; Based on the data features of the current data and the historical data in the historical database, if the matching is successful, the data conversion rule corresponding to the same type of historical data is called; Based on the data conversion rule, the data is converted, and the converted data is loaded onto the storage device.

[0013] The clustering matching adopts a K-means clustering algorithm.

[0014] The data is converted based on the data conversion rule, and the converted data is loaded onto the storage device, including: Based on the Flink stream processing framework, the data converted based on the data conversion rule is loaded in real time to the hot area on the storage device.

[0015] For real-time tasks, the hot area data is preferentially called; the cold area high-value data is preloaded and calculated in parallel with the hot area data; and the calculation result is output in real time to the SCADA system; For offline tasks, the warm / cold area data is loaded in batches; the loading speed is matched with the CPU / GPU calculation speed; and the result is stored in the cold area.

[0016] A pumped storage power station data processing device, comprising: A weight assignment module for assigning a data value weight and a scenario weight to pumped storage power station data on the storage device, wherein the data value weight is assigned based on the influence degree of the data on the business, and the scenario weight is assigned based on the pumped storage power station running scenario; The priority calculation module is used to calculate the comprehensive priority of each data every t time interval, based on the current business content and current operating scenario, combined with the data value weight and scenario weight of each data, as well as the access frequency in the most recent t time interval. The storage adjustment module is used to adjust the storage area of ​​each data on the storage device based on the comprehensive priority of each data. The storage area includes hot zone, warm zone and cold zone.

[0017] A storage medium storing a computer program executable by a processor, which, when executed, implements the steps of the pumped storage power station data development method.

[0018] A pumped-storage power station data processing device includes a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the pumped-storage power station data development method.

[0019] The beneficial effects of this invention are as follows: Every t time interval, based on the current business content and current operating scenario, combined with the data value weight and scenario weight of each data, as well as the access frequency within the most recent t time interval, the comprehensive priority of each data is calculated. Based on the comprehensive priority of each data, the storage area of ​​each data on the storage device is adjusted to achieve scenario-based dynamic hierarchical storage, increasing the storage resource utilization rate from 30% to 65%; and reducing the data access latency during peak periods from 2.5 seconds to 0.3 seconds.

[0020] This invention performs clustering and matching based on the data characteristics of current data and historical data in a historical database. Based on the matching results and data transformation rules, it establishes a dynamic mapping mechanism for multi-source data. Based on data time-series characteristics and scenario requirements, it automatically generates transformation rules, achieving real-time, adaptive integration of multi-source data. The multi-source data integration latency is reduced from 5 minutes in traditional ETL to within 10 seconds, meeting the real-time requirements of power grid peak shaving and fault early warning (latency requirement <30 seconds). When data format changes, the rule adjustment time is reduced from 4 hours to 0 (automatic adaptation).

[0021] In this invention, computing tasks call layered data on demand (real-time tasks prioritize hot zone data, and offline tasks preload cold zone data), reducing loading latency, improving the efficiency of data value mining, and increasing the accuracy of fault early warning from 75% to 92% (based on 100 fault simulations, the false negative rate decreased from 15% to 3%); the root cause analysis time for faults is shortened from 1 hour to 15 minutes (due to reduced loading latency of high-value data). Attached Figure Description

[0022] Figure 1 This is the architecture diagram corresponding to the embodiment. Detailed Implementation

[0023] Example 1: Figure 1 The diagram shows the architecture of the data development method for pumped storage power stations in this embodiment, which includes a data access layer, a dynamic mapping layer, a dynamic hierarchical storage layer, a storage-computing collaboration layer, and an application layer.

[0024] In this embodiment, the data access layer includes data sources such as sensors (vibration, temperature), SCADA systems, and power grid dispatching systems, which are accessed through an edge gateway (supporting multiple protocols); the dynamic mapping layer includes a database, an improved K-means matching module, and a Flink stream processing module to achieve automatic conversion and real-time loading of multi-source data; the dynamic hierarchical storage layer (storage device) is divided into Redis (hot zone), SSD (warm zone), and HDD (cold zone), which controls the dynamic migration of data in a scenario-based hierarchical manner; the storage-computing collaboration layer includes a task classification module (real-time / offline) and a resource scheduling module to achieve collaborative calling of computing tasks and hierarchical data; the application layer outputs six types of business application results, such as fault warning, peak-shaving efficiency optimization, and equipment life prediction, and connects to the pumped storage power station operation and maintenance system and the power grid dispatching platform.

[0025] The pumped storage power station data development method in this embodiment includes the following steps: S100: Multi-source data enters the dynamic mapping layer through the access layer and completes automatic conversion; S200: The converted data enters the dynamic hierarchical storage layer, and storage media are dynamically allocated according to the scenario and value. S300 and the collaboration layer call layered data as needed according to the task type to complete the calculation; S400: The calculation results are output through the application layer to support the business decisions of pumped storage power stations.

[0026] In some specific embodiments, step S100 specifically includes: S110. When accessing data in real time, extract the data features of the current data. The data features include time series features, scenario requirement features, and format features.

[0027] In this example, the timing characteristics include: sampling frequency of unit speed (10Hz), synchronization correlation with volute water pressure (0.85), and update cycle of load command (10 seconds); scenario demand characteristics include: matching accuracy threshold of unit output and load command in peak shaving scenario (95%), and acquisition accuracy threshold of equipment insulation value in maintenance scenario (0.1Ω); and format characteristics include: JSON format for vibration data, CSV format for temperature data, and XML format for dispatching command.

[0028] In this embodiment, data features are extracted from the historical data (1 year, 50TB) of the pumped storage power station data source, and a historical database is established by combining the data transformation rules of each data source.

[0029] S120. Based on the data characteristics of the current data (such as the sampling frequency of the newly added vibration sensor being 10Hz and the format being JSON), perform clustering matching with historical data in the historical database. If the matching is successful, call the data conversion rule corresponding to the same type of historical data; if the matching fails (such as adding a new data source), generate temporary conversion rules based on the scenario requirements (such as converting the insulation data of new equipment with a precision of 0.1Ω in the maintenance scenario), and update the historical database synchronously.

[0030] In this embodiment, an improved K-means clustering algorithm is used, which introduces temporal correlation as a weight to achieve automatic matching and transformation of multi-source data.

[0031] S130: Based on data transformation rules, transform the data and load the transformed data into a hotspot on the storage device in real time using the Flink stream processing framework, with processing latency controlled within 10 seconds (actual average latency of 8.2 seconds).

[0032] In some specific embodiments, step S200 specifically includes: S210. Assign data value weight and scenario weight to the pumped storage power station data on the storage device. The data value weight is assigned based on the degree of impact of the data on the business (0-1), such as vibration data 1 hour before the fault (0.9), real-time load command (0.8), and ambient temperature data (0.3). The scenario weight is assigned based on the operation scenario of the pumped storage power station (0-1), such as peak shaving scenario (19:00-22:00, 0.8), valley filling scenario (00:00-06:00, 0.6), flat period scenario (other time periods, 0.4), and maintenance scenario (0.7).

[0033] S220. Every t time interval, based on the current business content and current operating scenario, combined with the data value weight and scenario weight of each data, as well as the access frequency within the most recent t time interval, calculate the comprehensive priority of each data.

[0034] In this embodiment, a hierarchical adjustment is performed every 5 minutes to calculate the overall priority of each data item on the storage device. Overall Priority = (Data Value Weight × a) + (Access Frequency within the Last t Time Period × b) + (Current Scenario Weight × c) Where a is the priority calculation weight of data value weight, which is 0.6 in this example; b is the priority calculation weight of access frequency, which is 0.3 in this example; and c is the priority calculation weight of scenario weight, which is 0.1 in this example.

[0035] S230. Based on the overall priority of each data, adjust the storage area of ​​each data on the storage device. The storage area includes a hot zone, a warm zone, and a cold zone.

[0036] In this embodiment, the hierarchy is divided according to the overall priority: high priority (>0.7): stored in Redis (hot zone); medium priority (0.4-0.7): stored in SSD (warm zone); low priority (<0.4): stored in HDD (cold zone).

[0037] For example, during peak hours (19:00): the overall priority of load command data = 0.8×0.6 + 0.9×0.3 + 0.8×0.1 = 0.83 (hot zone). During off-peak hours (12:00): the overall priority of the same load command data = 0.8×0.6 + 0.2×0.3 + 0.4×0.1 = 0.58 (mild zone). The system automatically migrates from Redis to SSD, freeing up memory resources.

[0038] In some specific embodiments, step S300 specifically includes: The data application tasks of pumped storage power stations are divided into two categories, and different scheduling strategies are adopted: Task type Typical scenario Scheduling strategy Real-time task Fault early warning, real-time output control 1. Prioritize hot zone data (such as real-time vibration, speed); 2. Preload cold zone high-value data (such as historical fault vibration templates) and calculate in parallel with hot zone data; 3. Real-time output of calculation results to SCADA system. Offline task Annual efficiency analysis, fault tracing 1. Load warm / cold zone data by batch (such as monthly warm zone data, 3 years of cold zone data); 2. Load speed matches CPU / GPU calculation speed (such as loading 10GB data per batch, calculation time and loading time are both 2 minutes); 3. Results are stored on HDD for subsequent queries. The following example illustrates a fault warning task: Task trigger: When the vibration data of the hot zone exceeds the threshold (e.g., >0.15mm / s), a fault warning is triggered; Data retrieval: Two operations are executed in parallel: ① Read real-time rotational speed and water pressure data of the hot zone (0.3 seconds); ② Preload vibration data of similar faults in the cold zone over the past year (0.5 seconds). Collaborative computing: The GPU server processes two types of data simultaneously, calculates fault similarity (e.g., if the cosine similarity is >0.9, the faults are classified as the same type), and takes 0.4 seconds. Output: The warning information (fault type and suggested measures) will be pushed to the operation and maintenance terminal within 1.2 seconds to meet the real-time requirements.

[0039] Example 2: This example is a data processing device for a pumped storage power station, comprising: The feature extraction module is used to extract the data features of the current data when data is received in real time. The data features include time series features, scenario requirement features, and format features. The rule bar module is used to perform clustering matching between the current data and historical data in the historical database based on the data characteristics of the current data. If the match is successful, the data transformation rules corresponding to the same type of historical data are called. The data conversion module is used to convert data based on data conversion rules and load the converted data into a storage device; The weighting module is used to assign data value weights and scenario weights to the pumped storage power station data on the storage device. The data value weights are assigned based on the impact of the data on the business, and the scenario weights are assigned based on the operation scenario of the pumped storage power station. The priority calculation module is used to calculate the comprehensive priority of each data point every t time interval, based on the current business content and current operating scenario, combined with the data value weight and scenario weight of each data point. The storage adjustment module is used to adjust the storage area of ​​each data on the storage device based on the comprehensive priority of each data. The storage area includes hot zone, warm zone and cold zone. The collaborative scheduling module is used to prioritize hot zone data for real-time tasks; preload high-value cold zone data for parallel computation with hot zone data; output computation results to the SCADA system in real time; and load warm / cold zone data in batches for offline tasks, with loading speed matching CPU / GPU computation speed; and store results in the cold zone.

[0040] Example 3: This example is a storage medium that stores a computer program that can be executed by a processor. When the computer program is executed, it implements the steps of the pumped storage power station data development method described in Example 1.

[0041] Example 4: This example is a data processing device for pumped storage power stations, which has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the pumped storage power station data development method described in Example 1.

[0042] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the aforementioned functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0043] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0044] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0045] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the aforementioned program can be printed, because the aforementioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0046] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0047] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0048] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0049] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A pumped storage power plant data development method characterized by, The method comprises: assigning data value weights and scene weights to pumped storage power station data on a storage device, wherein the data value weights are assigned based on the degree of influence of the data on business, and the scene weights are assigned based on the running scene of the pumped storage power station; every t time, calculating the comprehensive priority of each data based on the current business content and the current running scene, in combination with the data value weights and the scene weights of each data, and the access frequency in the last t time; based on the comprehensive priority of each data, adjusting the storage area of each data on the storage device, the storage area including a hot area, a warm area, and a cold area.

2. The pumped storage power plant data development method according to claim 1, characterized by, The storage device includes Redis as the hot area, an SSD array as the warm area, and an HDD array as the cold area.

3. The pumped storage power plant data development method according to claim 1, characterized by, The calculation of the comprehensive priority of each data based on the current business content and the current running scene, in combination with the data value weights and the scene weights of each data, and the access frequency in the last t time, comprises: Comprehensive priority = (data value weight × a) + (access frequency in the last t time × b) + (current running scene weight × c) wherein a is the priority calculation weight of the data value weight; b is the priority calculation weight of the access frequency; c is the priority calculation weight of the scene weight; and a + b + c = 1.

4. The pumped storage power plant data development method according to claim 1, characterized by, Further comprising: when accessing data in real time, extracting data features of the current data, the data features including time sequence features, scene demand features, and format features; based on the clustering matching of the data features of the current data with historical data in a historical database, if the matching is successful, calling the data conversion rule corresponding to the same type of historical data; based on the data conversion rule, converting the data, and loading the converted data to the storage device.

5. The pumped storage power plant data development method according to claim 4, characterized by, The clustering matching adopts a K-means clustering algorithm.

6. The pumped storage power plant data development method according to claim 4, characterized by, The conversion of the data based on the data conversion rule and the loading of the converted data to the storage device comprise: based on a Flink stream processing framework, loading the data converted based on the data conversion rule to the hot area of the storage device in real time.

7. The pumped storage power plant data development method according to claim 1, characterized by, Further comprising: for real-time tasks, preferentially calling hot area data; preloading high-value data in the cold area and calculating in parallel with the hot area data; the calculation result is output to the SCADA system in real time; for offline tasks, loading warm / cold area data in batches; the loading speed matches the CPU / GPU calculation speed; the result is stored in the cold area.

8. A pumped storage power plant data processing device characterized by comprising: The method comprises: a weight assignment module for assigning data value weights and scene weights to pumped storage power station data on a storage device, wherein the data value weights are assigned based on the degree of influence of the data on business, and the scene weights are assigned based on the running scene of the pumped storage power station; a priority calculation module for calculating the comprehensive priority of each data every t time based on the current business content and the current running scene, in combination with the data value weights and the scene weights of each data; a storage adjustment module for adjusting the storage area of each data on the storage device based on the comprehensive priority of each data, the storage area including a hot area, a warm area, and a cold area.

9. A storage medium having stored thereon a computer program, which is capable of being executed by a processor, characterized in that The computer program, when executed, implements the steps of the pumped storage power station data development method of any one of claims 1-7.

10. A pumped storage power plant data processing device having a memory and a processor, the memory having stored thereon a computer program executable by the processor, characterized in that, The computer program, when executed, implements the steps of the pumped storage power plant data development method of any one of claims 1-7.

Images