An information processing method and system for a hydropower plant based on an industrial internet of things

Abstract

This invention provides a method and system for processing information from hydropower plants based on the Industrial Internet of Things (IIoT), belonging to the field of big data processing technology. The method includes: collecting IIoT sensor data during normal operation of the hydropower plant; weaving the data into time series data; performing data clustering; dividing each sensor data into N levels; mapping real-time data to the corresponding sensor levels; displaying a first coordinate axis on a first interface; displaying the average level over approximately M periods on the first coordinate axis; connecting the level points of all sensors in the clustering results; translating the clustering level lines so that they do not intersect; displaying a first bubble on the peak; and displaying detailed data of the corresponding sensor on a second interface. This solution allows for simple and convenient analysis of massive amounts of IIoT data from hydropower plants.

Description

Technical Field

[0001] This invention relates to the field of big data processing, and more specifically to a method and system for processing information in hydropower plants based on the Industrial Internet of Things. Background Technology

[0002] The Industrial Internet of Things (IIoT) refers to the technology and concept of connecting physical devices such as sensors, equipment, and instruments via the internet to enable data collection, transmission, analysis, and application, thereby optimizing industrial production and operation processes. By connecting various physical devices to the internet, the IIoT enables data sharing and collaborative work between devices, improving production efficiency.

[0003] By collecting various data in the production environment in real time through sensors and other devices, sensors are installed at key equipment and locations in the hydropower plant to collect data related to hydropower production in real time, such as water level, water quality, temperature, humidity, flow rate, and water pressure. By analyzing a large amount of production data, useful information can be extracted and potential faults in the hydropower plant can be detected in advance.

[0004] However, with the development of the industry, more and more sensors are connected to data centers through the Industrial Internet of Things (IIoT). In some hydropower projects, the number of sensors can reach tens of thousands. Furthermore, the complex environment of hydropower plants means that data collected by the same type of sensor can vary significantly. For example, a water pressure gauge may collect different data at different locations and water levels. The safety standards for data at different locations may also differ. For instance, the normal water pressure range at point A is 10-20 kPa, while at point B it is 20-40 kPa. Current technology typically requires setting separate alarm thresholds for points A and B. When there are many monitoring points and diverse data types, setting alarm thresholds for each point is extremely labor-intensive. Additionally, identifying anomalies from tens of thousands of monitoring data points is also very challenging. Summary of the Invention

[0005] To address the problems in the existing technology, this invention provides a method and system for processing information in hydropower plants based on the Industrial Internet of Things.

[0006] In one aspect of the present invention, a method for processing information in a hydropower plant based on the Industrial Internet of Things (IIoT) is provided: collecting IIoT sensor data during normal operation of the hydropower plant; organizing the sensor data into time series data at equal time intervals; clustering data with similar trends in the time series data to obtain multiple clustering results; for each sensor data, dividing each sensor data into N levels according to historical records, where N is a positive integer not less than 3; acquiring IIoT sensor data in real time and mapping the real-time data to the levels corresponding to the sensors; for each sensor data, taking the average level over the last M periods, where M is not less than 2 The first interface displays a positive integer; the vertical axis of the first coordinate axis represents the level, and the horizontal axis represents different sensors; for each sensor, the average level of its last M periods is represented on the first coordinate axis to obtain level points; for the same clustering result, the level points of all sensors in the clustering result are connected to obtain a clustering level line; the clustering level line is translated on the vertical axis of the first coordinate axis so that all clustering level lines do not intersect; for each clustering level line, when there is a peak on the clustering level line that is higher than a first threshold, a first bubble is displayed on the peak; clicking the first bubble displays the detailed data of the corresponding sensor on the second interface.

[0007] Furthermore, the step of dividing each sensor data into N levels based on historical records includes: calculating the average and standard deviation of the historical data; classifying the data as level 0 when it deviates from the average by one standard deviation; class 1 when it deviates from the average by between one and two standard deviations; class -1 when it deviates from the average by between one and two standard deviations; class 2 when it deviates from the average by more than two standard deviations; and class -2 when it deviates from the average by less than two standard deviations.

[0008] Furthermore, there are no fewer than two clustering level lines, and all of the clustering level lines are displayed on the first interface.

[0009] Furthermore, the method for determining whether there are peaks on the clustering level line that are higher than the first threshold is as follows: calculate the current average level of each cluster, calculate the distance of each point in the same cluster from the average level, and determine whether it exceeds the first threshold. If it exceeds the first threshold, it indicates that the peak is too high.

[0010] Furthermore, the second interface displays links to data information of other sensors in the same cluster as the anomaly sensor. When the user clicks a link, the corresponding sensor data details are displayed.

[0011] Another aspect of the present invention provides an information processing system for a hydropower plant based on the Industrial Internet of Things (IIoT). The system includes: a data collection module for collecting IIoT sensor data during normal operation of the hydropower plant; a data processing module for organizing the sensor data into time series data at equal time intervals; clustering data with similar trends in the time series data to obtain multiple clustering results; classifying each sensor data into N levels based on historical records, where N is a positive integer not less than 3; a real-time acquisition module for acquiring IIoT sensor data in real time and mapping the real-time data to the corresponding sensor level; and an analysis module for taking the nearest M values ​​from each sensor's data. The average grade over the period, where M is a positive integer not less than 2; a display module, used to display a first coordinate axis on the first interface, where the vertical axis of the first coordinate axis represents the grade and the horizontal axis represents different sensors; for each sensor, the average grade over its last M periods is represented on the first coordinate axis to obtain grade points; for the same clustering result, the grade points of all sensors in the clustering result are connected to obtain a clustering grade line; the clustering grade line is translated on the vertical axis of the first coordinate axis so that all clustering grade lines do not intersect; for each clustering grade line, when there is a peak on the clustering grade line that is higher than a first threshold, a first bubble is displayed on the peak; clicking the first bubble displays detailed data of the corresponding sensor on the second interface.

[0012] Furthermore, the step of dividing each sensor data into N levels based on historical records includes: calculating the average and standard deviation of the historical data; classifying the data as level 0 when it deviates from the average by one standard deviation; class 1 when it deviates from the average by between one and two standard deviations; class -1 when it deviates from the average by between one and two standard deviations; class 2 when it deviates from the average by more than two standard deviations; and class -2 when it deviates from the average by less than two standard deviations.

[0013] Furthermore, there are no fewer than two clustering level lines, and all of the clustering level lines are displayed on the first interface.

[0014] Furthermore, the method for determining whether there are peaks on the clustering level line that are higher than the first threshold is as follows: calculate the current average level of each cluster, calculate the distance of each point in the same cluster from the average level, and determine whether it exceeds the first threshold. If it exceeds the first threshold, it indicates that the peak is too high.

[0015] Furthermore, the second interface displays links to data information of other sensors in the same cluster as the anomaly sensor. When the user clicks a link, the corresponding sensor data details are displayed.

[0016] The present invention, through the above technical solution, can produce the following beneficial effects: Clustering sensor data according to similar trends and alerting users to outliers can avoid users having to set alarm thresholds for each sensor individually. When dealing with tens of thousands of sensors, this can greatly reduce the setup work.

[0017] Furthermore, by hierarchically classifying the data and displaying it as a straight line on a coordinate axis, the overall situation of tens of thousands of sensors can be displayed graphically on the same interface, making it convenient for maintenance personnel to observe the overall situation of the hydropower plant. At the same time, anomalies are displayed in bubbles, improving the ease of operation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 Here is an example of a level point; Figure 2 Here is an example of cluster rank lines; Figure 3 This is an example of a balanced clustering rank line; Figure 4 This is an example of an error message. Detailed Implementation

[0020] The invention will now be described in preferred form with reference to the accompanying drawings and specific embodiments.

[0021] This embodiment solves the above problem through the following steps: In one embodiment, the present invention provides a hydropower plant information processing method based on the Industrial Internet of Things, specifically including: Collect industrial IoT sensor data during normal operation of the hydropower plant.

[0022] During normal operation of a hydropower plant, all systems and equipment work together to maintain efficient power generation. Because the data during normal operation is specific and for reference only, this embodiment uses the data during normal operation as the standard to judge whether the data in subsequent operation processes is normal.

[0023] Industrial IoT sensor data from hydropower plants typically includes various parameters for monitoring and managing the power generation process, such as water level sensors, flow sensors, temperature sensors, humidity sensors, pressure sensors, vibration sensors, current sensors, voltage sensors, water quality sensors, and speed sensors. This sensor data can be collected in real-time through an industrial IoT platform and stored in the cloud or a local database for further analysis, monitoring, and control of the hydropower plant's operation.

[0024] The sensor data is organized into time series data at the same time intervals.

[0025] Different sensors may collect data at different periods. To synchronize the data, the sensor data is organized into a time-series data stream and sampled and recorded at the same time intervals (e.g., every minute, every hour). During serialization, a timestamp is assigned to each sensor data point, representing the time of data acquisition. This ensures that the data from all sensors are aligned on the same timeline. This is achieved by sampling at each timestamp, ensuring that each sensor records data at the same point in time.

[0026] Determine the sampling time interval, i.e., the time distance between data points. This can be seconds, minutes, hours, etc., depending on the needs of data analysis or monitoring; this embodiment does not impose a specific limitation. It is also necessary to ensure that each time interval has corresponding data points. If data for a certain time point is missing, interpolation, filling with default values, or other processing methods can be selected.

[0027] Organize the data into a time-series data structure, such as a data frame or a database table. Each row represents a time point, and each column represents a sensor measurement parameter. This can be implemented using existing technologies such as pandas' resample method.

[0028] For example, the format of time series data can be as follows: | Timestamp | Temperature (degrees Celsius) | |---------------------|-------------- | | 2024-01-05 00:00:00 | 20.5 | | 2024-01-05 00:01:00 | 21.0 | | 2024-01-05 00:02:00 | 20.8 | The data with similar trends in the time series data are clustered to obtain multiple clustering results.

[0029] Many sensor data points in hydropower plants are strongly correlated with water level, flow rate, and flow velocity. Although the absolute data from each sensor differ, their relative trends are similar when the external environment changes. For example, when the dam water level rises, the water pressure sensor at point A rises from 10 kPa to 20 kPa, while the water pressure sensor at point B rises from 20 kPa to 40 kPa. Many sensors generate similar data points. Clustering data with similar trends can identify data with similar changes. If all data trends are similar, it indicates that the data at all locations are normal. If there are outliers, there may be equipment malfunctions or leaks at that location, requiring further investigation. Based on this principle, this implementation further clusters data with similar trends in the time series data, obtaining multiple clustering results.

[0030] When performing clustering, clustering algorithms can be used to group time series with similar trends together. First, determine the method for measuring the similarity between time series, such as Euclidean distance, Manhattan distance, or Dynamic Time Warping (DTW). Then, select an appropriate clustering algorithm, such as K-means clustering, hierarchical clustering, or DBSCAN. Using the selected clustering algorithm, the time series data are divided into different clusters. The goal of clustering is to ensure that time series within the same cluster have similar trends. Any existing clustering method can be used; this embodiment does not impose specific limitations. Preferably, the tslearn library in Python is used for DTW clustering.

[0031] For each sensor data point, based on historical records, each sensor data point is divided into N levels.

[0032] Since normal sensor data is within a dynamic range, it is difficult to represent dynamic characteristics if the raw data is used directly. Therefore, the sensor data is classified into levels. In subsequent processing, most data are in the same level, the overall data performance is more linear, and it is easier to observe abnormal data.

[0033] Collect historical data from the sensors before classifying them to ensure that there is enough data to calculate quantiles.

[0034] Each sensor data point is divided into N levels, using quantiles or percentiles, methods commonly used in existing technologies. The levels are defined as 1, 2, 3, or more sub-levels. Each sensor data point is then mapped to its corresponding level. This can be achieved by comparing the data with quantiles.

[0035] N is a positive integer not less than 3. The specific choice of level is not limited in this embodiment. Experiments can be conducted based on field data to select a reasonable level.

[0036] Furthermore, in order to make the levels more representative of the data level, this embodiment preferably uses the following method for level classification: For each sensor data point, the average and standard deviation of the historical data are calculated. When the data deviates from the average by one standard deviation, it is classified as level 0; when the data deviates from the average by between one and two standard deviations, it is classified as level 1; when the data deviates from the average by between one and two standard deviations, it is classified as level -1; when the data deviates from the average by more than two standard deviations, it is classified as level 2; and when the data deviates from the average by less than two standard deviations, it is classified as level -2.

[0037] Real-time acquisition of industrial IoT sensor data, mapping the real-time data to the corresponding sensor classification levels.

[0038] When acquiring real-time industrial IoT sensor data and mapping it to corresponding levels, a method similar to that described in the previous step can be used. For each sensor's real-time data, the same level definition method is used to calculate the level to which the real-time data belongs. The real-time data is then mapped to the corresponding level. For example, when a data point is received, the level classification method described in the previous steps is used to determine whether the current data is level 0, level 1, or another level.

[0039] For each sensor's data, take the average level over the last M cycles.

[0040] The period is determined based on the data acquisition frequency. For each sensor, the level is calculated according to the aforementioned method within each period. The average level data of each sensor over the last M periods is taken. Since the data may fluctuate in the short term, using the average level over M periods can further smooth the data and reduce false alarms, where M is a positive integer not less than 2. The larger the value of M, the better the smoothing performance, but the slower the response to anomalies. When implementing this embodiment, the experimental selection can be carried out according to the specific data conditions. This embodiment does not impose specific limitations.

[0041] The first interface displays a first coordinate axis. The vertical axis of the first coordinate axis represents the level, and the horizontal axis represents different sensors. For each sensor, the average level of its last M cycles is represented on the first coordinate axis to obtain the level point.

[0042] like Figure 1 As shown, the first interface displays the first coordinate axis, and the level data of a sensor is marked on the coordinate axis at preset intervals. Each point represents a sensor, and the value of each point is the average level of the sensor over the past M cycles. Figure 1 The example shows eight different sensors, all of which had a rank of 1 for the last five cycles.

[0043] For the same clustering result, connect the level points of all sensors in the clustering result to obtain the clustering level line.

[0044] For the same clustering result, since all points show similar trends, connecting all sensor level points within the same cluster would normally result in a straight line. For example... Figure 2 As shown, there are two non-clustering rank lines, with the two clustering rank lines at the current ranks of 1 and 0, respectively.

[0045] The clustering level lines are translated along the vertical axis of the first coordinate axis so that all clustering level lines do not intersect.

[0046] Because the number of ranks is finite, while the number of clusters can be very large, it is necessary to further disperse the different clusters in order to distinguish between different rank lines. If Figure 2 If there are two other clusters where the rank lines are both 0, then only [the following information] can be displayed. Figure 2 The pattern is achieved by translating the clustering rank lines along the vertical axis of the first coordinate axis, ensuring that all clustering rank lines do not intersect. For example... Figure 3 As shown, the intersecting grade lines are translated on the vertical axis to obtain four non-intersecting lines, each representing a cluster. In this embodiment, subsequent steps only concern whether the lines are straight, not the specific data. Therefore, the specific result after translation (the position of the lines) does not need to be specifically specified, as long as the different grade lines can be distinguished.

[0047] For each clustering level line, when there is a peak on the clustering level line that is higher than a first threshold, a first bubble is displayed on that peak.

[0048] In this embodiment, since the values ​​of the same cluster change with the same trend, although the specific absolute data are different, the levels are similar after grading. This is represented on the graph as each cluster grading line being a straight line. If there is a prominent peak on a certain straight line, it indicates that the data at that point deviates from the group and may belong to problematic data. Therefore, for each cluster grading line, when there is a peak on the cluster grading line that is higher than a first threshold, a first bubble is displayed on that peak to indicate the data.

[0049] like Figure 4 As shown, there is a prominent peak on the third level line, which is different from the overall trend (i.e., not on a straight line). Therefore, the data is likely to be abnormal, and a bubble chart is used to indicate the abnormality on this peak.

[0050] The status of all sensors can be visually observed through the grade lines. There can be multiple grade lines, with each point at a certain interval representing a sensor. Therefore, this embodiment allows for the visual monitoring of the overall status of tens of thousands of sensors on the same screen. Maintenance personnel can judge the overall status of the hydropower plant based on the number and distribution of spikes, which is helpful for overall monitoring.

[0051] When identifying spikes, the data for the same cluster is generally a straight line. Therefore, the average value of the cluster can be calculated. For each point in the same cluster, the distance from the average value is calculated, and it is determined whether it exceeds the first threshold. If it exceeds the first threshold, it indicates that the spike is too high, and an abnormality is indicated.

[0052] Clicking the first bubble will display detailed data for the corresponding sensor on the second screen.

[0053] Furthermore, to facilitate maintenance personnel in viewing specific abnormal data, when a maintenance personnel clicks on the first bubble on the interface, detailed data of the sensor, such as sensor location and recent sensor data, is displayed on a second interface different from the first interface for maintenance personnel's reference. Furthermore, the second interface also displays links to data information of other sensors in the same cluster as the abnormal sensor; when the user clicks a link, the corresponding sensor data details are displayed.

[0054] By using the above method, sensor data can be clustered according to the same trend, and outlier data can be flagged as abnormal. This avoids the need for users to set alarm thresholds for each sensor, and can greatly reduce the setup work when users are dealing with tens of thousands of sensors.

[0055] Furthermore, by hierarchically classifying the data and displaying it as a straight line on a coordinate axis, the overall situation of tens of thousands of sensors can be displayed graphically on the same interface, making it convenient for maintenance personnel to observe the overall situation of the hydropower plant. At the same time, anomalies are displayed in bubbles, improving the ease of operation.

[0056] On the other hand, the present invention also provides a hydropower plant information processing system based on the Industrial Internet of Things, the system comprising the following modules: The data collection module is used to collect industrial IoT sensor data during normal operation of the hydropower plant; The data processing module is used to organize the sensor data into time series data at the same time intervals; cluster data with similar trends in the time series data to obtain multiple clustering results; and for each sensor data, divide each sensor data into N levels according to historical records, where N is a positive integer not less than 3. The real-time acquisition module is used to acquire industrial IoT sensor data in real time and map the real-time data to the corresponding sensor classification level. The analysis module is used to take the average level of the data from each sensor over the last M periods, where M is a positive integer not less than 2; The display module is used to display a first coordinate axis on the first interface. The vertical axis of the first coordinate axis represents the level, and the horizontal axis represents different sensors. For each sensor, the average level of its last M cycles is represented on the first coordinate axis to obtain a level point. For the same clustering result, the level points of all sensors in the clustering result are connected to obtain a clustering level line. The clustering level line is translated on the vertical axis of the first coordinate axis so that all clustering level lines do not intersect. For each clustering level line, when there is a peak on the clustering level line that is higher than the first threshold, a first bubble is displayed on the peak; clicking the first bubble displays the detailed data of the corresponding sensor on the second interface.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0058] For any module structures not specifically defined in this invention, the existing technical specifications shall prevail. The existing technical specifications mentioned in the foregoing background and specific embodiments sections are considered part of this invention and are used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual contents of the claims.

Claims

1. A method for processing information in hydropower plants based on the Industrial Internet of Things, characterized in that: Collect industrial IoT sensor data during normal operation of the hydropower plant; The sensor data is organized into time series data at the same time intervals; Clustering data with similar trends in the time series data yields multiple clustering results; For each sensor data point, based on historical records, each sensor data point is divided into N levels, where N is a positive integer not less than 3; Real-time acquisition of industrial IoT sensor data, mapping the real-time data to the corresponding sensor classification level; For the data from each sensor, take the average level over the last M periods, where M is a positive integer not less than 2; The first interface displays a first coordinate axis, where the vertical axis represents the level and the horizontal axis represents different sensors. For each sensor, the average grade over its last M cycles is plotted on the first coordinate axis to obtain the grade point; For the same clustering result, connect the rank points of all sensors in the clustering result to obtain the clustering rank line; The clustering level lines are translated along the vertical axis of the first coordinate axis so that all clustering level lines do not intersect. For each clustering level line, when there is a peak on the clustering level line that is higher than the first threshold, a first bubble is displayed on the peak; Clicking the first bubble will display detailed data for the corresponding sensor on the second screen; The process of classifying each sensor data into N levels based on historical records includes: calculating the average and standard deviation of the historical data; classifying the data as level 0 when it deviates from the average by one standard deviation; class 1 when it deviates from the average by between one and two standard deviations; class -1 when it deviates from the average by between one and two standard deviations; class 2 when it deviates from the average by more than two standard deviations; and class -2 when it deviates from the average by less than two standard deviations. 2.The industrial Internet of Things based hydropower plant information processing method according to claim 1, characterized in that: There are at least two clustering level lines, and all of them are displayed on the first interface. 3.The industrial Internet of Things based hydropower plant information processing method according to claim 1, characterized in that The method for determining whether there are peaks on the clustering level line that are higher than the first threshold is as follows: calculate the current average level of each cluster, calculate the distance of each point in the same cluster from the average level, and determine whether it exceeds the first threshold. If it exceeds the first threshold, it indicates that the peak is too high. 4.The industrial Internet of Things based hydropower plant information processing method according to claim 1, characterized in that: The second interface further displays links to data information of other sensors in the same cluster as the abnormal sensor. When the user clicks a link, the corresponding sensor data details are displayed.

5. An industrial internet of things-based hydropower plant information processing system, characterized by The system includes: The data collection module is used to collect industrial IoT sensor data during normal operation of the hydropower plant; The data processing module is used to organize the sensor data into time series data at the same time intervals; cluster data with similar trends in the time series data to obtain multiple clustering results; and for each sensor data, divide each sensor data into N levels according to historical records, where N is a positive integer not less than 3. The real-time acquisition module is used to acquire industrial IoT sensor data in real time and map the real-time data to the corresponding sensor classification level. The analysis module is used to take the average level of the data from each sensor over the last M periods, where M is a positive integer not less than 2; The display module is used to display a first coordinate axis on the first interface. The vertical axis of the first coordinate axis represents the level, and the horizontal axis represents different sensors. For each sensor, the average level of its last M cycles is represented on the first coordinate axis to obtain a level point. For the same clustering result, the level points of all sensors in the clustering result are connected to obtain a clustering level line. The clustering level line is translated on the vertical axis of the first coordinate axis so that all clustering level lines do not intersect. For each clustering level line, when there is a peak on the clustering level line that is higher than the first threshold, a first bubble is displayed on the peak; clicking the first bubble displays the detailed data of the corresponding sensor on the second interface; The process of classifying each sensor data into N levels based on historical records includes: calculating the average and standard deviation of the historical data; classifying the data as level 0 when it deviates from the average by one standard deviation; class 1 when it deviates from the average by between one and two standard deviations; class -1 when it deviates from the average by between one and two standard deviations; class 2 when it deviates from the average by more than two standard deviations; and class -2 when it deviates from the average by less than two standard deviations.

6. The industrial internet of things based hydropower plant information processing system of claim 5, wherein: There are at least two clustering level lines, and all of them are displayed on the first interface. 7.The industrial Internet of Things based hydropower plant information processing system of claim 5, wherein The method for determining whether there are peaks on the clustering level line that are higher than the first threshold is as follows: calculate the current average level of each cluster, calculate the distance of each point in the same cluster from the average level, and determine whether it exceeds the first threshold. If it exceeds the first threshold, it indicates that the peak is too high.

8. The hydropower plant information processing system based on the Industrial Internet of Things as described in claim 5, characterized in that: The second interface further displays links to data information of other sensors in the same cluster as the abnormal sensor. When the user clicks a link, the corresponding sensor data details are displayed.

Images