Wind farm power generation prediction method considering correlation of site wind speed and fan failure

Abstract

The application discloses a wind farm power generation capacity prediction method considering the correlation of site wind speed and fan failure, comprising: predicting the future wind speed distribution and generating long time series wind speed based on MCP method and LSTM; obtaining the initial wind speed of each fan site through vertical extrapolation and spatial correlation extrapolation combined with Copula function; calculating the wind speed reduction by using the wake model; innovatively introducing the fan random failure simulation based on Poisson distribution, and dynamically coupling the failure and the wake effect in the simulation, that is, when the upstream fan fails and stops, the wake effect of the downstream fan is simultaneously removed; finally, combining the power curve, the probability distribution and uncertainty of the future annual power generation capacity of the wind farm are obtained through multiple Monte Carlo simulations. The application overcomes the defects that the prior art does not consider the randomness of failure and the dynamic interaction with the wake, and significantly improves the accuracy of long-term power generation capacity prediction and the comprehensiveness of risk assessment.

Description

Technical Field

[0001] This invention belongs to the field of wind resource assessment technology, specifically a method for predicting wind farm power generation that considers the correlation between turbine location and wind speed and wind turbine failure. Background Technology

[0002] Wind energy, as a clean energy source with abundant reserves, wide distribution, and a high degree of commercialization, has become a core direction of global energy transition. Wind farms are the main form of large-scale wind energy utilization, and the accuracy of their power generation forecasts is crucial for farm planning, grid dispatching, operation management, electricity market trading, and investment risk assessment. The strong randomness and intermittency of wind speed, as well as the time-varying nature of wind turbine operating conditions, result in significant uncertainties in wind farm output. Therefore, developing methods to quantify uncertainties in future power generation calculations is of great significance for ensuring the safe and stable operation of the power system and improving the economic benefits of wind power.

[0003] Currently, wind farm future power generation forecasts can be categorized into ultra-short-term, short-term, medium-term, and long-term forecasts based on time scale. Long-term forecasts, especially the assessment of future annual power generation, primarily serve strategic decisions such as wind farm investment planning, annual electricity trading, and the formulation of long-term maintenance plans. Due to the long time span, the uncertainties involved are more complex, and probabilistic methods based on Monte Carlo simulations are typically used for assessment.

[0004] Existing technologies have made some progress in the long-term power generation probability assessment of wind farms, mainly focusing on uncertainty modeling and propagation in the following aspects: (1) Wind resource assessment: This includes using the Measurement-Correlation-Prediction (MCP) method combined with long-term reference data to extrapolate the long-term wind speed sequence of the wind tower, performing vertical extrapolation based on the wind shear index, and performing horizontal spatial extrapolation using wind flow models. This type of research aims to quantify the uncertainty of wind resources in time and space distribution.

[0005] (2) Wake effect calculation: The wake models of Jensen and Gauss are used to calculate the mutual influence between wind turbines and to analyze the uncertainty brought about by the model parameters. Some studies have integrated the wake model and its uncertainty into the Monte Carlo simulation framework to evaluate the uncertainty of the overall wind farm and the power generation of individual turbines.

[0006] (2) Wind turbine failure handling: Current research and practice mainly address the impact of wind turbine failures on power generation through two methods: Deterministic reduction based on historical statistics: This method involves statistically analyzing a large amount of wind turbine operating data to summarize fault patterns, or directly using a comprehensive empirical reduction factor in engineering calculations to approximate the power loss due to faults. However, this approach treats faults as a fixed, averaged loss, failing to reflect their random occurrence.

[0007] Fault estimation for short-term operations: In short-term profit forecasting or operations management, planned and unplanned downtime for a specific future period is estimated by calling historical maintenance logs. Although this method involves faults, it is essentially a deterministic estimation based on historical data and is not coupled with wind farm physical models (such as wake).

[0008] However, existing technologies have obvious limitations and gaps, as follows.

[0009] First, there is a lack of probabilistic modeling of wind turbine failures and their systematic integration in long-term power generation assessment. Current mainstream methods either use fixed reduction rates or perform deterministic estimations, neither of which can characterize the core stochastic characteristics of the random occurrence of failures—the timing, frequency, and duration—making it difficult to accurately assess the uncertainty in power generation introduced by failures.

[0010] Secondly, the dynamic coupling effect between wind turbine failure and wake effect is seriously neglected. In actual wind farms, the failure and shutdown of a wind turbine not only results in zero power generation for the turbine itself, but also alters the flow field structure in its downstream region. If the failed turbine was originally located upstream, its downstream turbines will no longer be affected by its wake, wind speed may recover, and power generation will increase accordingly. This "failure-wake" coupling effect has not been considered in existing power generation assessment methods. Existing wake models only apply to the state where all wind turbines are operating normally and do not include a dynamic correction mechanism for the wake field caused by a wind turbine randomly shutting down.

[0011] Finally, existing research has focused on the independent analysis of a single or a few sources of uncertainty (e.g., analyzing wind speed extrapolation uncertainty or wake model uncertainty separately), rather than establishing a unified comprehensive evaluation framework that can simultaneously handle the spatiotemporal uncertainty of wind resources, the uncertainty of random failures of wind turbines, and the complex coupling effects between the two. Summary of the Invention

[0012] In view of this, the purpose of this invention is to provide a wind farm power generation prediction method that considers the correlation between turbine location and wind speed and turbine failure. By establishing a Poisson stochastic model of turbine failure and dynamically correcting the wake effect downstream of the failed turbine in Monte Carlo simulation, the method can more realistically and comprehensively reflect the combined effect of equipment reliability and flow field interaction, and finally obtain a more accurate probability distribution and uncertainty assessment result of the wind farm's future annual power generation.

[0013] To achieve the above objectives, the present invention provides the following technical solution: A method for predicting wind farm power generation that considers the correlation between turbine location and wind speed, as well as turbine failure, includes the following steps: Step 1: Obtain the long-term reference wind speed time history U and the short-term measured wind speed time history V at the anemometer tower. Using the MCP method, establish the relationship between the long-term reference wind speed time history U and the short-term measured wind speed time history V through linear regression, and calculate the long-term wind speed time history at the anemometer tower location. ; Step 2: Based on long-term wind speed time history The distribution parameters of wind speed for the next year are predicted, and based on this, the wind speed time history at the meteorological tower for the next year is simulated and generated. ; Step 3: Based on wind shear index The wind speed time history Vertically pushed outward to the height of the wind turbine hub To obtain the wheel hub height Future wind speed timeline and its probability distribution ; Step 4: Based on the historical SCADA wind speed data of each wind turbine in the wind farm, calculate the wind speed correlation coefficient matrix R, and use the Copula function to convert the probability distribution... By incorporating spatial correlation, the initial wind speed time histories at each wind turbine location within the wind farm are simulated and generated for the next year. ,in Number the wind turbine; Step 5: Based on the wind farm layout, wind direction data, and the selected wake model, calculate the wake effect of each wind turbine and apply the initial wind speed time history. Wake reduction is performed to obtain the reduced wind speed time history for each fan. ; Step Six: Based on the Poisson distribution, simulate the number of failures of each wind turbine in the coming year, the maintenance duration of each failure, and the failure start time to determine the effective number of failures for each wind turbine; during the downtime period, the corresponding wind turbines... The wind speed time history is set to zero; and if the shut-down fan is an upstream fan, the wake reduction it causes to downstream fans is canceled during the corresponding shutdown period; after fault and wake coupling adjustment, the final usable wind speed time history of each fan is obtained. ; Step 7: Based on the available wind speed time history of each wind turbine Based on the wind turbine power curves and air density, calculate the predicted annual power generation of each wind turbine. Summing these values ​​yields the predicted annual power generation (AEP) of the wind farm. Step 8: Repeat steps 1 to 7 multiple times to obtain the statistical distribution and uncertainty of the future AEP prediction value of the wind farm.

[0014] Furthermore, in step one, the long-term wind speed time history at the anemometer tower location... for: in: The long-term wind speed time history is represented at the location of the wind tower; U represents the long-term wind speed time history at the reference location. and These are the linear regression coefficients and the intercept, calculated using the following formula: in: and These are the standard deviations of the short-term wind speed time histories at the same height of the meteorological tower and the reference location, respectively, within the corresponding time period; and These are the average short-term wind speed time histories at the wind measurement tower location and the reference location, respectively.

[0015] Furthermore, in step two, the wind speed time history at the wind measurement tower for the next year is simulated and generated. The method steps are as follows: 21) Regarding the long-term wind speed time history The time series data of shape parameter k and scale parameter c are obtained by segmenting the data by year and fitting a Weibull distribution year by year. 22) Using the time-series historical data of k and c as the training set, train the Long Short-Term Memory (LSTM) network model and predict the parameter values ​​for the next year. and ; 23) According to and The defined Weibull distribution is used to simulate and generate wind speed time histories for the next year using the Monte Carlo method. .

[0016] Furthermore, in step three, if the short-term measured wind speed time history V of the anemometer tower includes the long-term reference wind speed time history U after the duration of years t1-t2... Annual wind speed data This will allow us to calculate the long-term wind speed time history at the location of the meteorological tower. With wind speed data Combined into a new long-term wind speed time history If the wind measurement data from the anemometer tower all fall within the duration of the long-term reference wind speed time history U for the years t1-t2, then based on the long-term wind speed time history... .

[0017] Furthermore, in step three, the hub height is obtained by vertical extrapolation. Future wind speed timeline for: in: The wind measurement tower is based on vertical extrapolation. Wind speed at altitude; It is a wind measurement tower Wind speed at altitude; This refers to the height of the wind turbine hub. The reference wind speed is at altitude; α is the wind shear index.

[0018] Furthermore, in step four, the method for simulating and generating the initial wind speed time history of each wind turbine using the Copula function is as follows: 41) Through future wind speed time history and its probability distribution The correlation coefficient matrix was calculated using historical wind speed data from the wind turbine's SCADA system. ; 42) Generate a vector with a mean vector of 0 and a covariance matrix of... The numerical vector of an n-dimensional multivariate normal distribution ; 43) For numerical vectors Each element in The transformation is performed using the inverse cumulative distribution function of the standard normal distribution: in: Let be a random variable that follows a uniform distribution in the interval [0,1] after being transformed by the inverse cumulative distribution function of the standard normal distribution. Numerical vector The i-th element in; This is the inverse cumulative distribution function of the standard normal distribution; 44) For each of the transformed... The probability distribution of wind speed at hub height obtained through step three. Transform the inverse cumulative distribution function: in: To determine the probability distribution of wind speed at hub height The set of simulated future wind speed values ​​at the location of the i-th wind turbine is obtained after the inverse cumulative distribution function transformation. Wind speed probability distribution at hub height The inverse cumulative distribution function; 45) Repeat steps 42) to 44) to generate multiple sets of simulated values ​​over the next year, thus constructing the initial wind speed time history for each wind turbine location. .

[0019] Furthermore, in step five, the wake model adopts a Gaussian wake model, and the downstream wind turbine's wind speed loss rate... Calculated by the following formula: When considering the combined effect of multiple upstream wind turbines, the wind speed of the downstream wind turbines. Perform the superposition of squares using the following formula: in: This indicates the horizontal distance between the downstream and upstream wind turbines; This represents the position coordinates of the wind turbine in the xz plane; This refers to the wind speed at the height of the upstream wind turbine hub. ; The diameter of the wind turbine; Wake width; The radial distance from the downstream fan location to the centerline of the upstream fan wake; , , ; Indicates environmental turbulence; This represents the thrust coefficient of the upstream wind turbine.

[0020] Furthermore, in step six, the method for simulating wind turbine faults based on Poisson distribution comprises the following steps: 61) For wind turbines The number of failures in the next year Follows a Poisson distribution: ,in This represents the average annual failure rate. 62) If the simulation results Then for the first The repair time for this fault is... It follows another Poisson distribution: ,in For maintenance time rate parameter, ; 63) Within the simulation time range [0, T], randomly generate the first... The start time of the secondary fault , must meet , and when When >1, ; 64) Obtaining time Effective number of failures of the internal fan i ;in: Let be the number of faults occurring in wind turbine i during the simulation period T, obtained through Poisson distribution simulation.

[0021] Furthermore, in step six, the method for canceling the wake reduction caused to the downstream wind turbine is as follows: For all This makes the fan i in all The wind speed in the middle is 0, that is ; exist Within a given time period, determine whether fan j is located downstream of fan i. If so, calculate the time period. When wind turbine i is removed from its upstream wind turbine set, the wake loss caused by wind turbine i to wind turbine j is no longer considered.

[0022] Furthermore, in step seven, the predicted annual power generation value of wind turbine i... for: in: The wind speed in the j-th hour The corresponding power, It is a standardized air density obtained based on a uniform distribution.

[0023] The beneficial effects of this invention are as follows: This invention provides a wind farm power generation prediction method that considers the correlation between turbine location and wind speed and wind turbine failure. By integrating the probabilistic stochastic simulation of wind turbine failure with the dynamic coupling mechanism of wake effect into the long-term power generation Monte Carlo evaluation framework, the following technical effects are achieved.

[0024] First, it significantly improves the accuracy and physical realism of long-term power generation forecasts. Traditional methods use a fixed reduction rate to handle faults, while this invention simulates the random occurrence time, frequency, and maintenance duration of faults based on Poisson distribution, more accurately characterizing the uncertainty of equipment reliability. More importantly, the model dynamically eliminates the wake effect of upstream faulty wind turbines on downstream areas during simulation, reflecting the key physical process of "fault-flow field" interaction in long-term assessments, making the prediction model closer to actual operating scenarios.

[0025] Secondly, it achieves a more comprehensive and refined quantification of the uncertainty of power generation. The method of this invention simultaneously propagates the spatiotemporal uncertainty of wind resources, the uncertainty of random equipment failures, and the coupling effect between the two. The resulting probability distribution of annual power generation not only has a more reliable mean, but its distribution pattern (such as standard deviation and confidence interval) can also more realistically reflect the additional risks caused by random wind turbine outages, providing a deeper assessment dimension.

[0026] Finally, the decision-support value of the assessment results is greatly enhanced. More accurate probabilistic assessment results can directly serve the formulation of differentiated operation and maintenance strategies for wind farms, the optimization of preventive maintenance, the pricing of power generation contracts, and the analysis of return on investment, providing a more powerful quantitative tool for the refined and intelligent operation and risk management of wind farms. Attached Figure Description

[0027] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart of an embodiment of the wind farm power generation prediction method of the present invention, which considers the correlation between wind speed at turbine location and wind turbine failure. Figure 2 This is a layout diagram of a wind farm. Figure 3 The following are scatter plots and probability density plots of the future AEP calculation values ​​for wind farms; (a) is a scatter plot of the future AEP calculation values ​​for wind farms; (b) is a probability density plot. Detailed Implementation

[0028] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0029] like Figure 1 As shown, this embodiment of the wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure includes the following steps.

[0030] Step 1: Obtain the long-term reference wind speed time history U (usually over 20 years, with the duration denoted as t1-t2) and the short-term measured wind speed time history V of the anemometer tower (e.g., one year, and this year falls within the t1-t2 time period). Using the MCP method, establish the relationship between the long-term reference wind speed time history U and the short-term measured wind speed time history V through linear regression, and calculate the long-term wind speed time history at the anemometer tower location. .

[0031] In this embodiment, the long-term wind speed time history at the location of the wind measurement tower is calculated. for: in: The long-term wind speed time history is represented at the location of the wind tower; U represents the long-term wind speed time history at the reference location. and These are the linear regression coefficients and the intercept, calculated using the following formula: in: and These are the standard deviations of the short-term wind speed time histories at the same height of the meteorological tower and the reference location, respectively, within the corresponding time period; and These are the average short-term wind speed time histories at the wind measurement tower location and the reference location, respectively.

[0032] Step 2: Based on long-term wind speed time history The distribution parameters of wind speed for the next year are predicted, and based on this, the wind speed time history at the meteorological tower for the next year is simulated and generated. .

[0033] In this embodiment, the wind speed time history at the meteorological tower for the next year is simulated and generated. The method steps are as follows: 21) Regarding the long-term wind speed time history By segmenting the data by year and fitting a Weibull distribution year by year, time-series data of shape parameter k and scale parameter c are obtained.

[0034] 22) Using the time-series historical data of k and c as the training set, a Long Short-Term Memory (LSTM) network model is trained to predict the shape parameters for the next year. With scale parameters .

[0035] 23) Based on shape parameters With scale parameters The defined Weibull distribution is used to simulate and generate wind speed time histories for the next year using the Monte Carlo method. .

[0036] Specifically, if the short-term measured wind speed time history V of the anemometer tower includes the long-term reference wind speed time history U after the duration of years t1-t2... Annual wind speed data This will allow us to calculate the long-term wind speed time history at the location of the meteorological tower. With wind speed data Combined into a new long-term wind speed time history If the wind measurement data from the anemometer tower all fall within the duration of the long-term reference wind speed time history U for the years t1-t2, then based on the long-term wind speed time history... .

[0037] Step 3: Based on wind shear index The wind speed time history Vertically pushed outward to the height of the wind turbine hub To obtain the wheel hub height Future wind speed timeline and its probability distribution .

[0038] The hub height is obtained by vertical extrapolation. Future wind speed timeline for: in: The wind measurement tower is based on vertical extrapolation. Wind speed at altitude; It is a wind measurement tower Wind speed at altitude; This refers to the height of the wind turbine hub. The reference wind speed is at altitude; α is the wind shear index.

[0039] pass Wind speed distribution at hub height was obtained. .

[0040] Step 4: Based on the historical SCADA wind speed data of each wind turbine in the wind farm, calculate the wind speed correlation coefficient matrix R, and use the Copula function to convert the probability distribution... By incorporating spatial correlation, the initial wind speed time histories at each wind turbine location within the wind farm are simulated and generated for the next year. ,in This is the number for the wind turbine.

[0041] In this embodiment, the steps for simulating and generating the initial wind speed time history of each wind turbine using the Copula function are as follows.

[0042] 41) Through future wind speed time history and its probability distribution The correlation coefficient matrix was calculated using historical wind speed data from the wind turbine's SCADA system. .

[0043] 42) Generate a vector with a mean vector of 0 and a covariance matrix of... The numerical vector of an n-dimensional multivariate normal distribution .

[0044] 43) For numerical vectors Each element in The transformation is performed using the inverse cumulative distribution function of the standard normal distribution: in: Let be a random variable that follows a uniform distribution in the interval [0,1] after being transformed by the inverse cumulative distribution function of the standard normal distribution. Numerical vector The i-th element in; It is the inverse cumulative distribution function of the standard normal distribution.

[0045] 44) For each of the transformed... By applying the inverse cumulative distribution function of the Weibull distribution, and then performing a further inverse transformation with equal probability on the transformed vector, a set of predicted simulated wind speed values ​​for the next year is obtained. Specifically, in this embodiment, the wind speed probability distribution at wheel hub height is used... Transform the inverse cumulative distribution function: in: To determine the probability distribution of wind speed at hub height The set of simulated future wind speed values ​​at the location of the i-th wind turbine is obtained after the inverse cumulative distribution function transformation. Wind speed probability distribution at hub height The inverse cumulative distribution function.

[0046] 45) Repeat steps 42) to 44) to generate multiple sets of simulated values ​​over the next year, thus constructing the initial wind speed time history for each wind turbine location. .

[0047] Step 5: Based on the wind farm layout, wind direction data, and the selected wake model, calculate the wake effect of each wind turbine and analyze the initial wind speed time history. Wake reduction is performed to obtain the reduced wind speed time history for each fan. .

[0048] In this embodiment, the wake model adopts the Gaussian wake model, and the wind speed loss rate of the downstream wind turbine in the model is... Calculated by the following formula: When considering the combined effect of multiple upstream wind turbines, the wind speed of the downstream wind turbines. Perform the superposition of squares using the following formula: in: This indicates the horizontal distance between the downstream and upstream wind turbines; This represents the position coordinates of the wind turbine in the xz plane; This refers to the wind speed at the height of the upstream wind turbine hub. ; The diameter of the wind turbine; For the wake width, and ; Let be the wake expansion coefficient, and ; Let be the initial wake width coefficient, and ; The radial distance from the downstream wind turbine location to the centerline of the upstream wind turbine wake is given. ; This refers to the hub height of the upstream wind turbine; This refers to the height of the downstream wind turbine hub. , , ; Indicates environmental turbulence; This represents the thrust coefficient of the upstream wind turbine.

[0049] Step Six: Based on the Poisson distribution, simulate the number of failures of each wind turbine in the coming year, the maintenance duration of each failure, and the failure start time to determine the effective number of failures for each wind turbine; during the downtime period, the corresponding wind turbines... The wind speed time history is set to zero; and if the shut-down fan is an upstream fan, the wake reduction it causes to downstream fans is canceled during the corresponding shutdown period; after fault and wake coupling adjustment, the final usable wind speed time history of each fan is obtained. .

[0050] Assuming that the timing of failures and the corresponding maintenance time for each failure are independent of each other, the following steps are taken to calculate wind turbine failures, determine the downtime of each wind turbine, adjust for wake loss, and finally simulate wind turbine failures based on Poisson distribution.

[0051] 61) For wind turbines The number of failures in the next year Follows a Poisson distribution: ,in This represents the average annual failure rate.

[0052] 62) If the simulation results Then for the first The repair time for this fault is... It follows another Poisson distribution: ,in For maintenance time rate parameter, .like If the wind turbine generates electricity throughout the year, the wind speed time history is the complete wind speed time history obtained above.

[0053] 63) Within the simulation time range [0, T], randomly generate the first... The start time of the secondary fault Specifically, let's assume... and and from uniform distribution Extraction Note the timing of the next failure. Inevitably and This must occur between two points, i.e., it must satisfy the following conditions. , and when When >1, .like Stop this step and proceed to the next step.

[0054] 64) Obtaining time Effective number of failures of the internal fan i ;in: Let be the number of faults occurring in wind turbine i during the simulation period T, obtained through Poisson distribution simulation.

[0055] In this embodiment, the method to cancel the wake reduction caused to the downstream wind turbine is as follows: For all This makes the fan i in all The wind speed in the middle is 0, that is ; exist Within a given time period, determine whether fan j is located downstream of fan i. If so, calculate the time period. When wind turbine i is removed from its upstream wind turbine set, the wake loss caused by wind turbine i to wind turbine j is no longer considered.

[0056] Step 7: Based on the available wind speed time history of each wind turbine Based on the wind turbine power curves and air density, calculate the predicted annual power generation of each wind turbine. The summation yields the predicted annual power generation (AEP) of the wind farm.

[0057] The corresponding air density and output power are obtained through simulation using a probabilistic model of normalized air density and power: in: For normalized air density The probability density function; This represents the average output power of the fan corresponding to wind speed V; This represents the average output power of the fan corresponding to wind speed V; Let V be the standard deviation of the fan output power corresponding to the wind speed V; This refers to the rated wind speed of the fan; This refers to the cut-in wind speed of the fan.

[0058] Based on the available wind speed of each wind turbine And based on the wind turbine power curve calculation, the predicted annual power generation value of wind turbine i is... for: in: The wind speed in the j-th hour The corresponding power, It is a standardized air density obtained based on a uniform distribution.

[0059] Step 8: Repeat steps 1 to 7 multiple times for iterative calculations to obtain the statistical distribution and uncertainty of the future AEP prediction value of the wind farm. In this embodiment, the above process is finally repeated 1000 times to obtain the calculated future AEP value and its uncertainty.

[0060] The following detailed description, using specific examples, further illustrates the detailed implementation of the wind farm power generation prediction method of the present invention, which considers the correlation between turbine location and wind speed and wind turbine failure.

[0061] To verify the effectiveness of the method proposed in this embodiment, a calculation was performed on a wind farm. The wind farm is located in a flat area, and its layout is shown in the diagram below. Figure 2 The formula for calculating the long-term wind speed of a wind measuring tower is: Its probability distribution parameters and The prediction results are as follows: and .

[0062] In this example, the height of the wind turbine hub, the height of the wind measuring tower, and the height of the reference wind speed are all 100 m, which are consistent, so vertical extrapolation is not performed here.

[0063] The correlation coefficient matrix was obtained from the annual wind speed measurements of the six wind turbines in the wind farm. (The lower triangular matrix of the symmetric matrix) is shown in Table 1. It can be seen that the correlation coefficient is around 0.9, indicating a strong correlation. Finally, the wind speed of each wind turbine is obtained based on the predicted wind speed.

[0064] Table 1 Correlation coefficients of wind speed at the fan location Then, wake loss and reduction calculations, failure rate calculations, and wake loss corrections are performed to obtain the final usable wind speed for each turbine. Finally, combining the power curve and air density, the AEP (Advanced Power Efficiency Prediction) value for each turbine is calculated, thus obtaining the AEP prediction value and probability distribution of the wind farm, as shown in [reference needed]. Figure 3 .

[0065] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for predicting wind farm power generation considering the correlation between turbine location and wind speed, and wind turbine failure, characterized in that: Includes the following steps: Step 1: Obtain the long-term reference wind speed time history U and the short-term measured wind speed time history V at the anemometer tower. Using the MCP method, establish the relationship between the long-term reference wind speed time history U and the short-term measured wind speed time history V through linear regression, and calculate the long-term wind speed time history at the anemometer tower location. ; Step 2: Based on long-term wind speed time history The distribution parameters of wind speed for the next year are predicted, and based on this, the wind speed time history at the meteorological tower for the next year is simulated and generated. ; Step 3: Based on wind shear index The wind speed time history Vertically pushed outward to the height of the wind turbine hub To obtain the wheel hub height Future wind speed timeline and its probability distribution ; Step 4: Based on the historical SCADA wind speed data of each wind turbine in the wind farm, calculate the wind speed correlation coefficient matrix R, and use the Copula function to convert the probability distribution... By incorporating spatial correlation, the initial wind speed time histories at each wind turbine location within the wind farm are simulated and generated for the next year. ,in Number the wind turbine; Step 5: Based on the wind farm layout, wind direction data, and the selected wake model, calculate the wake effect of each wind turbine and apply the initial wind speed time history. Wake reduction is performed to obtain the reduced wind speed time history for each fan. ; Step Six: Based on the Poisson distribution, simulate the number of failures of each wind turbine in the coming year, the maintenance duration of each failure, and the failure start time to determine the effective number of failures for each wind turbine; during the downtime period, the corresponding wind turbines... The wind speed time history is set to zero; and if the shut-down fan is an upstream fan, the wake reduction caused by it to the downstream fan is canceled during the corresponding shutdown time period. After adjusting for fault and wake coupling, the final usable wind speed time history for each wind turbine was obtained. ; Step 7: Based on the available wind speed time history of each wind turbine Based on the wind turbine power curves and air density, calculate the predicted annual power generation of each wind turbine. Summing these values ​​yields the predicted annual power generation (AEP) of the wind farm. Step 8: Repeat steps 1 to 7 multiple times to obtain the statistical distribution and uncertainty of the future AEP prediction value of the wind farm.

2. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step one, the long-term wind speed time history at the anemometer tower location... for: in: The long-term wind speed time history is represented at the location of the wind tower; U represents the long-term wind speed time history at the reference location. and These are the linear regression coefficients and the intercept, calculated using the following formula: in: and These are the standard deviations of the short-term wind speed time histories at the same height of the meteorological tower and the reference location, respectively, within the corresponding time period; and These are the average short-term wind speed time histories at the wind measurement tower location and the reference location, respectively.

3. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step two, the wind speed time history at the wind measurement tower for the next year is simulated and generated. The method steps are as follows: 21) Regarding the long-term wind speed time history The time series data of shape parameter k and scale parameter c are obtained by segmenting the data by year and fitting a Weibull distribution year by year. 22) Using the time-series historical data of k and c as the training set, train the Long Short-Term Memory (LSTM) network model and predict the parameter values ​​for the next year. and ; 23) According to and The defined Weibull distribution is used to simulate and generate wind speed time histories for the next year using the Monte Carlo method. .

4. The wind farm power generation prediction method considering the correlation between turbine location wind speed and wind turbine failure as described in claim 1 or 3, characterized in that: In step three, if the short-term measured wind speed time history V of the wind measuring tower includes the long-term reference wind speed time history U after the duration of years t1-t2... Annual wind speed data This will allow us to calculate the long-term wind speed time history at the location of the meteorological tower. With wind speed data Combined into a new long-term wind speed time history If the wind measurement data from the anemometer tower all fall within the duration of the long-term reference wind speed time history U for the years t1-t2, then based on the long-term wind speed time history... .

5. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step three, the hub height is obtained by vertical extrapolation. Future wind speed timeline for: in: The wind measurement tower is based on vertical extrapolation. Wind speed at altitude; It is a wind measurement tower Wind speed at altitude; This refers to the height of the wind turbine hub. The reference wind speed is at altitude; α is the wind shear index.

6. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step four, the method for simulating and generating the initial wind speed time history of each wind turbine using the Copula function is as follows: 41) Through future wind speed time history and its probability distribution The correlation coefficient matrix was calculated using historical wind speed data from the wind turbine's SCADA system. ; 42) Generate a vector with a mean vector of 0 and a covariance matrix of... The numerical vector of an n-dimensional multivariate normal distribution ; 43) For numerical vectors Each element in The transformation is performed using the inverse cumulative distribution function of the standard normal distribution: in: Let be a random variable that follows a uniform distribution in the interval [0,1] after being transformed by the inverse cumulative distribution function of the standard normal distribution. Numerical vector The i-th element in; This is the inverse cumulative distribution function of the standard normal distribution; 44) For each of the transformed... The probability distribution of wind speed at hub height obtained through step three. Transform the inverse cumulative distribution function: in: To determine the probability distribution of wind speed at hub height The set of simulated future wind speed values ​​at the location of the i-th wind turbine is obtained after the inverse cumulative distribution function transformation. Wind speed probability distribution at hub height The inverse cumulative distribution function; 45) Repeat steps 42) to 44) to generate multiple sets of simulated values ​​over the next year, thus constructing the initial wind speed time history for each wind turbine location. .

7. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step five, the wake model adopts the Gaussian wake model, and the wind speed loss rate of the downstream wind turbine is... Calculated by the following formula: When considering the combined effect of multiple upstream wind turbines, the wind speed of the downstream wind turbines. Perform the superposition of squares using the following formula: in: This indicates the horizontal distance between the downstream and upstream wind turbines; This represents the position coordinates of the wind turbine in the xz plane; This refers to the wind speed at the height of the upstream wind turbine hub. ; The diameter of the wind turbine; Wake width; The radial distance from the downstream fan location to the centerline of the upstream fan wake; , , ; Indicates environmental turbulence; This represents the thrust coefficient of the upstream wind turbine.

8. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step six, the method for simulating wind turbine faults based on Poisson distribution consists of the following steps: 61) For wind turbines The number of failures in the next year Follows a Poisson distribution: ,in This represents the average annual failure rate. 62) If the simulation results Then for the first The repair time for this fault is... It follows another Poisson distribution: ,in For maintenance time rate parameter, ; 63) Within the simulation time range [0, T], randomly generate the first... The start time of the secondary fault , must meet , and when When >1, ; 64) Obtaining time Effective number of failures of the internal fan i ;in: Let be the number of faults occurring in wind turbine i during the simulation period T, obtained through Poisson distribution simulation.

9. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 8, characterized in that: In step six, the method to cancel the wake reduction caused to the downstream wind turbine is as follows: For all This makes the fan i in all The wind speed in the middle is 0, that is ; exist Within a given time period, determine whether fan j is located downstream of fan i. If so, calculate the time period. When wind turbine i is removed from its upstream wind turbine set, the wake loss caused by wind turbine i to wind turbine j is no longer considered.

10. The wind farm power generation prediction method considering the correlation between turbine location and wind speed and wind turbine failure as described in claim 1, characterized in that: In step seven, the predicted annual power generation value of wind turbine i for: in: The wind speed in the j-th hour The corresponding power, It is a standardized air density obtained based on a uniform distribution.

Images