Hydropower station oil level early warning method based on image recognition

Abstract

A hydropower station oil level early warning method based on image recognition comprises the steps that an oleophobic material is adopted in an oil tank to form scale values to display the real oil level, an intelligent camera is installed in an upper cover plate, the oil level is monitored in real time, and an actual oil level picture is transmitted regularly; forming an initial sample set, and forming a learning data set and a training algorithm model; the intelligent camera is set to transmit video pictures to the Faster R-CNN algorithm model at regular time, and the Faster R-CNN algorithm model marks and outputs the oil level of the pictures; signal transmission and judgment; the oil level of turbine oil in a hydropower station is monitored in real time, abnormal alarm signals of equipment such as an oil level sensor and an oil pump motor and alarm signals of higher and lower limit values of the oil level are reported in time, the operation reliability of the equipment is improved, and the safety production risk and the environmental protection problem caused by the fact that the equipment such as the sensor is not monitored in place are completely eradicated.

Description

Technical Field

[0001] The invention belongs to the technical field of oil level early warning for hydropower stations, and relates to an oil level early warning method for hydropower stations based on image recognition. Background Art

[0002] The hydraulic control system in the hydropower station uses oil as the working medium to operate the guide vane relay, electromagnetic control valve and other equipment components. The hydropower station is equipped with a foundation pit oil leakage tank, a technical supply oil leakage tank and other devices to collect the turbine oil after the equipment is operated, and the oil level is monitored in real time through the oil level sensor. When the collected oil level reaches the limit height, the oil pump control system will start the corresponding oil pump motor and pump the oil back into the oil collection tank for recycling.

[0003] At present, the oil control system of the hydropower station only monitors the oil level through the oil level sensor, and the data is single. Once the oil level sensor is stuck, fails or has other problems during operation, two unsafe phenomena will occur. First, it may cause the oil level in the foundation pit leakage tank, technical supply leakage tank and other devices to continue to rise, but the oil pump motor will not start, and the oil will eventually overflow and leak, causing unit operation safety problems and environmentally unfriendly incidents; second, it may cause the oil pump motor to run for a long time, resulting in burning the pump and damaging the equipment. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an oil level early warning method for a hydropower station based on image recognition. When the oil level monitoring sensor in the hydropower station becomes stuck or fails during operation, a fault signal is reported in time so that the operating personnel can discover the fault in the first time and remotely monitor the actual status of the oil level.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for early warning of oil level in a hydropower station based on image recognition, comprising the following steps: S1, preparatory preparation, using oleophobic material to form a scale value inside the oil tank to display the actual oil level, and installing a smart camera inside the upper cover to monitor the oil level in real time and transmit the actual oil level picture at regular intervals; S2, model training, collect pictures, form an initial sample set, and learn the data set to train the algorithm model; S3, oil level recognition, sets the smart camera to transmit the video image to the Faster R-CNN algorithm model at a fixed time, and the Faster R-CNN algorithm model marks and outputs the oil level in the image; S4, signal transmission and judgment; S4-1, using appropriate signal modules and communication methods to connect with the local control unit LCU and the image monitoring system; S4-2, the output marked value is transmitted to the local control unit LCU, and a logic judgment process is added inside the LCU; S4-3, the annotated pictures are transmitted to the picture acquisition and monitoring system. When the annotated pictures exceed the upper or lower limit, the picture acquisition and monitoring system will alarm, pop up the picture screen and continuously refresh the display, reminding the operating personnel to check in time and handle professionally.

[0006] In S2, the initial sample set is formed by manually controlling the oil level inside the mailbox when the unit is shut down for maintenance, obtaining initial images of different oil level values, and labeling the oil level in each image using LabelImg software. The labeling software calibrates the image oil level scale and annotates the oil level height value, and the labeled images are aggregated to form the initial sample set.

[0007] In S2, a learning data set is formed: the ImgAug software is used to perform random contrast, random rectangular occlusion, and salt and pepper noise processing on the annotated images in the initial sample set to expand the sample set and form a learning data set.

[0008] In S2, the algorithm model is trained: the Faster R-CNN algorithm model is used for recognition, and the formed learning data set is input into the algorithm model to train the positioning and recognition capabilities of the algorithm model.

[0009] In S4-2, the logic determination process includes the following steps: S4-2-1, compare the output oil level value with the value collected by the sensor. If the difference is within a certain range, it is normal. If it exceeds a certain range, the sensor is judged to be faulty and an alarm signal is sent to the monitoring system. S4-2-2, the output oil level value is lower than the too low limit, the oil level is judged to be too low limit fault, and an alarm signal is sent to the monitoring system; S4-2-3, after the output oil level value is lower than the too low limit, the difference between the initial too low oil level value and the oil level value after 10 minutes exceeds +10mm, it is determined that the oil level too low limit fault persists and there may be a risk of long-term operation of the motor, and an alarm signal is sent to the monitoring system; S4-2-4, the output oil level mark value is higher than the high limit, the oil level is judged to be high limit fault, and an alarm signal is sent to the monitoring system; S4-2-5, after the output oil level mark value is higher than the too low limit, the difference between the initial too high oil level value and the oil level value after 5 minutes exceeds -10mm, it is determined that the oil level too high limit fault persists and there may be a risk of oil level overflow, and an alarm signal is sent to the monitoring system.

[0010] The main beneficial effects of the present invention are: It effectively avoids oil tank overflow incidents such as oil pump motor damage and pit oil tank leakage and technical supply oil tank leakage when the sensor fails.

[0011] Transmit oil level in real time to improve the reliability of oil system monitoring.

[0012] Abnormal oil level signals are reported in a timely manner, facilitating timely processing by operating and maintenance personnel.

[0013] The invention can be extended to similar auxiliary equipment control systems in other power plants or industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0015] Figure 1 This is a diagram of the oil control system of the present invention.

[0016] Figure 2 This is the oil pump motor starting logic diagram of the present invention.

[0017] Figure 3 This is the oil pump motor stop logic diagram of the present invention.

[0018] Figure 4 This is a diagram of the model training process of the present invention.

[0019] Figure 5 This is the oil level fault judgment logic diagram. DETAILED DESCRIPTION

[0020] like Figure 1~Figure 5 In the present invention, a method for early warning of oil level in a hydropower station based on image recognition comprises the following steps: S1, preparatory preparation, using oleophobic material to form a scale value inside the oil tank to display the actual oil level, and installing a smart camera inside the upper cover to monitor the oil level in real time and transmit the actual oil level picture at regular intervals; S2, model training, collect pictures, form an initial sample set, and learn the data set to train the algorithm model; S3, oil level recognition, sets the smart camera to transmit the video image to the Faster R-CNN algorithm model at a fixed time, and the Faster R-CNN algorithm model marks and outputs the oil level in the image; S4, signal transmission and judgment.

[0021] Embodiment 1, like Figure 2 , Figure 3 In the process, the turbine oil after the operation of the guide vane relay, electromagnetic control valve and other equipment components of the hydropower station flows into the oil collection device a, which includes the foundation pit leakage tank, technical supply leakage tank and oil collecting tank.

[0022] An oil level sensor b is installed on the side of the oil collection device to transmit real-time oil level analog signals and switch signals of oil level being too low, too high, or too low to the oil control system c.

[0023] The PLC in the oil control system controls the start and stop of the oil pump according to the sensor signal, pumps the oil into the oil sump for recirculation, and at the same time sends the oil level and oil pump status to the monitoring system.

[0024] Embodiment 2, In order to avoid the oil level sensor failure or stuck at a certain position in Example 1, resulting in no change in the oil level causing oil leakage or pump burning, a hydropower station oil level early warning method based on image recognition and positioning is proposed. Figure 4 ~ Figure 5 includes: 1. Preliminary preparation: Based on the existing oil system of the hydropower station, oleophobic material is used inside the oil tank to form a scale value to display the actual oil level, and a smart camera is installed inside the upper cover to monitor the oil level in real time and transmit the actual oil level picture at regular intervals.

[0025] 2. Model training: 2.1 Forming the initial sample set: When the unit is shut down for maintenance, the oil level inside the mailbox is manually controlled to obtain initial images with different oil level values. The oil level in each image is labeled using LabelImg software, which can calibrate the image oil level scale and annotate the oil level height value. The labeled images are summarized to form the initial sample set.

[0026] 2.2 Forming a learning data set: ImgAug software is used to perform random contrast, random rectangular occlusion, salt and pepper noise processing on the annotated images in the initial sample set to expand the sample set and form a learning data set.

[0027] 2.3 Training algorithm model: Among the many algorithm models currently available, the Fast R-CNN model has a relatively faster training speed and a relatively better recognition and positioning effect, so this method uses the Faster R-CNN algorithm model for recognition. The formed learning data set is input into the algorithm model to train the positioning and recognition capabilities of the algorithm model.

[0028] 3. Oil level identification: The smart camera is set to transmit video images to the Faster R-CNN algorithm model at a fixed time. The Faster R-CNN algorithm model marks the oil level in the image and outputs it.

[0029] 4. Signal transmission and judgment: 4.1 Use appropriate signal modules and communication methods to connect with the local control unit LCU and image monitoring system.

[0030] 4.2 The output marked value is transmitted to the local control unit LCU, and a logic judgment process is added inside the LCU: a. Compare the output oil level value with the value collected by the sensor. If the difference is within a certain range, it is normal. If it exceeds a certain range, the sensor is judged to be faulty and an alarm signal is sent to the monitoring system; b. If the output oil level value is lower than the too low limit, the oil level is judged to be too low limit fault, and an alarm signal is sent to the monitoring system; c. After the output oil level value is lower than the too low limit, the difference between the initial too low oil level value and the oil level value after 10 minutes exceeds +10mm, and it is determined that the oil level too low limit fault persists and there may be a risk of long-term operation of the motor, and an alarm signal is sent to the monitoring system; d. If the output oil level value is higher than the high limit, the oil level is judged to be high limit fault, and an alarm signal is sent to the monitoring system; e. After the output oil level value is higher than the too low limit, the difference between the initial too high oil level value and the oil level value after 5 minutes exceeds -10mm, and it is determined that the oil level too high limit fault persists and there may be a risk of oil level overflow, and an alarm signal is sent to the monitoring system; 4.3 The output annotated images are sent to the image acquisition and monitoring system. When the annotated images exceed the upper or lower limit, the image acquisition and monitoring system will alarm, pop up the image screen and continuously refresh the display to remind the operator to check in time for professional processing.

[0031] The above method effectively avoids damage to the oil pump motor and oil overflow incidents such as oil tank leakage in the pit and technical supply tank when the sensor fails; transmits the oil level height in real time to improve the reliability of oil system monitoring; and reports abnormal position signals in a timely manner, which is convenient for operation and maintenance personnel to deal with in time and is easy to promote.

[0032] The above-mentioned embodiments are only preferred technical solutions of the present invention and should not be regarded as limitations of the present invention. The embodiments and features in the embodiments of the present application can be arbitrarily combined with each other without conflict. The protection scope of the present invention shall be the technical solutions recorded in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for early warning of oil level in a hydropower station based on image recognition, characterized in that: Its characteristics are: The steps include: S1, preparatory preparation, using oleophobic material to form a scale value inside the oil tank to display the actual oil level, and installing a smart camera inside the upper cover to monitor the oil level in real time and transmit the actual oil level picture at regular intervals; S2, model training, collect pictures, form an initial sample set, and learn the data set to train the algorithm model; S3, oil level recognition, sets the smart camera to transmit the video image to the Faster R-CNN algorithm model at a fixed time, and the Faster R-CNN algorithm model marks the oil level in the image and outputs it; S4, signal transmission and judgment; S4-1, using appropriate signal modules and communication methods to connect with the local control unit LCU and the image monitoring system; S4-2, the output marked value is transmitted to the local control unit LCU, and a logic judgment process is added inside the LCU; S4-3, the annotated pictures are transmitted to the picture acquisition and monitoring system. When the annotated pictures exceed the upper or lower limit, the picture acquisition and monitoring system will alarm, pop up the picture screen and continuously refresh the display, reminding the operating personnel to check in time and handle professionally.

2. The oil level early warning method for a hydropower station based on image recognition according to claim 1 is characterized in that: In S2, the initial sample set is formed by manually controlling the oil level inside the mailbox when the unit is shut down for maintenance, obtaining initial images of different oil level values, and labeling the oil level in each image using LabelImg software. The labeling software calibrates the image oil level scale and annotates the oil level height value, and the labeled images are aggregated to form the initial sample set.

3. The oil level early warning method for a hydropower station based on image recognition according to claim 1 is characterized in that: In S2, a learning data set is formed: the ImgAug software is used to perform random contrast, random rectangular occlusion, and salt and pepper noise processing on the annotated images in the initial sample set to expand the sample set and form a learning data set.

4. The oil level early warning method for a hydropower station based on image recognition according to claim 1 is characterized in that: In S2, the algorithm model is trained: the Faster R-CNN algorithm model is used for recognition, and the formed learning data set is input into the algorithm model to train the positioning and recognition capabilities of the algorithm model.

5. The oil level early warning method for a hydropower station based on image recognition according to claim 1 is characterized in that: In S4-2, the logic determination process includes the following steps: S4-2-1, compare the output oil level value with the value collected by the sensor. If the difference is within a certain range, it is normal. If it exceeds a certain range, the sensor is judged to be faulty and an alarm signal is sent to the monitoring system. S4-2-2, the output oil level value is lower than the too low limit, the oil level is judged to be too low limit fault, and an alarm signal is sent to the monitoring system; S4-2-3, after the output oil level value is lower than the too low limit, the difference between the initial too low oil level value and the oil level value after 10 minutes exceeds +10mm, it is determined that the oil level too low limit fault persists and there may be a risk of long-term operation of the motor, and an alarm signal is sent to the monitoring system; S4-2-4, the output oil level mark value is higher than the high limit, the oil level is judged to be high limit fault, and an alarm signal is sent to the monitoring system; S4-2-5, after the output oil level mark value is higher than the too low limit, the difference between the initial too high oil level value and the oil level value after 5 minutes exceeds -10mm, it is determined that the oil level too high limit fault persists and there may be a risk of oil level overflow, and an alarm signal is sent to the monitoring system.

Images