Oil filtering control method, device, equipment and storage medium for fan gearbox

Abstract

The present application discloses an oil filtration control method, device, equipment and storage medium for a wind turbine gearbox, and relates to the technical field of electrical digital data processing. The method relies on an existing electromagnetic adsorption system and performs real-time detection and electromagnetic control, forming a system from detection to response. The method utilizes the characteristics that the material of the wind turbine gear is usually low-carbon steel or medium-carbon alloy structural steel, and learns and trains the amount of debris of the aforementioned material adsorbed by the electromagnet to pursue more accurate control. The present embodiment has good robustness for wind turbines in different wind scenarios, and obtains different prediction models by analyzing the amount of debris under different wind forces, which is generally understood as one model for one machine, so that the present embodiment can accurately control the electromagnetic filter of each wind turbine gearbox respectively, and can integrate and display the control strategy of each electromagnetic filter at the terminal station, with a high degree of personalized customization and accurate control effect.

Description

Technical Field

[0001] The present application relates to the technical field of electronic digital data processing, and in particular to an oil filtering control method, device, equipment and storage medium for a fan gearbox. Background Art

[0002] Wind power generation refers to the conversion of wind kinetic energy into electrical energy. Wind energy is a clean, pollution-free renewable energy source. Wind power generation uses wind to drive the windmill blades to rotate, and then increases the speed of rotation through the speed increaser to drive the generator to generate electricity. Wind power generation does not require the use of fuel, and does not produce radiation or air pollution. It is a renewable energy source. Wind power generation mainly converts wind energy into mechanical work through the fan (wind turbine), which drives the rotor to rotate and finally outputs alternating current.

[0003] Usually the speed of the wind wheel is very low, far from the speed required by the generator. It must be achieved through the speed increase of the gearbox gear pair, so the gearbox is also called a speed increaser. According to the overall layout requirements of the unit, sometimes the transmission shaft (commonly known as the main shaft) directly connected to the wind wheel hub is integrated with the gearbox, and there is also a structure in which the main shaft and the gearbox are arranged separately, and the expansion sleeve device or coupling is used to connect them. Therefore, the gearbox is an important mechanical component that is widely used in wind turbines. Its main function is to transmit the power generated by the wind wheel under the action of wind to the generator and make it get the corresponding speed.

[0004] Since the gearbox is a mechanical transmission, there will inevitably be a certain degree of mechanical wear, such as grinding wear: This is because the abrasive particles on the surface of the grinding wheel gradually become passivated during the friction between the gears and bearings, resulting in wear marks on the surfaces of the gears, bearings and other parts in the gearbox. Fatigue wear: Under the action of alternating loads, pitting or shedding appears on the friction surface, usually because the components are subjected to cyclic loads for a long time, resulting in cracks, collapse and other problems on the surface of the gears, bearings and other components of the gearbox. Corrosion wear: In the corrosive medium, the metal surface reacts chemically or electrochemically with the corrosive medium, accompanied by mechanical wear, usually because the humidity inside the gearbox is too high or the oil is contaminated. Abrasive wear: When the hardness of one side of the friction pair is much greater than that of the other side, or when there are hard particles between the contact surfaces, abrasive wear will occur, resulting in abrasive particles and abrasive particle accumulation on the surface of the gears, bearings and other components of the gearbox.

[0005] At present, the mainstream cleaning method for gearbox wear is to clean the debris of the fan gearbox, which mainly includes daily maintenance and regular maintenance. Among them, daily maintenance is to clean the surface of the gearbox body, check whether there is leakage in the body and lubrication pipeline, and whether the external lubrication pipeline is loose; check whether the oil level and oil color are normal. If the oil color is obviously darkened and black, you should consider oil quality inspection and strengthen the operation monitoring of the unit; when there is a filter blockage alarm, it should be checked and handled in time, and the inside of the filter should be thoroughly cleaned. If conditions permit, it is best to remove the filter assembly and clean and inspect it in the workshop. Check whether the wiring of sensors such as gear oil level, temperature, pressure, pressure difference, bearing temperature, heater, and radiator is normal, and whether the wires are worn; during daily inspections, you should also pay attention to whether the noise of the unit is abnormal, and discover potential faults in time. Regular maintenance includes checking the torque of the gearbox connecting bolts, checking the gear meshing and tooth surface wear, testing the sensor function, and checking the lubrication and cooling system functions; regularly replacing the gear oil filter, collecting oil samples, and, if conditions permit, using relevant industrial testing equipment to test and analyze indicators such as vibration and noise in the gearbox operating status.

[0006] It can be seen that the current cleaning method is based on wear observation and regular maintenance based on subjective human experience, which makes the wear detection and control measures of the wind turbine gearbox prone to redundancy or deficiency, thus affecting the service life of the gearbox. Summary of the invention

[0007] The main purpose of the present application is to provide an oil filtering control method, device, equipment and storage medium for a fan gearbox to solve the problem that wear detection and control measures for fan gearboxes in the prior art are prone to redundancy or deficiency.

[0008] In order to achieve the above objectives, this application provides the following technical solutions:

[0009] A method for controlling oil filtration of a wind turbine gearbox, wherein the wind turbine gearbox has a wind wheel transmission end, a gear set, and a power generation transmission end that are mechanically connected in sequence, the gear set adopts splash or forced lubrication, wherein the other end of the wind wheel transmission end is mechanically connected to the wind wheel, and the other end of the power generation transmission end is mechanically connected to the generator, and the lubricating oil absorbs iron debris through an electromagnetic filter connected to the gearbox, and the oil filtration control method comprises:

[0010] Step S1, defining a rotation cycle of the wind wheel as a data acquisition cycle;

[0011] Step S2, acquiring historical rotation data of a plurality of the wind wheels and historical adsorption amounts of a plurality of the electromagnetic filters based on the data acquisition cycle;

[0012] Step S3, dividing all historical rotation data into a first training set and a first sample set according to a first preset ratio, and dividing all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio;

[0013] Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model;

[0014] Step S5, outputting the second training set to the preset neural network model and training it to obtain a second prediction model;

[0015] Step S6, outputting the first sample set to the first prediction model to obtain a first prediction value, and outputting the second sample set to the second prediction model to obtain a second prediction value;

[0016] Step S7, solving the linear regression function of all the second predicted values ​​and all the first predicted values ​​by LASSO linear regression;

[0017] Step S8, obtaining the future slope of the linear regression function based on a preset number of prediction steps;

[0018] Step S9, defining the future slope as a proportionality coefficient, and increasing or decreasing the real-time power of the electromagnetic filter according to the proportionality coefficient.

[0019] As a further improvement of the present application, step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, includes:

[0020] Step S41, defining a topological relationship of the preset neural network model, wherein the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected;

[0021] Step S42, outputting the first training set to the input layer, training the first training set for a first preset number of times by using a least squares estimation method, and obtaining a root mean square error between the first sample set and the current training result based on each training;

[0022] Step S43, obtaining a first training minimum value in a root mean square error between the first sample set and all training results;

[0023] Step S44: obtaining a training result corresponding to the first training minimum value as the first prediction model.

[0024] As a further improvement of the present application, step S5, outputting the second training set to the preset neural network model and training it to obtain a second prediction model, includes:

[0025] Step S51, outputting the second training set to the input layer, training the second training set for a second preset number of times by using the least squares estimation method, and obtaining a root mean square error between the second sample set and the current training result based on each training;

[0026] Step S52, obtaining a second training minimum value in the root mean square error between the second sample set and all training results;

[0027] Step S53: Obtain a training result corresponding to the second training minimum value as the second prediction model.

[0028] As a further improvement of the present application, the preset neural network model is represented by formula (1):

[0029] (1);

[0030] in, The preset neural network model; is the input layer and , is the first input nodes, each input node corresponds to one of the first training sets or one of the second training sets, is the first The input nodes to the hidden layer The weights of the input nodes; is the first The bias of the input node; is the transfer function, and In formula (1), the numbers in the brackets of the symbols are the number of layers. For example, the superscript (1) is the first layer, that is, the input layer, and the superscript (1, 2) is the first layer to the second layer, that is, the input layer to the hidden layer.

[0031] As a further improvement of the present application, the root mean square error is characterized by formula (2):

[0032] (2);

[0033] in, is the root mean square error, is the number of the first training sets or the number of the second training sets, For the The true value of the first training set or The true value of the second training set, For the The first training set or The training results after the second training set training is completed.

[0034] As a further improvement of the present application, step S7, solving the linear regression function of all second prediction values ​​and all first prediction values ​​by LASSO linear regression, includes:

[0035] Step S71, define the linear regression relationship between all second prediction values ​​and all first prediction values ​​according to formula (3):

[0036] (3);

[0037] in, For the The second predicted value; is the intercept of the linear regression relationship; For the The linear regression coefficient of the first predicted value; For the The first predicted value; is a random error;

[0038] Step S72, calculating the regression coefficient of the linear regression model by formula (4):

[0039] (4);

[0040] in, for An estimated value of is the matrix of all first prediction values ; For the matrix The transposed matrix of

[0041] Step S73: Substitute the regression coefficient into the linear regression model to obtain the linear regression function.

[0042] As a further improvement of the present application, in step S9, the future slope is defined as a proportional coefficient, and the real-time power of the electromagnetic filter is increased or decreased according to the proportional coefficient, and then, the method includes:

[0043] Step S10, determining whether the real-time power exceeds the maximum output power of the electromagnetic filter, if the real-time power exceeds the maximum output power of the electromagnetic filter, executing step S92;

[0044] Step S20, generating a gearbox wear abnormality signal;

[0045] Step S30, sending the gearbox wear abnormality signal to an external receiving end.

[0046] In order to achieve the above objectives, this application also provides the following technical solutions:

[0047] An oil filter control device for a fan gearbox, the oil filter control device for the fan gearbox is applied to the above-mentioned oil filter control method for the fan gearbox, and the oil filter control device for the fan gearbox comprises:

[0048] A data acquisition cycle definition module, used to define a rotation cycle of the wind wheel as a data acquisition cycle;

[0049] A historical rotation data and historical adsorption amount acquisition module, used for acquiring the historical rotation data of a plurality of the wind wheels and the historical adsorption amounts of a plurality of the electromagnetic filters based on the data acquisition cycle;

[0050] A historical rotation data and historical adsorption amount data division module, used to divide all historical rotation data into a first training set and a first sample set according to a first preset ratio, and divide all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio;

[0051] A historical rotation data training module, used for outputting the first training set to a preset neural network model and performing training to obtain a first prediction model;

[0052] A historical adsorption amount training module, used for outputting the second training set to the preset neural network model and performing training to obtain a second prediction model;

[0053] A rotation data and adsorption amount prediction module, used for outputting the first sample set to the first prediction model to obtain a first prediction value, and outputting the second sample set to the second prediction model to obtain a second prediction value;

[0054] A linear regression function calculation module, used for solving the linear regression function of all second prediction values ​​and all first prediction values ​​by LASSO linear regression;

[0055] A future slope prediction module, used to obtain the future slope of the linear regression function based on a preset prediction step number;

[0056] The electromagnetic filter control module is used to define the future slope as a proportionality coefficient and increase or decrease the real-time power of the electromagnetic filter according to the proportionality coefficient.

[0057] In order to achieve the above objectives, this application also provides the following technical solutions:

[0058] An electronic device comprises a processor and a memory coupled to the processor, wherein the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the oil filtering control method of the fan gearbox as described above is implemented.

[0059] In order to achieve the above objectives, this application also provides the following technical solutions:

[0060] A storage medium stores program instructions, and when the program instructions are executed by a processor, the oil filtering control method for a fan gearbox as described above can be implemented.

[0061] The present application relies on the existing electromagnetic adsorption system and performs real-time detection and electromagnetic control, forming a system from detection to response. The present application defines a rotation cycle of the wind wheel as a data acquisition cycle; based on the data acquisition cycle, the historical rotation data of several wind wheels and the historical adsorption amounts of several electromagnetic filters are acquired; all the historical rotation data are divided into a first training set and a first sample set according to a first preset ratio, and all the historical adsorption amounts are divided into a second training set and a second sample set according to a second preset ratio; the first training set is output to a preset neural network model and trained to obtain a first prediction model; the second training set is output to a preset neural network model and trained to obtain a second prediction model; the first sample set is output to the first prediction model to obtain a first prediction value, and the second sample set is output to the second prediction model to obtain a second prediction value; the linear regression function of all second prediction values ​​and all first prediction values ​​is solved by LASSO linear regression; the future slope of the linear regression function based on a preset prediction step number is acquired; the future slope is defined as a proportional coefficient, and the real-time power of the electromagnetic filter is increased or decreased according to the proportional coefficient. The present application utilizes the fact that the material of wind turbine gears is usually low-carbon steel or medium-carbon alloy structural steel, and learns and trains the amount of debris of the aforementioned material adsorbed by electromagnets in order to pursue more accurate control. The present application has good robustness for wind turbines in different wind scenarios, and obtains different prediction models by analyzing the amount of debris under different wind forces, which is generally understood as one model for one machine. This enables the present application to accurately control the electromagnetic filter of each wind turbine gearbox separately, and can integrate and display the control strategy of each electromagnetic filter at the terminal station, with a high degree of customization and accurate control effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic flow chart of the steps of an embodiment of the oil filtering control method for a fan gearbox of the present application;

[0063] Figure 2 This is a functional module diagram of an embodiment of an oil filter control device for a fan gearbox of the present application;

[0064] Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of the present application;

[0065] Figure 4 This is a schematic diagram of the structure of an embodiment of the storage medium of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0067] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0068] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0069] like Figure 1 As shown, this embodiment provides an embodiment of the oil filtering control method of the wind turbine gearbox. In this embodiment, the wind turbine gearbox has a wind wheel transmission end, a gear set, and a generator transmission end that are mechanically connected in sequence. The gear set adopts splash or forced lubrication, wherein the other end of the wind wheel transmission end is mechanically connected to the wind wheel, and the other end of the generator transmission end is mechanically connected to the generator. The lubricating oil absorbs iron debris through an electromagnetic filter connected to the gearbox.

[0070] Specifically, the structure of the wind turbine gearbox is mainly composed of gears, bearings, housing, lubrication system and cooling system. Among them, the gears, as the core components of the gearbox, are responsible for transmitting speed and torque; the bearings are used to support the gears and reduce friction; the housing is used to support all components inside the gearbox; the lubrication system is used to lubricate the gears and bearings to reduce friction and wear; the cooling system is used to cool the heat inside the gearbox.

[0071] In addition, the design of the wind turbine gearbox also takes into account the speed increase effect between the speed of the wind rotor and the speed required for the generator to generate electricity, and realizes power transmission through the speed increase effect of the gear pair. The lubrication methods of the gearbox include gear oil lubrication and semi-fluid grease lubrication, which are widely used in transportation equipment, chemical equipment, environmental protection machinery and other fields. In a wind turbine generator set, the gearbox is an important mechanical component. Its function is to transfer the power generated by the wind rotor under the action of wind to the generator and make it reach the speed required for power generation. The structural classification of wind turbine gearboxes includes one-stage planetary two-stage parallel shaft (doubly fed), two-stage planetary one-stage parallel shaft (doubly fed), gearbox with main shaft (doubly fed) and compact gearbox (semi-direct drive). These structural types are selected according to factors such as power level, load capacity, volume, weight, etc. to meet different application requirements.

[0072] It is worth noting that the wind gearbox in this embodiment is a conventional application, and the wind gearbox is a mature existing structure, and the detailed structure of the wind gearbox shown in the figure is not repeated in this embodiment.

[0073] Specifically, the oil filter control method includes the following steps:

[0074] Step S1, defining one rotation cycle of the wind wheel as one data acquisition cycle.

[0075] Step S2, acquiring historical rotation data of a plurality of wind wheels and historical adsorption amounts of a plurality of electromagnetic filters based on a data acquisition cycle.

[0076] Preferably, the historical rotation data of several wind wheels can be directly measured by themselves or directly obtained through the wind farm terminal station, and the historical adsorption amount of the electromagnetic filter can also be directly measured by themselves or directly obtained through the wind farm terminal station.

[0077] Preferably, the historical adsorption amount of the electromagnetic filter can be directly measured during each centralized collection, or can be obtained through existing visual detection, weight sensing of the electromagnetic filter itself, and the like.

[0078] Step S3, dividing all historical rotation data into a first training set and a first sample set according to a first preset ratio, and dividing all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio.

[0079] Preferably, the first preset ratio and the second preset ratio can be set to the same 8:2, that is, all historical rotation data are divided into a first training set and a first sample set at a ratio of 8:2, and all historical adsorption amounts are divided into a second training set and a second sample set at a ratio of 8:2.

[0080] Preferably, in actual application, a certain proportion of validation sets are required to verify the accuracy of the model, that is, verification is required after training is completed, and the sample set prediction is started only after the verification is successful. Usually, the image data is divided into training set, validation set, and sample set in a ratio of 70%:15%:15%, that is, 70% of the data is the training set, 15% of the data is the validation set, and 15% of the data is the sample set.

[0081] Step S4: output the first training set to a preset neural network model and perform training to obtain a first prediction model.

[0082] Step S5, outputting the second training set to a preset neural network model and performing training to obtain a second prediction model.

[0083] Preferably, the training model training a neural network usually requires providing a large amount of data, namely a data set; the data set is generally divided into three categories, namely the above-mentioned training set (training set), validation set (validation set) and test set (test set).

[0084] Among them, one epoch is equal to the process of training once with all the samples in the training set. The so-called training once refers to one forward pass and one back pass. When the number of samples in an epoch (i.e., training set) is too large, training once may consume too much time, and it is not necessary to use all the data in the training set for each training. In this case, the entire training set needs to be divided into multiple small blocks, that is, divided into multiple batches for training. An epoch consists of one or more batches, where a batch is a part of the training set. Each training process only uses a part of the data, i.e., a batch. The process of training a batch is an iteration.

[0085] Preferably, the neural network training specifically includes a perceptron, which is composed of two layers of neurons. The input layer receives external input signals and then transmits them to the output layer. The output layer is MP neurons, and the step function is .

[0086] Preferably, given a training data set, the weights ( =1,2,...,n) and training bias It can be obtained through learning, It can be understood as the weight corresponding to a fixed value of -1,0. .

[0087] It should be noted that the step function here has no interchangeable symbolic meaning with other formulas in the embodiment. This step function is only used for principle explanation and does not participate in the calculation of other formulas.

[0088] Preferably, in this embodiment, the number of neural network training times can be set to 1000 times.

[0089] Preferably, the learning rate from the 1st to the 500th epoch can be set to 0.01, the learning rate from the 501st to the 750th epoch can be set to 0.001, and the learning rate from the 751st to the 1000th epoch can be set to 0.0001.

[0090] It can be understood that the neural network training of this embodiment mainly includes the following ideas:

[0091] ① Initialize the weights and bias items in the network.

[0092] Initializing parameter values ​​(output unit weights, bias terms and hidden unit weights, bias terms are all model parameters) is to activate forward propagation, obtain the output value of each layer element, and then obtain the value of the loss function.

[0093] ②Activate forward propagation to obtain the output value of each layer and the expected value of the loss function of each layer.

[0094] ③According to the loss function, calculate the error term of the output unit and the error term of the hidden unit.

[0095] Calculate various errors, calculate the gradient of the parameters with respect to the loss function, or calculate partial derivatives according to the chain rule of calculus. For partial derivatives of vectors or matrices in a composite function, the partial derivative of the function inside the composite function is always multiplied on the left; for partial derivatives of scalars in a composite function, the partial derivative of the function inside the composite function can be multiplied on the left or on the right.

[0096] ④Update the weights and bias items in the neural network.

[0097] ⑤ Repeat ②~④ until the loss function is less than the preset bias or the number of iterations is used up, and the parameters output at this time are the current optimal parameters.

[0098] Step S6: output the first sample set to the first prediction model to obtain a first prediction value, and output the second sample set to the second prediction model to obtain a second prediction value.

[0099] Step S7, solving the linear regression function of all the second prediction values ​​and all the first prediction values ​​by LASSO linear regression.

[0100] Step S8, obtaining the future slope of the linear regression function based on a preset number of prediction steps.

[0101] Step S9, defining the future slope as a proportional coefficient, and increasing or decreasing the real-time power of the electromagnetic filter according to the proportional coefficient.

[0102] Preferably, the setting of the proportionality coefficient can normalize the future slope, and the value obtained after the processing is the proportionality coefficient.

[0103] Furthermore, in step S4, the first training set is output to a preset neural network model and trained to obtain a first prediction model, including:

[0104] Step S41, defining the topological relationship of the preset neural network model, the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected.

[0105] Step S42, outputting the first training set to the input layer, training the first training set for a first preset number of times using the least squares estimation method, and obtaining a root mean square error between the first sample set and the current training result based on each training.

[0106] Step S43, obtaining a first training minimum value in a root mean square error between the first sample set and all training results.

[0107] Step S44, obtaining a training result corresponding to the first training minimum value as a first prediction model.

[0108] Preferably, the least squares estimation method is a nonlinear least squares method, which is used to solve nonlinear optimization problems, such as the neural network model in this embodiment. The method uses the gradient information of the objective function to adjust the parameters in each iteration to find the best model parameter estimate. The method uses an adjustment factor called the least squares estimation method, which balances the gradient descent and Gauss-Newton method according to the current iteration step. In each iteration, the method calculates a parameter increment and then decides whether to accept the parameter increment by comparing the size of the current model residual and the model residual estimated using the new parameters.

[0109] Preferably, the iterative step size of the least squares estimation method is given by Representation.

[0110] in, is the iterative step length of the least squares estimation method, is the current model residual, is the Jacobian matrix of the current model residual, is the transposed matrix of the Jacobian matrix of the current model residual, is the above adjustment factor, is the identity matrix, is the covariance matrix.

[0111] Preferably, the order of the identity matrix is ​​the same as the order of the Jacobian matrix.

[0112] Preferably, when hour, is a positive number, Prefer the Gauss-Newton method; when hour, Tends to balance gradient descent.

[0113] It should be noted that the symbolic meanings of the preferred contents here are not interchangeable with the symbolic meanings of other formulas in the embodiment, and the least squares estimation method here is only used for principle explanation and does not participate in the calculation of other formulas.

[0114] Furthermore, in step S5, the second training set is output to a preset neural network model and trained to obtain a second prediction model, including:

[0115] Step S51, output the second training set to the input layer, train the second training set for a second preset number of times using the least squares estimation method, and obtain the root mean square error between the second sample set and the current training result based on each training.

[0116] Step S52, obtaining a second training minimum value in the root mean square error between the second sample set and all training results.

[0117] Step S53: Obtain a training result corresponding to the second training minimum value as a second prediction model.

[0118] Furthermore, the preset neural network model is represented by formula (1):

[0119] (1).

[0120] in, To preset the neural network model; is the input layer and , The input layer input nodes, each input node corresponds to a first training set or a second training set, The input layer The first input node to the hidden layer The weights of the input nodes; is connected to the hidden layer The bias of the input node; is the transfer function, and In formula (1), the numbers in the brackets of the symbols are the number of layers. For example, the superscript (1) is the first layer, that is, the input layer, and the superscript (1, 2) is the first layer to the second layer, that is, the input layer to the hidden layer.

[0121] Furthermore, the root mean square error is expressed by formula (2):

[0122] (2).

[0123] in, is the root mean square error, is the number of the first training set or the number of the second training set, For the The true value of the first training set or The true value of the second training set, For the The first training set or The training results after the second training set training is completed.

[0124] Further, step S7, solving the linear regression function of all second prediction values ​​and all first prediction values ​​by LASSO linear regression, includes:

[0125] Step S71, define the linear regression relationship between all second prediction values ​​and all first prediction values ​​according to formula (3):

[0126] (3).

[0127] in, For the The second predicted value; is the intercept of the linear regression relationship; For the The linear regression coefficient of the first predicted value; For the The first predicted value; is a random error.

[0128] Step S72, calculate the regression coefficient of the linear regression model by formula (4):

[0129] (4).

[0130] in, for An estimated value of is the matrix of all first prediction values ; For the matrix The transposed matrix of .

[0131] Step S73, substituting the regression coefficient into the linear regression model to obtain a linear regression function.

[0132] Preferably, the square of the residual Test the significance of the linear regression function. If , it means that the linear regression function is more significant; if , then the significance of the linear regression function is general; if , it means that the significance of the linear regression function is poor, and the significance of the linear regression function is linearly positively correlated with the square of the residual.

[0133] Further, in step S9, the future slope is defined as a proportional coefficient, and the real-time power of the electromagnetic filter is increased or decreased according to the proportional coefficient, and then, the method includes:

[0134] Step S10, determining whether the real-time power exceeds the maximum output power of the electromagnetic filter. If the real-time power exceeds the maximum output power of the electromagnetic filter, executing step S92.

[0135] Step S20, generating a gearbox wear abnormality signal.

[0136] Step S30, sending a gearbox wear abnormality signal to an external receiving end.

[0137] Preferably, the real-time wind force in the area where the wind turbine is located, the real-time power of the electromagnetic filter, and the real-time power generation of the wind turbine can also be integrated into the terminal station for unified display, wherein the measurement of each real-time data can be achieved through existing equipment.

[0138] The present embodiment relies on the existing electromagnetic adsorption system and performs real-time monitoring and electromagnetic control, forming a system from detection to response. The present embodiment defines a rotation cycle of the wind wheel as a data acquisition cycle; based on the data acquisition cycle, the historical rotation data of several wind wheels and the historical adsorption amounts of several electromagnetic filters are acquired; all the historical rotation data are divided into a first training set and a first sample set according to a first preset ratio, and all the historical adsorption amounts are divided into a second training set and a second sample set according to a second preset ratio; the first training set is output to a preset neural network model and trained to obtain a first prediction model; the second training set is output to a preset neural network model and trained to obtain a second prediction model; the first sample set is output to the first prediction model to obtain a first prediction value, and the second sample set is output to the second prediction model to obtain a second prediction value; the linear regression function of all the second prediction values ​​and all the first prediction values ​​is solved by LASSO linear regression; the future slope of the linear regression function based on the preset prediction step number is acquired; the future slope is defined as a proportional coefficient, and the real-time power of the electromagnetic filter is increased or decreased according to the proportional coefficient. This embodiment utilizes the fact that the material of the wind turbine gear is usually low-carbon steel or medium-carbon alloy structural steel. The amount of debris of the aforementioned material adsorbed by the electromagnet is learned and trained to pursue more accurate control. This embodiment has good robustness for wind turbines in different wind scenarios. Different prediction models are obtained by analyzing the amount of debris under different wind forces, which is generally understood as one model for one machine. This enables this embodiment to accurately control the electromagnetic filter of each wind turbine gearbox separately, and the control strategy of each electromagnetic filter can be integrated and displayed at the terminal station, with a high degree of customization and accurate control effect.

[0139] like Figure 2 As shown, this embodiment provides an embodiment of an oil filter control device for a fan gearbox. In this embodiment, the oil filter control device is applied to the oil filter control method for a fan gearbox as in the above embodiment. The oil filter control device includes a data acquisition cycle definition module 1, a historical rotation data and historical adsorption amount acquisition module 2, a historical rotation data and historical adsorption amount data division module 3, a historical rotation data training module 4, a historical adsorption amount training module 5, a rotation data and adsorption amount prediction module 6, a linear regression function calculation module 7, a future slope prediction module 8, and an electromagnetic filter control module 9, which are electrically connected in sequence.

[0140] Among them, the data acquisition cycle definition module 1 is used to define a rotation cycle of the wind wheel as a data acquisition cycle; the historical rotation data and historical adsorption amount acquisition module 2 is used to acquire the historical rotation data of several wind wheels and the historical adsorption amounts of several electromagnetic filters based on the data acquisition cycle; the historical rotation data and historical adsorption amount data division module 3 is used to divide all historical rotation data into a first training set and a first sample set according to a first preset ratio, and divide all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio; the historical rotation data training module 4 is used to output the first training set to a preset neural network model and train it to obtain a first prediction model; the historical adsorption amount ... The training module 5 is used to output the second training set to the preset neural network model and perform training to obtain the second prediction model; the rotation data and adsorption amount prediction module 6 is used to output the first sample set to the first prediction model to obtain the first prediction value, and output the second sample set to the second prediction model to obtain the second prediction value; the linear regression function calculation module 7 is used to solve the linear regression function of all second prediction values ​​and all first prediction values ​​through LASSO linear regression; the future slope prediction module 8 is used to obtain the future slope of the linear regression function based on a preset prediction step number; the electromagnetic filter control module 9 is used to define the future slope as a proportional coefficient, and increase or decrease the real-time power of the electromagnetic filter according to the proportional coefficient.

[0141] Furthermore, the historical rotation data training module 4 specifically includes a first historical rotation data training sub-module, a second historical rotation data training sub-module, a third historical rotation data training sub-module, and a fourth historical rotation data training sub-module, which are electrically connected in sequence; the first historical rotation data training sub-module is electrically connected to the historical rotation data and historical adsorption amount data division module, and the fourth historical rotation data training sub-module is electrically connected to the historical adsorption amount training module.

[0142] Among them, the first historical rotation data training submodule is used to define the topological relationship of the preset neural network model, and the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected; the second historical rotation data training submodule is used to output the first training set to the input layer, train the first training set for a first preset number of times through the least squares estimation method, and obtain the root mean square error between the first sample set and the current training result based on each training; the third historical rotation data training submodule is used to obtain the first training minimum value in the root mean square error between the first sample set and all training results; the fourth historical rotation data training submodule is used to obtain the training result corresponding to the first training minimum value as the first prediction model.

[0143] Furthermore, the first historical rotation data training submodule is equipped with a preset neural network model represented by formula (1):

[0144] (1).

[0145] in, To preset the neural network model; is the input layer and , The input layer input nodes, each input node corresponds to a first training set or a second training set, The input layer The first input node to the hidden layer The weights of the input nodes; is connected to the hidden layer The bias of the input node; is the transfer function, and In formula (1), the numbers in the brackets of the symbols are the number of layers. For example, the superscript (1) is the first layer, that is, the input layer, and the superscript (1, 2) is the first layer to the second layer, that is, the input layer to the hidden layer.

[0146] Furthermore, the second historical rotation data training submodule is equipped with the root mean square error represented by formula (2):

[0147] (2).

[0148] in, is the root mean square error, is the number of the first training set or the number of the second training set, For the The true value of the first training set or The true value of the second training set, For the The first training set or The training results after the second training set training is completed.

[0149] Furthermore, the historical adsorption amount training module 5 specifically includes a first historical adsorption amount training submodule, a second historical adsorption amount training submodule, and a third historical adsorption amount training submodule, which are electrically connected in sequence; the first historical adsorption amount training submodule is electrically connected to the fourth historical rotation data training submodule, and the third historical adsorption amount training submodule is electrically connected to the rotation data and adsorption amount prediction module.

[0150] Among them, the first historical adsorption amount training submodule is used to output the second training set to the input layer, train the second training set for a second preset number of times through the least squares estimation method, and obtain the root mean square error between the second sample set and the current training result based on each training; the second historical adsorption amount training submodule is used to obtain the second training minimum value in the root mean square error between the second sample set and all training results; the third historical adsorption amount training submodule is used to obtain the training result corresponding to the second training minimum value as the second prediction model.

[0151] Furthermore, the linear regression function calculation module 7 specifically includes a first linear regression function calculation submodule, a second linear regression function calculation submodule, and a third linear regression function calculation submodule, which are electrically connected in sequence; the first linear regression function calculation submodule is electrically connected to the rotation data and adsorption amount prediction module, and the third linear regression function calculation submodule is electrically connected to the future slope prediction module.

[0152] The first linear regression function calculation submodule is used to define the linear regression relationship between all second prediction values ​​and all first prediction values ​​according to formula (3):

[0153] (3).

[0154] in, For the The second predicted value; is the intercept of the linear regression relationship; For the The linear regression coefficient of the first predicted value; For the The first predicted value; is a random error.

[0155] The second linear regression function calculation submodule is used to calculate the regression coefficient of the linear regression model through formula (4):

[0156] (4).

[0157] in, for An estimated value of is the matrix of all first prediction values ; For the matrix The transposed matrix of .

[0158] The third linear regression function calculation submodule is used to substitute the regression coefficient into the linear regression model to obtain the linear regression function.

[0159] Furthermore, the oil filter control device also includes an electromagnetic filter real-time power judgment module, a gear box wear abnormality signal generating module, and a gear box wear abnormality signal sending module which are electrically connected in sequence; the electromagnetic filter real-time power judgment module is electrically connected to the electromagnetic filter control module.

[0160] Among them, the electromagnetic filter real-time power judgment module is used to judge whether the real-time power exceeds the maximum output power of the electromagnetic filter; the gear box wear abnormality signal generation module is used to generate a gear box wear abnormality signal if the real-time power exceeds the maximum output power of the electromagnetic filter; the gear box wear abnormality signal sending module is used to send the gear box wear abnormality signal to an external receiving end.

[0161] It should be noted that this embodiment is a functional module embodiment based on the above method embodiment. Additional contents such as the preference, expansion, limitation, and example illustration of this embodiment can be found in the above method embodiment, and will not be repeated in this embodiment.

[0162] The present embodiment relies on the existing electromagnetic adsorption system and performs real-time monitoring and electromagnetic control, forming a system from detection to response. The present embodiment defines a rotation cycle of the wind wheel as a data acquisition cycle; based on the data acquisition cycle, the historical rotation data of several wind wheels and the historical adsorption amounts of several electromagnetic filters are acquired; all the historical rotation data are divided into a first training set and a first sample set according to a first preset ratio, and all the historical adsorption amounts are divided into a second training set and a second sample set according to a second preset ratio; the first training set is output to a preset neural network model and trained to obtain a first prediction model; the second training set is output to a preset neural network model and trained to obtain a second prediction model; the first sample set is output to the first prediction model to obtain a first prediction value, and the second sample set is output to the second prediction model to obtain a second prediction value; the linear regression function of all the second prediction values ​​and all the first prediction values ​​is solved by LASSO linear regression; the future slope of the linear regression function based on the preset prediction step number is acquired; the future slope is defined as a proportional coefficient, and the real-time power of the electromagnetic filter is increased or decreased according to the proportional coefficient. This embodiment utilizes the fact that the material of the wind turbine gear is usually low-carbon steel or medium-carbon alloy structural steel. The amount of debris of the aforementioned material adsorbed by the electromagnet is learned and trained to pursue more accurate control. This embodiment has good robustness for wind turbines in different wind scenarios. Different prediction models are obtained by analyzing the amount of debris under different wind forces, which is generally understood as one model for one machine. This enables this embodiment to accurately control the electromagnetic filter of each wind turbine gearbox separately, and the control strategy of each electromagnetic filter can be integrated and displayed at the terminal station, with a high degree of customization and accurate control effect.

[0163] like Figure 3As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 10 includes a processor 101 and a memory 102 coupled to the processor 101 .

[0164] The memory 102 stores program instructions for implementing the oil filtering control method for the wind turbine gearbox according to any of the above embodiments.

[0165] The processor 101 is used to execute program instructions stored in the memory 102 to perform oil filtering control of the fan gearbox.

[0166] The processor 101 may also be referred to as a CPU (Central Processing Unit). The processor 101 may be an integrated circuit chip having data processing capabilities. The processor 101 may also be a general-purpose processor, a digital data processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0167] Further, Figure 4 The schematic diagram of the structure of the storage medium of an embodiment of the present application is that the storage medium 11 of the embodiment of the present application stores program instructions 111 that can implement all the above methods, wherein the program instructions 111 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0168] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0169] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0170] The specific implementation methods of the present application are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the present application is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A method for controlling oil filtration in a fan gearbox, wherein the fan gearbox has a wind wheel transmission end, a gear set, and a power generation transmission end that are mechanically connected in sequence, and the gear set adopts splash or forced lubrication, wherein: The other end of the wind wheel transmission end is mechanically connected to the wind wheel, and the other end of the power generation transmission end is mechanically connected to the generator. The lubricating oil absorbs iron debris through an electromagnetic filter connected to the gear box. It is characterized in that the oil filtering control method includes: Step S1, defining a rotation cycle of the wind wheel as a data acquisition cycle; Step S2, acquiring historical rotation data of a plurality of the wind wheels and historical adsorption amounts of a plurality of the electromagnetic filters based on the data acquisition cycle; Step S3, dividing all historical rotation data into a first training set and a first sample set according to a first preset ratio, and dividing all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio; Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model; Step S5, outputting the second training set to the preset neural network model and training it to obtain a second prediction model; Step S6, outputting the first sample set to the first prediction model to obtain a first prediction value, and outputting the second sample set to the second prediction model to obtain a second prediction value; Step S7, solving the linear regression function of all the second predicted values ​​and all the first predicted values ​​by LASSO linear regression; Step S8, obtaining the future slope of the linear regression function based on a preset number of prediction steps; Step S9, defining the future slope as a proportionality coefficient, and increasing or decreasing the real-time power of the electromagnetic filter according to the proportionality coefficient.

2. The oil filtering control method for a fan gearbox according to claim 1, characterized in that: Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, including: Step S41, defining a topological relationship of the preset neural network model, wherein the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected; Step S42, outputting the first training set to the input layer, training the first training set for a first preset number of times by using a least squares estimation method, and obtaining a root mean square error between the first sample set and the current training result based on each training; Step S43, obtaining a first training minimum value in a root mean square error between the first sample set and all training results; Step S44: obtaining a training result corresponding to the first training minimum value as the first prediction model.

3. The oil filtering control method for a fan gearbox according to claim 2, characterized in that: Step S5, outputting the second training set to the preset neural network model and training it to obtain a second prediction model, including: Step S51, outputting the second training set to the input layer, training the second training set for a second preset number of times by using the least squares estimation method, and obtaining a root mean square error between the second sample set and the current training result based on each training; Step S52, obtaining a second training minimum value in the root mean square error between the second sample set and all training results; Step S53: Obtain a training result corresponding to the second training minimum value as the second prediction model.

4. The oil filtering control method for a fan gearbox according to claim 3, characterized in that: The preset neural network model is represented by formula (1): （1）； in, The preset neural network model; is the input layer and , is the first input nodes, each input node corresponds to one of the first training sets or one of the second training sets, is the first The input nodes to the hidden layer The weights of the input nodes; is the first The bias of the input node; is the transfer function, and In formula (1), the numbers in the brackets of the symbols are the number of layers. For example, the superscript (1) is the first layer, that is, the input layer, and the superscript (1, 2) is the first layer to the second layer, that is, the input layer to the hidden layer.

5. The oil filtering control method for a fan gearbox according to claim 3, characterized in that: The root mean square error is represented by formula (2): （2）； in, is the root mean square error, is the number of the first training sets or the number of the second training sets, For the The true value of the first training set or The true value of the second training set, For the The first training set or The training results after the second training set training is completed.

6. The oil filtering control method for a fan gearbox according to claim 1, characterized in that: Step S7, solving the linear regression function of all second prediction values ​​and all first prediction values ​​by LASSO linear regression, including: Step S71, define the linear regression relationship between all second prediction values ​​and all first prediction values ​​according to formula (3): （3）； in, For the The second predicted value; is the intercept of the linear regression relationship; For the The linear regression coefficient of the first predicted value; For the The first predicted value; is a random error; Step S72, calculating the regression coefficient of the linear regression model by formula (4): （4）； in, for An estimated value of is the matrix of all first prediction values ; For the matrix The transposed matrix of Step S73: Substitute the regression coefficient into the linear regression model to obtain the linear regression function.

7. The oil filtering control method for a fan gearbox according to claim 1, characterized in that: Step S9, defining the future slope as a proportionality coefficient, and increasing or decreasing the real-time power of the electromagnetic filter according to the proportionality coefficient, and then comprising: Step S10, determining whether the real-time power exceeds the maximum output power of the electromagnetic filter, if the real-time power exceeds the maximum output power of the electromagnetic filter, executing step S92; Step S20, generating a gearbox wear abnormality signal; Step S30, sending the gearbox wear abnormality signal to an external receiving end.

8. An oil filter control device for a fan gearbox, the oil filter control device for a fan gearbox being applied to the oil filter control method for a fan gearbox as claimed in any one of claims 1 to 7, characterized in that: The oil filter control device of the fan gear box comprises: A data acquisition cycle definition module, used to define a rotation cycle of the wind wheel as a data acquisition cycle; A historical rotation data and historical adsorption amount acquisition module, used for acquiring the historical rotation data of a plurality of the wind wheels and the historical adsorption amounts of a plurality of the electromagnetic filters based on the data acquisition cycle; A historical rotation data and historical adsorption amount data division module, used to divide all historical rotation data into a first training set and a first sample set according to a first preset ratio, and divide all historical adsorption amounts into a second training set and a second sample set according to a second preset ratio; A historical rotation data training module, used for outputting the first training set to a preset neural network model and performing training to obtain a first prediction model; A historical adsorption amount training module, used for outputting the second training set to the preset neural network model and performing training to obtain a second prediction model; A rotation data and adsorption amount prediction module, used for outputting the first sample set to the first prediction model to obtain a first prediction value, and outputting the second sample set to the second prediction model to obtain a second prediction value; A linear regression function calculation module, used for solving the linear regression function of all second prediction values ​​and all first prediction values ​​by LASSO linear regression; A future slope prediction module, used to obtain the future slope of the linear regression function based on a preset prediction step number; The electromagnetic filter control module is used to define the future slope as a proportionality coefficient and increase or decrease the real-time power of the electromagnetic filter according to the proportionality coefficient.

9. An electronic device, characterized in that: It includes a processor and a memory coupled to the processor, wherein the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the oil filtering control method for the fan gearbox as described in any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium stores program instructions, and when the program instructions are executed by the processor, the oil filtering control method for the wind turbine gearbox according to any one of claims 1 to 7 can be implemented.

Images