A Wind Turbine Power Prediction Method and System Based on Bayesian Deep Learning

Abstract

The present invention discloses a method and system for predicting the power of a wind turbine based on Bayesian deep learning. The method includes: screening and considering the hybrid sample features related to the wind turbine power prediction under icing conditions according to a physical model; using the hybrid sample features as inputs to construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network; generating an expert experience sample database for the wind turbine power output according to the physical model, and using the expert experience sample database for the wind turbine power output as a prior to pre-train the constructed adaptive wind power prediction model, and collecting the historical operation data of the wind farm as a posterior, and regularly re-training the adaptive wind power prediction model based on a stochastic variational inference method to obtain a wind turbine power prediction model; inputting the real-time operation data of the wind turbine into the wind turbine power prediction model, and outputting the wind turbine power prediction result. The comprehensive utilization of expert knowledge and historical operation data greatly improves the accuracy of the wind power prediction under icing conditions of the wind turbine.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power prediction, and particularly relates to a method and system for predicting the power of a wind turbine based on Bayesian deep learning. Background Art

[0002] When a wind turbine operates in cold or high-altitude areas, icing is likely to occur on the blade surface, resulting in a decrease in aerodynamic performance, which in turn causes a significant reduction in power generation, an increase in structural load, and even safety accidents. Wind turbine icing not only affects its own operating performance but also may have an adverse impact on the stability of the power grid. Therefore, accurately predicting the power output of a wind turbine under icing conditions is of great significance for ensuring the safe and stable operation of a wind farm and power grid dispatching.

[0003] Currently, the methods for predicting wind turbine power mainly include physical modeling methods, statistical methods, and intelligent algorithms based on machine learning. Physical modeling methods are based on the aerodynamic principles of wind turbines and consider factors such as wind speed and direction for power calculation. However, under icing conditions, the performance of the blades changes, and it is difficult for the model to accurately describe it. Statistical methods use historical data for modeling, but when the number of icing samples is small, the reliability of the model decreases. Intelligent algorithms based on machine learning can improve the prediction accuracy to a certain extent by learning a large amount of historical data. However, when the icing data is scarce, the model is prone to overfitting and cannot quantify the uncertainty of the prediction results.

[0004] Traditional machine learning methods often ignore the uncertainty of the model prediction results when dealing with the problem of predicting wind turbine power under icing conditions, which may lead to incorrect dispatching decisions. In addition, wind turbine icing mainly occurs in winter, and the number of data samples is small. Especially for newly built wind farms, directly applying traditional deep learning models may have the risk of overfitting, insufficient model generalization ability, and inability to provide a credibility assessment of the prediction results. Summary of the Invention

[0005] The present invention provides a method and system for predicting the power of a wind turbine based on Bayesian deep learning to solve the technical problem that the change in aerodynamic performance caused by blade icing makes it difficult for traditional physical modeling methods to accurately describe the characteristics of the ice-covered blades, resulting in a decrease in power prediction accuracy.

[0006] In a first aspect, the present invention provides a method for predicting the power of a wind turbine based on Bayesian deep learning, including:

[0007] Construct a physical model of the influence of the icing process on the output of the wind turbine;

[0008] Determine the sample characteristics of the measurement system according to real-time measurement data, and then screen out the mixed sample characteristics related to wind turbine power prediction considering icing conditions according to the physical model;

[0009] Taking the mixed sample features as input, an adaptive wind power prediction model based on a Bayesian multi-layer neural network is constructed;

[0010] According to the physical model, an expert experience sample database of fan power output is generated, and the expert experience sample database of fan power output is used as a prior to pre-train the constructed adaptive wind power prediction model. The historical operation data of the wind farm is collected as a posterior, and the adaptive wind power prediction model is regularly re-trained based on the stochastic variational inference method to obtain a fan power prediction model;

[0011] Obtain the real-time operation data of the fan, and input the real-time operation data of the fan into the fan power prediction model. The fan power prediction model outputs a fan power prediction result.

[0012] In a second aspect, the present invention provides a fan power prediction system based on Bayesian deep learning, including:

[0013] A first construction module configured to construct a physical model of the influence of the icing process on the output of the fan;

[0014] A screening module configured to determine the sample features of the measurement system according to the real-time measurement data, and then screen the mixed sample features related to the fan power prediction under the icing condition according to the physical model;

[0015] A second construction module configured to take the mixed sample features as input and construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network;

[0016] A training module configured to generate an expert experience sample database of fan power output according to the physical model, and use the expert experience sample database of fan power output as a prior to pre-train the constructed adaptive wind power prediction model. The historical operation data of the wind farm is collected as a posterior, and the adaptive wind power prediction model is regularly re-trained based on the stochastic variational inference method to obtain a fan power prediction model;

[0017] An output module configured to obtain the real-time operation data of the fan, and input the real-time operation data of the fan into the fan power prediction model. The fan power prediction model outputs a fan power prediction result.

[0018] In a third aspect, an electronic device is provided, which includes: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the steps of the fan power prediction method based on Bayesian deep learning according to any embodiment of the present invention.

[0019] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program instructions are executed by a processor, the processor is caused to execute the steps of the method for predicting wind turbine power based on Bayesian deep learning according to any embodiment of the present invention.

[0020] In the method and system for predicting wind turbine power based on Bayesian deep learning of the present application, the physical models of wind turbine icing, natural ice melting of blades, and wind turbine output are used as prior knowledge, and the historical operation data of wind turbines collected by a wind farm are used as posterior. A probability mapping relationship between wind turbine operation data and wind power under icing conditions is constructed using a Bayesian multi-layer neural network. Finally, the neural network parameters are trained by combining Bayesian statistical inference theory and stochastic variational inference methods, thereby comprehensively using expert knowledge and historical operation data to greatly improve the accuracy of wind power prediction under wind turbine icing conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following-described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 is a flowchart of a method for predicting wind turbine power based on Bayesian deep learning according to an embodiment of the present invention;

[0023] Figure 2 is a structural block diagram of a system for predicting wind turbine power based on Bayesian deep learning according to an embodiment of the present invention;

[0024] Figure 3 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0026] Please refer to Figure 1 , which shows a flowchart of a method for predicting wind turbine power based on Bayesian deep learning of the present application.

[0027] As shown in Figure 1As shown, the fan power prediction method based on Bayesian deep learning specifically includes the following steps:

[0028] Step S101, construct a physical model of the icing process affecting the fan output.

[0029] In this step, construct a fan power output model based on the wind energy conversion theory, a fan blade icing model based on the FENSAP software, and a natural ice removal model of the fan blade based on the convective radiation effect and blade rotation ice throwing. Among them, the expression of the fan power output model is:

[0030] ,

[0031] ,

[0032] ,

[0033] ,

[0034] ,

[0035] ,

[0036] In the formula, is the fan power output value, is the mechanical loss coefficient, is the electrical loss coefficient, is the rotation coefficient, is the yaw error, is the air density, is the rotor swept area, is the inflow wind speed, is the distance from the blade element to the hub center, is the fan blade profile, is the tip, is the wake effect, is the tip speed ratio, is the tip shape factor, is the lift-drag ratio, is the lift per unit length of the airfoil along the span, is the drag per unit length of the airfoil along the span, is the inflow angle of the local combined wind direction blade element, represents the local inflow angle of the blade, represents related to the number of blades , the tip speed ratio and the tip shape factor is related, is the rotor radius, is the number of blades.

[0037] The expression of the airflow in the ice accretion model of the wind turbine blade is:

[0038] ,

[0039] In the formula, is the laminar kinematic viscosity, is the dimensionless laminar viscosity, , are both cut-off functions, is the vorticity, is the Reynolds number at infinity, , , and are all closure coefficients, is the closure function, is the distance from the wall, is the Stefan-Boltzmann constant, is the Hamiltonian operator, is the velocity change.

[0040] The Euler model is used for the impact calculation, and the expressions of the continuity and momentum equations related to the droplets are:

[0041]

[0042] ,

[0043] In the formula, is the droplet volume fraction, is the airflow rate, is the relative velocity of the droplet, is the droplet drag coefficient, is the droplet Reynolds number, is the droplet inertia parameter, is the air density, is the density of water, is the Froude number, is the acceleration due to gravity;

[0044] According to the droplet collection efficiency Solve the icing process, and the expressions of the partial differential equations of mass conservation and energy conservation obtained are:

[0045] ,

[0046] In the formula, is the density of the water film, is the height, is the symbol of divergence, is the flow velocity, is the reference velocity, is the liquid water content, is the evaporation rate of liquid water, is the freezing rate of liquid water;

[0047] ,

[0048] wherein, is the specific heat capacity of the water film, is the latent heat of evaporation, is the latent heat of sublimation, is the latent heat of fusion, is the specific heat capacity of ice, is the solid heat dissipation coefficient, is the reference temperature, is the reference droplet temperature in Kelvin, is the reference droplet temperature in degrees Celsius, is the water-ice interface temperature in degrees Celsius, is the heat transferred by convection, is the water-ice interface temperature in Kelvin;

[0049] ,

[0050] wherein, is the normal vector of the blade surface;

[0051] The expression for the ice melting rate of the natural ice removal model of the wind turbine blade is:

[0052] ,

[0053] wherein, is the diameter of the frozen object, is the convective exchange coefficient, , are the air temperature and the ice surface temperature respectively, is the radiation constant, is the specific heat capacity of ice, is the latent heat of ice, is the density of ice;

[0054] The expression for the ice thickness at ice shedding is:

[0055] ,

[0056] wherein, is the ice thickness at ice shedding, is the viscous shear stress, is the wind turbine diameter, is the blade rotation rate.

[0057] In step S102, determine the sample characteristics of the measurement system according to the real-time measurement data, and then screen the hybrid sample characteristics related to the wind turbine power prediction under icing conditions according to the physical model.

[0058] In this step, the expression of the hybrid sample characteristics is:

[0059] ,

[0060] In the formula, is the air flow rate, is the blade rotation rate, is the tip speed ratio, is the pitch angle, is the number of blades, is the rotor radius, is the yaw error, is the air density, is the incoming wind speed, is the liquid water content, is the droplet volume fraction, is the relative velocity of the droplet, is the water-ice interface temperature in degrees Celsius, is the water-ice interface temperature in Kelvin, is the convective exchange coefficient, is the air surface, is the ice surface temperature, is the density of ice.

[0061] In step S103, use the hybrid sample characteristics as the input to construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network.

[0062] In this step, the adaptive wind power prediction model includes:

[0063] A variational Bayesian gated recurrent unit, which is used to process the time-series observation historical data in SCADA, where the time-series observation historical data in SCADA contains hybrid sample characteristics;

[0064] A variational Bayesian neuron, the basic unit of which is an artificial neuron, and the expression is:

[0065] ,

[0066] In the formula, is the deviation, is the output of the variational Bayesian neuron, is the activation function, is the weight assigned to the network connection, is the input of the variational Bayesian neuron, is the number of inputs.

[0067] GRU unit, and the expression of the GRU unit is:

[0068] ,

[0069] ,

[0070] ,

[0071] ,

[0072] ,

[0073] In the formula, is the reset gate, is the update gate, 、 are the hidden state of the GRU unit at the t-th iteration and the hidden state of the GRU unit at the (t - 1)-th iteration respectively, is the Hadamard product, is the sigmoid function, is the candidate hidden state, is the output of the GRU unit, is the input of the GRU unit at time t, is the weight of the previous hidden state in the output of the reset gate, is the weight of the current input in the output of the reset gate, is the weight of the previous hidden state in the output of the update gate, is the weight of the current input in the output of the update gate, is the weight of the Hadamard product of the previous hidden state and the reset gate in the candidate hidden state, is the weight of the current input in the candidate hidden state, is the hyperbolic tangent function.

[0074] Step S104: Generate an expert experience sample database of the fan power output according to the physical model, and use the expert experience sample database of the fan power output as the adaptive wind power prediction model constructed by prior pre-training, and collect the historical operation data of the wind farm as the posterior. Based on the stochastic variational inference method, the adaptive wind power prediction model is retrained regularly to obtain the fan power prediction model.

[0075] In this step, the wind turbine geographical data, meteorological forecast data, and historical operation data observed by SCADA are input into the physical model to calculate the wind turbine power, and an expert experience sample database of wind turbine power output is obtained; the geographical information data includes longitude, latitude, and altitude data. The meteorological prediction data includes: time, temperature, atmospheric pressure, relative humidity, weather type, wind speed and direction, and air density at different heights of 10m, 50m, and 100m.

[0076] Pre-train the constructed adaptive wind power prediction model based on the expert experience sample database, where The input of the adaptive wind power prediction model at time is:

[0077] ,

[0078] In the formula, is the historical operation data at time is the wind turbine geographical data, is the meteorological forecast data at time

[0079] The output of the adaptive wind power prediction model at time is the power output of the wind turbine at time , and the expression is:

[0080] ,

[0081] In the formula, is the wind power output of the w-th wind turbine at time is the wind power output of the -th wind turbine at time

[0082] It should be noted that based on the Bayesian statistical inference theory for model re-training, the above-mentioned wind turbine power prediction model based on the Bayesian multi-layer neural network can be denoted as:

[0083] ,

[0084] In the formula, is the set of random parameters, is the probability density function of the corresponding distribution, is the set of variational parameters, is the sample data set.

[0085] According to the Bayesian theory, the posterior inference problem can be expressed as:

[0086] ,

[0087] The SVI introduces a series of distributions and determines the detailed type and initial parameters of the distribution by using prior knowledge and finds the distribution closest to through optimization by means of optimization

[0088] In step S105, real-time data of the fan operation is obtained and input into the fan power prediction model, and the fan power prediction model outputs a fan power prediction result

[0089] In summary, in the method of this application, the physical models of fan icing, natural ice melting of blades and fan output are used as prior knowledge, and the historical operation data of the fan collected by the wind farm is used as posterior. A probability mapping relationship between the fan operation data and the wind power under the icing state is constructed by using a Bayesian multi-layer neural network. Finally, the neural network parameters are trained by combining the Bayesian statistical inference theory and the stochastic variational derivation method, so as to comprehensively utilize expert knowledge and historical operation data to greatly improve the accuracy of wind power prediction under fan icing conditions

[0090] Please refer to Figure 2 which shows a structural block diagram of a fan power prediction system based on Bayesian deep learning according to this application

[0091] As Figure 2 shown, the fan power prediction system 200 includes a first construction module 210, a screening module 220, a second construction module 230, a training module 240 and an output module 250

[0092] Among them, the first construction module 210 is configured to construct a physical model of the influence of the icing process on the fan output; the screening module 220 is configured to determine the sample characteristics of the measurement system according to the real-time measurement data, and then screen the mixed sample characteristics related to the fan power prediction under the icing condition according to the physical model; the second construction module 230 is configured to use the mixed sample characteristics as input to construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network; the training module 240 is configured to generate a fan power output expert experience sample database according to the physical model, and use the fan power output expert experience sample database as a prior to pre-train the constructed adaptive wind power prediction model, and collect the historical operation data of the wind farm as a posterior, and re-train the adaptive wind power prediction model regularly based on the stochastic variational derivation method to obtain a fan power prediction model; the output module 250 is configured to obtain the real-time data of the fan operation and input the real-time data of the fan operation into the fan power prediction model, and the fan power prediction model outputs a fan power prediction result

[0093] It should be understood thatFigure 2 The various modules described in Figure 1 correspond to the respective steps in the method described in Figure 2 . Therefore, the operations, features, and corresponding technical effects described above for the method also apply to

[0094] the various modules in

[0095] and will not be elaborated here.

[0096] Build a physical model of the influence of the icing process on the output of the wind turbine;

[0097] Determine the sample characteristics of the measurement system based on real-time measurement data, and then screen out the mixed sample characteristics related to wind turbine power prediction under icing conditions according to the physical model;

[0098] Use the mixed sample characteristics as input to build an adaptive wind power prediction model based on a Bayesian multi-layer neural network;

[0099] Generate an expert experience sample database of wind turbine power output according to the physical model, and use the expert experience sample database of wind turbine power output as the prior to pre-train the constructed adaptive wind power prediction model, and collect the historical operation data of the wind farm as the posterior, and regularly re-train the adaptive wind power prediction model based on the stochastic variational inference method to obtain a wind turbine power prediction model;

[0100] Obtain the real-time operation data of the wind turbine, and input the real-time operation data of the wind turbine into the wind turbine power prediction model, and the wind turbine power prediction model outputs a wind turbine power prediction result.

[0101] A computer-readable storage medium may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the fan power prediction system based on Bayesian deep learning, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the computer-readable storage medium may optionally include a memory remotely provided with respect to the processor, and these remote memories may be connected to the fan power prediction system based on Bayesian deep learning through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0102] Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention, as Figure 3 shown, the device includes: a processor 310 and a memory 320. The electronic device may further include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330, and the output device 340 may be connected through a bus or other means, Figure 3 taking the connection through the bus as an example. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running non-volatile software programs, instructions, and modules stored in the memory 320, that is, implements the fan power prediction method based on Bayesian deep learning in the above method embodiment. The input device 330 may receive input digital or character information, and generate key signal inputs related to user settings and function controls of the fan power prediction system based on Bayesian deep learning. The output device 340 may include a display device such as a display screen.

[0103] The above electronic device may execute the method provided by the embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method. For technical details not described in detail in this embodiment, reference may be made to the method provided by the embodiment of the present invention.

[0104] As an implementation manner, the above electronic device is applied to a fan power prediction system based on Bayesian deep learning and is used for a client, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0105] Construct a physical model of the influence of the icing process on the fan output;

[0106] Determine the sample characteristics of the measurement system based on real-time measurement data, and then screen out the hybrid sample characteristics related to wind turbine power prediction under icing conditions according to the physical model;

[0107] Take the hybrid sample characteristics as input and construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network;

[0108] Generate an expert experience sample database of wind turbine power output according to the physical model, and use the expert experience sample database of wind turbine power output as a prior for pre-training the constructed adaptive wind power prediction model. Collect historical operation data of the wind farm as a posterior, and regularly re-train the adaptive wind power prediction model based on the stochastic variational inference method to obtain a wind turbine power prediction model;

[0109] Obtain the real-time operation data of the wind turbine and input the real-time operation data of the wind turbine into the wind turbine power prediction model, and the wind turbine power prediction model outputs a wind turbine power prediction result.

[0110] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course also by hardware. Based on such an understanding, the above technical solutions, in essence, or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of each embodiment or some parts of the embodiments.

[0111] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A method for predicting the power of a wind turbine based on Bayesian deep learning, characterized in that, Including: Constructing a physical model for the impact of the icing process on the output of a wind turbine. Constructing a physical model for the impact of the icing process on the output of a wind turbine includes: Constructing a wind turbine power output model based on the wind energy conversion theory, a wind turbine blade icing model based on the FENSAP software, and a wind turbine blade natural de-icing model based on the convective radiation effect and blade rotation ice throwing. Among them, the expression of the wind turbine power output model is: ， ， ， ， ， ， Wherein, is the fan power output value, is the mechanical loss coefficient, is the electrical loss coefficient, is the rotation coefficient, is the yaw error, is the air density, is the rotor swept area, is the inflow wind speed, is the distance from the blade element to the hub center, is the fan blade profile, is the blade tip, is the wake effect, is the tip speed ratio, is the tip shape factor, is the lift-drag ratio, is the lift per unit length of the airfoil along the span direction, is the drag per unit length of the airfoil along the span direction, is the inflow angle of the local resultant wind direction blade element, representing the local inflow angle of the blade, represents related to the number of blades and the tip speed ratio as well as the tip shape factor is related, is the rotor radius, is the number of blades; The expression of the air flow in the wind turbine blade icing model is: ， In the formula, is the laminar kinematic viscosity, is the dimensionless laminar viscosity, , are both cut-off functions, is the vorticity, is the Reynolds number at infinity, , , and are all closure coefficients, is the closure function, is the distance from the wall, is the Stefan-Boltzmann constant, is the Hamiltonian operator, is the velocity change; Using the Euler model for impact calculation, where the expressions of the continuity and momentum equations related to droplets are: ， In the formula, is the droplet volume fraction, is the air flow rate, is the relative velocity of the droplet, is the drag coefficient of the droplet, is the Reynolds number of the droplet, is the inertial parameter of the droplet, is the air density, is the density of water, is the Froude number, is the acceleration due to gravity; According to the droplet collection efficiency Solving the icing process, the expressions of the partial differential equations of mass conservation and energy conservation obtained are as follows: ， In the formula, is the water film density, is the height, is the symbol of divergence, is the flow velocity, is the reference velocity, is the liquid water content, is the liquid water evaporation rate, is the liquid water freezing rate; ， In the formula, is the specific heat capacity of the water film, is the latent heat of evaporation, is the latent heat of sublimation, is the latent heat of fusion, is the specific heat capacity of ice, is the solid heat dissipation coefficient, is the reference temperature, is the reference droplet temperature in Kelvin, is the reference droplet temperature in degrees Celsius, is the water-ice interface temperature in degrees Celsius, is the heat transferred by convection, is the water-ice interface temperature in Kelvin; ， In the formula, is the normal vector of the blade surface; The expression of the ice melting rate in the wind turbine blade natural de-icing model is: ， In the formula, is the diameter of the frozen object, is the convective heat transfer coefficient, and are the air temperature and the ice surface temperature respectively, is the radiation constant, is the specific heat capacity of ice, is the latent heat of ice, is the density of ice; The expression of the ice thickness when the ice falls off is: ， In the formula, is the ice thickness when ice shedding occurs, is the viscous shear stress, is the diameter of the wind turbine, is the blade rotation rate; Determining the sample characteristics of the measurement system according to the real-time measurement data, and then screening the mixed sample characteristics related to wind turbine power prediction under icing conditions according to the physical model; Taking the mixed sample characteristics as the input, constructing an adaptive wind power prediction model based on the Bayesian multi-layer neural network; Generating a wind turbine power output expert experience sample database according to the physical model, and using the wind turbine power output expert experience sample database as the prior for pre-training the constructed adaptive wind power prediction model, and collecting the historical operation data of the wind farm as the posterior, and regularly re-training the adaptive wind power prediction model based on the stochastic variational inference method to obtain a wind turbine power prediction model; Obtaining the real-time operation data of the wind turbine, and inputting the real-time operation data of the wind turbine into the wind turbine power prediction model, and the wind turbine power prediction model outputs a wind turbine power prediction result.

2. The method for predicting the power of a wind turbine based on Bayesian deep learning according to claim 1, wherein The expression of the mixed sample characteristics is: ， In the formula, is the air flow rate, is the blade rotation rate, is the tip speed ratio, is the pitch angle, is the number of blades, is the rotor radius, is the yaw error, is the air density, is the inflow wind speed, is the liquid water content, is the droplet volume fraction, is the droplet relative velocity, is the water-ice interface temperature in degrees Celsius, is the water-ice interface temperature in Kelvin, is the convective exchange coefficient, is the air surface, is the ice surface temperature, is the density of ice.

3. A method for predicting the power of a wind turbine based on Bayesian deep learning according to claim 1, characterized in that, The adaptive wind power prediction model includes: A variational Bayesian gated recurrent unit, which is used to process the time-series observation historical data in the SCADA. Among them, the time-series observation historical data in the SCADA contains mixed sample characteristics; A variational Bayesian neuron, the basic unit of which is an artificial neuron, and the expression is: ， wherein, is the deviation, is the output of the variational Bayesian neuron, is the activation function, is the weight assigned to the network connection, is the input of the variational Bayesian neuron, is the number of inputs; A GRU unit, and the expression of the GRU unit is: ， ， ， ， ， Wherein, is the reset gate, is the update gate, , are respectively the hidden state of the GRU cell at the t-th iteration and the hidden state of the GRU cell at the (t - 1)-th iteration, is the Hadamard product, is the sigmoid function, is the candidate hidden state, is the output of the GRU cell, is the input of the GRU cell at time t, is the weight of the output of the reset gate occupied by the previous hidden state, is the weight of the output of the reset gate occupied by the current input, is the weight of the output of the update gate occupied by the previous hidden state, is the weight of the output of the update gate occupied by the current input, is the weight of the candidate hidden state occupied by the Hadamard product of the previous hidden state and the reset gate, is the weight of the candidate hidden state occupied by the current input, is the hyperbolic tangent function.

4. A method for predicting the power of a wind turbine based on Bayesian deep learning according to claim 1, characterized in that, The generating a wind turbine power output expert experience sample database according to the physical model, and using the wind turbine power output expert experience sample database as the prior for pre-training the constructed adaptive wind power prediction model includes: Inputting the wind turbine geographical data, meteorological forecast data, and historical operation data observed by the SCADA into the physical model to calculate the wind turbine power, and obtaining a wind turbine power output expert experience sample database; Pre-train the constructed adaptive wind power prediction model based on the expert experience sample database of the database, where the input of the adaptive wind power prediction model at a moment is: ， In the formula, is historical operation data at time is the geographical data of the fan, is weather forecast data at time t The output of the time-adaptive wind power prediction model is t the power output of the wind turbine at a certain moment , and the expression is: ， In the formula, is the wind power output of the w-th fan at time is the wind power output of the -th fan at time 5. A wind turbine power prediction system based on Bayesian deep learning, characterized in that, Including: A first construction module configured to construct a physical model for the impact of the icing process on the output of a wind turbine. Constructing a physical model for the impact of the icing process on the output of a wind turbine includes: Constructing a wind turbine power output model based on the wind energy conversion theory, a wind turbine blade icing model based on the FENSAP software, and a wind turbine blade natural de-icing model based on the convective radiation effect and blade rotation ice throwing. Among them, the expression of the wind turbine power output model is: ， ， ， ， ， ， Wherein, is the fan power output value, is the mechanical loss coefficient, is the electrical loss coefficient, is the rotation coefficient, is the yaw error, is the air density, is the rotor swept area, is the inflow wind speed, is the distance from the blade element to the hub center, is the fan blade profile, is the blade tip, is the wake effect, is the tip speed ratio, is the tip shape factor, is the lift-to-drag ratio, is the lift per unit length of the airfoil along the span direction, is the drag per unit length of the airfoil along the span direction, is the inflow angle of the local resultant wind direction blade element, representing the local inflow angle of the blade, represents related to the number of blades , the tip speed ratio and the tip shape factor is related to, is the rotor radius, is the number of blades; The expression of the air flow in the wind turbine blade icing model is: ， In the formula, is the laminar kinematic viscosity, is the dimensionless laminar viscosity, , are both cut-off functions, is the vorticity, is the Reynolds number at infinity, , , and are all closure coefficients, is the closure function, is the distance from the wall, is the Stefan-Boltzmann constant, is the Hamiltonian operator, is the velocity change; Using the Euler model for impact calculation, where the expressions of the continuity and momentum equations related to droplets are: ， In the formula, is the droplet volume fraction, is the air flow rate, is the relative velocity of the droplet, is the drag coefficient of the droplet, is the Reynolds number of the droplet, is the inertial parameter of the droplet, is the air density, is the density of water, is the Froude number, is the acceleration due to gravity; According to the droplet collection efficiency Solving the icing process, the expressions of the partial differential equations of mass conservation and energy conservation obtained are as follows: ， In the formula, is the water film density, is the height, is the symbol of divergence, is the flow velocity, is the reference velocity, is the liquid water content, is the liquid water evaporation rate, is the liquid water freezing rate; ， In the formula, is the specific heat capacity of the water film, is the latent heat of evaporation, is the latent heat of sublimation, is the latent heat of fusion, is the specific heat capacity of ice, is the solid heat dissipation coefficient, is the reference temperature, is the reference droplet temperature in Kelvin, is the reference droplet temperature in degrees Celsius, is the water-ice interface temperature in degrees Celsius, is the heat transferred by convection, is the water-ice interface temperature in Kelvin; ， In the formula, is the normal vector of the blade surface; The expression for the ice melting rate of the natural ice removal model of the fan blade is as follows: ， In the formula, is the diameter of the frozen object, is the convective heat transfer coefficient, , are the air temperature and the ice surface temperature respectively, is the radiation constant, is the specific heat capacity of ice, is the latent heat of ice, is the density of ice; The expression for the ice thickness when the ice falls off is as follows: ， In the formula, is the ice thickness when the ice detaches, is the viscous shear stress, is the diameter of the wind turbine, is the blade rotation rate; A screening module, configured to determine the sample characteristics of the measurement system according to the real-time measurement data, and then screen the mixed sample characteristics related to the fan power prediction under icing conditions according to the physical model; A second construction module, configured to use the mixed sample characteristics as input to construct an adaptive wind power prediction model based on a Bayesian multi-layer neural network; A training module, configured to generate a fan power output expert experience sample database according to the physical model, and use the fan power output expert experience sample database as a prior for pre-training the constructed adaptive wind power prediction model, and collect the historical operation data of the wind farm as a posterior, and regularly re-train the adaptive wind power prediction model based on the stochastic variational inference method to obtain a fan power prediction model; An output module, configured to obtain the real-time operation data of the fan, and input the real-time operation data of the fan into the fan power prediction model, and the fan power prediction model outputs a fan power prediction result.

6. An electronic device, characterized in that, Including: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the method according to any one of claims 1 to 4.

Images