Method and system for detecting internal oil leakage of speed regulator of hydraulic turbine set

Abstract

The invention discloses a hydraulic turbine set governor internal oil leakage detection method and system, and the method comprises the steps: S1, obtaining the historical oil leakage state data of a hydraulic turbine set governor oil pressure device, and building the oil leakage risk evaluation data of the oil pressure device; s2, preprocessing the oil leakage risk assessment data of the oil pressure device to obtain an oil leakage risk assessment data set; s3, constructing an oil leakage comprehensive prediction model according to the oil leakage risk assessment data set; and S4, detecting the oil leakage amount of the target hydraulic turbine set speed regulator through the oil leakage comprehensive prediction model, constructing an oil leakage comprehensive prediction model by collecting historical oil leakage state data of the hydraulic turbine set speed regulator oil pressure device, and calculating the oil leakage amount of the target hydraulic turbine set speed regulator by searching the weight and the threshold of the minimum error of the BP neural network real oil leakage amount and the prediction output result. According to the method, the accuracy of predicting the predicted oil leakage amount of the target hydraulic turbine set speed regulator is improved, workers do not need to conduct field detection in the process, and the time and energy of the workers are greatly saved.

Description

Technical Field

[0001] The invention belongs to the technical field of oil leakage fault detection, and in particular relates to a method and system for detecting internal oil leakage of a water turbine speed governor. Background Art

[0002] The turbine speed governor is an important part of the turbine. The stability of the turbine speed governor directly affects whether the turbine can operate normally, and further determines whether the hydropower station can provide safe electricity. Therefore, it is very necessary to do a good job of speed governor inspection and maintenance. The oil pressure device is an important part of the speed governor. Once the oil pressure device leaks oil, it is easy to have guide vane problems. It is necessary to strengthen the control of oil pressure, oil temperature and oil volume in the oil pressure device to provide guarantee for the subsequent work of the speed governor. However, various faults often occur in the oil pressure device during operation, and dynamic supervision and detection must be carried out to keep it within the normal range at all times to maintain the safety and stability of the turbine.

[0003] There are many existing leakage detection methods, which can only detect whether leakage occurs or not, but lack the estimation of leakage situation based on system status. In this case, how to accurately predict the oil leakage of turbine governor has become a crucial and urgent problem to be solved. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a method and system for detecting internal oil leakage in a turbine speed governor provided by the present invention solves the problem of low accuracy in predicting the oil leakage amount of a turbine speed governor at present.

[0005] In order to achieve the above-mentioned invention object, in a first aspect, the technical solution adopted by the present invention is: a method for detecting oil leakage inside a turbine speed governor, comprising the following steps:

[0006] S1. Obtain the historical oil leakage status data of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device;

[0007] S2. Preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set;

[0008] S3, constructing a comprehensive oil spill prediction model based on the oil spill risk assessment dataset;

[0009] S4. Detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

[0010] Further: in said S1, the risk assessment data of oil leakage of the oil pressure device includes oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level, return oil tank liquid level and oil leakage amount.

[0011] Further: In S2, the method for preprocessing the risk assessment data of oil leakage of the hydraulic device is specifically as follows:

[0012] The risk assessment data is normalized and the normalized data The specific expression is:

[0013]

[0014] In the formula, x i is the i-th risk assessment data, x min is the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

[0015] Further: S3 includes the following sub-steps:

[0016] S31, inputting the oil spill risk assessment data set into the BP neural network, optimizing the BP neural network using a particle swarm algorithm, and outputting a better weight and threshold of the BP neural network;

[0017] S32. Calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

[0018] Further: S31 includes the following sub-steps:

[0019] S311, inputting the oil spill risk assessment data set into a BP neural network, wherein the BP neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes, and the output layer is provided with 1 neuron node;

[0020] The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity;

[0021] S312, initializing the particle population and the number of iterations, and taking the absolute value of the error between the actual output and the expected output of the BP neural network as the fitness value of the individual;

[0022] S313, calculating the fitness of each particle in the particle population under the current number of iterations;

[0023] S314, updating the position of the particle in the particle population according to the fitness of the particle;

[0024] S315, judging whether the current number of iterations reaches the iteration number threshold, if so, taking the position of the updated particle population as the global optimal position and proceeding to S316; if not, adding 1 to the number of iterations and returning to S313;

[0025] S316. Obtaining a relatively optimal weight and threshold of the BP neural network according to the global optimal position.

[0026] The beneficial effect of the above further scheme is: the present invention optimizes the BP neural network through the particle swarm algorithm, and searches for the weights and thresholds of the BP neural network with the smallest error between the actual output and the expected output, which can improve the global search capability and overcome the defect that it is easy to stop iteration at the local optimum, so as to reduce the prediction error.

[0027] Further: In S314, the position x of the lth particle in the particle population under the current number of iterations is calculated. l t+1 The specific expression is:

[0028] x l t+1 =x l t +v l t+1

[0029] In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number;

[0030] v l t+1 =ωv l t +c(P l -x l t )+σ(X g -x l t )

[0031] Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

[0032] The beneficial effects of the above further scheme are as follows: the present invention learns and optimizes the speed of particles according to the relationship between the extreme value of the particle population and the individual extreme value of the particles, which significantly improves the search speed, introduces random perturbations in the updated speed of the particles to further expand the search range, and prevents the early position search from converging to the local optimal value.

[0033] Further: S32 includes the following sub-steps:

[0034] S321, calculating the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtaining the error between the predicted output result of the oil leakage amount and the actual oil leakage amount;

[0035] S322, judging whether the error is lower than the error threshold, if so, saving the weight and threshold of the current BP neural network, and establishing a comprehensive oil leakage prediction model; if not, proceeding to S323;

[0036] S323. Update the weights and thresholds of the current BP neural network according to the error, generate updated weights and thresholds of the BP neural network, and return to S321.

[0037] The beneficial effect of the above further scheme is: the present invention learns the relationship between the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level and the return tank liquid level and the oil leakage by adjusting the weights and thresholds in the network, calculates the error between the predicted output result of the oil leakage and the actual oil leakage, and further optimizes the weights and thresholds of the BP neural network, so that the final prediction accuracy is higher. Compared with the traditional BP neural network optimization method, it can ensure the prediction accuracy while the calculation speed is faster, and has a stronger ability to solve prediction problems.

[0038] Further: In the above S321, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically:

[0039] E=YZ

[0040] In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is:

[0041]

[0042] In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer;

[0043]

[0044] Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

[0045] Further: In said S323, the expression for generating the weight of the updated BP neural network is specifically:

[0046]

[0047] In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update;

[0048] The expression for generating the threshold of the updated BP neural network is specifically:

[0049]

[0050] b 1 =b+E

[0051] In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

[0052] Further: S4 includes the following sub-steps:

[0053] S41, obtaining the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor;

[0054] S42, inputting the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level and the return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage amount of the target turbine unit governor.

[0055] In a second aspect, the present invention provides a system for detecting internal oil leakage of a turbine speed governor, comprising:

[0056] A data acquisition module is used to obtain the status data of the historical oil leakage of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device;

[0057] A data processing module, used for preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set;

[0058] Model training module, used to build a comprehensive oil spill prediction model based on the oil spill risk assessment dataset;

[0059] The oil leakage prediction module is used to detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

[0060] Furthermore: the risk assessment data of oil leakage of the hydraulic device includes oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level, return oil tank liquid level and oil leakage amount.

[0061] Further: the data processing module is provided with a normalization processing submodule, and the method for the normalization processing submodule to preprocess the risk assessment data of oil leakage of the oil pressure device is specifically:

[0062] The risk assessment data is normalized and the normalized data The specific expression is:

[0063]

[0064] In the formula, x i is the i-th risk assessment data, x min is the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

[0065] Further: The model training module includes:

[0066] The particle swarm algorithm submodule is used to input the oil spill risk assessment data set into the BP neural network, optimize the BP neural network using the particle swarm algorithm, and output the better weights and thresholds of the BP neural network;

[0067] The BP neural network optimization submodule is used to calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

[0068] Further: The method for the particle swarm algorithm submodule to output a relatively optimal BP neural network weight and threshold is specifically:

[0069] A1. Input the oil spill risk assessment data set into the BP neural network, wherein the BP neural network includes an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes and the output layer is provided with 1 neuron node;

[0070] The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity;

[0071] A2, initialize the particle population and the number of iterations, and take the absolute value of the error between the actual output and the expected output of the BP neural network as the individual fitness value;

[0072] A3. Calculate the fitness of each particle in the particle population under the current number of iterations;

[0073] A4. Update the position of particles in the particle population according to the fitness of the particles;

[0074] A5, determine whether the current number of iterations reaches the iteration threshold. If so, take the updated position of the particle population as the global optimal position and proceed to A6; if not, add 1 to the number of iterations and return to A3;

[0075] A6. Obtain the better weights and thresholds of the BP neural network based on the global optimal position.

[0076] Further: In A4, the position x of the lth particle in the particle population at the current number of iterations is calculated. l t+1 The specific expression is:

[0077] x l t+1 =x l t +v l t+1

[0078] In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number;

[0079] v l t+1 =ωv l t +c(P l -x l t )+σ(X g -x l t )

[0080] Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

[0081] Further: The method for establishing the comprehensive oil leakage prediction model by the BP neural network optimization submodule is specifically as follows:

[0082] B1. Calculate the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtain the error between the predicted output result of the oil leakage amount and the actual oil leakage amount;

[0083] B2, determine whether the error is lower than the error threshold. If so, save the weight and threshold of the current BP neural network and establish a comprehensive oil leakage prediction model; if not, enter B3;

[0084] B3. Update the weights and thresholds of the current BP neural network according to the error, generate the updated weights and thresholds of the BP neural network, and return to B1.

[0085] Further: In B1, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically:

[0086] E=YZ

[0087] In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is:

[0088]

[0089] In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer;

[0090]

[0091] Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

[0092] Further: In B3, the expression for generating the updated weights of the BP neural network is specifically:

[0093]

[0094] In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update;

[0095] The expression for generating the threshold of the updated BP neural network is specifically:

[0096]

[0097] b1 =b+E

[0098] In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

[0099] Further: the oil leakage prediction module includes:

[0100] A data acquisition submodule, used to obtain the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor;

[0101] The oil leakage prediction submodule is used to input the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage of the target turbine unit governor.

[0102] The beneficial effects of the present invention are:

[0103] (1) The present invention provides a method and system for detecting oil leakage inside a turbine governor. The method collects the status data of historical oil leakage of the hydraulic device of the turbine governor, constructs a comprehensive oil leakage prediction model, and searches for the weight and threshold with the smallest error between the actual oil leakage amount and the predicted output result of the BP neural network to improve the accuracy of predicting the oil leakage amount of the target turbine governor. The process does not require staff to conduct on-site inspection, which greatly saves staff time and energy.

[0104] (2) The present invention optimizes the BP neural network through the particle swarm algorithm, takes the absolute value of the error between the actual output and the expected output of the BP neural network as the individual fitness value, continuously optimizes the particle position, improves the global search ability of the BP neural network training and improves its convergence efficiency. On this basis, the weights and thresholds of the BP neural network are further optimized according to the error between the predicted output result and the actual oil leakage amount, ensuring that the weights and thresholds of the searched BP neural network are globally optimal, which can improve the accuracy of oil leakage prediction.

[0105] (3) The present invention constructs a comprehensive oil leakage prediction model based on a particle swarm optimization BP neural network. By obtaining the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level of the target turbine unit governor, the oil leakage of the target turbine unit governor can be quickly predicted, and safety accidents caused by oil leakage can be effectively avoided during the operation of the target turbine unit, thereby achieving the purpose of improving safety management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 The present invention is a flow chart of a method for detecting oil leakage inside a turbine speed governor.

[0107] Figure 2 The present invention is a schematic diagram of an internal oil leakage detection system for a water turbine speed governor. DETAILED DESCRIPTION

[0108] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0109] Embodiment 1:

[0110] like Figure 1 As shown, in one embodiment of the present invention, a method for detecting oil leakage inside a turbine speed governor comprises the following steps:

[0111] S1. Obtain the historical oil leakage status data of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device;

[0112] S2. Preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set;

[0113] S3, constructing a comprehensive oil spill prediction model based on the oil spill risk assessment dataset;

[0114] S4. Detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

[0115] In S1, the risk assessment data of oil leakage of the oil pressure device includes the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level, the return oil tank liquid level and the oil leakage amount.

[0116] In this embodiment, corresponding sensors are arranged on the oil pressure device to collect the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level, and the oil leakage is collected through the on-site oil leakage tank.

[0117] In S2, the method for preprocessing the risk assessment data of oil leakage of the hydraulic device is specifically as follows:

[0118] The risk assessment data is normalized and the normalized data The specific expression is:

[0119]

[0120] In the formula, x i is the i-th risk assessment data, x minis the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

[0121] In this embodiment, the risk assessment data used for BP neural network training needs to be normalized and preprocessed so that the input value range is within the interval [0, 1].

[0122] The S3 comprises the following sub-steps:

[0123] S31, inputting the oil spill risk assessment data set into the BP neural network, optimizing the BP neural network using a particle swarm algorithm, and outputting a better weight and threshold of the BP neural network;

[0124] S32. Calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

[0125] The S31 comprises the following sub-steps:

[0126] S311, inputting the oil spill risk assessment data set into a BP neural network, wherein the BP neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes, and the output layer is provided with 1 neuron node;

[0127] The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity;

[0128] In this embodiment, during the network training process, the interval range of the data is taken into account, the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are used as input quantities, and the oil leakage amount is used as output quantity, and the ultimate goal is to predict the oil leakage amount.

[0129] S312, initializing the particle population and the number of iterations, and taking the absolute value of the error between the actual output and the expected output of the BP neural network as the fitness value of the individual;

[0130] S313, calculating the fitness of each particle in the particle population under the current number of iterations;

[0131] S314, updating the position of the particle in the particle population according to the fitness of the particle;

[0132] S315, judging whether the current number of iterations reaches the iteration number threshold, if so, taking the position of the updated particle population as the global optimal position and proceeding to S316; if not, adding 1 to the number of iterations and returning to S313;

[0133] S316. Obtaining a relatively optimal weight and threshold of the BP neural network according to the global optimal position.

[0134] In this embodiment, the present invention optimizes the BP neural network through a particle swarm algorithm to search for the weights and thresholds of the BP neural network with the smallest error between the actual output and the expected output, which can improve the global search capability and overcome the defect that it is easy to stop iteration at the local optimum, so as to reduce the prediction error.

[0135] In S314, the position x of the lth particle in the particle population under the current number of iterations is calculated. l t+1 The specific expression is:

[0136] x l t+1 =x l t +v l t+1

[0137] In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number;

[0138] v l t+1 =ωv l t +c(P l -x l t )+σ(X g -x l t )

[0139] Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

[0140] In this embodiment, during the iteration of the particle population, each particle will interact and share information. The best global position reached during the iteration of the particle population is taken as the extreme value of the particle population, and the best global position reached by each particle is taken as the individual extreme value of the particle. The speed of the particles is optimized based on the relationship between the two, which significantly improves the search speed. Random disturbances are introduced into the updated speed of the particles to further expand the search range and prevent the early position search from converging to the local optimal value.

[0141] The S32 comprises the following sub-steps:

[0142] S321, calculating the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtaining the error between the predicted output result of the oil leakage amount and the actual oil leakage amount;

[0143] S322, judging whether the error is lower than the error threshold, if so, saving the weight and threshold of the current BP neural network, and establishing a comprehensive oil leakage prediction model; if not, proceeding to S323;

[0144] S323. Update the weights and thresholds of the current BP neural network according to the error, generate updated weights and thresholds of the BP neural network, and return to S321.

[0145] In this embodiment, the relationship between the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank level, the return tank level and the oil leakage is learned by adjusting the weights and thresholds in the network, the error between the predicted output result of the oil leakage and the actual oil leakage is calculated, and the weights and thresholds of the BP neural network are further optimized, so that the final prediction accuracy is higher. Compared with the traditional BP neural network optimization method, it can ensure the prediction accuracy while the calculation speed is faster, and has a stronger ability to solve prediction problems.

[0146] In S321, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically:

[0147] E=YZ

[0148] In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is:

[0149]

[0150] In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer;

[0151]

[0152] Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

[0153] In S323, the expression for generating the updated weight of the BP neural network is specifically:

[0154]

[0155] In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update;

[0156] The expression for generating the threshold of the updated BP neural network is specifically:

[0157]

[0158] b 1 =b+E

[0159] In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

[0160] The S4 comprises the following sub-steps:

[0161] S41, obtaining the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor;

[0162] S42, inputting the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level and the return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage amount of the target turbine unit governor.

[0163] Embodiment 2:

[0164] like Figure 2 As shown, in this embodiment, a system for detecting internal oil leakage of a turbine speed governor includes:

[0165] A data acquisition module is used to obtain the status data of the historical oil leakage of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device;

[0166] A data processing module, used for preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set;

[0167] Model training module, used to build a comprehensive oil spill prediction model based on the oil spill risk assessment dataset;

[0168] The oil leakage prediction module is used to detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

[0169] The risk assessment data for oil leakage in hydraulic devices include oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level, return tank liquid level and oil leakage amount.

[0170] The data processing module is provided with a normalization processing submodule, and the method for the normalization processing submodule to preprocess the risk assessment data of oil leakage of the oil pressure device is specifically as follows:

[0171] The risk assessment data is normalized and the normalized data The specific expression is:

[0172]

[0173] In the formula, x i is the i-th risk assessment data, x min is the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

[0174] The model training module includes:

[0175] The particle swarm algorithm submodule is used to input the oil spill risk assessment data set into the BP neural network, optimize the BP neural network using the particle swarm algorithm, and output the better weights and thresholds of the BP neural network;

[0176] The BP neural network optimization submodule is used to calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

[0177] The method for the particle swarm algorithm submodule to output the optimal weights and thresholds of the BP neural network is specifically as follows:

[0178] A1. Input the oil spill risk assessment data set into the BP neural network, wherein the BP neural network includes an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes and the output layer is provided with 1 neuron node;

[0179] The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity;

[0180] A2, initialize the particle population and the number of iterations, and take the absolute value of the error between the actual output and the expected output of the BP neural network as the individual fitness value;

[0181] A3. Calculate the fitness of each particle in the particle population under the current number of iterations;

[0182] A4. Update the position of particles in the particle population according to the fitness of the particles;

[0183] A5, determine whether the current number of iterations reaches the iteration threshold. If so, take the updated position of the particle population as the global optimal position and proceed to A6; if not, add 1 to the number of iterations and return to A3;

[0184] A6. Obtain the better weights and thresholds of the BP neural network based on the global optimal position.

[0185] In A4, the position x of the lth particle in the particle population at the current number of iterations is calculated. l t+1 The specific expression is:

[0186] x l t+1 =x l t +v l t+1

[0187] In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number;

[0188] v l t+1 =ωv l t +c(P l -x l t )+σ(X g -x l t )

[0189] Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

[0190] The method for establishing the comprehensive oil leakage prediction model by the BP neural network optimization submodule is specifically as follows:

[0191] B1. Calculate the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtain the error between the predicted output result of the oil leakage amount and the actual oil leakage amount;

[0192] B2, determine whether the error is lower than the error threshold. If so, save the weight and threshold of the current BP neural network and establish a comprehensive oil leakage prediction model; if not, enter B3;

[0193] B3. Update the weights and thresholds of the current BP neural network according to the error, generate the updated weights and thresholds of the BP neural network, and return to B1.

[0194] In B1, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically:

[0195] E=YZ

[0196] In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is:

[0197]

[0198] In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer;

[0199]

[0200] Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

[0201] In B3, the expression for generating the updated weights of the BP neural network is specifically:

[0202]

[0203] In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update;

[0204] The expression for generating the threshold of the updated BP neural network is specifically:

[0205]

[0206] b 1 =b+E

[0207] In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

[0208] The oil leakage prediction module comprises:

[0209] A data acquisition submodule, used to obtain the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor;

[0210] The oil leakage prediction submodule is used to input the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage of the target turbine unit governor.

[0211] The beneficial effects of the present invention are as follows: the present invention provides a method and system for detecting internal oil leakage of a turbine set speed governor, by collecting the status data of historical oil leakage of the oil pressure device of the turbine set speed governor, constructing an oil leakage comprehensive prediction model, and searching for the weights and thresholds with the smallest error between the actual oil leakage amount and the predicted output result of the BP neural network to improve the accuracy of predicting the oil leakage amount of the target turbine set speed governor. This process does not require staff to conduct on-site inspections, greatly saving the time and energy of the staff.

[0212] The present invention optimizes the BP neural network through the particle swarm algorithm, takes the absolute value of the error between the actual output and the expected output of the BP neural network as the individual fitness value, continuously optimizes the particle position, improves the global search ability of the BP neural network training and improves its convergence efficiency, and further optimizes the weights and thresholds of the BP neural network according to the error between the predicted output result and the actual oil leakage amount on this basis, ensures that the weights and thresholds of the searched BP neural network are globally optimal, and can improve the accuracy of oil leakage prediction.

[0213] The present invention constructs a comprehensive oil leakage prediction model based on a particle swarm optimization BP neural network. By obtaining the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level of the target turbine unit governor, the oil leakage of the target turbine unit governor can be quickly predicted, and safety accidents caused by oil leakage can be effectively avoided during the operation of the target turbine unit, thereby achieving the purpose of improving safety management efficiency.

[0214] In the description of the present invention, it is necessary to understand that the orientation or positional relationship indicated by the terms "center", "thickness", "upper", "lower", "horizontal", "top", "bottom", "inner", "outer", "radial", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", and "third" are used only for descriptive purposes, and cannot be understood as indicating or implying the relative importance or the number of implicitly specified technical features. Therefore, the features defined by "first", "second", and "third" may explicitly or implicitly include one or more of the features.

Claims

1. A method for detecting oil leakage inside a turbine speed governor, characterized in that: The following steps are involved: S1. Obtain the historical oil leakage status data of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device; S2. Preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set; S3, constructing a comprehensive oil spill prediction model based on the oil spill risk assessment dataset; S4. Detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

2. The method for detecting oil leakage inside a turbine speed governor according to claim 1, characterized in that: In S1, the risk assessment data of oil leakage of the oil pressure device includes the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level, the return oil tank liquid level and the oil leakage amount.

3. The method for detecting oil leakage inside a turbine speed governor according to claim 1, characterized in that: In S2, the method for preprocessing the risk assessment data of oil leakage of the hydraulic device is specifically as follows: The risk assessment data is normalized and the normalized data The specific expression is: In the formula, x i is the i-th risk assessment data, x min is the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

4. The method for detecting oil leakage inside a turbine speed governor according to claim 1, characterized in that: The S3 comprises the following sub-steps: S31, inputting the oil spill risk assessment data set into the BP neural network, optimizing the BP neural network using a particle swarm algorithm, and outputting a better weight and threshold of the BP neural network; S32. Calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

5. The method for detecting oil leakage inside a turbine speed governor according to claim 4, characterized in that: The S31 comprises the following sub-steps: S311, inputting the oil spill risk assessment data set into a BP neural network, wherein the BP neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes, and the output layer is provided with 1 neuron node; The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity; S312, initializing the particle population and the number of iterations, and taking the absolute value of the error between the actual output and the expected output of the BP neural network as the fitness value of the individual; S313, calculating the fitness of each particle in the particle population under the current number of iterations; S314, updating the position of the particle in the particle population according to the fitness of the particle; S315, judging whether the current number of iterations reaches the iteration number threshold, if so, taking the position of the updated particle population as the global optimal position and proceeding to S316; if not, adding 1 to the number of iterations and returning to S313; S316. Obtaining a relatively optimal weight and threshold of the BP neural network according to the global optimal position.

6. The method for detecting oil leakage inside a turbine speed governor according to claim 5, characterized in that: In S314, the position x of the lth particle in the particle population under the current number of iterations is calculated. l t+1 The specific expression is: x l t+1 =x l t +v l t+1 In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number; v l t+1 =ωv l t +c(P l -x l t )+σ(X g -x l t ) Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

7. The method for detecting oil leakage inside a turbine speed governor according to claim 6, characterized in that: The S32 comprises the following sub-steps: S321, calculating the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtaining the error between the predicted output result of the oil leakage amount and the actual oil leakage amount; S322, judging whether the error is lower than the error threshold, if so, saving the weight and threshold of the current BP neural network, and establishing a comprehensive oil leakage prediction model; if not, proceeding to S323; S323. Update the weights and thresholds of the current BP neural network according to the error, generate updated weights and thresholds of the BP neural network, and return to S321.

8. The method for detecting oil leakage inside a turbine speed governor according to claim 7, characterized in that: In S321, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically: E=YZ In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is: In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer; Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

9. The method for detecting oil leakage inside a turbine speed governor according to claim 8, characterized in that: In S323, the expression for generating the updated weight of the BP neural network is specifically: In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update; The expression for generating the threshold of the updated BP neural network is specifically: b 1 =b+E In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

10. The method for detecting oil leakage inside a turbine speed governor according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41, obtaining the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor; S42, inputting the oil pump output flow, the oil pump motor signal, the oil tank pressure, the oil tank liquid level and the return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage amount of the target turbine unit governor.

11. A system for detecting internal oil leakage of a turbine speed governor, characterized in that: include: A data acquisition module is used to obtain the status data of the historical oil leakage of the hydraulic device of the turbine speed governor and establish the risk assessment data of the oil leakage of the hydraulic device; A data processing module, used for preprocessing the risk assessment data of oil leakage of the hydraulic device to obtain an oil leakage risk assessment data set; Model training module, used to build a comprehensive oil spill prediction model based on the oil spill risk assessment dataset; The oil leakage prediction module is used to detect the oil leakage of the target turbine speed governor through the comprehensive oil leakage prediction model.

12. The hydraulic turbine governor internal oil leakage detection system according to claim 11, characterized in that: The risk assessment data for oil leakage in hydraulic devices include oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level, return tank liquid level and oil leakage amount.

13. The hydraulic turbine governor internal oil leakage detection system according to claim 11, characterized in that: The data processing module is provided with a normalization processing submodule, and the method for the normalization processing submodule to preprocess the risk assessment data of oil leakage of the oil pressure device is specifically as follows: The risk assessment data is normalized and the normalized data The specific expression is: In the formula, x i is the i-th risk assessment data, x min is the minimum value of the risk assessment data, x max is the maximum value of the risk assessment data.

14. The hydraulic turbine governor internal oil leakage detection system according to claim 11, characterized in that: The model training module includes: The particle swarm algorithm submodule is used to input the oil spill risk assessment data set into the BP neural network, optimize the BP neural network using the particle swarm algorithm, and output the better weights and thresholds of the BP neural network; The BP neural network optimization submodule is used to calculate the error between the predicted output result of the oil leakage and the actual oil leakage according to the weights and thresholds of the better BP neural network, further optimize the weights and thresholds of the BP neural network according to the error, and establish a comprehensive oil leakage prediction model.

15. The internal oil leakage detection system of the turbine speed governor according to claim 14, characterized in that: The method for the particle swarm algorithm submodule to output the optimal weights and thresholds of the BP neural network is specifically as follows: A1. Input the oil spill risk assessment data set into the BP neural network, wherein the BP neural network includes an input layer, a hidden layer and an output layer, wherein the input layer is provided with 5 neuron nodes and the output layer is provided with 1 neuron node; The oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return tank liquid level are taken as input quantities, and the oil leakage quantity is taken as output quantity; A2, initialize the particle population and the number of iterations, and take the absolute value of the error between the actual output and the expected output of the BP neural network as the individual fitness value; A3. Calculate the fitness of each particle in the particle population under the current number of iterations; A4. Update the position of particles in the particle population according to the fitness of the particles; A5, determine whether the current number of iterations reaches the iteration number threshold. If so, take the updated position of the particle population as the global optimal position and proceed to A6; If not, the number of iterations is increased by 1 and A3 is returned; A6. Obtain the optimal weights and thresholds of the BP neural network based on the global optimal position.

16. The hydraulic turbine governor internal oil leakage detection system according to claim 15, characterized in that: In A4, the position x of the lth particle in the particle population at the current number of iterations is calculated. l t+1 The specific expression is: x l t+1 =x l t +v l t+1 In the formula, x l t is the position of the lth particle before update, v l t+1 is the updated velocity of the lth particle, and t is the current iteration number; Where ω is the inertia weight, v l t is the speed of the lth particle before updating, c is the learning factor, P l is the individual extreme value of the lth particle, X g is the extreme value of the particle population, and σ is the random disturbance.

17. The hydraulic turbine governor internal oil leakage detection system according to claim 16, characterized in that: The method for establishing the comprehensive oil spill prediction model by the BP neural network optimization submodule is specifically as follows: B1. Calculate the predicted output result of the oil leakage amount according to the weight and threshold of the current BP neural network, and obtain the error between the predicted output result of the oil leakage amount and the actual oil leakage amount; B2, determine whether the error is lower than the error threshold. If so, save the weight and threshold of the current BP neural network and establish a comprehensive oil leakage prediction model; if not, enter B3; B3. Update the weights and thresholds of the current BP neural network according to the error, generate the updated weights and thresholds of the BP neural network, and return to B1.

18. The hydraulic turbine governor internal oil leakage detection system according to claim 17, characterized in that: In B1, the expression for calculating the error E between the predicted output result of the oil leakage amount and the actual oil leakage amount is specifically: E=YZ In the formula, Z is the actual oil leakage, and Y is the predicted output result of the oil leakage. The specific expression is: In the formula, h k is the output of the kth neuron node in the hidden layer, W k is the weight between the kth neuron node in the hidden layer of the current BP neural network and the neuron node in the output layer, N is the total number of neuron nodes in the hidden layer, and b is the threshold of the neuron node in the output layer; Where f(·) is the output function of the hidden layer, W jk is the weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer of the current BP neural network, x j The risk assessment data input to the jth neuron node in the input layer, a k is the threshold of the kth neuron node in the hidden layer.

19. The hydraulic turbine governor internal oil leakage detection system according to claim 18, characterized in that: In B3, the expression for generating the updated weights of the BP neural network is specifically: In the formula, is the updated weight between the jth neuron node in the input layer and the kth neuron node in the hidden layer, is the weight between the kth neuron node in the hidden layer and the neuron node in the output layer after update; The expression for generating the threshold of the updated BP neural network is specifically: b 1 =b+E In the formula, is the threshold of the kth neuron node in the hidden layer after update, b 1 is the threshold of the updated output layer neuron node.

20. The hydraulic turbine governor internal oil leakage detection system according to claim 11, characterized in that: The oil leakage prediction module comprises: A data acquisition submodule, used to obtain the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level of the target turbine unit governor; The oil leakage prediction submodule is used to input the oil pump output flow, oil pump motor signal, oil tank pressure, oil tank liquid level and return oil tank liquid level into the comprehensive oil leakage prediction model to obtain the predicted oil leakage of the target turbine unit governor.

Images