Method and device for analyzing, diagnosing and optimizing power generation capacity of wind turbine generator

Abstract

The invention discloses a wind turbine generator power generation capacity analysis, diagnosis and optimization method and device, and relates to the field of wind turbine generator power generation capacity monitoring, and the method comprises the steps: obtaining a plurality of groups of to-be-processed data including wind turbine generator operation data, meteorological data and wind field operation data, carrying out the preprocessing, and inputting the data into a plurality of analysis models, comprising a power curve fitting analysis model, a wind deviation analysis model, a temperature-power analysis model, a running time analysis model, a power loss analysis model, a pitch angle analysis model, a micro-siting analysis model and / or an AGC curve deviation analysis model, and obtaining multiple groups of analysis data; inputting the multiple groups of analysis data into a diagnosis model constructed based on machine learning to obtain a diagnosis result; and inputting the diagnosis result into the expert optimization suggestion model to obtain an optimization suggestion scheme. The method can improve the power generation efficiency of the wind power plant, improves the system efficiency and accuracy, reduces the operation cost, and facilitates the realization of the digital transformation of the wind power industry.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind turbine generator set power generation capacity monitoring, and in particular relates to a method and device for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine generator set. Background Art

[0002] Wind power generation is an important clean energy with great development potential. However, due to factors such as wind resources, wind turbine performance, operation and maintenance, there are large differences in the power generation capacity of wind turbines, making it difficult to fully realize their power generation potential. The existing wind farm management system mainly relies on experience judgment and manual analysis. However, manual analysis requires a large amount of data and is time-consuming, and it is impossible to timely discover and solve potential problems that affect the power generation capacity of wind turbines; experience judgment is highly subjective and lacks scientific basis, making it difficult to accurately evaluate the power generation capacity of wind turbines; moreover, the existing management system cannot monitor the operating status of wind turbines in real time, making it difficult to timely discover and handle abnormal situations; therefore, the existing wind farm management system leads to low power generation efficiency, high operating costs, and difficulty in achieving sustainable development. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a method and device for analyzing, diagnosing and optimizing the power generation capacity of wind turbines, which can improve the power generation efficiency of wind farms, enhance system efficiency and accuracy, reduce operating costs, and facilitate the digital transformation of the wind power industry.

[0004] In a first aspect, the present invention provides a method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine generator set, comprising:

[0005] Acquire multiple groups of data to be processed, wherein the multiple groups of data to be processed include wind turbine operating data, meteorological data and / or wind farm operating data;

[0006] Preprocessing the multiple groups of data to be processed to obtain multiple groups of preprocessed data;

[0007] Inputting the plurality of groups of preprocessed data into corresponding plurality of analysis models respectively to obtain plurality of groups of analysis data; the plurality of analysis models include power curve fitting analysis model, wind deviation analysis model, temperature-power analysis model, operation time analysis model, power loss analysis model, pitch angle analysis model, micro-site selection analysis model and / or AGC curve deviation analysis model;

[0008] Inputting the plurality of sets of analysis data into the trained diagnostic model respectively to obtain a diagnostic result; the diagnostic model is constructed based on a machine learning model;

[0009] The diagnosis results are input into a pre-built expert optimization suggestion model to obtain an optimization suggestion plan.

[0010] In an optional embodiment, the wind turbine operating data includes a combination of one or more of wind turbine power, wind speed, wind direction, temperature, pitch angle, wind angle, speed and status code; the meteorological data includes a combination of one or more of ambient wind speed, ambient wind direction, ambient temperature, ambient air pressure and ambient humidity; the wind farm operating data includes a combination of one or more of wind turbine grid-connected status, wind turbine power limitation and maintenance records.

[0011] In an optional implementation, the power curve fitting analysis model includes:

[0012] A wind turbine power prediction module, used to predict the corresponding wind turbine power under different ambient wind speeds based on the preprocessed data and a pre-built power curve sub-model;

[0013] A deviation rate prediction module is used to determine the power curve deviation rate according to the actual power corresponding to different environmental wind speeds, the rated power and theoretical output power of the wind turbine generator set;

[0014] An updating module, configured to update the power curve sub-model according to the power curve deviation rate;

[0015] The power curve sub-model is constructed by the following method:

[0016] Performing curve fitting on the historical data of the wind turbine generator set using a plurality of machine learning algorithms to obtain a plurality of primary power curve models of the wind turbine generator set; the machine learning algorithms include support vector machines, neural networks and / or random forests;

[0017] The fitting effects of the multiple power curve primary models are evaluated by using evaluation indicators, and the power curve primary model with the best fitting effect is determined from the multiple power curve primary models according to the fitting effect, and the power curve primary model with the best fitting effect is used as the power curve sub-model; wherein the evaluation indicators include an R square value indicator and / or a root mean square error indicator. .

[0018] In an optional embodiment, the wind deviation analysis model includes:

[0019] A yaw error data module is used to obtain static yaw error data according to the wind speed preprocessing data of the wind turbine, the wind angle preprocessing data of the wind turbine and a pre-built wind deviation sub-model in the pre-processing data, wherein the wind deviation sub-model is built based on a support vector regression algorithm;

[0020] An error correction module, used for adjusting the yaw control strategy of the wind turbine according to the static yaw error data to optimize the wind energy utilization efficiency;

[0021] The yaw error compensation module is used to compensate the yaw error data into the power curve sub-model to improve the accuracy of the power curve sub-model.

[0022] In an optional embodiment, the temperature-power analysis model includes:

[0023] A power generation capacity prediction module is used to determine the influence coefficient of temperature on power according to the ambient temperature preprocessing data, the fan power preprocessing data and the pre-built temperature-power relationship sub-model in the preprocessing data, and predict the fan power at different temperatures according to the influence coefficient of temperature on power;

[0024] A temperature compensation module, used to adjust the power curve sub-model according to the influence coefficient of the temperature on the power, so as to improve the accuracy of the power curve sub-model;

[0025] The temperature-power relationship sub-model is determined according to the maximum power variation of the wind turbine generator set under different ambient temperatures by using linear regression, polynomial regression or neural network model.

[0026] In an optional implementation, the power loss analysis model includes:

[0027] A loss type module, used to determine the loss electricity corresponding to multiple loss types according to the pre-processed data and the pre-built loss electricity sub-module; wherein the multiple loss types include one or more of maintenance, failure, power restriction, performance, and involvement;

[0028] A loss ratio module, used to determine the power loss ratio of each loss type according to the power loss;

[0029] A loss cause module, used to determine the loss cause corresponding to each loss type;

[0030] The power loss submodule is constructed by the following method:

[0031] Determine the power generation amount of the wind turbine set in multiple time periods according to the wind turbine set operation data and the meteorological data;

[0032] Compare the power generation that the wind turbines should generate in multiple time periods with the actual power generation in the corresponding time periods to determine the power loss in each time period;

[0033] Determine the loss type corresponding to the lost power in each time period.

[0034] In an optional embodiment, the micro-site selection analysis model is used to obtain meteorological preprocessing data from multiple groups of preprocessed data of the target location, input the meteorological preprocessing data of the target location into a pre-constructed wind field model, and output wind resource distribution data; wherein the wind field model is constructed based on computational fluid dynamics software combined with terrain data of the target location.

[0035] In an optional implementation, the AGC curve deviation analysis model includes:

[0036] A deviation evaluation module, used to determine the AGC curve deviation rate of the wind turbine according to the preprocessed wind turbine power and the pre-constructed AGC model; wherein the AGC model is constructed based on a time series analysis method;

[0037] A control optimization module, used for adjusting the wind turbine operation strategy according to the AGC curve deviation rate of the wind turbine group, wherein the wind turbine operation strategy includes a shutdown strategy, a startup strategy and a power control strategy;

[0038] The AGC model correction module is used to correct the AGC model according to the AGC curve deviation rate of the wind turbine generator set.

[0039] In an optional embodiment, the optimization suggestion scheme includes adjusting fan control parameters, optimizing fan operation strategy and formulating a maintenance plan;

[0040] The adjusting of the wind turbine control parameters includes adjusting the yaw angle, the pitch angle and / or the rotation speed;

[0041] Optimizing wind turbine operation strategies including shutdown strategies and / or startup strategies;

[0042] Develop a maintenance schedule that includes regular inspections and / or breakdown repairs.

[0043] In a second aspect, the present invention provides a device for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine generator set, comprising:

[0044] An acquisition module, used for acquiring multiple groups of data to be processed, wherein the multiple groups of data to be processed include wind turbine operating data, meteorological data and / or wind farm operating data;

[0045] A preprocessing module, used for preprocessing the plurality of groups of data to be processed to obtain a plurality of groups of preprocessed data;

[0046] An analysis module, used for inputting the plurality of groups of preprocessed data into corresponding plurality of analysis models respectively to obtain plurality of groups of analysis data; the plurality of analysis models include a power curve fitting analysis model, a wind deviation analysis model, a temperature-power analysis model, a running time analysis model, a power loss analysis model, a pitch angle analysis model, a micro-site selection analysis model and / or an AGC curve deviation analysis model;

[0047] A diagnosis module, used to input the plurality of sets of analysis data into the trained diagnosis model to obtain a diagnosis result; the diagnosis model is constructed based on a machine learning model;

[0048] The optimization suggestion module is used to input the diagnosis results into a pre-built expert optimization suggestion model to obtain an optimization suggestion plan.

[0049] The beneficial effects of the technical solution provided by the embodiment of the present invention are as follows: the present invention can integrate and analyze multi-source data including wind turbine operating data, meteorological data and wind farm operating data, perform fault diagnosis in combination with a diagnostic model constructed by a machine learning model, and input the diagnostic results into an expert optimization suggestion model to obtain an optimization suggestion plan; the present invention can improve the power generation efficiency of wind turbines and increase annual power generation; it can reduce operating costs and improve economic benefits; it uses machine learning algorithms to achieve improved intelligence levels and intelligent diagnosis and optimization suggestions, and improve the intelligence level of wind farms; it promotes the digital transformation of the wind power industry and provides technical support for the sustainable development of the new energy industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of a flow chart of a method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine provided by an embodiment of the present invention;

[0051] Figure 2 A system schematic diagram of a wind turbine generator system power generation capacity analysis, diagnosis and optimization device provided in an embodiment of the present invention.

[0052] In the figure: 100 - acquisition module; 200 - preprocessing module; 300 - analysis module; 400 - diagnosis module; 500 - optimization suggestion module. DETAILED DESCRIPTION

[0053] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0054] See also Figure 1 This embodiment provides a method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine, including the following steps S100 to S500.

[0055] Step S100, obtaining multiple groups of data to be processed, the multiple groups of data to be processed include wind turbine operating data, meteorological data and / or wind farm operating data. Specifically, wind turbine operating data include wind turbine power, wind turbine wind speed, wind turbine wind direction, wind turbine temperature, wind turbine pitch angle, wind turbine speed and wind turbine status code, etc., meteorological data include ambient wind speed, ambient wind direction, ambient temperature, ambient air pressure and ambient humidity, etc.; wind farm operating data include wind turbine grid connection status, wind turbine power limit status and maintenance records, etc. It should be noted that, as needed, the data to be processed may also include other data.

[0056] The device for acquiring the data to be processed includes one or more of a data collector, a SCADA system (Supervisory Control And Data Acquisition), and a cloud platform. For example, wind turbine operation data is collected by a data collector, and some meteorological data and wind farm operation data are collected by SCADA (wind turbine operation data can also be collected using this system). Usually, the data collected by the data collector is sent to the SCADA system, and the cloud platform is usually used to integrate data from multiple sources, which may include data from the SCADA system, additional meteorological data (such as forecast data from meteorological services and satellite or radar meteorological data), more detailed wind farm operation data (such as complete maintenance records and historical performance data), power grid data (such as regional power demand and electricity price information), and other external data sources (such as environmental monitoring data and bird activity data).

[0057] Among them, it should be noted that wind turbine operation data is different from meteorological data. Wind turbine operation data is usually directly measured by sensors installed on wind turbines, with high measurement accuracy and frequency (such as every second or every minute), which can reflect the local conditions of a single wind turbine location and is mainly used to monitor and analyze the performance of a single wind turbine. Meteorological data usually comes from meteorological stations or meteorological measurement towers, which may be located in or near wind farms. Its accuracy may be low and the measurement interval may be long (such as every 10 minutes or every hour). It can represent the average meteorological conditions in a larger range and is often used for wind resource assessment and power generation forecasting of the entire wind farm. The status code refers to the operating status identification of the wind turbine, which is used to indicate the current working status or fault condition of the wind turbine. It is usually represented by numbers or codes. For example, "0" indicates normal operation, "1" indicates standby state, "2" indicates the startup process, "3" indicates the shutdown process, "4" indicates emergency shutdown, "5" indicates regular maintenance, and "10" to "99" indicate various specific faults (such as yaw fault, pitch fault, overspeed fault, etc.). The grid-connected status of wind turbines is grid-connected and not grid-connected. Grid-connected means that the wind turbine is connected to the grid. Power restriction refers to the situation where the power output is restricted due to grid load restrictions, wind farm output restrictions, wind farm safety control, policy-based power restrictions, etc.

[0058] This embodiment integrates and analyzes multi-source data such as wind turbine operating data, meteorological data, and wind farm operating data, thereby improving the accuracy and reliability of the analysis results.

[0059] Step S200, preprocessing multiple groups of data to be processed to obtain multiple groups of preprocessed data.

[0060] The preprocessing method of this embodiment includes data cleaning (eliminating erroneous, missing, and abnormal data), data conversion (unifying different types of data into a processable format), data normalization (scaling the data to a uniform range to improve data consistency and comparability to facilitate model training and analysis), and data interpolation (using an interpolation algorithm to fill in missing data to ensure data integrity).

[0061] Step S300, inputting the multiple groups of preprocessed data into the corresponding multiple analysis models respectively to obtain multiple groups of analysis data; the multiple analysis models include a power curve fitting analysis model, a wind deviation analysis model, a temperature-power analysis model, an operating time analysis model, a power loss analysis model, a pitch angle analysis model, a micro-site selection analysis model and an AGC curve deviation analysis model, as described in detail below (1)-(8).

[0062] Power curve fitting analysis model

[0063] The power curve fitting analysis model includes a wind turbine power prediction module, a wind turbine power prediction module and an update module.

[0064] The wind turbine power prediction module is used to predict the corresponding wind turbine power under different ambient wind speeds based on the preprocessed data and the pre-built power curve sub-model. The preprocessed data input to the power curve sub-model can be wind speed preprocessed data, temperature preprocessed data, or multiple sets of preprocessed data such as wind speed, temperature, wind deviation, pitch angle, etc. can be integrated into a power curve sub-model to establish a more comprehensive power model. The deviation rate prediction module is used to determine the power curve deviation rate based on the actual wind turbine power, rated power and theoretical output power of the wind turbine set corresponding to different ambient wind speeds. The update module is used to dynamically update the power curve sub-model according to the power curve deviation rate to timely reflect changes in wind turbine performance.

[0065] Here, the deviation rate is determined by the following formula (1).

[0066]

[0067] In formula (1), P hi represents the actual power of the fan at time i, P ri Indicates the wind speed at time i corresponding to the power value on the standard power curve, P drepresents the design power of the fan, and n represents the number of sampling points.

[0068] The power curve sub-model is constructed through the following steps (1-1) to (1-2).

[0069] Step (1-1), using a variety of machine learning algorithms to curve fit the historical data of the wind turbine to obtain multiple primary models of the power curve of the wind turbine; the machine learning algorithm includes support vector machine, neural network and / or random forest. Support vector machine is good at processing high-dimensional data, has good robustness to small sample data, and is suitable for classification and regression problems. In this embodiment, support vector machine can be used to fit the power data of the wind turbine at different wind speeds. Neural network can learn complex nonlinear relationships, is suitable for processing large amounts of data, and is suitable for classification, regression and prediction problems. In this embodiment, neural network can be used to fit the power data of the wind turbine under different wind speeds, temperatures, wind deviations and other conditions to establish a more refined power model. Random forest is an integrated learning algorithm that improves the prediction accuracy of the model by combining multiple decision trees, and is suitable for classification and regression problems; in this embodiment, random forest can be used to fit the power data of the wind turbine under different wind speeds, temperatures, wind deviations and other conditions to establish a more stable power model.

[0070] Step (1-2), using evaluation indicators to evaluate the fitting effects of multiple power curve primary models, determining the power curve primary model with the best fitting effect from the multiple power curve primary models according to the fitting effects, and using the power curve primary model with the best fitting effect as the power curve sub-model; wherein the evaluation indicators include the R-squared value indicator and the root mean square error indicator.

[0071] It should be noted that in actual applications, the required operating data is selected as needed to establish a power curve sub-model. For example, only wind speed and temperature data can be used to establish a power curve sub-model, or different power curve sub-models can be established according to different application scenarios.

[0072] The power curve fitting analysis model of this embodiment uses a machine learning algorithm to fit historical data, establishes a power curve sub-model of the wind turbine, and uses evaluation indicators such as the R square value index and the root mean square error index to evaluate the fitting effect to select the best fitting model, thereby calculating the power curve deviation rate based on the best fitting model, and using the deviation rate to update the power curve sub-model. An update module is also designed to realize the dynamic update of the power curve sub-model, so as to timely reflect the changes in wind turbine performance and improve the real-time and accuracy of the analysis results.

[0073] (1) Wind deviation analysis model

[0074] The wind deviation analysis model includes a yaw error data module, an error correction module and a yaw error compensation module. The yaw error data module is used to obtain static yaw error data based on the wind direction preprocessing data, wind angle preprocessing data and a pre-built wind deviation sub-model in the preprocessing data, wherein the wind deviation sub-model is constructed based on a support vector regression algorithm.

[0075] Specifically, in order to maximize the use of wind energy, wind turbines need to always maintain the best angle to the wind, that is, the wind turbine blades are parallel to the wind direction. The yaw system that adjusts the direction of the wind turbine has a static yaw error, including errors in the sensor that measures the wind direction, controller errors, actuator errors, and installation errors. Static yaw error affects power generation efficiency, causes blade fatigue, and increases the risk of equipment failure. The wind deviation sub-model calculates the wind deviation through the wind direction and wind angle of the wind turbine to obtain output data at different wind deviation angles. The yaw error data module identifies the static yaw error based on the output data output by the wind deviation sub-model. It should be noted that the output data refers to the average output of the wind turbine in a specific time period.

[0076] The error correction module is used to adjust the yaw control strategy of the wind turbine according to the static yaw error data to optimize the efficiency of wind energy utilization. Among them, the yaw control strategy is to adjust the yaw angle of the wind turbine so that the wind turbine always maintains the best wind-facing state, improves power generation efficiency and reduces blade fatigue risk. The yaw control strategy includes fixed-point yaw control strategy, wind direction tracking control strategy, optimal power tracking control strategy and comprehensive control strategy.

[0077] The fixed-point yaw control strategy refers to adjusting the wind turbine to the corresponding angle according to the preset yaw angle and keeping it stable; the fixed-point yaw control strategy can be used when the wind direction changes slightly or the installation position of the wind turbine is relatively fixed; for example, if the wind turbine is installed in an area where the wind direction is mainly northeast, the yaw angle can be set to the northeast, and the control system can be used to keep the wind turbine always pointing to the northeast. The wind direction tracking control strategy refers to real-time monitoring of wind direction changes through a wind direction sensor, and adjusting the yaw angle of the wind turbine according to the wind direction changes, always keeping the wind turbine parallel to the wind direction; when the wind direction changes greatly or more precise control is required, the wind direction tracking control strategy can be used; for example, the wind direction gradually changes from the northeast to the southeast, and the wind direction tracking control system will monitor the wind direction changes in real time and adjust the yaw angle of the wind turbine so that it always points to the current wind direction, ensuring that the wind turbine maintains the best wind-facing state. The optimal power tracking control strategy is to monitor the wind turbine power generation and adjust the wind turbine yaw angle according to the power change so that the wind turbine is always in the optimal power output state; when the wind speed changes greatly or the power generation efficiency needs to be maximized, the optimal power tracking control strategy can be used; for example, the wind turbine has the highest power generation when the wind speed is 8m / s, then the optimal power tracking control system will monitor the wind speed change in real time and adjust the wind turbine yaw angle to the optimal wind angle to ensure that the wind turbine always operates at a wind speed of 8m / s, thereby achieving maximum power output. The comprehensive control strategy refers to combining multiple control strategies, such as wind direction tracking control strategy and optimal power tracking control strategy, and selecting the appropriate control strategy according to different operating conditions to achieve the best control effect; for complex wind turbine operating environments, it is necessary to combine multiple control strategies to achieve an efficient and stable control strategy; for example, when the wind speed is low, the wind direction tracking control strategy can be used to ensure that the wind turbine always maintains the optimal wind angle; when the wind speed is high, the optimal power tracking control strategy can be used to achieve maximum power output.

[0078] The yaw error compensation module is used to compensate the yaw error data into the power curve sub-model to improve the accuracy of the power curve sub-model.

[0079] This embodiment

[0080] (2) Temperature-power analysis model

[0081] The temperature-power analysis model includes a power generation capacity prediction module and a temperature compensation module. The power generation capacity prediction module is used to determine the influence coefficient of temperature on power based on the ambient temperature preprocessing data, the fan power preprocessing data and the pre-built temperature-power relationship sub-model in the preprocessing data, and predict the fan power at different temperatures based on the influence coefficient of temperature on power. The temperature compensation module is used to adjust the power curve sub-model according to the influence coefficient of temperature on power to improve the accuracy of the power curve sub-model. Among them, the temperature-power relationship sub-model is determined by linear regression, polynomial regression or neural network model based on the maximum power change of the wind turbine at different ambient temperatures.

[0082] Specifically, temperature and power are usually nonlinearly related, and polynomial regression can capture this relationship. Therefore, this embodiment uses polynomial regression to construct a temperature-power relationship sub-model. This embodiment uses the temperature-power relationship sub-model to predict the wind turbine power generation capacity at different temperatures, optimize the wind turbine operation strategy, increase power generation, and provide a reference for wind turbine operation control.

[0083] (3) Runtime Analysis Model

[0084] The operation time analysis model includes a problem diagnosis module and a maintenance strategy optimization module. The problem diagnosis module is used to determine the operation time and non-operation time of the fan in different wind speed ranges based on the fan wind speed preprocessing data, operation time preprocessing data and the pre-built wind speed-operation time sub-model in the preprocessing data. According to the operation time and non-operation time of the fan in different wind speed ranges, the causes that may cause the fan to shut down are diagnosed, such as failure, maintenance, power restriction, etc. In addition, the power restriction is judged based on the measurement point data, such as 1 represents maintenance, 2 represents operation, and 3 represents maintenance. In some cases, the power restriction is judged by a combination of multiple measurement points, such as the AGC active setting value and output reduction.

[0085] The maintenance strategy optimization module is used to formulate corresponding maintenance strategies based on the diagnosis results, reduce downtime and improve wind turbine utilization. Among them, the wind speed-operating time sub-model can analyze the operating time and non-operating time of the wind turbine in different wind speed ranges to identify potential problems that affect power generation capacity.

[0086] Among them, the reason for the fan shutdown is identified according to the wind speed-operating time relationship model. The short operating time in certain wind speed ranges may be caused by a fault. According to the output result of the wind speed-operating time relationship model, combined with the alarm records, maintenance records, etc., analysis and diagnosis are performed, and then combined with the elimination method and logical reasoning, the cause of the shutdown is inferred. If necessary, experts in related fields can be consulted for professional opinions. In the maintenance strategy optimization module, corresponding measures are taken according to the determined cause of the shutdown, such as repairing the fault, adjusting the control parameters, optimizing the operation strategy, etc. Corresponding preventive measures can also be formulated according to the cause of the shutdown to avoid similar problems from happening again. This embodiment analyzes the operating time and non-operating time in different wind speed ranges, and formulates corresponding maintenance strategies based on the analysis results to improve the utilization rate of the fan.

[0087] (4) Power loss analysis model

[0088] The power loss analysis model includes a loss type module, a loss ratio module and a loss cause module. The loss type module is used to determine the power loss corresponding to multiple loss types based on preprocessed data and pre-built loss power sub-modules; wherein the multiple loss types include one or more of maintenance, failure, power restriction, performance, and involvement. The loss ratio module is used to determine the power loss ratio of each loss type based on the power loss to provide a reference for the optimization strategy. The loss cause module is used to determine the loss cause corresponding to each loss type in order to formulate targeted improvement measures. Among them, the improvement measures can be: if the fault downtime is relatively long, the training of fault repair and the reserve of spare parts can be strengthened. In addition, the assessment management of maintenance personnel can be strengthened to improve the efficiency of fault repair and reduce the length of fault downtime; if the power restriction time is long, it is necessary to actively communicate with the power grid to strive for less power restriction.

[0089] Among them, the power loss sub-module is constructed through the following steps (5-1) to (5-3).

[0090] Step (5-1), determining the power generation of the wind turbine in multiple time periods based on the wind turbine operation data and meteorological data.

[0091] Step (5-2) compares the power generation that the wind turbine should generate in multiple time periods with the actual power generation in the corresponding time periods to determine the power loss in each time period.

[0092] Step (5-3), determining the loss type corresponding to the lost electricity in each time period.

[0093] This embodiment analyzes the power loss caused by various factors through the power loss analysis model, and takes measures to reduce the power loss according to the loss analysis results, thereby improving the overall power generation efficiency of the wind farm.

[0094] (5) Pitch angle analysis model

[0095] The pitch angle analysis model includes a pitch angle error correction submodule and a pitch angle compensation submodule. The pitch angle error correction submodule is used to determine the pitch angle error based on the pitch angle preprocessing data in the preprocessing data and the pre-built pitch angle identification submodel, and adjust the pitch angle control strategy according to the pitch angle error to improve the wind energy utilization efficiency. The pitch angle compensation submodule is used to supplement the power curve submodel according to the pitch angle error to improve the accuracy of the power curve submodel. Among them, the pitch angle identification submodel is determined based on the actual pitch angle and the theoretical pitch angle in combination with the machine learning model. Specifically, machine learning algorithms such as random forests, support vector machines or neural networks may be used to establish a relationship model between the pitch angle and wind turbine performance (such as power output, power generation efficiency).

[0096] The pitch angle control strategy is adjusted based on the analysis results and the root cause of the problem is targeted to propose a practical solution.

[0097] The improvement strategy based on the root cause of the problem is to find the root cause of the problem through in-depth research on the analysis results, and formulate solutions for the root cause; this strategy is applicable to various problems, such as power curve deviation, wind deviation, operation time analysis, power loss analysis, etc. For example, the power curve deviation rate of a certain wind turbine is high, and the analysis found that it is caused by dust accumulation on the surface of the blades. The improvement strategy is to formulate a blade cleaning and maintenance plan, clean the blades regularly, improve blade efficiency, and reduce the power curve deviation rate.

[0098] Data-driven improvement refers to the use of wind turbine operation data and historical fault records, through data analysis and machine learning and other methods, to discover potential problems and improvement directions, and formulate corresponding measures; this strategy is applied to scenarios with large data volumes and complex information, such as intelligent diagnosis, optimization suggestions, etc. For example, through data analysis, it was found that a certain wind turbine has a short operating time in a specific wind speed range, resulting in low power generation efficiency. The improvement strategy is to adjust the wind turbine control strategy according to the data analysis results, such as optimizing the startup strategy and shutdown strategy, to increase the operating time of the wind turbine in a specific wind speed range and improve power generation efficiency.

[0099] Improvement based on expert experience refers to the combination of the experience and expertise of experts in the wind power field to propose improvement plans for specific problems; this strategy is suitable for some problems that are difficult to solve through data analysis or machine learning, such as complex fault diagnosis and special case handling. For example, a wind turbine fails, but the analysis results cannot determine the specific cause; the improvement measures are to consult wind power experts, troubleshoot and diagnose the fault based on expert experience, and formulate corresponding maintenance plans.

[0100] Cost-effectiveness-based improvement refers to evaluating the cost and benefit of improvement measures and selecting the most cost-effective solution for implementation. This strategy is suitable for scenarios where costs and benefits need to be weighed, such as replacing equipment, upgrading systems, etc. For example, if the power curve deviation rate of a wind turbine is high, it can be solved by replacing the blades or upgrading the control system, but the costs and benefits of the two solutions are different; the improvement measure is to compare the costs, benefits and implementation difficulties of the two solutions and select the most cost-effective solution for implementation. For example, if the cost of replacing the blades is low and the effect is obvious, you can choose to replace the blades; if the cost of upgrading the control system is high, but it can improve the overall operating efficiency, you can choose to upgrade the control system.

[0101] The following further illustrates an improvement strategy based on cost-effectiveness. First, a threshold value (e.g., 10%) is set for the deviation rate to determine whether the deviation is significant. If the deviation rate exceeds the threshold, the cost-effectiveness analysis phase is entered, including blade replacement and control system upgrade solutions.

[0102] a) Blade replacement plan.

[0103] Cost calculation: Consider the price of new blades, installation costs, downtime losses, etc.

[0104] Benefit estimation: Based on the new blade characteristics, a power model is used to predict the increase in power generation after the improvement.

[0105] b) Upgrade control system solution.

[0106] Cost calculation: Consider the price of the new control system, installation and commissioning costs, downtime losses, etc.

[0107] Benefit estimation: Use power models to predict potential increases in power generation based on improved control strategies.

[0108] Calculate the payback period and long-term benefits of each solution, use decision trees or other algorithms to compare the cost-effectiveness of two solutions, and select the solution with the highest cost-effectiveness as the final recommendation. Conduct sensitivity analysis, for example, consider uncertain factors such as electricity price fluctuations and changes in wind resources; use methods such as Monte Carlo simulation to evaluate the feasibility of solutions under different scenarios. In addition, regularly re-evaluate the power curve and deviation rate, and continuously optimize the cost-effectiveness analysis model based on the actual implementation results.

[0109] In this way, the machine learning model is combined with the traditional cost-benefit analysis, multiple influencing factors are taken into account, the accuracy of decision-making is improved, the robustness of decision-making is enhanced through sensitivity analysis, and the automation and intelligence of the decision-making process are realized. This data-driven intelligent decision-making method can help wind farm managers make equipment transformation decisions more scientifically and efficiently. This embodiment identifies the pitch angle error and improves the efficiency of wind energy utilization by correcting the pitch angle control strategy.

[0110] (6) Micro-site selection analysis model

[0111] The micro-site selection analysis model is used to obtain meteorological preprocessing data from multiple sets of preprocessing data of the target location, input the meteorological preprocessing data of the target location into the pre-built wind field model, and output wind resource distribution data; wherein, the wind field model is constructed based on computational fluid dynamics software combined with the terrain data of the target location. Computational fluid dynamics software can accurately simulate the wind field distribution under complex terrain, and the meteorological and processing data include terrain preprocessing data and meteorological preprocessing data.

[0112] Specifically, the micro-site selection analysis model includes a wind resource assessment module, a layout optimization module, and a site selection recommendation module. The wind resource assessment module uses a pre-built wind farm model to perform wind resource assessment. The model displays information such as wind turbine location, wind rose diagram, and energy rose diagram through a map to evaluate the efficiency of wind resource utilization. The layout optimization module is used to optimize the layout of wind turbines based on the analysis results of the wind resource assessment module to improve the efficiency of wind energy utilization, while the site selection recommendation module also uses the analysis results of the wind resource assessment module to provide reference suggestions for future wind farm site selection. The site selection recommendation module can provide site selection suggestions based on many factors such as the existing wind turbine layout, power grid facilities, topography, and environmental factors.

[0113] The micro-site selection analysis model is further explained below through examples.

[0114] First, preliminary data collection is required, including the location of wind turbines, operating data, meteorological data, power grid data, topographic data, etc. of wind farms that have been put into operation; evaluate the current operating status of wind turbines, such as power generation efficiency, failure rate, blade fatigue, etc.; analyze the mutual influence between wind turbines, such as wind shear effect, eddy effect, etc.; analyze the capacity and transmission capacity of power grid facilities, such as grid connection points, substations, transmission lines, etc. Then, data analysis is carried out, using wind energy resource assessment software to evaluate the wind energy utilization efficiency of the existing wind turbine layout and analyze whether there is room for improvement; use CFD simulation software to simulate the mutual influence between wind turbines, analyze wind shear effect and eddy effect, and propose an optimized layout plan; analyze the carrying capacity of power grid facilities and evaluate the feasibility of wind farm expansion.

[0115] For the layout optimization module, the goals are to improve wind energy utilization efficiency, reduce mutual interference between wind turbines, reduce operating costs, and ensure grid security; for wind farms that have been put into operation, layout optimization needs to consider factors such as the operating conditions of existing wind turbines, limitations of grid facilities, topography, and formulate targeted optimization plans.

[0116] Common optimization solutions include adjusting the spacing between wind turbines, adjusting the orientation of wind turbines, adding wind turbines, and replacing wind turbines. Adjust the spacing between wind turbines based on factors such as wind turbine type, wind speed, and wind direction to avoid mutual interference between wind turbines and improve wind energy utilization efficiency. Adjust the orientation of wind turbines based on wind rose diagrams and energy rose diagrams to make them more adaptable to wind energy distribution and improve wind energy utilization efficiency. Add wind turbines Based on the carrying capacity of power grid facilities and wind energy resource assessment results, add wind turbines in appropriate areas to increase the power generation capacity of wind farms. Replace wind turbines Based on the operating status of existing wind turbines, consider replacing old wind turbines or lower-power wind turbines to improve the overall power generation efficiency of wind farms.

[0117] For the site selection recommendation module, based on the analysis results of wind farms that have been put into operation, reference suggestions can be provided for future expansion, such as determining the appropriate expansion area, wind turbine type, layout plan, etc.; the experience of wind farms that have been put into operation can also be applied to the site selection of new wind farms to avoid similar mistakes and improve the design efficiency and operation efficiency of new wind farms.

[0118] The principle of micro-site selection mainly considers factors such as wind energy resources, topography, power grid facilities, and environmental impacts in order to select the optimal layout plan so that wind turbines can efficiently utilize wind energy resources and minimize the impact on the environment.

[0119] For wind energy resource assessment, data such as wind speed, wind direction, and wind energy density are used to evaluate the wind energy resource potential of the wind farm area. In mountainous areas or by the sea, wind energy resources are usually abundant due to the influence of topography and are suitable for the construction of wind farms. For example, the wind rose diagram of a certain area shows that northeast wind accounts for a high proportion and the wind energy density is high. Based on the analysis, it is possible to choose to build a wind farm in this area and arrange wind turbines according to the wind direction to maximize the use of wind energy resources.

[0120] For topographic analysis, consider the impact of topography on wind energy resources, select areas that are conducive to wind energy collection, and avoid areas that are not conducive to wind energy collection. Hillsides, seaside and other topography are conducive to wind energy collection due to the wind energy acceleration effect and are suitable for the construction of wind farms. For example, arranging wind turbines on a hillside can take advantage of the wind energy acceleration effect of the hillside and improve the efficiency of wind energy utilization.

[0121] Grid facilities are matched, and wind turbines are placed in locations that match grid facilities to ensure grid access and power transmission of wind farms. Wind turbines need to be close to grid facilities, such as substations and transmission lines, to be successfully connected to the grid for power generation. For example, choosing to build a wind farm in an area close to a substation can reduce transmission line losses and improve power generation efficiency.

[0122] Environmental impact assessment: assess the impact of wind turbine construction on the environment, such as the impact on wild animals such as birds and bats, the impact on the landscape, etc., and take measures to mitigate them. The construction of wind turbines needs to consider the impact on the surrounding environment, such as the impact on bird migration routes, the impact on the landscape, etc. For example, choose to build wind farms in areas far away from bird migration routes and use environmentally friendly wind turbines to reduce the impact on birds.

[0123] This embodiment evaluates the utilization efficiency of wind resources and optimizes the layout plan of wind turbines according to the analysis results to improve the utilization efficiency of wind energy.

[0124] (7) AGC curve deviation analysis model

[0125] The AGC curve deviation analysis model includes a deviation evaluation module, a control optimization module and an AGC model correction module. The deviation evaluation module is used to determine the AGC curve deviation rate of the wind turbine according to the preprocessed wind turbine power and the pre-built AGC model; wherein the AGC model is constructed based on a time series analysis method (such as ARIMA). The control optimization module is used to adjust the wind turbine operation strategy according to the AGC curve deviation rate of the wind turbine, and the wind turbine operation strategy includes a shutdown strategy, a startup strategy and a power control strategy. The AGC model correction module is used to correct the AGC model according to the AGC curve deviation rate of the wind turbine.

[0126] Specifically, the time series analysis method is suitable for processing time series data with periodicity and trend. This embodiment uses the time series analysis method to predict the power fluctuation characteristics and obtain the power output result. The deviation evaluation module compares the power output result with the actual power output of the wind farm to evaluate the deviation rate of the site AGC curve.

[0127] The function of the control optimization module is to propose suggestions for improving the AGC control strategy, such as adjusting control parameters, optimizing response algorithms, etc.; to propose optimization of the wind turbine group control plan to improve the overall AGC response capability; to propose suggestions for improving the wind speed prediction model to improve the foresight of AGC control; if the problem involves hardware limitations, it may be recommended to upgrade or modify related equipment. The control optimization module mainly includes model correction, parameter adjustment and control strategy improvement. Model correction mainly corrects the AGC model and improves the AGC control accuracy based on the results of AGC deviation analysis. Parameter adjustment mainly adjusts AGC control parameters such as control cycle, power regulation amplitude, etc. based on the results of AGC deviation analysis to optimize AGC control performance. Control strategy improvement mainly improves AGC control strategy based on the results of AGC deviation analysis, such as introducing prediction algorithms, improving the response speed of the control system, etc., to improve AGC control efficiency.

[0128] For the AGC model correction module, the main causes of deviation are wind speed fluctuations, wind direction changes, wind turbine failures, grid load changes and control system errors. For wind speed fluctuations, the power control strategy can be adjusted to timely adjust the wind turbine power generation according to wind speed changes to reduce AGC deviations. For wind direction changes, the yaw control strategy can be adjusted to keep the wind turbine at the best angle to the wind and reduce AGC deviations. For wind turbine failures, the faulty wind turbine needs to be repaired in a timely manner, and the wind turbine operation strategy needs to be adjusted according to the fault situation to avoid affecting AGC control. For grid load changes, it is necessary to coordinate with the grid operator to adjust the wind farm power generation in a timely manner according to grid load changes to reduce AGC deviations. For control system errors, the AGC control system needs to be optimized to improve control accuracy and response speed and reduce AGC deviations.

[0129] Assume that the AGC deviation analysis found that the AGC deviation is large when the wind speed fluctuates greatly in the wind farm. The AGC model does not fully consider the impact of wind speed fluctuations, resulting in insufficient control accuracy. In this regard, the AGC model is improved, wind speed fluctuations are incorporated into the model, and the power generation of the wind farm is predicted based on wind speed changes to improve the AGC control accuracy.

[0130] In some embodiments, the output results of the AGC curve deviation analysis model include the AGC deviation rate, AGC response time, AGC control accuracy, power fluctuation characteristics and possible influencing factors, wherein the AGC deviation rate refers to the percentage difference between the actual power output and the AGC required output; the AGC response time refers to the time required for the wind farm to adjust the output to meet the AGC requirements; the AGC control accuracy is an assessment of the ability of the wind farm to maintain within the AGC required output range; the power fluctuation characteristics refer to the fluctuation frequency, amplitude, etc.; the possible influencing factors refer to wind speed changes, equipment response delays, etc.

[0131] The following further illustrates the application scenarios of the AGC curve deviation analysis model. The AGC curve deviation analysis model can be applied to the optimization of large-scale wind farm grid-connected operation. For example, in order to optimize the AGC response capability of the wind farm and improve the quality of grid-connected operation, a large wind farm needs to accurately respond to AGC instructions to ensure the stability of the grid when operating under the grid dispatch requirements.

[0132] First, collect operation data, including real-time power output data of wind farms through SCADA system, obtain AGC command data from power grid dispatch center, and collect meteorological data such as wind speed and wind direction. Then perform data preprocessing, including time series alignment to ensure the consistency of AGC command and actual output data timestamps; perform data smoothing, and use moving average and other methods to handle instantaneous fluctuations. After that, perform AGC curve deviation analysis, use dynamic time warping (DTW) algorithm to compare AGC command curve and actual output curve, calculate deviation rate, response time and other indicators, and use Fourier transform to analyze power fluctuation characteristics. Perform intelligent diagnosis, use decision tree algorithm to analyze the cause of deviation, input AGC deviation index, wind speed data, equipment status, etc., and output the main factors causing AGC deviation (such as wind speed prediction error, improper control strategy, etc.). Finally, according to historical AGC response data and diagnosis results, the AGC control strategy is optimized based on reinforcement learning algorithm, and the optimized AGC control parameters and wind turbine group control strategy adjustment suggestions are obtained. The output result can be to provide a detailed AGC response capability analysis report, give specific AGC control strategy optimization suggestions, and predict the optimized AGC response improvement effect.

[0133] Through this application scenario, we can see how AGC curve deviation analysis works with the intelligent diagnosis and optimization suggestion modules to improve the grid-connected operation quality and grid-friendliness of wind farms, which is of great significance for large-scale wind power grid connection and stable operation of the power grid.

[0134] In addition, for the selection of algorithm models, support vector machines can be used to analyze the operating data of wind turbines and identify potential faults. Neural networks can be used to analyze historical wind speed data and predict future wind speeds to provide a reference for wind farm operations. Neural networks can be used to analyze wind turbine operating data, predict future operating status of wind turbines, and prevent faults in advance. Random forests can be used to analyze different types of power losses and identify the main causes of losses. Random forests can be used to analyze wind turbine operating data and maintenance records, optimize maintenance strategies, and reduce downtime.

[0135] Step S400, inputting multiple groups of analysis data into the trained diagnosis model respectively to obtain the diagnosis result; the diagnosis model is constructed based on the machine learning model.

[0136] Based on the analysis results of the aforementioned step S300, the power curve deviation rate reflecting the degree of deviation between the wind turbine power output and the theoretical model, the yaw error reflecting the degree of deviation of the wind turbine from the optimal wind angle, the temperature-power relationship reflecting the relationship between the wind turbine power and temperature, the operating time analysis reflecting the operating time and downtime under different wind speed ranges, the loss power analysis reflecting the power loss caused by various factors, the pitch angle error reflecting the deviation between the actual pitch angle and the theoretical pitch angle, and the micro-site selection analysis results reflecting the utilization efficiency of wind energy resources and the rationality of the wind turbine layout are analyzed.

[0137] The AGC curve deviation rate, which reflects the deviation between the actual power output of the wind farm and the expected power output of the AGC curve, is used as a feature and input into the pre-trained diagnostic model for identification to obtain the key factors affecting the power generation capacity of the wind turbine.

[0138] For diagnostic models, if you need to predict future operating status, you can use supervised learning algorithms, such as random forests, gradient boosting trees, etc.; if you need to discover hidden patterns in the data, you can use unsupervised learning algorithms, such as clustering, dimensionality reduction, etc.; if you need to process complex data, you can use deep learning algorithms, such as convolutional neural networks, recurrent neural networks, etc.

[0139] The output of the diagnostic model can be used to identify key factors that affect the wind turbine's power generation capacity, predict possible future failures, and evaluate the wind turbine's operating status.

[0140] Step S500, input the diagnosis result into the pre-built expert optimization suggestion model to obtain the optimization suggestion plan. According to the diagnosis result of step S400, provide targeted optimization suggestions, such as adjusting the wind turbine control parameters, optimizing the operation strategy, formulating a maintenance plan, etc., to improve the power generation capacity and operating efficiency of the wind turbine. The optimization suggestion plan includes adjusting the wind turbine control parameters, optimizing the wind turbine operation strategy and formulating a maintenance plan. Adjusting the wind turbine control parameters includes adjusting the yaw angle, pitch angle and / or speed, optimizing the wind turbine operation strategy includes the shutdown strategy and / or the startup strategy, and formulating a maintenance plan includes regular inspection and / or fault repair.

[0141] The expert optimization recommendation model can use neural network models such as multi-layer perceptron (MLP), convolutional neural network (CNN), recurrent neural network (RNN), etc., and can be selected according to specific circumstances. Tree models such as random forest, gradient boosting tree, or other machine learning models can also be used. The model structure needs to be designed according to the data characteristics and analysis objectives, and the model performance should be optimized through model training and evaluation.

[0142] This embodiment uses a diagnostic model and an expert optimization suggestion model constructed using algorithms such as machine learning to provide targeted optimization suggestions and improve the power generation capacity and operational efficiency of wind turbines.

[0143] The principle of the method of this embodiment is further explained below through Example 1 and Example 2.

[0144] Embodiment 1: Optimizing wind turbine power curve.

[0145] The wind farm operator finds that the actual power generation of a wind turbine is lower than expected, so it is necessary to optimize the wind turbine power curve to improve power generation efficiency. The method includes the following steps (11) to (15).

[0146] Step (11), collect multi-source data.

[0147] Wind turbine operation data: power output, speed, pitch angle, yaw angle, etc. are collected through the SCADA system.

[0148] Meteorological data: wind speed, wind direction, temperature, air pressure, etc. are collected through weather stations.

[0149] Wind farm operation data: Grid connection status, power restriction, etc. are collected through the SCADA system.

[0150] Step (12), perform data preprocessing.

[0151] Perform data cleaning, data normalization, and time processing. Perform data cleaning to remove outliers, such as data that is beyond the physically possible range. Unify data of different dimensions into the [0, 1] interval for model processing. Perform time alignment to ensure that data from different sources have consistent timestamps.

[0152] Step (13), perform model selection and processing.

[0153] Select the power curve fitting analysis model, temperature-power analysis model and wind deviation analysis model. For the power curve fitting analysis model, select the random forest algorithm to establish the power curve fitting analysis model. This is because random forest can handle nonlinear relationships, has strong anti-noise ability, and is not easy to overfit. The input data includes wind speed, temperature, air pressure, etc. The model outputs predicted power. Temperature and power usually have a nonlinear relationship, and polynomial regression can capture this relationship. Therefore, polynomial regression is used to establish a temperature-power analysis model. The temperature data is used as the model data, and the model outputs the influence coefficient of temperature on power. Support vector regression also has good generalization ability for small samples and is suitable for handling delicate relationships such as yaw error. Therefore, support vector regression is used to establish a wind deviation analysis model. The wind direction and yaw angle data are used as model inputs, and the model outputs yaw error estimation.

[0154] Step (14), perform intelligent diagnosis.

[0155] Ensemble learning can integrate the advantages of multiple models and improve the accuracy of diagnosis. Therefore, an ensemble learning method (such as gradient boosting decision tree) is used to construct a diagnosis module and comprehensively analyze the outputs of each module. In this embodiment, the diagnosis module outputs: the main reasons for the power curve deviation (such as temperature influence, yaw error, etc.).

[0156] Step (15), make optimization suggestions.

[0157] Expert systems can encode domain knowledge into rules and generate explainable suggestions. The expert optimization suggestion model of this embodiment uses a rule-based expert system to generate optimization suggestions.

[0158] The expert optimization suggestion model outputs specific optimization suggestions, such as adjusting the yaw control strategy, optimizing the cooling system, etc. It also generates a wind turbine power curve optimization plan, including specific parameter adjustment suggestions and expected improvement effects.

[0159] Embodiment 2: Micro-site optimization of wind farms.

[0160] In order to optimize the site selection of new wind turbines and maximize the efficiency of wind energy utilization, the wind farm developer plans to expand the existing wind farm, including the following steps (21) to (25).

[0161] Step (21), collect multi-source data.

[0162] Wind turbine operation data: SCADA data of existing wind turbines.

[0163] Meteorological data: On-site weather tower and numerical weather forecast data.

[0164] Topographic data: high-resolution topographic maps, satellite imagery.

[0165] Step (22), perform data preprocessing.

[0166] Perform data interpolation and time series processing; interpolate terrain data to generate a continuous terrain model; perform time series analysis on meteorological data to extract long-term trends.

[0167] Step (23), perform model selection and processing.

[0168] Select the micro-site selection analysis model, power loss analysis model and AGC curve deviation analysis model.

[0169] For the micro-site selection analysis model, a computational fluid dynamics (CFD) model is used to simulate the wind farm, and terrain data and meteorological data are input; a high-precision wind resource distribution map is output. For the power loss analysis model, neural networks can capture complex nonlinear relationships and are suitable for simulating wake effects. Therefore, artificial neural networks are used to predict possible wake losses. The model inputs wind turbine location, wind speed, and wind direction data. The model outputs predicted wake losses. For the AGC curve deviation analysis model, a time series analysis method (such as ARIMA) is used to predict power fluctuations; the model inputs historical power output data, and the model outputs predicted power fluctuation characteristics.

[0170] Step (24), perform intelligent diagnosis.

[0171] Genetic algorithms can find a balance between multiple objectives and are suitable for complex site selection problems. Therefore, the diagnostic module of this embodiment adopts a multi-objective optimization algorithm (such as a genetic algorithm); the input of the diagnostic module includes wind resource distribution, wake loss prediction, power grid constraints, etc. The output of the diagnostic module is multiple possible wind turbine layout solutions.

[0172] Step (25), make optimization suggestions.

[0173] The decision support system can consider multiple factors and assist in complex decision making. Therefore, the expert optimization suggestion model of this embodiment uses the decision support system to evaluate various plans. The input of the expert optimization suggestion model is the layout plan generated by intelligent diagnosis; the output is the optimal wind turbine layout suggestion, including the expected power generation and economic benefit analysis of each location. The results of the expert optimization suggestion model can provide a detailed wind farm expansion plan, including the optimal location of the new wind turbines, the expected power generation and economic benefit evaluation.

[0174] In summary, this embodiment can effectively improve the power generation efficiency of wind turbines and increase annual power generation through analysis, diagnosis and optimization; reduce operating costs and improve economic benefits by optimizing wind turbine operation and maintenance; use machine learning algorithms to achieve intelligent diagnosis and optimization suggestions, and improve the intelligence level of wind farms; promote the digital transformation of the wind power industry and provide technical support for the sustainable development of the new energy industry.

[0175] See also Figure 2, this embodiment provides a wind turbine generator generation capacity analysis, diagnosis and optimization device, including: an acquisition module 100, a preprocessing module 200, an analysis module 300, a diagnosis module 400 and an optimization suggestion module 500. The acquisition module 100 is used to acquire multiple groups of data to be processed, and the multiple groups of data to be processed include wind turbine generator operation data, meteorological data and / or wind farm operation data. The preprocessing module 200 is used to preprocess the multiple groups of data to be processed to obtain multiple groups of preprocessed data. The analysis module 300 is used to input the multiple groups of preprocessed data into corresponding multiple analysis models respectively to obtain multiple groups of analysis data; the multiple analysis models include power curve fitting analysis model, wind deviation analysis model, temperature-power analysis model, operation time analysis model, power loss analysis model, pitch angle analysis model, micro-site selection analysis model and / or AGC curve deviation analysis model. The diagnosis module 400 is used to input the multiple groups of analysis data into the trained diagnosis model respectively to obtain the diagnosis result; the diagnosis model is constructed based on the machine learning model. The optimization suggestion module 500 is used to input the diagnosis result into a pre-built expert optimization suggestion model to obtain an optimization suggestion plan.

[0176] In an optional embodiment, the wind turbine operating data includes a combination of one or more of wind turbine power, wind turbine speed, wind turbine direction, wind turbine temperature, wind turbine pitch angle, wind turbine angle to wind, wind turbine speed and wind turbine status code; the meteorological data includes a combination of one or more of ambient wind speed, ambient wind direction, ambient temperature, ambient air pressure and ambient humidity; the wind farm operating data includes a combination of one or more of the wind turbine grid-connected status, wind turbine power limitation and maintenance records.

[0177] In an optional embodiment, the power curve fitting analysis model includes a wind turbine power prediction module, a deviation rate prediction module and an update module. The wind turbine power prediction module is used to predict the corresponding wind turbine power under different environmental wind speeds based on preprocessed data and a pre-built power curve sub-model. The deviation rate prediction module is used to determine the power curve deviation rate based on the actual power, rated power and theoretical output power corresponding to different environmental wind speeds. The update module is used to update the power curve sub-model according to the power curve deviation rate; wherein the power curve sub-model includes a curve fitting module and a sub-model module. The curve fitting module is used to use a variety of machine learning algorithms to curve fit the historical data of the wind turbine to obtain multiple power curve primary models of the wind turbine; the machine learning algorithm includes a support vector machine, a neural network and / or a random forest; the sub-model module is used to evaluate the fitting effect of multiple power curve primary models using evaluation indicators, determine the power curve primary model with the best fitting effect from multiple power curve primary models according to the fitting effect, and use the power curve primary model with the best fitting effect as the power curve sub-model; wherein the evaluation indicators include an R square value indicator and / or a root mean square error indicator.

[0178] In an optional embodiment, the wind deviation analysis model includes a yaw error data module, an error correction module and a yaw error compensation module. The yaw error data module is used to obtain static yaw error data based on the wind speed preprocessing data of the wind turbine, the wind angle preprocessing data of the wind turbine and the pre-constructed wind deviation sub-model in the preprocessing data, wherein the wind deviation sub-model is constructed based on the support vector regression algorithm. The error correction module is used to adjust the yaw control strategy of the wind turbine according to the static yaw error data to optimize the wind energy utilization efficiency. The yaw error compensation module is used to compensate the yaw error data into the power curve sub-model to improve the accuracy of the power curve sub-model.

[0179] In an optional embodiment, the temperature-power analysis model includes a power generation capacity prediction module and a temperature compensation module. The power generation capacity prediction module is used to determine the influence coefficient of temperature on power based on the ambient temperature preprocessing data, the wind turbine power preprocessing data and the pre-built temperature-power relationship sub-model in the preprocessing data, and predict the wind turbine power at different temperatures based on the influence coefficient of temperature on power. The temperature compensation module is used to adjust the power curve sub-model according to the influence coefficient of temperature on power to improve the accuracy of the power curve sub-model. Among them, the temperature-power relationship sub-model is determined by linear regression, polynomial regression or neural network model based on the maximum power change of the wind turbine set at different ambient temperatures.

[0180] In an optional embodiment, the power loss analysis model includes a loss type module, a loss ratio module and a loss cause module. The loss type module is used to determine the power loss corresponding to multiple loss types based on preprocessed data and a pre-built power loss submodule; wherein the multiple loss types include one or more of maintenance, failure, power restriction, performance, and involvement. The loss ratio module is used to determine the power loss ratio of each loss type based on the power loss. The loss cause module is used to determine the loss cause corresponding to each loss type. Among them, the power loss submodule includes a power generation module, a power generation module and a loss type module. The power generation module is used to determine the power generation that the wind turbine should generate in multiple time periods based on the wind turbine operation data and meteorological data. The power loss module is used to compare the power generation that the wind turbine should generate in multiple time periods with the actual power generation in the corresponding time periods, and determine the power loss in each time period. The loss type module is used to determine the loss type corresponding to the power loss in each time period.

[0181] In an optional embodiment, the micro-site selection analysis model is used to obtain meteorological preprocessed data from multiple groups of preprocessed data of the target location, input the meteorological preprocessed data of the target location into a pre-constructed wind field model, and output wind resource distribution data; wherein the wind field model is constructed based on computational fluid dynamics software combined with terrain data of the target location.

[0182] In an optional embodiment, the optimization suggestion includes adjusting wind turbine control parameters, optimizing wind turbine operation strategies, and formulating maintenance plans;

[0183] Adjusting wind turbine control parameters includes adjusting yaw angle, pitch angle and / or rotation speed;

[0184] Optimizing wind turbine operation strategies including shutdown strategies and / or startup strategies;

[0185] Develop a maintenance schedule that includes regular inspections and / or breakdown repairs.

[0186] In an optional embodiment, the optimization suggestion includes adjusting wind turbine control parameters, optimizing wind turbine operation strategies, and formulating maintenance plans;

[0187] Adjusting wind turbine control parameters includes adjusting yaw angle, pitch angle and / or rotation speed;

[0188] Optimizing wind turbine operation strategies including shutdown strategies and / or startup strategies;

[0189] Develop a maintenance schedule that includes regular inspections and / or breakdown repairs.

[0190] The device provided in the embodiment of the present application is adopted. Since the device adopts the same inventive concept as the above-mentioned method provided in the embodiment of the present application, on the premise that the method can solve the technical problem, the device can also solve the technical problem, which will not be repeated here.

[0191] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

Claims

1. A method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine, characterized in that: include: Acquire multiple groups of data to be processed, wherein the multiple groups of data to be processed include wind turbine operating data, meteorological data and / or wind farm operating data; Preprocessing the multiple groups of data to be processed to obtain multiple groups of preprocessed data; Inputting the plurality of groups of preprocessed data into corresponding plurality of analysis models respectively to obtain plurality of groups of analysis data; the plurality of analysis models include power curve fitting analysis model, wind deviation analysis model, temperature-power analysis model, operation time analysis model, power loss analysis model, pitch angle analysis model, micro-site selection analysis model and / or AGC curve deviation analysis model; Inputting the plurality of sets of analysis data into the trained diagnostic model respectively to obtain a diagnostic result; the diagnostic model is constructed based on a machine learning model; The diagnosis results are input into a pre-built expert optimization suggestion model to obtain an optimization suggestion plan.

2. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 1, characterized in that: The wind turbine operation data includes one or more combinations of wind turbine power, wind turbine speed, wind turbine direction, wind turbine temperature, wind turbine pitch angle, wind turbine angle to wind, wind turbine speed and wind turbine status code; the meteorological data includes one or more combinations of ambient wind speed, ambient wind direction, ambient temperature, ambient air pressure and ambient humidity; the wind farm operation data includes one or more combinations of wind turbine grid-connected status, wind turbine power limitation status and maintenance records.

3. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 1, characterized in that: The power curve fitting analysis model includes: A wind turbine power prediction module, used to predict the corresponding wind turbine power under different ambient wind speeds based on the preprocessed data and a pre-built power curve sub-model; A deviation rate prediction module is used to determine the power curve deviation rate according to the actual power corresponding to different environmental wind speeds, the rated power and theoretical output power of the wind turbine generator set; An updating module, configured to update the power curve sub-model according to the power curve deviation rate; The power curve sub-model is constructed by the following method: Performing curve fitting on the historical data of the wind turbine generator set using a plurality of machine learning algorithms to obtain a plurality of primary power curve models of the wind turbine generator set; the machine learning algorithms include support vector machines, neural networks and / or random forests; The fitting effects of the multiple power curve primary models are evaluated by using evaluation indicators, and the power curve primary model with the best fitting effect is determined from the multiple power curve primary models according to the fitting effects, and the power curve primary model with the best fitting effect is used as the power curve sub-model; wherein the evaluation indicators include an R-square value indicator and / or a root mean square error indicator.

4. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 3, characterized in that: The wind deviation analysis model includes: A yaw error data module is used to obtain static yaw error data according to the wind speed preprocessing data of the wind turbine, the wind angle preprocessing data of the wind turbine and a pre-built wind deviation sub-model in the pre-processing data, wherein the wind deviation sub-model is built based on a support vector regression algorithm; An error correction module, used for adjusting the yaw control strategy of the wind turbine according to the static yaw error data to optimize the wind energy utilization efficiency; The yaw error compensation module is used to compensate the yaw error data into the power curve sub-model to improve the accuracy of the power curve sub-model.

5. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 3, characterized in that: The temperature-power analysis model includes: A power generation capacity prediction module is used to determine the influence coefficient of temperature on power according to the ambient temperature preprocessing data, the fan power preprocessing data and the pre-built temperature-power relationship sub-model in the preprocessing data, and predict the fan power at different temperatures according to the influence coefficient of temperature on power; A temperature compensation module, used for adjusting the power curve sub-model according to the influence coefficient of the temperature on the power, so as to improve the accuracy of the power curve sub-model; The temperature-power relationship sub-model is determined according to the maximum power variation of the wind turbine generator set under different ambient temperatures by using linear regression, polynomial regression or neural network model.

6. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 1, characterized in that: The power loss analysis model includes: A loss type module, used to determine the loss electricity corresponding to multiple loss types according to the pre-processed data and the pre-built loss electricity sub-module; wherein the multiple loss types include one or more of maintenance, failure, power restriction, performance, and involvement; A loss ratio module, used to determine the power loss ratio of each loss type according to the power loss; A loss cause module, used to determine the loss cause corresponding to each loss type; The power loss submodule is constructed by the following method: Determine the power generation amount of the wind turbine set in multiple time periods according to the wind turbine set operation data and the meteorological data; Compare the power generation that the wind turbines should generate in multiple time periods with the actual power generation in the corresponding time periods to determine the power loss in each time period; Determine the loss type corresponding to the lost power in each time period.

7. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 1, characterized in that: The micro-site selection analysis model is used to obtain meteorological preprocessing data from multiple groups of preprocessing data of the target location, input the meteorological preprocessing data of the target location into a pre-constructed wind field model, and output wind resource distribution data; wherein the wind field model is constructed based on computational fluid dynamics software combined with terrain data of the target location.

8. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 2, characterized in that: The AGC curve deviation analysis model includes: A deviation evaluation module, used to determine the AGC curve deviation rate of the wind turbine according to the preprocessed wind turbine power and the pre-constructed AGC model; wherein the AGC model is constructed based on a time series analysis method; A control optimization module, used for adjusting the wind turbine operation strategy according to the AGC curve deviation rate of the wind turbine group, wherein the wind turbine operation strategy includes a shutdown strategy, a startup strategy and a power control strategy; The AGC model correction module is used to correct the AGC model according to the AGC curve deviation rate of the wind turbine generator set.

9. The method for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine according to claim 1, characterized in that: The optimization proposal includes adjusting fan control parameters, optimizing fan operation strategy and formulating maintenance plan; The adjusting of the wind turbine control parameters includes adjusting the yaw angle, the pitch angle and / or the rotation speed; Optimizing wind turbine operation strategies including shutdown strategies and / or startup strategies; Develop a maintenance schedule that includes regular inspections and / or breakdown repairs.

10. A device for analyzing, diagnosing and optimizing the power generation capacity of a wind turbine, characterized in that: include: An acquisition module, used for acquiring multiple groups of data to be processed, wherein the multiple groups of data to be processed include wind turbine operating data, meteorological data and / or wind farm operating data; A preprocessing module, used for preprocessing the plurality of groups of data to be processed to obtain a plurality of groups of preprocessed data; An analysis module, used for inputting the plurality of groups of preprocessed data into corresponding plurality of analysis models respectively to obtain plurality of groups of analysis data; the plurality of analysis models include a power curve fitting analysis model, a wind deviation analysis model, a temperature-power analysis model, a running time analysis model, a power loss analysis model, a pitch angle analysis model, a micro-site selection analysis model and / or an AGC curve deviation analysis model; A diagnosis module, used to input the plurality of sets of analysis data into the trained diagnosis model to obtain a diagnosis result; the diagnosis model is constructed based on a machine learning model; The optimization suggestion module is used to input the diagnosis results into a pre-built expert optimization suggestion model to obtain an optimization suggestion plan.

Images