Method and device for evaluating energy efficiency level of wind turbine generator set

Abstract

The present invention provides a method and device for evaluating the energy efficiency level of a wind turbine generator set. The evaluation method is based on a neural network and an adaptive threshold, and includes the following steps: S1, establishing a neural network model based on historical data for training to obtain a predicted value of a core parameter; S2, calculating the reliability loss of the core parameter according to the predicted value and actual value of the core parameter; S3, setting the weight of the core parameter; S4, calculating the energy efficiency level of the subsystem corresponding to the core parameter according to the weight and the reliability loss; S5, based on the extreme value theory, using the exponential probability density form to directly fit the historical data of the reliability loss, and obtaining the starting value of the adaptive threshold of the reliability loss; S6, establishing a sliding window to calculate the adaptive threshold at each time point; S7, judging whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold. The present invention accurately obtains the wind energy conversion level of the subsystem and the whole machine by calculating the reliability loss of the core parameter, and can more accurately obtain the predicted value of the core parameter.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular to a method and a device for evaluating the energy efficiency level of a wind power generator set. Background Art

[0002] As the operating time of wind power stock assets increases, operators have an increasing demand for equipment performance evaluation and monitoring technology. Wind farms use a variety of equipment, and there are many factors affecting energy efficiency. Wind power operators lack technical support and tools in the process of equipment monitoring and maintenance. At present, the energy efficiency evaluation of wind turbines is mainly based on the analysis of unit power characteristics. Standards such as IEC 61400-12-1 clearly define the prerequisites and calculation methods for related analysis.

[0003] In recent years, domestic OEMs, scientific research institutions, and certification agencies have carried out a large number of related research. On the one hand, they are committed to studying the energy efficiency evaluation methods of the whole machine or subsystems (such as yaw, gearbox, and bearing health status). On the other hand, they focus on the specific mathematical models used in the evaluation methods, such as PCA, neural grid, similarity analysis and other models.

[0004] (1) Current status of wind turbine operation status monitoring and evaluation research

[0005] There are many technical forms for monitoring the operating status of wind turbines. Different monitoring methods are given for different equipment. For example, for the status of blades, vibration, acoustics and optical fiber strain analysis can be used; for transmission components, acoustics and vibration analysis can be used; for electronics, thermal spectrum analysis can be used. There are also many signal processing methods for wind turbine status monitoring, including data mining, wavelet analysis, fuzzy mathematics, state space theory, etc.

[0006] The mainstream monitoring technology for wind turbine status monitoring mainly monitors the single signal of wind turbine components, and the use of multi-signal fusion is only limited to the level of single signal monitoring and evaluation and fusion of multi-source signal decision results. Due to the complex and changeable operating conditions of wind turbines, monitoring and analysis using only a single signal is likely to cause distortion of the evaluation results. Therefore, it is urgent to carry out operating status monitoring and analysis based on multi-source information fusion.

[0007] (2) Current status of research on energy efficiency loss of wind turbines

[0008] Research on the energy efficiency loss mechanism of wind turbines is a necessary condition for improving the energy utilization rate of wind turbines. In terms of the aerodynamic performance of wind turbine impellers in capturing wind energy, it is necessary to comprehensively use relevant theoretical methods such as momentum theory, momentum-blade element theory and Betz theory to establish an impeller model. The core issue of studying the aerodynamic performance of impellers is how to obtain accurate axial velocity induction factors and tangential velocity induction factors. Usually, the iterative calculation method is mainly used to obtain the accurate value of the induction factor. Most studies have shown that when the impeller is in a vortex ring state, the impeller model needs to be modified to meet actual needs. Currently common correction methods include Wilson method, Glauert and Prandtl loss factor method.

[0009] Research on the impeller mechanism found that the wind energy utilization coefficient directly reflects the impeller's ability to absorb and convert wind energy. Some domestic scholars have compared the differences between different blade tip loss models and the degree of proximity between their wind energy utilization coefficient values ​​and actual values. In terms of the mechanical transmission chain system that converts wind energy into kinetic energy, some domestic scholars have studied gearboxes, using the actual shaft speed as input to determine the actual gearbox loss, and defined the rated loss. Some domestic scholars have also determined the relationship between the gearbox model and operating parameters by studying the loss conditions of different gearboxes.

[0010] In the generator system of wind turbines that convert kinetic energy into electrical energy, the generator loss index can usually be divided into fixed loss and variable loss. Of course, some scholars also divide it into generator loss, inverter loss and self-use power loss, and calculate and simulate the losses of various components. Domestic and foreign scholars have conducted in-depth analysis and correction of the calculation model of variable losses such as copper loss, iron loss, mechanical loss, etc. in the generator body loss, and have now achieved a relatively accurate research method.

[0011] However, these studies only conduct in-depth analysis and research on a certain component or a single system. Few domestic and foreign scholars have conducted research from the perspective of the energy flow conversion efficiency between the wind turbine's capture performance, transmission performance, and power generation performance.

[0012] (3) Current status of research on wind turbine power characteristics

[0013] The power generation of wind turbines can be improved by studying the power characteristics of wind turbines. According to the IEC power test standard, some domestic scholars have proposed a data correction method using the measured data of wind turbines, and conducted a comparative analysis of the power characteristics of wind turbines. By studying the power characteristics of different units in the same time period and the same unit in different time periods, the reasons affecting the power characteristics of the units were found. Some scholars used aerodynamic simulation experiments to obtain the relationship between the wind speed monitored by the unit anemometer and the wind speed of the wind tower, and successfully corrected the measured power curve of the wind turbine. Most of the above research methods are based on laboratory data, and the authenticity of these data has not been verified, which is still a long way from the actual application of engineering.

[0014] (4) Current status of research on wind turbine energy efficiency monitoring

[0015] Foreign wind turbine manufacturers such as GE, Gamesa, and Vestas all provide their own wind turbine equipment with unit monitoring systems, but these systems are only useful for specific units. Therefore, a number of third-party companies have emerged, which are committed to the design and development of wind turbine monitoring systems to monitor the energy efficiency of wind turbines from different manufacturers. Compared with foreign manufacturers, domestic wind power manufacturers' monitoring systems were applied later, with single monitoring functions, limited to completing traditional data collection, statistical analysis, and display tasks.

[0016] Many scientific research institutions and wind turbine manufacturers in the market have developed monitoring systems to target different business pain points, but these only evaluate and diagnose the performance status of a single component of a wind turbine. There is a lack of comprehensive research and system development in the evaluation and diagnosis of wind turbine energy efficiency performance.

[0017] In the prior art, the current methods for evaluating the energy efficiency level of wind turbines have the following problems:

[0018] 1. The mainstream monitoring technology for wind turbine status monitoring mainly monitors the single signal of wind turbine components, and the use of multi-signal fusion is only limited to the level of single signal monitoring and evaluation and multi-source signal decision-making result fusion. Due to the complex and changeable operating conditions of wind turbines, only using a single signal for monitoring and analysis is likely to cause distortion of the evaluation results.

[0019] Second, the research on energy efficiency loss of wind turbines is only an in-depth analysis of a certain component or a single system. Domestic and foreign scholars rarely conduct research from the perspective of the conversion efficiency of energy flow between the wind turbine's capture performance, transmission performance, and power generation performance.

[0020] 3. Most of the methods for studying the power characteristics of wind turbines are based on laboratory data, and the authenticity of these data has not been verified, which is still a long way from practical application in engineering.

[0021] 4. The unit monitoring systems provided by foreign wind turbine manufacturers for their own wind turbine equipment are only useful for specific units.

[0022] 5. Domestic wind power manufacturers applied monitoring systems relatively late, with single monitoring functions limited to completing traditional data collection, statistical analysis, and display tasks.

[0023] 6. The wind turbine monitoring systems on the market only evaluate and diagnose the performance status of a single component of a wind turbine, and lack comprehensive research and system development in the evaluation and diagnosis of wind turbine energy efficiency performance.

[0024] Therefore, the inventors of the present application have designed a method and device for evaluating the energy efficiency level of a wind turbine generator set in order to overcome the above-mentioned technical problems. Summary of the invention

[0025] The technical problem to be solved by the present invention is to overcome the defect that the evaluation effect of the energy efficiency level of a wind turbine generator set in the prior art is not ideal, and to provide a method and device for evaluating the energy efficiency level of a wind turbine generator set.

[0026] The present invention solves the above technical problems through the following technical solutions: The present invention provides a method for evaluating the energy efficiency level of a wind turbine generator set, which is characterized in that the evaluation method is based on a neural network and an adaptive threshold, and comprises the following steps: S 1 , establish a neural network model based on historical data for training and obtain the predicted values ​​of core parameters; S 2 , calculating the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter; S 3 , set the weight of the core parameters; S 4 , calculating the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter; S 5 Based on the extreme value theory, the historical data of reliability loss are fitted in the form of exponential probability density to obtain the starting value of the adaptive threshold of reliability loss; S 6 , establish a sliding window to calculate the adaptive threshold at each time point; S 7 , judging whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold.

[0027] According to one embodiment of the present invention, the step S 1 The steps of establishing a neural network model based on historical data are as follows: 11 , clarify the core parameters of each subsystem; S 12 , clarify the associated parameters of the core parameters and obtain the associated parameter values; S 13, using the associated parameter value as input and the predicted value of the core parameter as output, to establish a BP neural network model.

[0028] According to one embodiment of the present invention, the step S 2 In the above formula, the reliability loss calculation formula of the core parameter is: Among them: i is the core parameter; loss i is the loss value of the core parameter i; j is the time series of the imported data; N represents the number of time series data points; is the predicted value of core parameter i under the condition of time series j; x i,j is the actual value of core parameter i under the condition of time series j.

[0029] According to one embodiment of the present invention, the step S 4 In the equation, the energy efficiency level of the subsystem = 1-weight of the core parameter * reliability loss of the core parameter of the subsystem

[0030] According to one embodiment of the present invention, the step S 4 Then the following steps are included: 41 , calculate the energy efficiency level of the overall unit according to the energy efficiency level of each subsystem; the energy efficiency level of the overall unit = the multiplication value of the energy efficiency levels of each subsystem; the step S 7 The energy efficiency level mentioned in the table is the energy efficiency level of the entire unit.

[0031] The present invention also provides a device for evaluating the energy efficiency level of a wind turbine generator set, the device adopts the method for evaluating the energy efficiency level of a wind turbine generator set as described above, the device comprising: a training module, used to establish a neural network model based on historical data for training, and obtain a predicted value of a core parameter; a calculation module, used to calculate the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter, set the weight of the core parameter, calculate the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter, calculate the starting point of the threshold of the reliability loss based on extreme value theory and historical data, and establish a sliding window to calculate the threshold at each time point; a detection module, used to determine whether the energy efficiency level meets the standard according to the threshold.

[0032] According to an embodiment of the present invention, the training module uses the associated parameter value as input and the predicted value of the core parameter as output to establish a BP neural network model.

[0033] According to one embodiment of the present invention, the calculation module is also used to calculate the energy efficiency level of the overall unit based on the energy efficiency levels of each subsystem, and the energy efficiency level of the overall unit = the multiplied value of the energy efficiency levels of each subsystem; the detection module is used to determine whether the energy efficiency level of the overall unit meets the standard.

[0034] The present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a program or instruction, and the processor executes the program or instruction, so that the electronic device executes the steps in the above method embodiment.

[0035] The present invention also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed on an electronic device, the electronic device executes the steps in the above method embodiment.

[0036] The positive and progressive effects of the present invention are:

[0037] 1. Since the method of the present invention is an energy efficiency level assessment method for wind turbine subsystems and the entire machine based on the reliability of associated parameters, the wind energy conversion level of the subsystem and the entire machine can be accurately obtained by calculating the reliability loss of core parameters.

[0038] 2. Since the method of the present invention is an extreme value adaptive threshold calculation method based on a single-parameter exponential distribution, it is used to evaluate whether the energy efficiency level is qualified. It has good robustness. The algorithm can automatically calculate the threshold of the energy efficiency level under the current working conditions, and judge whether the energy efficiency level is qualified based on whether it exceeds the limit.

[0039] 3. Since the method of the present invention establishes a BP neural network model for prediction based on the historical operating data of the wind turbine generator set in a healthy state, the predicted values ​​of the core parameters can be obtained more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein:

[0041] Figure 1 A logical schematic diagram of energy efficiency evaluation of a wind turbine generator set provided for one or more embodiments of the present invention;

[0042] Figure 2a A schematic diagram of a relationship for calculating the energy efficiency level of a wind turbine generator set provided by one or more embodiments of the present invention;

[0043] Figure 2b A schematic diagram of core parameters of energy efficiency level of a wind turbine generator set provided by one or more embodiments of the present invention;

[0044] Figure 3 A schematic diagram of a neural network modeling result of a generator speed prediction value provided by one or more embodiments of the present invention;

[0045] Figure 4aA schematic diagram of the probability density distribution of a full data set provided for one or more embodiments of the present invention;

[0046] Figure 4b A schematic diagram of probability density distribution of extreme value data subsets provided for one or more embodiments of the present invention;

[0047] Figure 5 A schematic diagram of the tail distribution of energy efficiency data of a real wind turbine generator set provided by one or more embodiments of the present invention;

[0048] Figure 6 A schematic diagram of the extreme value dynamic exponential distribution slip calculation steps provided for one or more embodiments of the present invention;

[0049] Figure 7 Schematic diagram of an example of adaptive threshold calculation results provided for one or more embodiments of the present invention DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0051] Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Reference will now be made in detail to preferred embodiments of the present invention, examples of which are shown in the accompanying drawings. Wherever possible, the same reference numerals will be used throughout the drawings to represent the same or similar parts.

[0052] Furthermore, although the terms used in the present invention are selected from well-known and commonly used terms, some terms mentioned in the present invention specification may be selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant parts of the description of this document.

[0053] Furthermore, it is required that the present invention be understood not only by the actual terms used but also by the meanings connoted by each term.

[0054] The present invention provides a method for evaluating the energy efficiency level of a wind turbine generator set. The method is based on a neural network and an adaptive threshold and comprises the following steps:

[0055] S 1 , Establish a neural network model based on historical data for training and obtain the predicted values ​​of core parameters;

[0056] S 2 , calculating the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter;

[0057] S 3 , set the weights of core parameters;

[0058] S4 , calculating the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter;

[0059] S 5 , Based on extreme value theory, the historical data of reliability loss are directly fitted in the form of exponential probability density to obtain the starting value of the adaptive threshold of reliability loss;

[0060] S 6 , establish a sliding window, continuously add the data in the window to the historical data set, and based on the extreme value theory, use the exponential probability density form to directly fit the reliability loss data set to calculate the adaptive threshold at each time point;

[0061] S 7 , judging whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold.

[0062] In the above method, step S 1 The establishment of a neural network model based on historical data can also include the following steps:

[0063] S 11 , clarify the core parameters of each subsystem;

[0064] S 12 , clarify the associated parameters of the core parameters and obtain the associated parameter values;

[0065] S 13 , using the associated parameter value as input and the predicted value of the core parameter as output, to establish a BP neural network model.

[0066] In the above method, step S 4 The following steps may also be included:

[0067] S 41 , calculate the energy efficiency level of the overall unit based on the energy efficiency level of each subsystem;

[0068] Figure 1 The figure shows the energy efficiency evaluation logic diagram of the wind turbine generator set of the present invention. As an embodiment of the present invention, the core calculation logic of the present invention is as follows:

[0069] According to the wind energy conversion link, the wind turbine generator set is divided into a wind energy capture system, a mechanical energy transmission system, and an electrical energy conversion system. Combined with the core parameters of each subsystem, the energy efficiency related parameters and energy efficiency influencing factors are extracted, and the reliability of each subsystem, that is, the energy efficiency level, is calculated separately. Because the three subsystems are in series in the energy flow of the wind turbine set, the energy efficiency level of each system is multiplied to obtain the energy efficiency level of the whole machine:

[0070] R 整机 =R风能捕获系统 *R 机械能传递系统 *R 电能转换系统

[0071] After obtaining the energy efficiency level of the whole machine, the threshold of the reliability of the whole machine or subsystem, that is, the threshold of the energy efficiency level, can be calculated; based on the reliability, that is, whether the energy efficiency level exceeds the threshold, it can be judged whether the energy efficiency level of the whole machine or subsystem is qualified.

[0072] The above-mentioned core parameters refer to the characteristic indicators of the reliability of a certain component of a wind turbine generator set, such as wind energy conversion rate, power deviation, generator speed, etc.; associated parameters refer to indicators that have a significant impact on the core parameters, such as wind speed, power, etc.

[0073] exist Figure 1 In step I, it is necessary to obtain the predicted value of the associated parameter through historical data modeling. The modeling method can use different neural networks without affecting the realization of the effect of the present invention.

[0074] The reliability of the present invention, that is, the specific method of calculating the energy efficiency level is as follows:

[0075] The energy efficiency level of each subsystem is obtained by subtracting the monitoring value loss of the core parameter and the weight of the parameter from 1:

[0076]

[0077]

[0078]

[0079] Among them: i represents the core parameter; loss i is the reliability loss of core parameter i; n represents the number of core parameters; w i Represents the weight of core parameter i.

[0080] The reliability loss is calculated by the following formula:

[0081]

[0082] Among them: i is the core parameter; loss i is the reliability loss of core parameter i; j is the SCADA time series data of wind turbine generator set; N represents the number of time series data points; is the predicted value of core parameter i under the condition of time series j; x i,j is the actual value of core parameter i under the condition of time series j.

[0083] The predicted value of the core parameter i is calculated based on the historical operating data of the healthy unit through the neural network modeling method. When using the neural network, all input parameters need to be normalized first, and the value interval is mapped to the interval of [-1, 1].

[0084] In one embodiment, the core parameters of each system are as follows: Figure 2a , 2b As shown in the figure. Different wind turbines have different technical routes and use different equipment. At the same time, the SCADA data tag points of equipment manufacturers are also different. Each model needs to establish a core parameter model based on the actual situation.

[0085] When calculating the reliability loss of some core parameters, it is necessary to first calculate the predicted value of the parameter. The calculation of the predicted value is based on the neural network model trained with historical data. In this method, in order to avoid overfitting and underfitting, on the basis of ensuring nonlinear fitting, the hidden layer is set to 3 layers, and the number of neurons in each hidden layer of the neural network = 2 / 3*(the number of neurons in the input layer + the number of neurons in the output layer).

[0086] Taking the impeller speed as an example, whether it is a direct-drive or gearbox wind turbine, in the maximum power tracking control of the wind turbine, the tip speed ratio is adjusted by controlling the impeller speed, that is, the generator speed, so that the wind turbine maintains the maximum wind energy utilization coefficient. The reliability loss of the generator speed is the absolute value of the difference between the actual speed and the predicted speed under this condition:

[0087]

[0088] Where: loss generator_rotor_speed is the reliability loss value of the generator speed; j represents the jth time point in the time series; N represents the number of data points in the time series; ω 实际,j is the actual engine speed at the jth time point in the time series; ω 预测,j is the predicted engine speed at the jth time point in the time series, ω 预测,j It is obtained by modeling based on historical data through neural network algorithm.

[0089] An embodiment of establishing a BP neural network model taking the impeller speed as an example is as follows:

[0090] The generator speed is the core parameter, and the predicted value of the generator speed is used as the output value. Wind speed and power are related parameters, and the values ​​of wind speed and power are used as input values. After normalizing the data, the BP neural network, that is, the multi-layer feedforward neural network trained by the error back propagation algorithm, is used for model training. Two hidden layers are used, and the number of neurons in each layer is 2. The activation function of each layer is 'tansig', and the training function uses the 'traingd' gradient descent algorithm. The learning rate is set to 0.1, the upper limit of the cycle is 500 times, and the target value is set to 1.0*10^(-3).

[0091] As the actual working conditions are complex and changeable, prediction errors are inevitable. Figure 3 As shown, the generator speed predicted by the neural network model is roughly consistent with the actual generator speed, which proves that the prediction results of the neural network model are effective and the prediction is relatively accurate.

[0092] The calculation principle and method of the reliability threshold, that is, the energy efficiency level threshold, of the present invention are as follows:

[0093] Since the working conditions of wind turbines change with time and environment, if the threshold is set manually, users need to have sufficient experience, and this setting method is very random. Therefore, an adaptive threshold based on extreme value theory is proposed as the basis for judging whether a subcomponent is in an abnormal state.

[0094] The Extreme Value Theory (EVT) believes that the distribution of extreme values ​​is similar to the central limit theorem and is independent of the original data distribution. This distribution is an approximate exponential distribution. This distribution has the following form:

[0095]

[0096] Taking Gaussian distribution as an example, assuming that a data set has a mean of 1.0, a standard deviation of 2.0, and the numerical distribution conforms to Gaussian distribution, then the probability density distribution of the data set is as follows: Figure 4a Assume that the threshold of extreme value (t) is 5.652, then the probability density distribution of the data subset with a value greater than 5.652 in the original data set, that is, the extreme value data subset, is as follows Figure 4b As shown, it is close to an exponential distribution.

[0097] The threshold is actually to find the extreme value in the extreme value data subset. By fitting the extreme value data subset distribution, we can artificially set a probability q and find the value z corresponding to the probability. q (i.e., threshold value), so that P(x>z q ) <q。

[0098] In this method, a single-parameter exponential distribution is used to fit the distribution of extreme value data subsets. The single-parameter exponential distribution form is:

[0099] PDF: f(x) = γe -γx

[0100] In the formula, γ is the parameter of the exponential distribution function. To simplify the calculation model, the extreme value data subset (M) is further changed, M * =Mt, so that the minimum value of the data in the extreme data subset is 0, which is convenient for function fitting.

[0101] After completing the fitting, the shape parameter γ is obtained, and then the probability density function (PDF) is integrated to obtain the cumulative distribution function (CDF, ​​which is the probability):

[0102]

[0103] When a threshold corresponding to probability q is manually specified, the extreme value b in the extreme value data subset can be obtained:

[0104] b={-ln[1-F(0≤x≤b)]} / γ

[0105] After obtaining b, the threshold of the original data set is:

[0106] Z q =t+b

[0107] Among them, t is the threshold value of the extreme value, b is the extreme value in the extreme value data subset, and Z q is the threshold value.

[0108] This ensures that the probability density distribution of the tail data must meet the requirements (the first integral is the full probability, i.e. 100%, and the second integral is the mean), and the distribution form must conform to the exponential distribution. Figure 5 The figure shows the tail distribution of energy efficiency data of a real wind turbine generator set in one embodiment.

[0109] At the same time, considering that the unit operation data is dynamically recorded at a time point, in this embodiment, it is one point every 10 minutes, and the working conditions at each time point are different, so the threshold should also change with the working conditions. d To reflect the changes in the data value itself.

[0110] An example of an implementation of the extreme value dynamic exponential distribution slip calculation process is Figure 6 shown.

[0111] Using the SCADA data of a certain manufacturer's unit, the loss degree of the power generation system is calculated; the loss degree of the first 2000 time series points is sorted, the 1.5% loss degree is the starting value of t, q is set to 0.01, and 6 data points are used as the sliding window to calculate the threshold of each time point in turn. The test results are as follows Figure 7,In the figure, “*” indicates the point where the reliability loss is too large, indicating that the energy efficiency level of the unit at that time point is poor.

[0112] In the calculation method of the adaptive threshold, the fitting function for the distribution of the data points at the tail of the data may use different forms and methods without affecting the realization of the effect of the present invention.

[0113] Since the method of the present invention is a method for evaluating the energy efficiency level of wind turbine subsystems and the whole machine based on the reliability of associated parameters, the wind energy conversion level of the subsystem and the whole machine can be accurately obtained by calculating the reliability loss of core parameters.

[0114] Since the method of the present invention is an extreme value adaptive threshold calculation method based on a single-parameter exponential distribution, it is used to evaluate whether the energy efficiency level is qualified. It has good robustness and can automatically calculate the threshold of the energy efficiency level under the current working conditions by the algorithm, and judge whether the energy efficiency level is qualified based on whether it exceeds the limit.

[0115] Since the method of the present invention establishes a BP neural network model for prediction based on historical operating data of the wind turbine generator set in a healthy state, the predicted values ​​of the core parameters can be obtained more accurately.

[0116] The present invention further provides an embodiment of a device to implement the method for evaluating the energy efficiency level of a wind turbine generator set of the present invention, comprising:

[0117] The training module is used to establish a neural network model based on historical data for training and obtain the predicted values ​​of core parameters;

[0118] A calculation module, used to calculate the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter, set the weight of the core parameter, calculate the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter, calculate the starting point of the adaptive threshold of the reliability loss based on extreme value theory and historical data, and establish a sliding window to calculate the adaptive threshold at each time point;

[0119] The detection module is used to determine whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold.

[0120] The training module takes the associated parameter value as input and the predicted value of the core parameter as output to establish a BP neural network model.

[0121] The calculation module is further used to calculate the energy efficiency level of the overall unit according to the energy efficiency levels of each subsystem, and the energy efficiency level of the overall unit = the multiplied value of the energy efficiency levels of each subsystem; the detection module is used to determine whether the energy efficiency level of the overall unit meets the standard.

[0122] Since the above device embodiment uses a method for evaluating the energy efficiency level of wind turbine subsystems and the whole machine based on the reliability of associated parameters, the wind energy conversion level of the subsystem and the whole machine can be accurately obtained by calculating the reliability loss of core parameters.

[0123] Since the above-mentioned device embodiment uses an extreme value adaptive threshold calculation method based on a single-parameter exponential distribution to evaluate whether the energy efficiency level is qualified, it has good robustness. The algorithm can automatically calculate the threshold of the energy efficiency level under the current working conditions, and judge whether the energy efficiency level is qualified based on whether it exceeds the limit.

[0124] Since the above device embodiment establishes a BP neural network model for prediction based on the historical operating data of the wind turbine generator set in a healthy state, the predicted values ​​of the core parameters can be obtained more accurately.

[0125] The present invention also provides an embodiment of an electronic device, including a processor and a memory, wherein the memory stores programs or instructions, and the processor executes the programs or instructions so that the electronic device executes the method for evaluating the energy efficiency level of a wind turbine generator set in the method embodiment of the present invention.

[0126] Since the electronic device embodiment uses an energy efficiency level assessment method for wind turbine subsystems and the entire machine based on the reliability of associated parameters, the wind energy conversion level of the subsystem and the entire machine can be accurately obtained by calculating the reliability loss of core parameters.

[0127] Since the above-mentioned electronic device embodiment uses an extreme value adaptive threshold calculation method based on a single-parameter exponential distribution to evaluate whether the energy efficiency level is qualified, it has good robustness. The algorithm can automatically calculate the threshold of the energy efficiency level under the current working conditions, and judge whether the energy efficiency level is qualified based on whether it exceeds the limit.

[0128] Since the above electronic device embodiment establishes a BP neural network model for prediction based on the historical operating data of the wind turbine generator set in a healthy state, the predicted values ​​of the core parameters can be obtained more accurately.

[0129] The present invention also provides an embodiment of a readable storage medium, on which a program or instruction is stored. When the program or instruction is run on an electronic device, the electronic device executes the method for evaluating the energy efficiency level of a wind turbine generator set in the method embodiment of the present invention.

[0130] Since the above-mentioned readable storage medium embodiment uses a method for evaluating the energy efficiency level of wind turbine subsystems and the whole machine based on the reliability of associated parameters, the wind energy conversion level of the subsystem and the whole machine can be accurately obtained by calculating the reliability loss of core parameters.

[0131] Since the above-mentioned readable storage medium embodiment uses an extreme value adaptive threshold calculation method based on a single-parameter exponential distribution to evaluate whether the energy efficiency level is qualified, it has good robustness. The algorithm can automatically calculate the threshold of the energy efficiency level under the current working conditions, and judge whether the energy efficiency level is qualified based on whether it exceeds the limit.

[0132] Since the above-mentioned readable storage medium embodiment establishes a BP neural network model for prediction based on the historical operating data of the wind turbine generator set in a healthy state, the predicted values ​​of the core parameters can be obtained more accurately.

[0133] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for evaluating the energy efficiency level of a wind turbine generator set. It is characterized in that The evaluation method is based on a neural network and an adaptive threshold and comprises the following steps: S 1 , Establish a neural network model based on historical data for training and obtain the predicted values ​​of core parameters; S 2 , calculating the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter; S 3 , set the weights of core parameters; S 4 , calculating the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter; S 5 , Based on extreme value theory, the historical data of reliability loss is fitted in the form of exponential probability density to obtain the starting value of the adaptive threshold of reliability loss; S 6 , establishing a sliding window to calculate the adaptive threshold at each time point; S 7 , judging whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold; The step S 4 In the equation, the energy efficiency level of the subsystem = 1-weight of the core parameter*reliability loss of the core parameter of the subsystem; The step S 4 The following steps are then included: S 41 , calculate the energy efficiency level of the overall unit based on the energy efficiency level of each subsystem; The energy efficiency level of the whole unit = the multiplication value of the energy efficiency levels of each subsystem; The step S 7 The energy efficiency level mentioned in the table is the energy efficiency level of the entire unit.

2. The method for evaluating the energy efficiency level of a wind turbine generator set according to claim 1, It is characterized in that The step S 1 The steps of building a neural network model based on historical data are as follows: S 11 , clarify the core parameters of each subsystem; S 12 , clarify the associated parameters of the core parameters and obtain the associated parameter values; S 13 , using the associated parameter value as input and the predicted value of the core parameter as output, to establish a BP neural network model.

3. The method for evaluating the energy efficiency level of a wind turbine generator set according to claim 1, It is characterized in that The step S 2 In the above formula, the reliability loss calculation formula of the core parameter is: Among them: i is the core parameter; loss i is the loss value of the core parameter i; j is the time series of the imported data; N represents the number of time series data points; is the predicted value of core parameter i under the condition of time series j; x i,j is the actual value of core parameter i under the condition of time series j.

4. A device for evaluating the energy efficiency level of a wind turbine generator set, It is characterized in that The evaluation device adopts the evaluation method for the energy efficiency level of a wind turbine generator set as claimed in claims 1 to 3, and the evaluation device comprises: The training module is used to establish a neural network model based on historical data for training and obtain the predicted values ​​of core parameters; A calculation module, used to calculate the reliability loss of the core parameter according to the predicted value of the core parameter and the actual value of the core parameter, set the weight of the core parameter, calculate the energy efficiency level of the subsystem corresponding to the core parameter according to the weight of the core parameter and the reliability loss of the core parameter, calculate the starting point of the adaptive threshold of the reliability loss based on extreme value theory and historical data, and establish a sliding window to calculate the adaptive threshold at each time point; The detection module is used to determine whether the energy efficiency level of the wind turbine generator set meets the standard according to the adaptive threshold.

5. The device for evaluating the energy efficiency level of a wind turbine generator set according to claim 4, It is characterized in that The training module takes the associated parameter value as input and the predicted value of the core parameter as output to establish a BP neural network model.

6. The device for evaluating the energy efficiency level of a wind turbine generator set according to claim 4 or 5, It is characterized in that The calculation module is further used to calculate the energy efficiency level of the overall unit according to the energy efficiency levels of each subsystem, and the energy efficiency level of the overall unit = the multiplied value of the energy efficiency levels of each subsystem; the detection module is used to determine whether the energy efficiency level of the overall unit meets the standard.

7. An electronic device, It is characterized in that The electronic device comprises a processor and a memory, wherein the memory stores programs or instructions, and the processor executes the programs or instructions, so that the electronic device executes the method according to any one of claims 1 to 3.

8. A readable storage medium, It is characterized in that The readable storage medium stores a program or instruction. When the program or instruction is executed on an electronic device, the electronic device executes the method according to any one of claims 1 to 3.

Images