A method for evaluating the dynamic performance of offshore wind power structures

Abstract

The present invention discloses a method for evaluating the dynamic performance of an offshore wind power structure. The evaluation method integrates the wind power structure dynamic performance evaluation technology based on a root system morphology model. The root system morphology model includes a main root layer, a lateral root layer and a fibrous root layer. The number of lateral root layers is greater than 1. The main root layer is integrated and evaluated according to the data of a first-level lateral root layer, the upper-level lateral root layer is integrated and evaluated according to the data of a lower-level lateral root layer, and the lateral root layer is integrated and evaluated according to the corresponding fibrous root layer data. When the variation range of structural dynamic performance evaluation data classified based on environmental conditions is very small, there is a phenomenon that the root system morphology model and the structural dynamic performance evaluation data based on data fusion are close to each other in peaks and troughs. It can be seen that the structural dynamic performance evaluation based on the environmental root system morphology model comprehensively considers the environmental monitoring data and the vibration response data, effectively avoids the uncertainty of a single type of sensor, and improves the reliability of the evaluation result.

Description

Technical Field

[0001] The present invention belongs to the technical field of marine power generation, and in particular relates to a method for evaluating the dynamic performance of an offshore wind power structure. Background Art

[0002] Offshore wind power structure refers to wind power generation equipment and its supporting structure installed in the ocean, including wind turbines, towers, foundations, etc. Among various renewable energy sources, wind energy has attracted much attention due to its high technical readiness, abundant resources and low environmental impact. Compared with land-based wind power, offshore wind power has more stable wind speeds and greater power generation potential, which can better meet energy needs. As wind turbines continue to grow in size, offshore wind power structures are more susceptible to damage when serving in harsh marine environments, which affects their dynamic performance and carrying capacity, and brings harm to the safe operation of offshore wind power and social economy. Therefore, it is an effective way to ensure the safe operation of wind power and reduce the cost of safe operation and maintenance by monitoring data in real time, grasping the state of structural dynamic performance in real time, and conducting maintenance inspections in a timely manner. Traditional offshore wind power structure dynamic performance evaluation methods are usually based on reanalysis technology based on damage models, which belong to detection and evaluation. Therefore, the industry urgently needs to develop a monitoring and evaluation technology with high real-time and reliability to grasp the changes in the dynamic performance and service status of the structure in real time.

[0003] The dynamic performance of a structure refers to its ability to resist various external dynamic loads. When a structure is damaged or destroyed, its dynamic performance usually decreases, or even fails to serve normally. The most intuitive physical spatial manifestation is that the performance decreases and the flexibility increases. In recent years, with the continuous development of structural monitoring technology, structural monitoring has developed from the manual detection stage to the intelligent structural maintenance stage, which has greatly increased the ability to identify changes in structural dynamic performance and provided an important basis for the evaluation of structural dynamic performance. With the rapid development of sensor technology, the Internet of Things, and information technology, the types of collected structural monitoring data are becoming more and more abundant, and the amount of data is becoming larger and larger. Many researchers have begun to focus on using the collected structural dynamic response data to evaluate the dynamic performance and damage degree of the structure. The current methods for evaluating the dynamic performance and damage degree of structures using structural response data can be roughly divided into two categories: evaluation methods based on finite element models and evaluation methods based on monitoring data.

[0004] The evaluation method based on the finite element model mainly evaluates the performance change and damage degree of the structure by inverting the parameter changes of the finite element model of the structure. First, the finite element model of the structure is established, and then the model is analyzed according to the changes in parameters or indicators to evaluate the performance and damage degree of the structure. The more commonly used methods are the model matching index method and the model correction method. The model matching index method requires first obtaining the damage results of the structure based on the possible damage conditions of the structure using the finite element model, establishing a damage mode database, and then matching the measured data with different damage mode data to evaluate the damage of the structure; the vibration characteristics of the structure can also be used as an indicator to evaluate whether the structure is damaged, including vibration characteristic evaluation methods based on frequency, curvature mode and strain mode, but this method has problems such as low damage sensitivity and susceptibility to noise interference. The model correction method continuously corrects the finite element model of the structure through the structural response data, and evaluates the performance of the actual structure through the analysis results of the finite element model. After decades of development, this method has been widely used in the fields of aviation, machinery, automobiles, civil engineering, and bridges.

[0005] The structural dynamic performance evaluation method based on the finite element model has a large amount of calculation and low calculation efficiency. The current finite element calculation results can only have a good consistency with the deterministic test results, and the effect in actual engineering application is not good.

[0006] The traditional structural performance evaluation method based on monitoring data features has higher recognition efficiency than the model-based evaluation method, but a series of defects in time-frequency analysis and the influence of factors such as data quality also lead to poor reliability of the evaluation results. Therefore, in the structural performance evaluation method based on monitoring data, improving the discrimination method of structural performance and using statistical pattern recognition to evaluate structural performance have received widespread attention. Summary of the invention

[0007] The purpose of the present invention is to provide a method for evaluating the dynamic performance of an offshore wind power structure to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for evaluating the dynamic performance of offshore wind turbine structures integrates the wind turbine structure dynamic performance evaluation technology based on the root morphology model;

[0010] The root morphology model includes the main root layer, lateral root layer and fibrous root layer, and the number of lateral root layers is greater than 1;

[0011] The main root layer is fused and evaluated based on the data of the first-level lateral root layer, and the upper-level lateral root layer is fused and evaluated based on the data of the lower-level lateral root layer; the lateral root layer is fused and evaluated based on the data of the corresponding fibrous root layer.

[0012] Furthermore, the main root layer is for the dynamic performance evaluation of offshore wind power structures; the first-level lateral root layer includes structural performance dynamic performance evaluation based on data fusion and structural dynamic performance evaluation based on environmental condition classification.

[0013] Furthermore, the secondary lateral roots of the structural performance dynamic evaluation based on data fusion include the evaluation values ​​of the wind turbine foundation and tower, and the fibrous root layer under the secondary lateral root layer includes the acceleration monitoring data fusion of the wind turbine foundation and tower.

[0014] The secondary lateral roots based on the dynamic evaluation of structural performance fusion with environmental data include the evaluation of wave inclination angle and wind speed inclination angle. The tertiary lateral roots based on the dynamic evaluation of structural performance based on environmental data fusion include the evaluation of wind turbine foundation and tower. The fibrous root layer under the tertiary lateral root layer includes different inclinometers of the wind turbine foundation and tower and inclination data in different directions.

[0015] Furthermore, the evaluation method is performed based on an evaluation value of the dynamic performance of the wind power structure, and the evaluation value of the dynamic performance of the wind power structure is the sum of the dynamic performance evaluation value based on data fusion and the evaluation value based on environmental condition classification;

[0016] The steps for obtaining the power performance evaluation value based on data fusion include:

[0017] A1: Collect sensor acceleration monitoring data of wind turbine foundation and tower;

[0018] A2: Fusion of multi-sensor acceleration data;

[0019] A3: De-noise the fused acceleration data using the GJO-VMD-based data de-noising technology;

[0020] A4: Use random reduction technology to extract free response signals that are unrelated to environmental excitation from the denoised data, use the autoregressive moving average model to model and analyze the free response signals, and extract the AR coefficient of the model to characterize the dynamic performance characteristics of the structure;

[0021] A5: Calculate the Mahalanobis square distance between AR coefficients, evaluate the dynamic performance of the wind power structure, and obtain the dynamic performance evaluation value based on data fusion;

[0022] The steps for obtaining the power performance evaluation value based on environmental condition classification include:

[0023] B1: Collect inclination data of different inclinometers and different directions of wind turbine foundation and tower;

[0024] B2: The real-time environmental monitoring data is divided into bins in chronological order. Each bin contains monitoring data of a fixed time length, and the average value of the monitoring data replaces the bin data.

[0025] B3: Divide the structure into different states according to the monitoring data;

[0026] B4: Obtain the prediction interval corresponding to the data to be evaluated of the monitoring indicator, and calculate its correlation with the level interval of each state of the structure through the correlation function of the set extension theory;

[0027] B5: Use Pearson distance to assign values ​​to the dynamic performance level intervals, analyze the status within the group, achieve quantitative evaluation of the structural dynamic performance, and obtain the dynamic performance evaluation value based on environmental condition classification.

[0028] Furthermore, the fusion of acceleration data in step A2 is specifically implemented as follows:

[0029] Assuming that there are n vibration sensors used to monitor the vibration of the OWT, the vibration data sequence obtained from the n sensors is denoted as {X 1 ,X 2 ,…X i …,X n}, where X i (i=1,2,...,n) represents the vibration data obtained by the i-th sensor, and the weight value w of each sensor i Calculated by formula (1):

[0030]

[0031] Among them, H i represents the information entropy value of the i-th sensor, H i The smaller the value, the more effective information the corresponding sensor provides, and the greater the corresponding weight.

[0032] Furthermore, the noise reduction in step A3 is implemented as follows:

[0033] The Golden Leopard optimization algorithm is used to optimize the selection process of VMD parameters with the sum of the envelope entropy of the intrinsic mode function components as the fitness function. The optimal decomposition layer number K and the quadratic penalty factor are selected autonomously. Then, the structural feature information is extracted. According to the optimal VMD parameter combination, the field monitoring data is decomposed into multiple IMF components. Finally, the correlation coefficient between each IMF component and the original data is calculated, and the IMF with a higher correlation coefficient is selected to reconstruct the data.

[0034] Furthermore, other swarm intelligence optimization algorithms such as spectral algorithm, spider bee optimization algorithm, energy valley optimization algorithm, zebra optimization algorithm, etc. can also be used, or traditional swarm intelligence optimization algorithms can be improved to replace the GJO algorithm to achieve autonomous optimization of VMD parameters.

[0035] Furthermore, step A4 also includes:

[0036] In the modeling process, the Akaike information criterion and the Bayesian information criterion are used to jointly determine the order of the ARMA model, and the AIC and BIC values ​​after modeling with different parameters p and q are calculated. When the calculated value is the smallest, p and q at this time are considered to be the best parameters of the model, and the model established by the best parameters is the optimal model; at the same time, in order to reduce the uncertainty of parameter estimation and improve the stability of model coefficients, a regularization factor α is added in the modeling process, and the L-curve criterion method is used to select the regularization coefficient α, and the solution norm ln||x α || 2 and the residual norm ln||y-Ax α || 2 Plotted on the horizontal and vertical axes to form an L curve, the regularization parameter α corresponding to the maximum curvature value is the best choice;

[0037] Furthermore, step A5 also includes:

[0038] The Mahalanobis square distance between the AR coefficients is calculated by formula (2). When the Mahalanobis square distance between the AR coefficient of the ARMA model of the data to be tested and the AR coefficient of the health status is close, it means that the performance of the structure has not changed significantly. On the contrary, when the Mahalanobis square distance between the AR coefficient of the ARMA model of the data to be tested and the AR coefficient of the health status is large, it means that the performance of the structure has changed significantly.

[0039]

[0040] Furthermore, the specific implementation of the warehouse division method in step B2 is as follows:

[0041] The monitored objects are wind speed and wave height. The average environmental monitoring data level R is calculated based on the range of the average environmental monitoring data under the healthy dynamic performance state and the preset wind speed change range Δv and wave height change range Δh, which is obtained by formula (3):

[0042]

[0043] In the formula, R ν and R h Respectively represent the level of wind speed and wave height monitoring data; Represents the maximum and minimum bin data of wind speed and wave height respectively;

[0044] By calculating formula (3), the environmental monitoring data can be divided into several level intervals. According to formula (4), the inclination response data corresponding to the environmental monitoring data of different partitions in the same horizontal interval are clustered into the same sub-data partition:

[0045]

[0046] In the formula, C represents the clustering method, x ij represents the structural response data in the environmental monitoring data bin corresponding to the i-th response index, k ν,h The number of categories that represent the number of environments;

[0047] Furthermore, the specific implementation of the state division in step B3 is:

[0048] The monitoring data obtained in step B2 is divided into a healthy state data segment and a to-be-evaluated data segment. The healthy state data segment uses historical monitoring data, and the to-be-evaluated monitoring data segment uses real-time monitoring data. The healthy state data segment is further divided into n data subgroups, and the structural response monitoring data in each data subgroup is represented as x ij.l , (l=1,2,…,n), where n represents the number of subgroups, and the mean sequence corresponding to the health status data segment is calculated The average and variance σ, and the monitoring data segment to be evaluated is processed in the same way to calculate the monitoring data y of the structural response to be evaluated ij.l In the corresponding sub-database and Deviation Δ l , standardize the data distribution of the subgroups corresponding to the monitored data segment to be evaluated:

[0049]

[0050] right Take the average and maximum values ​​to form a prediction interval, expressed as [ψ l.mean ,ψ l.max ]; Taking the healthy state data segment as the baseline data, the change of the structural dynamic performance is evaluated by judging the deviation between the prediction interval and the baseline data. The greater the deviation, the greater the change of the structural dynamic performance; the smaller the deviation, the structural dynamic performance has not changed;

[0051] According to the Laida criterion, the upper limit of the maximum interval of the healthy state structural response monitoring data is defined as 3σ, and the structural dynamic performance state is further divided into four levels, namely healthy [0, 1.5σ]; sub-healthy [1.5σ, 2σ]; slightly abnormal [2σ, 3σ]; abnormal [3σ, INF], and the upper limit INF is not included in the calculation;

[0052] Furthermore, step B4 further includes:

[0053] Redefine the prediction interval to be X i = <x 1 ,x 2 >, the health level interval is defined as A= 1 ,a 2 >, when x 2 -x​1 2 -a 1 Time, interval<x,y> and interval<a,b> The distance ρ between them is expressed as:

[0054]

[0055] Interval correlation K(X i ) is calculated as follows:

[0056]

[0057] Where D(Xi,A,B) is the position value, indicating the position relationship between the intervals. Its calculation formula is:

[0058]

[0059] For the case that does not satisfy x 2 -x 1 2 -a 1 In the case of , swap the order of X and A.

[0060] The information entropy and network analytic hierarchy process (ANP) are combined to determine the initial weight of the bottom index according to the information entropy value. This method is used to calculate the mutual influence and weight distribution A of each inclination index of the root layer. Finally, the final interval correlation F(X) is obtained according to formula (9).

[0061]

[0062] Furthermore, the specific implementation of step B5 is as follows:

[0063] Assume that the measured data to be evaluated is X = {x i}, i = l, 2, ..., n corresponds to the healthy data in the load group as Y = {y i}, i = I, 2, ..., n; the Pearson correlation coefficient of the two sets of data is shown in formula (10):

[0064]

[0065] The length of the health data segment is m times that of the data to be evaluated. There will be m Pearson correlation coefficients between the measured data to be evaluated and the health data. The weighted average Pearson correlation coefficient is taken. Measures the degree of correlation between two sets of data. The range of the correlation coefficient is [-1,1]. When the value is a negative number, it means that the two sets of data are negatively correlated, otherwise it is positively correlated.

[0066] Pearson distance is defined as ​​Its value range is [0,1], and the smaller the value, the higher the correlation;

[0067] A method combining state interval assignment and Pearson distance is used to quantitatively evaluate the degree of data difference. First, different evaluation intervals are assigned values, and the evaluation intervals are divided into (0, 1.5σ), (1.5σ, 2.5σ), (2.5σ, 3σ), and (3σ, ∞), and the values ​​are assigned as 1, 2, 3, and 4 respectively. Then, the Pearson distance P is summed with the corresponding evaluation interval assignment to obtain the final quantitative evaluation value of the single indicator. Finally, the correlation degree of each indicator and its participation weight are combined to obtain the final structural dynamic performance evaluation value.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] The offshore wind power structure monitoring data denoising technology based on GJO-VMD overcomes the technical defects of traditional denoising methods, improves the accuracy of extracting effective information, and realizes the autonomy of the denoising process.

[0070] The offshore wind power structure dynamic performance evaluation technology based on data fusion solves the engineering and technical problem of poor coupling correlation between the sensor data of offshore wind power structures. This technical method combines the ARMA model, has a simple and fast calculation process, only requires response signals for modeling, and does not require external excitation information. The autoregressive coefficient of the model can well reflect the characteristics of the structure, realizing real-time evaluation of the dynamic performance of offshore wind power structures.

[0071] The offshore wind power structure dynamic performance assessment technology based on environmental condition classification takes into account the relationship between environmental loads and responses. It pre-processes the structural response monitoring data by binning the environmental monitoring data, which can effectively avoid the impact of environmental changes on the structural dynamic performance assessment results. This technical method, combined with the interval probability model, can reduce the impact of outliers on the structural dynamic performance assessment results. At the same time, the interval correlation calculation method can further improve the probability model of performance assessment and improve the reliability of structural dynamic performance assessment.

[0072] The offshore wind power structure dynamic performance assessment technology based on the root morphology model is a technology that further integrates the two technical methods. Compared with the structural dynamic performance assessment technology based on data fusion, it not only considers the structural acceleration response data, but also the inclination response data and wind speed and wave height monitoring data, effectively avoiding the uncertainty of a single type of sensor and improving the reliability of the assessment results. Compared with the structural dynamic performance assessment technology based on environmental condition classification, it effectively avoids the problem of being unable to conduct an assessment due to the environmental monitoring data exceeding the existing maximum value of the health database.

[0073] The offshore wind power structure dynamic performance evaluation method based on the root morphology model of the present application integrates the dynamic performance evaluation technology based on data fusion and the dynamic performance evaluation technology based on environmental condition classification. The former uses an autoregressive sliding average model to perform time series analysis on the acceleration response data of the offshore wind power structure, and the latter comprehensively considers the environmental monitoring data and the structural response data.

[0074] The dynamic performance evaluation method of offshore wind power structures based on the root morphology model makes up for the defects of the two dynamic performance evaluation technologies in practical engineering applications and improves the reliability of the dynamic performance evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 This is a schematic diagram of the root morphology model of the present invention;

[0076] Figure 2 It is a schematic diagram of the structural dynamic performance evaluation process based on the root morphology model of the present invention;

[0077] Figure 3 It is a schematic diagram of the L-curve criterion method of the present invention;

[0078] Figure 4 It is a schematic diagram of the steps of the improved ANP model for evaluating the dynamic performance of offshore wind power structures of the present invention;

[0079] Figure 5 The structural dynamic performance evaluation results based on data fusion

[0080] Figure 6 Dynamic evaluation of structural dynamic performance based on wave height-wind speed environmental condition classification

[0081] Figure 7 The results of dynamic performance evaluation based on the root morphology model. DETAILED DESCRIPTION

[0082] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0083] See also Figure 1 , the present invention provides a technical solution.

[0084] A method for evaluating the dynamic performance of offshore wind turbine structures integrates the wind turbine structure dynamic performance evaluation technology based on the root morphology model;

[0085] The root morphology model includes the main root layer, lateral root layer and fibrous root layer, and the number of lateral root layers is greater than 1;

[0086] The main root layer is fused and evaluated based on the data of the first-level lateral root layer, and the upper-level lateral root layer is fused and evaluated based on the data of the lower-level lateral root layer; the lateral root layer is fused and evaluated based on the data of the corresponding fibrous root layer.

[0087] The main root layer is for the dynamic performance evaluation of offshore wind power structures; the first-level lateral root layer includes structural performance dynamic performance evaluation based on data fusion and structural dynamic performance evaluation based on environmental condition classification.

[0088] The secondary lateral roots of the structural performance dynamic evaluation based on data fusion include the evaluation values ​​of the wind turbine foundation and tower, and the fibrous root layer under the secondary lateral root layer includes the acceleration monitoring data fusion of the wind turbine foundation and tower.

[0089] The secondary lateral roots based on the dynamic evaluation of structural performance fusion with environmental data include the evaluation of wave inclination angle and wind speed inclination angle. The tertiary lateral roots based on the dynamic evaluation of structural performance based on environmental data fusion include the evaluation of wind turbine foundation and tower. The fibrous root layer under the tertiary lateral root layer includes different inclinometers of the wind turbine foundation and tower and inclination data in different directions.

[0090] See also Figures 2 to 4 , an offshore wind power structure performance evaluation method is evaluated according to an evaluation value of the wind power structure dynamic performance, and the evaluation value of the wind power structure dynamic performance is the sum of a dynamic performance evaluation value based on data fusion and an evaluation value based on environmental condition classification;

[0091] The steps for obtaining the power performance evaluation value based on data fusion include:

[0092] A1: Collect sensor acceleration monitoring data of wind turbine foundation and tower;

[0093] A2: Fusion of multi-sensor acceleration data. The fusion of acceleration data is specifically implemented as follows:

[0094] Assuming that there are n vibration sensors used to monitor the vibration of the OWT, the vibration data sequence obtained from the n sensors is denoted as {X 1 ,X 2 ,…X i …,X n}, where X i (i=1,2,...,n) represents the vibration data obtained by the i-th sensor, and the weight value w of each sensor i Calculated by formula (1):

[0095]

[0096] Among them, H irepresents the information entropy value of the i-th sensor, H i The smaller the value, the more effective information the corresponding sensor provides, and the greater the corresponding weight.

[0097] A3: The fused acceleration data is denoised using the GJO-VMD data denoising technology. The denoising is achieved as follows:

[0098] The Golden Leopard optimization algorithm is used to optimize the selection process of VMD parameters with the sum of the envelope entropy of the intrinsic mode function components as the fitness function. The optimal decomposition layer number K and the quadratic penalty factor are selected autonomously. Then, the structural feature information is extracted. According to the optimal VMD parameter combination, the field monitoring data is decomposed into multiple IMF components. Finally, the correlation coefficient between each IMF component and the original data is calculated, and the IMF with a higher correlation coefficient is selected to reconstruct the data.

[0099] Other swarm intelligence optimization algorithms such as spectral algorithm, spider bee optimization algorithm, energy valley optimization algorithm, zebra optimization algorithm, etc. can also be used, or traditional swarm intelligence optimization algorithms can be improved to replace the GJO algorithm to achieve autonomous optimization of VMD parameters.

[0100] A4: Use random reduction technology to extract free response signals that are unrelated to environmental excitation from the denoised data, use the autoregressive moving average model to model and analyze the free response signals, and extract the AR coefficient of the model to characterize the dynamic performance characteristics of the structure;

[0101] In the modeling process, the Akaike information criterion and the Bayesian information criterion are used to jointly determine the order of the ARMA model, and the AIC and BIC values ​​after modeling with different parameters p and q are calculated. When the calculated value is the smallest, p and q at this time are considered to be the best parameters of the model, and the model established by the best parameters is the optimal model; at the same time, in order to reduce the uncertainty of parameter estimation and improve the stability of model coefficients, a regularization factor α is added in the modeling process, and the L-curve criterion method is used to select the regularization coefficient α, and the solution norm ln||x α || 2 and the residual norm ln||y-Ax α || 2 Plotted on the horizontal and vertical axes to form an L curve, the regularization parameter α corresponding to the maximum curvature value is the best choice;

[0102] A5: Finally, the Mahalanobis square distance between the AR coefficients is calculated by formula (2) to evaluate the dynamic performance of the wind power structure and obtain the dynamic performance evaluation value based on data fusion;

[0103] When the Mahalanobis distance between the AR coefficient of the ARMA model of the data to be tested and the AR coefficient of the health state is relatively close, it means that the performance of the structure has not changed significantly. On the contrary, when the Mahalanobis distance between the AR coefficient of the ARMA model of the data to be tested and the AR coefficient of the health state is relatively large, it means that the performance of the structure has changed significantly.

[0104]

[0105] The steps for obtaining the power performance evaluation value based on environmental condition classification include:

[0106] B1: Collect inclination data of different inclinometers and different directions of wind turbine foundation and tower;

[0107] B2: The real-time environmental monitoring data is divided into bins in chronological order. Each bin contains monitoring data of a fixed time length, and the average value of the monitoring data replaces the bin data.

[0108] The monitored objects are wind speed and wave height. The average environmental monitoring data level R is calculated based on the range of the average environmental monitoring data under the healthy dynamic performance state and the preset wind speed change range Δv and wave height change range Δh, which is obtained by formula (3):

[0109]

[0110] In the formula, R ν and R h Respectively represent the level of wind speed and wave height monitoring data; Represents the maximum and minimum bin data of wind speed and wave height respectively;

[0111] By calculating formula (3), the environmental monitoring data can be divided into several level intervals. According to formula (4), the inclination response data corresponding to the environmental monitoring data of different partitions in the same horizontal interval are clustered into the same sub-data partition:

[0112]

[0113] In the formula, C represents the clustering method, x ij represents the structural response data in the environmental monitoring data bin corresponding to the i-th response index, k ν,h The number of categories that represent the number of environments;

[0114] B3: Divide the structure into different states according to the monitoring data;

[0115] The monitoring data obtained in step B2 is divided into a healthy state data segment and a to-be-evaluated data segment. The healthy state data segment uses historical monitoring data, and the to-be-evaluated monitoring data segment uses real-time monitoring data. The healthy state data segment is further divided into n data subgroups, and the structural response monitoring data in each data subgroup is represented as x ij.l , (l=1,2,…,n), where n represents the number of subgroups, and the mean sequence corresponding to the health status data segment is calculated The average and variance σ, and the monitoring data segment to be evaluated is processed in the same way to calculate the monitoring data y of the structural response to be evaluated ij.l In the corresponding sub-database and Deviation Δ l , standardize the data distribution of the subgroups corresponding to the monitored data segment to be evaluated:

[0116]

[0117] right Take the average and maximum values ​​to form a prediction interval, expressed as [ψ l.mean ,ψ l.max ]; Taking the healthy state data segment as the baseline data, the change of the structural dynamic performance is evaluated by judging the deviation between the prediction interval and the baseline data. The greater the deviation, the greater the change of the structural dynamic performance; the smaller the deviation, the structural dynamic performance has not changed;

[0118] According to the Laida criterion, the upper limit of the maximum interval of the healthy state structural response monitoring data is defined as 3σ, and the structural dynamic performance state is further divided into four levels, namely healthy [0, 1.5σ]; sub-healthy [1.5σ, 2σ]; slightly abnormal [2σ, 3σ]; abnormal [3σ, INF], and the upper limit INF is not included in the calculation;

[0119] B4: Obtain the prediction interval corresponding to the data to be evaluated of the monitoring indicator, and calculate its correlation with the level interval of each state of the structure through the correlation function of the set extension theory;

[0120] Redefine the prediction interval to be X i = <x 1 ,x 2 >, the health level interval is defined as A= 1 ,a 2 >, when x 2 -x 1 2 -a 1 Time, interval<x,y> and interval<a,b> The distance ρ between them is expressed as:

[0121] ​​

[0122] Interval correlation K(X i ) is calculated as follows:

[0123]

[0124] Where D(Xi,A,B) is the position value, indicating the position relationship between the intervals. Its calculation formula is:

[0125]

[0126] For the case that does not satisfy x 2 -x 1 2 -a 1 In the case of , swap the order of X and A.

[0127] The information entropy and network analytic hierarchy process are combined to determine the initial weight of the bottom index according to the information entropy value. This method is used to calculate the mutual influence and weight distribution A of each inclination index of the root layer. Finally, the final interval correlation F(X) is obtained according to formula (9).

[0128]

[0129] B5: Use Pearson distance to assign values ​​to the dynamic performance level intervals, analyze the status within the group, achieve quantitative evaluation of the structural dynamic performance, and obtain the dynamic performance evaluation value based on environmental condition classification.

[0130] Assume that the measured data to be evaluated is X = {x i}, i = l, 2, ..., n corresponds to the healthy data in the load group as Y = {y i}, i = I, 2, ..., n; the Pearson correlation coefficient of the two sets of data is shown in formula (10):

[0131]

[0132] The length of the health data segment is m times that of the data to be evaluated. There will be m Pearson correlation coefficients between the measured data to be evaluated and the health data. The weighted average Pearson correlation coefficient is taken. Measures the degree of correlation between two sets of data. The range of the correlation coefficient is [-1,1]. When the value is a negative number, it means that the two sets of data are negatively correlated, otherwise it is positively correlated.

[0133] Pearson distance is defined as Its value range is [0,1], and the smaller the value, the higher the correlation;

[0134] ​A method combining state interval assignment and Pearson distance is used to quantitatively evaluate the degree of data difference. First, different evaluation intervals are assigned values, and the evaluation intervals are divided into (0, 1.5σ), (1.5σ, 2.5σ), (2.5σ, 3σ), and (3σ, ∞), and the values ​​are assigned as 1, 2, 3, and 4 respectively. Then, the Pearson distance P is summed with the corresponding evaluation interval assignment to obtain the final quantitative evaluation value of the single indicator. Finally, the correlation degree of each indicator and its participation weight are combined to obtain the final structural dynamic performance evaluation value.

[0135] The experimental method is as follows:

[0136] In order to verify the effectiveness of the dynamic performance evaluation method based on the root morphology model, an offshore wind farm project in the coastal waters of Rudong, Jiangsu, 40 kilometers offshore, was selected as a typical project. A total of 50 wind turbines were built in the project, and one of them was selected as a demonstration wind turbine of the typical project. The data during the period when Typhoon "Yanhua" passed through the demonstration wind turbine were analyzed in detail.

[0137] ① Five dynamic acceleration sensors, two dynamic tilt sensors, one wave radar and one ultrasonic wind speed and direction sensor were arranged on the demonstration wind turbine structure to monitor the structure in real time.

[0138] ② According to the site conditions, five triaxial acceleration sensors are arranged at different positions on the inner wall of the structure tower, and dynamic inclination sensors are arranged at the foundation flange layer and the twist cable layer. The dynamic acceleration response signal and the inclination response signal are collected by the CANSAS-SC8I collector produced by IMC of Germany. Wind speed, wind direction and water level wave radars are arranged on the platform outside the foundation flange layer to monitor the wind and wave environmental conditions near the demonstration wind turbine, and video surveillance equipment is set up to monitor the site. The dynamic acceleration sensor, dynamic inclination sensor, wave radar, ultrasonic wind speed and direction sensor and other real-time monitoring equipment of the demonstration wind turbine structure were arranged on May 19, 2021, and data collection began. The collection time lasted until April 16, 2022. During this period, wind power maintenance, power outages, equipment restarts and other reasons would cause the monitoring equipment to fail to operate normally, resulting in data loss and inability to collect data. In the final data collection period, a total of about 299 days of monitoring data were collected. Among them, Typhoon "Fireworks" passed through the demonstration wind farm on July 24 and 25, 2021.

[0139] The monitoring data were analyzed using a dynamic performance evaluation method based on the root morphology model, and combined with the structural dynamic performance evaluation results based on data fusion and wave height-wind speed environmental conditions classification, the dynamic performance evaluation value of the wind turbine foundation changed significantly during the passage of the typhoon on July 25 and 26. After the typhoon passed, the change in the dynamic performance evaluation value became smaller, proving that the dynamic performance of the wind turbine foundation will change under the influence of the typhoon. At the same time, this change is within the range of the wind turbine foundation's elastic recovery, and no irreversible damage has occurred, which is consistent with the situation that the actual project can still serve normally after the typhoon.

[0140] In addition, when the variation range of the structural dynamic performance evaluation data based on environmental condition classification is very small, the peaks and troughs of the structural dynamic performance evaluation data based on the root morphology model and data fusion are close. It can be seen that the structural dynamic performance evaluation based on the environmental root morphology model comprehensively evaluates the environmental monitoring data and vibration response data, effectively avoiding the uncertainty of a single type of sensor, and can further improve the reliability and accuracy of the evaluation results.

[0141] Although embodiments of the present invention have been shown and described (see the above detailed description for details), it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the dynamic performance of an offshore wind power structure, characterized in that: The evaluation method integrates the wind power structure dynamic performance evaluation technology based on the root morphology model; The root system morphology model includes a main root layer, a lateral root layer and a fibrous root layer, and the level of the lateral root layer is greater than 1; The main root layer is fused and evaluated according to the data of the first-level lateral root layer, and the upper-level lateral root layer is fused and evaluated according to the data of the lower-level lateral root layer; the lateral root layer is fused and evaluated according to the corresponding fibrous root layer data; The main root layer is for the dynamic performance evaluation of offshore wind power structures; the first-level lateral root layer includes structural performance dynamic evaluation based on data fusion and structural dynamic performance evaluation based on environmental condition classification; The secondary lateral roots of the structural performance dynamic evaluation based on data fusion include the evaluation values ​​of the wind turbine foundation and the tower, and the fibrous root layer under the secondary lateral root layer includes the acceleration monitoring data fusion of the wind turbine foundation and the tower; The secondary lateral root based on the environmental data fusion structural performance dynamic evaluation includes the evaluation of wave inclination angle and wind speed inclination angle, and the tertiary lateral root based on the environmental data fusion structural performance dynamic evaluation includes the evaluation of wind turbine foundation and tower; the fibrous root layer under the tertiary lateral root layer includes different inclinometers of wind turbine foundation and tower and inclination data in different directions; The evaluation method performs evaluation based on an evaluation value of the dynamic performance of the wind power structure, wherein the evaluation value of the dynamic performance of the wind power structure is the sum of the dynamic performance evaluation value based on data fusion and the evaluation value based on environmental condition classification; The step of obtaining the power performance evaluation value based on data fusion includes: A1: Collect sensor acceleration monitoring data of wind turbine foundation and tower; A2: Fusion of multi-sensor acceleration data; A3: De-noise the fused acceleration data using the GJO-VMD-based data de-noising technology; A4: Use random reduction technology to extract free response signals that are unrelated to environmental excitation from the denoised data, use the autoregressive moving average model to model and analyze the free response signals, and extract the AR coefficient of the model to characterize the dynamic performance characteristics of the structure; A5: Calculate the Mahalanobis square distance between AR coefficients, evaluate the dynamic performance of the wind power structure, and obtain the dynamic performance evaluation value based on data fusion; The step of obtaining the power performance evaluation value based on environmental condition classification includes: B1: Collect inclination data of different inclinometers and different directions of wind turbine foundation and tower; B2: The real-time environmental monitoring data is divided into bins in chronological order. Each bin contains monitoring data of a fixed time length, and the average value of the monitoring data replaces the bin data. B3: Divide the structure into different states according to the monitoring data; B4: Obtain the prediction interval corresponding to the data to be evaluated of the monitoring indicator, and calculate its correlation with the level interval of each state of the structure through the correlation function of the set extension theory; B5: Use Pearson distance to assign values ​​to the dynamic performance level intervals, analyze the status within the group, achieve quantitative evaluation of the structural dynamic performance, and obtain the dynamic performance evaluation value based on environmental condition classification.

2. A method for evaluating the dynamic performance of an offshore wind power structure according to claim 1, characterized in that: The fusion of acceleration data in step A2 is specifically implemented as follows: Assuming that there are n vibration sensors used to monitor the vibration of the OWT, the vibration data sequence obtained from the n sensors is recorded as {X1,X2,…X i …,X n }, X i represents the vibration data obtained by the i-th sensor, where i = 1, 2, ..., n, and the weight value w of each sensor i Calculated by formula (1): Among them, H i represents the information entropy value of the i-th sensor, H i The smaller the value, the more effective information the corresponding sensor provides, and the greater the corresponding weight.

3. The method for evaluating the dynamic performance of an offshore wind power structure according to claim 1, characterized in that: The implementation method of noise reduction described in step A3 is as follows: Using the golden leopard optimization algorithm, with the sum of the envelope entropy of the intrinsic mode function components as the fitness function, optimize the selection process of VMD parameters, autonomously select the optimal decomposition layer number K and the quadratic penalty factor, then extract the structural feature information, decompose the on-site monitoring data into multiple IMF components according to the optimal VMD parameter combination, and finally calculate the correlation coefficient between each IMF component and the original data, and select the IMF with a higher correlation coefficient to reconstruct the data; Step A4 also includes: In the modeling process, the Akaike information criterion and the Bayesian information criterion are used to jointly determine the order of the ARMA model, and the AIC and BIC values ​​after modeling with different parameters p and q are calculated. When the calculated value is the smallest, p and q at this time are considered to be the best parameters of the model, and the model established by the best parameters is the optimal model; at the same time, in order to reduce the uncertainty of parameter estimation and improve the stability of model coefficients, a regularization factor α is added in the modeling process, and the L-curve criterion method is used to select the regularization coefficient α, and the solution norm ln||x α ||2 and residual norm ln||y-Ax α ||2 is plotted on the horizontal and vertical axes to form an L curve, and the regularization parameter α corresponding to the maximum curvature value is the best choice; Step A5 also includes: Calculate the Mahalanobis square distance between AR coefficients through formula (2). When the Mahalanobis distance between the AR coefficients of the ARMA model of the measured data and the AR coefficients of the healthy state is relatively close, it indicates that the performance of the structure has not changed significantly. On the contrary, when the Mahalanobis distance between the AR coefficients of the ARMA model of the measured data and the healthy state coefficients has a large gap, it indicates that the performance of the structure has changed significantly 。 4. The method for evaluating the dynamic performance of an offshore wind power structure according to claim 1, characterized in that: The specific implementation of the binning method in step B2 is as follows: The monitored objects are wind speed and wave height. Calculate the average environmental monitoring data level R according to the range of the average environmental monitoring data under the healthy dynamic performance state and the preset wind speed change range Δv and wave height change range Δh, which is obtained from formula (3): In the formula, R ν and R h Respectively represent the level of wind speed and wave height monitoring data; Represents the maximum and minimum bin data of wind speed and wave height respectively; Through the calculation of formula (3), the environmental monitoring data can be divided into several grade intervals. According to formula (4), the inclination response data corresponding to the environmental monitoring data in different partitions within the same horizontal interval are clustered into the same sub-data partition: In the formula, C represents the clustering method, x ij represents the structural response data in the environmental monitoring data bin corresponding to the i-th response index, k ν,h The number of categories that represent the number of environments; The specific implementation method of state division in step B3 is: The monitoring data obtained in step B2 is divided into a healthy state data segment and a to-be-evaluated data segment. The healthy state data segment uses historical monitoring data, and the to-be-evaluated monitoring data segment uses real-time monitoring data. The healthy state data segment is further divided into n data subgroups, and the structural response monitoring data in each data subgroup is represented as x ij.l , where l = 1, 2, ..., n, n represents the number of subgroups, and the mean sequence corresponding to the health status data segment is calculated The average and variance σ, and the monitoring data segment to be evaluated is processed in the same way to calculate the monitoring data y of the structural response to be evaluated ij.l In the corresponding sub-database and Deviation Δ l , standardize the data distribution of the subgroup corresponding to the monitored data segment to be evaluated: right Take the average and maximum values ​​to form a prediction interval, expressed as [ψ l.mean ,ψ l.max ]; Taking the healthy state data segment as the baseline data, evaluate the change of the structural dynamic performance by judging the deviation degree between the prediction interval and the baseline data. The greater the deviation degree, the greater the change of the structural dynamic performance; the smaller the deviation degree, the structural dynamic performance has not changed; According to the Pauta criterion, define the upper limit of the maximum interval of the healthy state structure response monitoring data as 3σ, and further divide the structural dynamic performance state into four grades, namely healthy [0, 1.5σ]; sub-healthy [1.5σ, 2σ]; mildly abnormal [2σ, 3σ]; abnormal [3σ, INF], and the upper limit INF does not participate in the calculation.

5. A method for evaluating the dynamic performance of an offshore wind power structure according to claim 1, characterized in that Step B4 also includes: Redefine the prediction interval as X i =<x1,x2>, the health level interval is defined as A = <a1,a2>. When x2 - x1 < a2 - a1, the distance ρ between the interval <x,y> and the interval <a,b> is expressed as: Interval correlation K(X i ) is calculated as follows: In the formula, D(Xi, A, B) is the position value, indicating the position relationship between intervals, and its calculation formula: In the formula, for the case where x2 - x1 < a2 - a1 is not satisfied, swap the order of X and A; Combine information entropy and the analytic hierarchy process. Determine the initial weights of the underlying indicators according to the information entropy value. Calculate the mutual influence and weight distribution A of each inclination index of the fibrous root layer through this method, and finally obtain the final interval correlation degree F(X) according to formula (9) 。 6. A method for evaluating the dynamic performance of an offshore wind power structure according to claim 3, characterized in that: The specific implementation of step B5 is as follows: Assume that the measured data to be evaluated is X = x i ,i=l,2,…,nThe corresponding health data in the load group is Y=y i ,i=I,2,…,n; the Pearson correlation coefficient of the two sets of data is shown in formula (10): The length of the health data segment is m times that of the data to be evaluated. There will be m Pearson correlation coefficients between the measured data to be evaluated and the health data. The weighted average Pearson correlation coefficient is taken. Measures the degree of correlation between two sets of data. The range of the correlation coefficient is [-1,1]. When the value is a negative number, it means that the two sets of data are negatively correlated, otherwise it is positively correlated. Pearson distance is defined as Its value range is [0,1], and the smaller the value, the higher the correlation; A method combining state interval assignment and Pearson distance is used to quantitatively evaluate the degree of data difference. First, different evaluation intervals are assigned values, and the evaluation intervals are divided into (0, 1.5σ), (1.5σ, 2.5σ), (2.5σ, 3σ), and (3σ, ∞), and the values ​​are assigned as 0, 1, 2, and 3 respectively. Then, the Pearson distance P is summed with the corresponding evaluation interval assignment to obtain the final quantitative evaluation value of the single indicator. Finally, the correlation degree of each indicator and its participation weight are combined to obtain the final structural dynamic performance evaluation value.

Images