Method and system for monitoring oil quality of phase modifier lubricating oil system
Through the combination of BP neural network and the generative adversarial network GAN, the microwater content and particle number of the lubricant oil system are monitored in real time, solving the real-time monitoring problem of the lubricant oil system in the dual-water cooling camera, avoiding lubricant deterioration and equipment wear, and improving system reliability.
Patent Information
- Application Number
- CN202510589686.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2025-08-08
AI Technical Summary
In the prior art, the lubricant oil system of the dual-water cooling camera lacks real-time monitoring, resulting in deterioration of the lubricant oil, which in turn causes wear and scratches on the various journals and bearing shells of the camera.
Using machine learning models, especially BP neural networks, combined with the generation adversarial network GAN, the microwater content and particle number of lubricating oil are monitored in real time through capacitive humidity sensors, automatic particle counters and other equipment, and the mapping relationship between lubricating oil detection index data and oil quality monitoring results is established.
Real-time monitoring of lubricating oil quality is achieved, which avoids lubricating oil deterioration, reduces wear and scratches of journals and bearing shells, and improves the reliability of the dual-water cooling camera.
Smart Images
Figure CN120446220A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil quality monitoring for phase condenser, and in particular to an oil quality monitoring method and system for a lubricating oil system of a phase condenser. Background Art
[0002] Excessive water content in the lubricating oil system of a dual-water internally cooled condenser is primarily due to cooling water leaks from the lubricating oil cooler; improper configuration of the main oil tank cover; and air condensation, which can lead to water ingress into the oil. The presence of water in the lubricating oil system of a dual-water internally cooled condenser accelerates oil degradation and emulsification. Furthermore, water reacts with oil additives, causing them to decompose and leading to rust in the lubricating oil system. Rust products enter the oil system as hard particles, increasing particulate contamination, thinning the lubricating oil film, and potentially accelerating wear and scratching of the journals and bearings. Currently, water quality testing of the lubricating oil system of a dual-water internally cooled condenser in a converter station typically involves monthly oil sampling and testing, lacking real-time monitoring of lubricating oil system quality. Summary of the Invention
[0003] Technical problem to be solved by the present invention: In view of the above-mentioned problems of the prior art, a method and system for monitoring the oil quality of the lubricating oil system of a phase shifter are provided. The present invention aims to realize real-time monitoring of the lubricating oil quality of the lubricating oil system. There is no need to send the lubricating oil quality for inspection. The inspection is accurate and real-time, which can avoid the deterioration of the lubricating oil due to the inability to monitor the oil quality in time, and the wear and scratches of the various journals and bearings of the phase shifter due to the failure to timely handle the oil quality.
[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for monitoring the oil quality of a phase shifter lubricating oil system comprises the following steps: obtaining detection index data of the lubricating oil in the phase shifter lubricating oil system, including part or all of trace water content, temperature, and particle count; inputting the detection index data of the lubricating oil into a pre-trained machine learning model to obtain an oil quality monitoring result of the lubricating oil in the phase shifter lubricating oil system, wherein the machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring result, wherein the oil quality monitoring result includes part or all of the oil quality grade of the lubricating oil and whether it has deteriorated.
[0005] Optionally, the trace water content is detected by a capacitive humidity sensor, the particle number is detected by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer; and the temperature is detected by a temperature sensor.
[0006] Optionally, the result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil, and after obtaining the relative humidity of the lubricating oil, the method further includes performing temperature compensation on the relative humidity of the lubricating oil according to the following formula to obtain the trace water content of the lubricating oil: , in, The water content is trace, is the relative humidity of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient.
[0007] Optionally, when detecting the number of particles by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer, the detection window of the lubricating oil is divided into multiple sub-windows along the direction of gravity and the overlapping coefficients are set for each sub-window, and the overlapping coefficient of the sub-window closer to the bottom is larger. The number of particles in each sub-window is detected respectively, and the overlapping coefficient is used as a weight coefficient to perform weighted summation on the number of particles in each sub-window to obtain the number of particles in the entire detection window.
[0008] Optionally, the machine learning model is a BP neural network, which consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, the hidden layer, and the output layer are respectively 、 and , where the input vector of the input layer is The function expression is: , in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: , in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: , in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold value of each neuron node; the training process of the BP neural network includes: obtaining a detection index data sample of the lubricating oil in the lubricating oil system of the phase shifter; adding the detection index data sample based on the generative adversarial network (GAN); the generative adversarial network (GAN) is composed of a generator and a discriminator, the generator is used to generate the detection index data sample through continuous iteration through random noise, the discriminator is responsible for identifying whether the generated detection index data sample is derived from real data or generated data and giving a corresponding probability score, and the generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data sample generated by the generator is close to the distribution of real data and serves as the added detection index data sample: , , in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the detection index data samples are labeled, divided into training sets and test sets, the BP neural network is trained using the training sets and test sets, and the model parameters in the BP neural network are obtained using the gradient descent method 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results; in the training process of the BP neural network, the gradient descent method is used to find the model parameters in the BP neural network 、 、 、 The function expression of the adjustment amount is: , , , , in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of a neuron node.
[0009] In addition, the present invention also provides an oil quality monitoring system for a phase shifter lubricating oil system, comprising: A data acquisition program unit is used to obtain detection index data of lubricating oil in the phase shifter lubricating oil system, including part or all of trace water content, temperature and particle number; The result prediction program unit is used to input the detection index data of the lubricating oil into a pre-trained machine learning model to obtain the oil quality monitoring results of the lubricating oil in the phase shifter lubricating oil system. The machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results. The oil quality monitoring results include the oil quality grade of the lubricating oil and part or all of whether it has deteriorated.
[0010] Optionally, the trace water content is detected by a capacitive humidity sensor, the particle number is detected by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer; and the temperature is detected by a temperature sensor.
[0011] Optionally, the result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil, and after obtaining the relative humidity of the lubricating oil, the method further includes performing temperature compensation on the relative humidity of the lubricating oil according to the following formula to obtain the trace water content of the lubricating oil: , in, The water content is trace, is the relative humidity of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient.
[0012] Optionally, when detecting the number of particles by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer, the detection window of the lubricating oil is divided into multiple sub-windows along the direction of gravity and the overlapping coefficients are set for each sub-window, and the overlapping coefficient of the sub-window closer to the bottom is larger. The number of particles in each sub-window is detected respectively, and the overlapping coefficient is used as a weight coefficient to perform weighted summation on the number of particles in each sub-window to obtain the number of particles in the entire detection window.
[0013] Optionally, the machine learning model is a BP neural network, which consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, the hidden layer, and the output layer are respectively 、 and , where the input vector of the input layer is The function expression is: , in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: , in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: , in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold value of each neuron node; the training process of the BP neural network includes: obtaining a detection index data sample of the lubricating oil in the lubricating oil system of the phase shifter; adding the detection index data sample based on the generative adversarial network (GAN); the generative adversarial network (GAN) is composed of a generator and a discriminator, the generator is used to generate the detection index data sample through continuous iteration through random noise, the discriminator is responsible for identifying whether the generated detection index data sample is derived from real data or generated data and giving a corresponding probability score, and the generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data sample generated by the generator is close to the distribution of real data and serves as the added detection index data sample: , , in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the detection index data samples are labeled, divided into training sets and test sets, the BP neural network is trained using the training sets and test sets, and the model parameters in the BP neural network are obtained using the gradient descent method 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results; in the training process of the BP neural network, the gradient descent method is used to find the model parameters in the BP neural network 、 、 、 The function expression of the adjustment amount is: , , , , in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of a neuron node.
[0014] In addition, the present invention also provides an oil quality monitoring system for a phase shifter lubricating oil system, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the oil quality monitoring method for the phase shifter lubricating oil system.
[0015] In addition, the present invention also provides a computer-readable storage medium, which stores a computer program or instruction. The computer program or instruction is programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system through a processor.
[0016] In addition, the present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system through a processor.
[0017] Compared with the prior art, the present invention can mainly achieve the following beneficial effects: the method of the present invention includes obtaining detection index data of lubricating oil in the lubricating oil system of the phase shifter, and using a pre-trained machine learning model to obtain the oil quality monitoring result of the lubricating oil in the lubricating oil system of the phase shifter. The machine learning model is pre-trained to establish a mapping relationship between the detection index data of lubricating oil and the oil quality monitoring result. The present invention can realize real-time monitoring of the lubricating oil quality of the lubricating oil system, without the need for lubricating oil quality inspection, and the detection is accurate and real-time. It can avoid the deterioration of lubricating oil due to the inability to monitor oil quality in time, and the wear and scratches of various journals and bearings of the phase shifter caused by the failure to timely handle the oil quality, solves the problem that the micro-water content of the lubricating oil system of the double-water internally cooled phase shifter in the converter station cannot be monitored in real time, reduces the wear and scratches of the journals and bearings, and improves the reliability of the double-water internally cooled phase shifter. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 Schematic diagram of the basic process of the method of the embodiment of the present invention.
[0019] Figure 2 Detailed flowchart of the method according to the embodiment of the present invention. DETAILED DESCRIPTION
[0020] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments of the present invention.
[0021] like Figure 1 As shown, the oil quality monitoring method of the phase shifter lubricating oil system of this embodiment includes the following steps: obtaining detection index data of the lubricating oil in the phase shifter lubricating oil system, including part or all of the trace water content, temperature and particle number; inputting the detection index data of the lubricating oil into a pre-trained machine learning model to obtain the oil quality monitoring results of the lubricating oil in the phase shifter lubricating oil system, the machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results, the oil quality monitoring results including part or all of the oil quality grade of the lubricating oil and whether it has deteriorated.
[0022] When obtaining the detection index data of the lubricating oil in the lubricating oil system of the phase shifter, it is necessary to obtain the lubricating oil sample in the lubricating oil system of the phase shifter. The lubricating oil sample can be obtained through the lubricating oil pipeline sampling valve of the lubricating oil system.
[0023] like Figure 2 As shown, in this embodiment, the trace water content is detected by a capacitive humidity sensor. The humidity sensing mechanism of the capacitive humidity sensor is based on the fact that when the capacitive humidity sensor adsorbs gaseous water molecules in the environment, its dielectric constant also changes accordingly. The relationship between the capacitance and the relative pressure of water vapor in the environment (P / P0) can be expressed by the following formula: , , in, is the capacitance, is the dielectric constant of vacuum; Relative humidity Dielectric constant of polymer humidity-sensitive material at %RH; is the effective electrode area of the capacitive sensor; It is the thickness of the polymer wet-sensitive film. The dielectric constant at 0% RH; is the structural constant; for The mass of water molecules adsorbed per unit mass of the polymer at %RH; is the dielectric constant of adsorbed water. Polyimide polymer material as a humidity-sensitive film has higher sensitivity and a wider measurement range. When the relative humidity is U%RH, the mass of water molecules adsorbed by the polyimide polymer film humidity-sensitive capacitor is as follows: , in, is the structural constant, The relative pressure of water vapor in the oil environment; The mass of water molecules adsorbed by the polymer film. The number of particles is detected by an automatic particle counter, a dynamic image particle sensor, a capacitive sensor or a laser particle size analyzer, and the temperature is detected by a temperature sensor. In this embodiment, the result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil, and the monitored relative humidity analog quantity is converted into a digital signal by a transmitter. The water content in the oil of the main oil tank of the lubricating oil container is closely related to the chemical composition of the oil, the temperature, and the degree of aging of the oil. Different types of oils have different amounts of dissolved water at different temperatures (i.e., the solubility of water in the oil). When the temperature rises, the water content in the oil increases; when the temperature drops, the water dissolved in the oil will separate due to supersaturation and settle to the bottom of the main oil tank. Relative humidity of the lubricating oil To directly measure the moisture content of the oil, it is proportional to the water saturation of the individual oil products. The physical explanation of the relative humidity (water activity) of lubricating oil is the ratio of the pressure of the aqueous solution to the maximum water vapor pressure (pure water) at that temperature. The scenario in this embodiment is to measure the free water content in the lubricating oil sample, which is defined as the ratio of the actual value of the water content in the oil to the saturated water content in the oil at that temperature. The focus of the relative humidity (water activity) of lubricating oil is non-chemically bound water, and its measurement is based on the equilibrium water content. Therefore, the relative humidity (water activity) of lubricating oil is not expressed as a relative humidity in the range of 0 to 100%, but is expressed as 0 to 1. =0 represents anhydrous substances, =1 represents pure water. The water content of oil refers to the percentage of water in the total mass of the oil. The absolute water content is used in oil measurement. The value (ppm) indicates the average mass concentration of water in the oil. As the water content increases, the water activity also increases accordingly. The relationship between water content and water activity at a given temperature is called a hygroscopic isotherm, and these curves are determined experimentally. Experimental findings show that the relationship between water content and water activity is linear before the saturation point of water solubility in the oil and exponentially correlates with temperature. Therefore, after obtaining the relative humidity of the lubricating oil, this embodiment also includes temperature compensation of the relative humidity of the lubricating oil according to the following formula to determine the trace water content of the lubricating oil: , in, The water content is trace, is the relative humidity (water activity) of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient. For example, in this embodiment, the lubricating oil system of the dual-water internally cooled condenser uses a mineral-based lubricant (L-TSA32 turbine oil). The experiment used Lanlian L-TSA32 turbine oil, and the water content was varied by adding water. The water saturation of Lanlian turbine oil at different temperatures is shown in Table 1 below.
[0024] Table 1 Water saturation of lubricating oil at different temperatures Unit: ppm
[0025] It can be seen from Table 1 that the higher the temperature, the higher the water saturation.
[0026] The presence of water in the lubricating oil system of a dual-water internally cooled condenser accelerates oil degradation and emulsification. Furthermore, water reacts with oil additives, causing them to decompose and leading to corrosion in the lubricating oil system. When these corrosion products enter the lubricating oil as hard particles, they accumulate, resulting in inaccurate particle count detection. Furthermore, gravity causes variations in the distribution of hard particles at different heights, particularly during testing when the lubricating oil is stationary. In this case, the degree of hard particle accumulation varies at different heights. Therefore, to improve particle count accuracy, in this embodiment, particle count detection using an automatic particle counter, a dynamic image particle sensor, a capacitive sensor, or a laser particle size analyzer involves dividing the lubricating oil detection window into multiple subwindows along the direction of gravity and assigning overlapping coefficients to each subwindow. Lower subwindows have larger overlapping coefficients, and the particle counts within each subwindow are measured separately. The particle counts within each subwindow are then weighted and summed using the overlapping coefficients as weights to obtain the particle count for the entire detection window. The overlap coefficient can be obtained through experiments. For example, a certain number of hard particles are added to the lubricating oil in a container with a self-circulating function, and the number of particles is detected by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer. The overlap coefficient corresponding to the number of particles detected can be obtained by dividing the number of added particles by the number of particles detected. By adjusting the number of added hard particles, the overlap coefficient corresponding to different numbers of particles detected can be obtained. When setting the overlap coefficient, the corresponding overlap coefficient can be obtained by looking up the table according to the number of particles detected in each sub-window.
[0027] The key to the machine learning model is to use its learnable characteristics to learn the mapping relationship between the lubricant detection index data and the oil quality monitoring results. Therefore, the required machine learning model can be adopted as needed. For example, as an optional implementation, the machine learning model in this embodiment is a BP neural network. The BP (Back Propagation) neural network is a multi-layer feedforward neural network. The BP neural network consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, hidden layer, and output layer are respectively 、 and , where the input vector of the input layer is The function expression is: , in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: , in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: , in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold of the neuron node, the output vector can be expressed as: , in, ~ are the outputs of the 1st to pth neuron nodes in the output layer respectively. As an optional implementation, the activation function in this embodiment is a nonlinear Sigmoid function, which is generally used .
[0028] According to the actual output With the desired output mode , calculate the generalized error of each unit in the output layer : The goal of network learning is to make the error function Minimum, error function is defined as follows: . BP neural network can adjust the weights and thresholds of the network through the back propagation algorithm to minimize the error between the predicted value and the true value. Figure 2 , the training process of the BP neural network in this embodiment includes: Step 1: Obtain a sample of test index data of lubricating oil in a phase shifter lubricating oil system; Step 2: Add detection index data samples based on the generative adversarial network (GAN); the generative adversarial network (GAN) consists of a generator and a discriminator. The generator is used to generate detection index data samples through continuous iteration through random noise. The discriminator is responsible for identifying whether the generated detection index data samples are derived from real data or generated data and giving corresponding probability scores. The generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data samples generated by the generator are close to the distribution of real data and serve as additional detection index data samples: , , in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the generator and the discriminator are trained adversarially using the loss function shown in the following formula, specifically minimizing the loss function of the generator and maximizing the loss function of the discriminator, which can be expressed as: , in, is the objective function, Represents the real data distribution Samples on In other words, it is a weighted average of all true data points, with the weights determined by the data distribution. Represents the distribution of the latent space The expected value of the sample z on the latent space is the input space used by the generator to generate data; the generator G (Generator) continuously learns the probability distribution of the real data in the training set, with the goal of converting the input random noise data into data that can be mistaken for real data. The discriminator D (Discriminator) determines whether the data is real data, with the goal of distinguishing the generated data generated by the generator G from the real data in the training set. During the training process, the two networks are enhanced simultaneously through mutual competition, and the generator G and the discriminator D constitute a dynamic "game process"; over time, the generator and the discriminator continue to compete, and eventually the two reach a dynamic balance: the data generated by the generator Close to the real data distribution, and the discriminator cannot recognize the data generated by the generating network Finally, the generator can be used to generate micro-water content data for the phase shifter lubricating oil system. Fault data on micro-water content in phase shifter lubricating oil systems is relatively scarce, and data from early failure stages is limited. The Generative Adversarial Network (GAN) simulates the distribution of real data to generate micro-water content data for the lubricating oil system during both failures and normal operation, improving the quality and quantity of data. This enhances the accuracy and robustness of the BP neural network, effectively solving the problem of insufficient fault sample size and enabling accurate identification of micro-water content in phase shifter lubricating oil systems. Step 3: Add label values to the detection index data samples and divide them into training set and test set; Step 4: Use the training set and test set to train the BP neural network, and use the gradient descent method to find the model parameters in the BP neural network. 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results.
[0029] In this embodiment, during the training process of the BP neural network, the model parameters of the BP neural network are obtained using the gradient descent method. 、 、 、 The function expression of the adjustment amount is: , , , , in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of the neuron node. In some cases, the constant coefficient and The variable coefficient method is used to modify the connection weights to speed up the convergence of the network. According to the learned network weights and the input vector to be judged, the output vector corresponding to the input vector is obtained through the network to achieve the purpose of prediction.
[0030] In summary, the present embodiment of the method for monitoring the oil quality of a condenser lubricating oil system involves sampling lubricating oil from the condenser lubricating oil system, monitoring the relative humidity and temperature of the lubricating oil sample, converting the analog relative humidity into a digital signal, calculating the temperature-compensated water content, adding water content data to the condenser lubricating oil system using a generative adversarial network (GAN), constructing and training a BP neural network model, and predicting the water content in the condenser lubricating oil system. The present embodiment of the method for monitoring the oil quality of a condenser lubricating oil system utilizes a polyimide capacitive humidity sensor and temperature compensation to monitor the water content in the lubricating oil system of a dual-water-cooled condenser. By continuously iterating and training the water content data through a generative adversarial network (GAN), the method effectively addresses the issues of insufficient fault sample size and scarcity of fault data. By optimizing parameters to adapt to different input samples, the performance of the BP neural network is improved, addressing the inability to monitor the water content in the lubricating oil system of a dual-water-cooled condenser within a converter station in real time. This reduces wear and scratches on the journal and bearing, and improves the reliability of the dual-water-cooled condenser.
[0031] In addition, this embodiment also provides an oil quality monitoring system for a phase shifter lubricating oil system, comprising: A data acquisition program unit is used to obtain detection index data of lubricating oil in the phase shifter lubricating oil system, including part or all of trace water content, temperature and particle number; The result prediction program unit is used to input the detection index data of the lubricating oil into a pre-trained machine learning model to obtain the oil quality monitoring results of the lubricating oil in the phase shifter lubricating oil system. The machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results. The oil quality monitoring results include the oil quality grade of the lubricating oil and part or all of whether it has deteriorated.
[0032] In this embodiment, the trace water content is detected by a capacitive humidity sensor, the particle number is detected by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer; and the temperature is detected by a temperature sensor.
[0033] In this embodiment, the result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil. After obtaining the relative humidity of the lubricating oil, the relative humidity of the lubricating oil is temperature compensated according to the following formula to obtain the trace water content of the lubricating oil: , in, The water content is trace, is the relative humidity of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient.
[0034] In this embodiment, when detecting the number of particles by using an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer, the detection window of the lubricating oil is divided into multiple sub-windows along the direction of gravity and the overlapping coefficients are set for each sub-window, and the overlapping coefficient of the sub-window closer to the bottom is larger. The number of particles in each sub-window is detected respectively, and the overlapping coefficient is used as a weight coefficient to perform weighted summation on the number of particles in each sub-window to obtain the number of particles in the entire detection window.
[0035] In this embodiment, the machine learning model is a BP neural network, which consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, the hidden layer, and the output layer are respectively 、 and , where the input vector of the input layer is The function expression is: , in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: , in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: , in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold value of each neuron node; the training process of the BP neural network includes: obtaining a detection index data sample of the lubricating oil in the lubricating oil system of the phase shifter; adding the detection index data sample based on the generative adversarial network (GAN); the generative adversarial network (GAN) is composed of a generator and a discriminator, the generator is used to generate the detection index data sample through continuous iteration through random noise, the discriminator is responsible for identifying whether the generated detection index data sample is derived from real data or generated data and giving a corresponding probability score, and the generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data sample generated by the generator is close to the distribution of real data and serves as the added detection index data sample: , , in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the detection index data samples are labeled, divided into training sets and test sets, the BP neural network is trained using the training sets and test sets, and the model parameters in the BP neural network are obtained using the gradient descent method 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results; in the training process of the BP neural network, the gradient descent method is used to find the model parameters in the BP neural network 、 、 、 The function expression of the adjustment amount is: , , , , in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of a neuron node.
[0036] In addition, this embodiment also provides an oil quality monitoring system for a phase shifter lubricating oil system, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the oil quality monitoring method for a phase shifter lubricating oil system.
[0037] In addition, this embodiment also provides a computer-readable storage medium, which stores a computer program or instruction. The computer program or instruction is programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system through a processor.
[0038] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system through a processor.
[0039] Those skilled in the art should understand that the technical solution provided by the present invention may be in the form of a method, a system, or a computer program product. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions described in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.
[0040] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.
Claims
1. A method for monitoring the oil quality of a phase shifter lubricating oil system, characterized in that: The method includes the following steps: obtaining detection index data of lubricating oil in a phase shifter lubricating oil system, including part or all of the trace water content, temperature and particle number; inputting the detection index data of the lubricating oil into a pre-trained machine learning model to obtain oil quality monitoring results of the lubricating oil in the phase shifter lubricating oil system, wherein the machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results, and the oil quality monitoring results include part or all of the oil quality grade of the lubricating oil and whether it has deteriorated.
2. The method for monitoring oil quality of a phase condenser lubricating oil system according to claim 1, characterized in that: The trace water content is detected by a capacitive humidity sensor, the particle number is detected by an automatic particle counter, a dynamic image particle sensor, a capacitive sensor or a laser particle size analyzer; and the temperature is detected by a temperature sensor.
3. The method for monitoring oil quality of a phase condenser lubricating oil system according to claim 2, characterized in that: The result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil. After obtaining the relative humidity of the lubricating oil, the relative humidity of the lubricating oil is temperature compensated according to the following formula to obtain the trace water content of the lubricating oil: ， in, The water content is trace, is the relative humidity of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient.
4. The method for monitoring oil quality of a phase condenser lubricating oil system according to claim 2, characterized in that: When detecting the number of particles by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer, the detection window of the lubricating oil is divided into multiple sub-windows along the direction of gravity and the overlapping coefficients are set for each sub-window, and the overlapping coefficients of the sub-windows closer to the bottom are larger. The number of particles in each sub-window is detected respectively, and the overlapping coefficient is used as a weight coefficient to perform weighted summation on the number of particles in each sub-window to obtain the number of particles in the entire detection window.
5. The method for monitoring oil quality of a phase condenser lubricating oil system according to claim 1, characterized in that: The machine learning model is a BP neural network, which consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, the hidden layer, and the output layer are respectively 、 and , where the input vector of the input layer is The function expression is: ， in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: ， in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: ， in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold value of each neuron node; the training process of the BP neural network includes: obtaining a detection index data sample of the lubricating oil in the lubricating oil system of the phase shifter; adding the detection index data sample based on the generative adversarial network (GAN); the generative adversarial network (GAN) is composed of a generator and a discriminator, the generator is used to generate the detection index data sample through continuous iteration through random noise, the discriminator is responsible for identifying whether the generated detection index data sample is derived from real data or generated data and giving a corresponding probability score, and the generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data sample generated by the generator is close to the distribution of real data and serves as the added detection index data sample: ， ， in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the detection index data samples are labeled, divided into training sets and test sets, the BP neural network is trained using the training sets and test sets, and the model parameters in the BP neural network are obtained using the gradient descent method 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results; in the training process of the BP neural network, the gradient descent method is used to find the model parameters in the BP neural network 、 、 、 The function expression of the adjustment amount is: ， ， ， ， in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of a neuron node.
6. An oil quality monitoring system for a phase shifter lubricating oil system, characterized in that: include: A data acquisition program unit is used to obtain detection index data of lubricating oil in the phase shifter lubricating oil system, including part or all of trace water content, temperature and particle number; The result prediction program unit is used to input the detection index data of the lubricating oil into a pre-trained machine learning model to obtain the oil quality monitoring results of the lubricating oil in the phase shifter lubricating oil system. The machine learning model is pre-trained to establish a mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results. The oil quality monitoring results include the oil quality grade of the lubricating oil and part or all of whether it has deteriorated.
7. The oil quality monitoring system for the phase condenser lubricating oil system according to claim 6 is characterized in that: The trace water content is detected by a capacitive humidity sensor, the particle number is detected by an automatic particle counter, a dynamic image particle sensor, a capacitive sensor or a laser particle size analyzer; and the temperature is detected by a temperature sensor.
8. The oil quality monitoring system for the phase condenser lubricating oil system according to claim 7 is characterized in that: The result obtained by the capacitive humidity sensor is the relative humidity of the lubricating oil. After obtaining the relative humidity of the lubricating oil, the relative humidity of the lubricating oil is temperature compensated according to the following formula to obtain the trace water content of the lubricating oil: ， in, The water content is trace, is the relative humidity of the lubricating oil, is the temperature of the lubricating oil, and is a constant coefficient.
9. The oil quality monitoring system for the phase condenser lubricating oil system according to claim 7, characterized in that: When detecting the number of particles by an automatic particle counter, a dynamic image particle sensor capacitive sensor or a laser particle size analyzer, the detection window of the lubricating oil is divided into multiple sub-windows along the direction of gravity and the overlapping coefficients are set for each sub-window, and the overlapping coefficients of the sub-windows closer to the bottom are larger. The number of particles in each sub-window is detected respectively, and the overlapping coefficient is used as a weight coefficient to perform weighted summation on the number of particles in each sub-window to obtain the number of particles in the entire detection window.
10. The oil quality monitoring system for the phase condenser lubricating oil system according to claim 6, characterized in that: The machine learning model is a BP neural network, which consists of an input layer, a hidden layer, and an output layer. The number of neuron nodes in the input layer, the hidden layer, and the output layer are respectively 、 and , where the input vector of the input layer is The function expression is: ， in, ~ Respectively of samples detection index data; any hidden layer The function expression of a neuron node is: ， in, The hidden layer The output of a neuron node, is the activation function, The input layer neuron nodes to the hidden layer The connection weights of the neuron nodes, For the of samples Detection indicator data, The hidden layer The threshold of the neuron node; any of the output layers The function expression of a neuron node is: ， in, The output layer The output of a neuron node, The hidden layer neuron nodes to the output layer The connection weights of the neuron nodes, The output layer The threshold value of each neuron node; the training process of the BP neural network includes: obtaining a detection index data sample of the lubricating oil in the lubricating oil system of the phase shifter; adding the detection index data sample based on the generative adversarial network (GAN); the generative adversarial network (GAN) is composed of a generator and a discriminator, the generator is used to generate the detection index data sample through continuous iteration through random noise, the discriminator is responsible for identifying whether the generated detection index data sample is derived from real data or generated data and giving a corresponding probability score, and the generator and the discriminator are both adversarially trained through the loss function shown in the following formula to achieve the effect that the detection index data sample generated by the generator is close to the distribution of real data and serves as the added detection index data sample: ， ， in, is the loss function of the discriminator, For the discriminator to input data The judgment result, is the detection index data sample generated by the generator for the input noise z, For the discriminator pair The judgment result, is the loss function of the generator; the detection index data samples are labeled, divided into training sets and test sets, the BP neural network is trained using the training sets and test sets, and the model parameters in the BP neural network are obtained using the gradient descent method 、 、 、 Adjust the model parameters by adjusting the amount 、 、 、 , until the number of iterations is equal to the preset threshold or the prediction accuracy of the BP neural network on the test set meets the requirements, so that the BP neural network is trained to establish the mapping relationship between the detection index data of the lubricating oil and the oil quality monitoring results; in the training process of the BP neural network, the gradient descent method is used to find the model parameters in the BP neural network 、 、 、 The function expression of the adjustment amount is: ， ， ， ， in, 、 、 and They are 、 、 、 The adjustment amount, and is a constant coefficient, The output layer The output error of the neuron node is ,in The output layer The label value corresponding to the output of a neuron node.
11. A phase shifter lubricating oil system oil quality monitoring system, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the oil quality monitoring method for the lubricating oil system of a phase shifter according to any one of claims 1 to 5.
12. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system according to any one of claims 1 to 5 through a processor.
13. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the oil quality monitoring method of the phase shifter lubricating oil system according to any one of claims 1 to 5 through a processor.