A method for predicting the icing status of wind turbine blades in wind farms

Abstract

The present invention discloses a method for predicting the icing state of wind turbine blades in a wind farm, including: obtaining historical operation data and grid-connected power data of the wind farm, preprocessing the historical operation data and grid-connected power data of the wind farm, and eliminating invalid data; calculating and determining the time series of wind speed, wind direction, temperature, humidity, and pressure data in the first time period of the wind tower, and the time series of grid-connected power data, and integrating the two sets of data according to the intersection of time as historical data for predicting the icing state of wind turbine blades in the wind farm; obtaining the historical records of icing conditions of wind turbine blades in the on-site wind farm; establishing an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm; obtaining the time when icing occurs on the wind turbine blades in the wind farm according to the icing prediction model, and recording the state; if icing occurs, it is marked as 1, otherwise it is marked as zero. A corresponding system, electronic device, and computer-readable storage medium are also disclosed.

Description

Technical Field

[0001] The present invention relates to the technical field of wind farm equipment monitoring and management, and in particular to a method for predicting the icing state of wind turbine blades in a wind farm. Background Art

[0002] Wind power generation has a high resource utilization rate and relatively mature technology. With the development of technology, the cost is also decreasing, so the installed capacity will increase. However, wind power generation is very sensitive to changes in meteorological conditions and has a certain degree of instability. In areas with humid climates in winter and spring, icing of wind turbine blades is the main factor affecting the normal operation of wind farms. After the blades are covered with ice, the aerodynamic characteristics of the blades change, and the dynamic load of the unit increases, which will have a negative impact on the safe operation and benefits of the wind farm. Due to the influence of different climatic environments and terrain conditions, wind turbine blades will form ice of different textures and shapes during operation. There are great differences in the icing process and characteristics of wind turbine blades in different regions, and the corresponding de-icing technical measures are also different.

[0003] In addition, due to the daily operation system maintenance of the wind farm, power grid power restriction plan, data acquisition equipment failure and other reasons, some historical operation data of the wind farm cannot accurately reflect its actual power generation capacity at the corresponding time. This part of the data will cause great interference in accurately identifying the overall freezing state of the wind farm during the icing prediction process, thereby affecting the accuracy of the wind farm icing prediction and the reliability of the icing prediction model. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides the following technical solutions.

[0005] In one aspect, the present invention provides a method for predicting the ice coating state of wind turbine blades in a wind farm, comprising:

[0006] Step 1, obtaining historical operation data and grid-connected power data of the wind farm, performing data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminating invalid data;

[0007] Step 2: Calculate and determine the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrate the two sets of data according to the time intersection as the historical data for predicting the icing status of wind turbine blades in the wind farm;

[0008] Step 3, obtaining the historical records of ice coverage of wind turbine blades in the on-site wind farm;

[0009] Step 4, establishing an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm;

[0010] Step 5: According to the icing prediction model, the time when the wind turbine blades in the wind farm are covered with ice is obtained, and the status is recorded; if icing occurs, it is marked as 1, otherwise it is marked as zero.

[0011] Preferably, the removing of invalid data comprises:

[0012] (1) Eliminate data from the fan scheduled maintenance period;

[0013] (2) Eliminate shutdown data caused by various factors of the fan, including shutdown due to faults, manual shutdown, etc.;

[0014] (3) Eliminate data corruption caused by abnormal factors, including communication interruption, SCADA system failure, etc.;

[0015] (4) Eliminate the power reduction operation data caused by power rationing of the power grid.

[0016] Preferably, the first time period is 15 minutes.

[0017] Preferably, the step 4 establishes an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm, including:

[0018] Step 41, determining multiple index factors that affect blade icing;

[0019] Step 42, according to the grey correlation analysis, quantitatively calculate the values ​​of the correlation between the multiple index factors and the degree of icing on the wind turbine blades, determine the importance of the factors affecting the icing on the blades, and arrange the multiple index factors affecting the icing on the wind turbine blades in descending order of influence through the grey correlation analysis, and select the values ​​of the correlation to establish the parameters used in the blade icing prediction model;

[0020] Step 43, collecting weather forecast data;

[0021] Step 44, determining key element data in the weather forecast data according to the value of the correlation degree;

[0022] Step 45, obtaining an icing state sequence according to the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind measuring tower in the first time period and the time series of grid-connected power data;

[0023] Step 46: Based on the key element data in the weather forecast data and the icing state sequence, a statistical method and a probability method are used to obtain critical conditions of key meteorological elements when the wind farm is predicted to produce icing.

[0024] Preferably, the multiple indicator factors that affect blade icing determined in step 41 are freezing meteorological conditions and meteorological elements related to freezing weather, the freezing meteorological conditions include rime, rime and mixed rime of rain and fog, and the meteorological elements related to freezing weather include air temperature, relative humidity and wind speed.

[0025] Preferably, the method further comprises:

[0026] Step 6, when predicting the occurrence of icing on the fan blades, the icing degree is specifically divided based on the actual output data of the fan blades in the generated time period, and the influence coefficient of different degrees of icing on the actual output of the fan blades is calculated.

[0027] Preferably, the specific classification of the ice coverage degree includes predicting the degree of freezing and the power correction coefficient of the blade output corresponding to the freezing degree through fuzzy cluster analysis and particle swarm optimization methods.

[0028] A second aspect of the present invention provides a wind farm wind turbine blade icing state prediction system, comprising:

[0029] A wind farm data acquisition and preprocessing module is used to acquire historical operation data and grid-connected power data of the wind farm, perform data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminate invalid data;

[0030] The icing state prediction data calculation module is used to calculate and determine the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrate the two sets of data according to the time intersection as the historical data for the prediction of icing state of wind turbine blades in the wind farm;

[0031] An ice coverage history acquisition module is used to acquire the history records of ice coverage of wind turbine blades in the on-site wind farm;

[0032] An icing prediction model establishment module is used to establish an icing prediction model based on a reference wind turbine power curve and in combination with the historical records of icing conditions on wind turbine blades in the on-site wind farm;

[0033] A blade icing state prediction module is used to obtain the time when the wind turbine blades in the wind farm are covered with ice according to the icing prediction model, and record the state; if icing occurs, it is marked as 1, otherwise it is marked as zero;

[0034] The output correction module is used to specifically classify the degree of icing when predicting the occurrence of icing on the fan blades, combined with the actual output data of the fan blades within the generated time period, and calculate the influence coefficient of different degrees of icing on the actual output of the fan blades.

[0035] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is configured to read the instructions and execute the method described in the first aspect.

[0036] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor and execute the method described in the first aspect.

[0037] The present invention has the following beneficial effects:

[0038] 1) According to weather forecast data, before conducting icing prediction analysis on wind farms, it is necessary to preprocess the historical operation data of wind farms and eliminate invalid data to make the prediction more accurate.

[0039] 2) Grey correlation analysis can quantitatively calculate the influence of various factors on the icing degree of wind turbine blades, determine the importance of factors affecting blade icing, and lay the foundation for establishing an icing prediction model.

[0040] 3) According to the results of grey correlation analysis, the key element data in the numerical weather forecast data are sorted out, and combined with the icing state sequence data, statistical methods and probability methods are used to obtain the critical conditions of the key meteorological elements when predicting icing in wind farms.

[0041] 4) When predicting the occurrence of icing, it is necessary to combine the actual output data of the corresponding time period, give a specific classification of the degree of icing, and give the influence coefficient of different degrees of icing on the actual output. Then, the icing influence coefficient is introduced into the short-term wind power prediction model. Through fuzzy clustering analysis, particle swarm optimization and other methods, the degree of freezing and the power correction coefficient corresponding to the freezing degree are predicted, which improves the prediction accuracy and reduces the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a schematic flow chart of the method of the present invention.

[0043] Figure 2 A schematic structural diagram of an electronic device embodiment provided by the present invention. DETAILED DESCRIPTION

[0044] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0045] The method provided by the present invention can be implemented in the following terminal environment, and the terminal may include one or more of the following components: a processor, a memory, and a display screen. The memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method described in the following embodiment.

[0046] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the entire terminal, and executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory.

[0047] The memory may include random access memory (RAM) or read-only memory (ROM). The memory may be used to store instructions, programs, codes, code sets or instructions.

[0048] The display screen is used to display the user interface of each application.

[0049] In addition, those skilled in the art can understand that the structure of the above terminal does not constitute a limitation on the terminal, and the terminal may include more or fewer components, or combine certain components, or arrange the components differently. For example, the terminal also includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, and a power supply, which will not be described in detail here.

[0050] 1. Mechanism and type of ice coating on wind turbine blades

[0051] 1. Basic meteorological conditions for wind turbine blade icing

[0052] According to the analysis of the icing area, the reasonable forecast of the wind farm weather and the understanding of the operating conditions, the blade icing problem should meet the following conditions: the ambient temperature is lower than 0 degrees Celsius, the blade surface is lower than -5 degrees Celsius, and the air humidity is above 85%. Although the range of these meteorological conditions is relatively narrow, wind farms are usually located in the intersection of cold and hot air currents in the atmosphere, and are seriously affected by micro-topography and micro-meteorology, so the blade icing phenomenon of some wind farm units is relatively common.

[0053] 2. Analysis of the icing principle on blade surface

[0054] (1) Mechanism of ice formation on blade surface

[0055] The phenomenon of ice on the surface of the blade is a phase change of water under specific conditions. As long as the temperature is less than 0 degrees Celsius, water will freeze. This is a natural law that cannot be changed. Under specified conditions, water droplets caused by supercooled air collide with the fan blades and condense into ice. The ice on the surface of the blade is generally supercooled water droplets. Supercooled water droplets mainly refer to water droplets that still exist in liquid form when the environment is below 0 degrees Celsius. The state of water vapor, liquid, and solid will also change significantly due to different air pressures in the environment. The freezing temperature of water at standard atmospheric pressure is 0 degrees Celsius. The freezing temperature will gradually decrease as the ambient air pressure increases, so liquid water will also exist below 0 degrees Celsius. Supercooled water droplets in the atmosphere have poor stability. If the environment changes, supercooled water droplets will quickly condense and turn into ice. When the surface of a wind turbine blade hits the air with supercooled water droplets, the water droplets will hit the windward side of the leading edge of the blade, destroying the internal balance of the supercooled water droplets, making the freezing temperature of the supercooled water droplets higher, making it easier to freeze on the surface of the wind turbine blade. The occurrence of ice on the surface of wind turbine blades is a stage of coupled phase change and complex heat transfer. Because the surface temperature of the blade is lower than the temperature of the cold water droplets, when the supercooled water droplets hit the blades, they will quickly absorb water and release heat after solidification, and then the supercooled water droplets will quickly freeze on the surface of the blades.

[0056] (2) Icing types and classifications

[0057] There are three main types of ice on the surface of blades: rime, rime and mixed rime: rime. If the diameter of the supercooled water droplets is relatively large, after colliding with the blades, the water droplets first spread into a water film and then quickly condense into icicles. This type of ice generally appears when the ambient temperature is 0-5℃. Under conditions of high air humidity, a transparent ice layer with high hardness will appear on the surface of the blades. This type of ice layer has a large density and strong adhesion; rime. The diameter of the supercooled water droplets is small, floating and moving in the air, and quickly forming icicles after hitting the blades, existing in a dry growth mode. Generally, when the ambient temperature is lower than -5℃, there is very little water content in the air, and the shape of the crystals is irregular. It is very easy to form an ice surface on the rough blade surface. Its texture is loose, the density is low, and the adhesion is insufficient; mixed rime. The diameter of the supercooled water droplets is different. They float and move in the air. After contacting the blades, they quickly become ice. Some are dry growth and some are wet growth. The ice inside exists in a translucent form with medium density. It usually appears on the water-facing surface of the blade and has a strong adhesion. Under high wind speed conditions in mountainous areas, ice crystals from the clouds will form some medium-sized ground fog, which usually occurs when the ambient temperature is between -10 and -3°C. Therefore, rime and mixed rime are very likely to appear on the blade surface. In addition, after long-term operation of the blade, there will be more stains on the surface, leading edge corrosion, excessive blade roughness, and other problems, which will also cause ice on the blade.

[0058] 2. Analysis of factors affecting icing

[0059] The icing on wind turbine blades is directly related to the natural environment and terrain conditions of the wind farm, while the type of icing is directly related to the ambient temperature, wind speed, diameter of supercooled water droplets in the impeller swept area, and liquid water content in the air.

[0060] (1) Meteorological conditions affecting blade icing

[0061] Ambient temperature. It has the most direct and obvious impact on the ice on the leaves. Ice generally occurs when the ambient temperature is between -1℃ and -8℃. If the ambient temperature is too low, the supercooled water droplets will appear as snowflakes, and ice will not occur. Therefore, in winter, when the environment is low, the incidence of ice in the northern region is low, while in southern Yunnan, Guizhou, Hunan and other regions with high humidity, the problem of ice is more serious; air humidity. Its level has an absolute influence on the formation of leaf ice. Usually, when the relative humidity of the air exceeds 85%, it is very easy to produce leaf ice and rime. The highest incidence of ice is in Hunan, Hubei, Jiangxi and other places. It is rainy in winter and early spring, and the air humidity is very high. In many cases, it can even reach more than 90%, so the incidence of leaf ice is high, and it mainly exists in the form of rime; wind speed. Since wind can directly transport supercooled water droplets, it will also have a direct impact on the ice on the leaves. When the wind speed is low, it effectively promotes the formation of rime, and when the wind speed is high, it will produce granular rime. Generally speaking, the greater the wind speed, the faster ice will form on the blades.

[0062] The diameter of supercooled water droplets. When the diameter of water droplets is relatively large, an icing reaction will occur when they come into contact with the blades, and the latent heat release rate is very slow. On the contrary, if the latent heat release rate is too fast, the icing will also be faster, so the characteristics of ice formation vary greatly. During the icing process of rime on blades, the diameter of supercooled water droplets is relatively large, about 10 to 40 μm, and it is mostly drizzle weather. When rime is icing, the diameter of supercooled water droplets is between 1 and 20 μm, while the diameter of mixed rime water droplets is between 5 and 35 μm, and it is mostly dense fog weather; Condensation height. It is the altitude at which supercooled water droplets in the atmospheric environment form ice crystals or snowflakes, which will change to a certain extent according to different ground temperatures and dew point temperatures. The condensation height is mainly the height at which air droplets freeze after collision with the ground as the starting point. Its numerical value will have a direct impact on the icing of wind turbine blades in high-altitude areas. When the height of the wind turbine blade sweep area exceeds the condensation height, the units in this area are mostly in severe icing conditions.

[0063] (2) Topography and geographical environment that affect blade icing

[0064] Icing on wind turbine blades is affected by local terrain conditions, meteorological conditions and the operating status of wind turbines. Icing is directly affected by the direction, slope, watershed and other aspects of mountainous areas. Icing on wind turbine blades in mountainous areas is most affected by terrain conditions. According to the current research conclusions, the severity of icing is directly affected by terrain conditions. The absolute altitude does not have a direct impact on icing, but the terrain and measurement points in the surrounding environment and the relative height of the surrounding area will also affect icing. The icing will be more serious in areas with prominent terrain conditions under relatively good wind conditions, such as mountain tops, passes, windward slopes, etc., as well as areas where liquid ice is prone to appear in rivers and mountain tops with high air moisture.

[0065] (3) Impact of blade icing on unit power output

[0066] The incidence of blade icing is relatively high on the windward side, and ice on the blade tip is mainly deposited at the blade root. Ice accumulation and irregular shedding will also affect the unit output power. Micro-icing on the blade surface will cause its surface roughness to be high, resulting in low blade aerodynamic performance and low unit operating power; severe blade icing will cause zero blade torque and no output power. Natural icing will also cause severe vibration and cause shutdown problems.

[0067] Blade icing will directly affect the output power of the system, which is affected by the weight of ice, the distribution of the rear wing after ice, blade design and wind turbine control. A simulation analysis of the Vestas V80-1.8MW wind turbine shows that the blade icing can reach 709kg under rime conditions, accounting for 11% of the blade weight (total blade weight 6500kg); under rime conditions, the blade icing can reach 434kg, accounting for 6.7% of the blade weight. Research and analysis found that the icing problems caused by the following two situations will affect the blade resistance: the blade resistance increases exponentially along the blade axis, the blade tip resistance increases by 365% in rime, and increases by 250% in rime, while the lift at the blade tip is reduced by 40% under both conditions.

[0068] In summary, the blade icing combined with the severity of the ice disaster will cause a large change in the wind turbine output power. Light icing will cause a partial loss of blade power, while severe icing will cause its output power to be zero and it will stop running.

[0069] The hazards caused by blade icing to the unit include imbalance between static and dynamic loads, excessive vibration of the unit, changes in blade frequency, increased fatigue load, increased blade bending moment, and personal injury.

[0070] 3. Wind turbine blade ice detection and blade anti-icing and de-icing system

[0071] The existing technology cannot fundamentally eliminate the icing problem. It can only take necessary measures to remove the icing immediately after the icing occurs, which can effectively reduce the economic losses.

[0072] (1) Ice measurement method and necessity

[0073] Ice measurement can be divided into three categories: direct method, indirect method and mathematical model method. The direct method uses the changes in weight and reflection characteristics caused by ice to measure; the indirect method estimates ice based on ice-related factors such as wind speed, humidity and temperature and the change in wind turbine power generation under ice conditions. Ice detection can be carried out by taking measures to actively implement antifreeze and deicing systems to meet requirements, but it is also prone to negative effects. Under conditions of slight ice on the blades, the output power of the blades will drop by 5-15%. If the ice detector cannot accurately detect the ice, it will generally be detected through video and other means within a few hours of the ice appearing. In this process, not only the efficiency of wind power generation is reduced, but the blade heating system will also cause excessive energy consumption, and the ice will be dissolved by heating.

[0074] (2) Blade ice sensor based on blade vibration detection

[0075] Accurate ice detection must directly measure the amount of ice on the blades. The blades are running in cold air, and the tip speed reaches 250km / h. The icing conditions at the blade tip are very different from the icing conditions monitored by the sensor installed on the top of the nacelle. It is recommended that the unit use a blade vibration detection blade ice sensor, which is designed based on a simple physical principle: the increase in ice will cause the natural vibration frequency of the blade to change, and the natural frequency will decrease as the solid weight increases. Based on a highly sensitive sensor system and a special algorithm, the measurement resolution of the ice thickness within the millimeter range can be achieved. Measurements can be performed when the fan is running or in standby mode, which means that the fan is being tested at all times before starting to ensure that the amount of ice on the blade is within the range of operating conditions. Currently, there are blade ice detection systems based on fiber optic vibration sensors and blade ice detection systems based on fiber optic vibration sensors. Fan anti-freeze and de-icing mainly include two types: anti-icing and de-icing. The former is to take measures to prevent ice and snow from adhering to the blade surface, and the latter is to directly remove ice. These two methods can also be understood as passive and active. The former uses physical principles to avoid icing problems, while the latter eliminates ice by heating through an external system.

[0076] 4. Benefits and costs of antifreeze and deicing systems

[0077] At present, there is no commercial wind turbine de-icing system with relatively mature technology. Whether it is active or passive, the de-icing system cannot fundamentally avoid the occurrence of blade icing problems. Some de-icing systems can only ensure normal power output after the blades are covered with ice, prevent shutdown problems, avoid serious vibration effects on the blades, and protect people's lives. If the fan power is reduced by heating, the maintenance effect can be improved. Under low climate ambient temperature, the power consumption of the de-icing system is only 3-8% of the total power generation, and the power consumption of the anti-icing system is less than 3% of the total power generation. Assuming that the cost of the wind turbine blade de-icing system accounts for about 5% of the total investment, the proportion of the anti-icing system cost will decrease with the increase of blade size. The anti-icing system should be heated continuously to ensure that it is always kept above 0°C, while the de-icing system can effectively remove the ice. Both methods will cause power loss. Icing detection can reduce the power loss caused by ice melting. In order to reduce energy consumption, it can be divided into multiple heating areas. Therefore, ensuring that 30% of the length of the blade tip is free of ice can achieve 90% of the fan's aerodynamic performance. If an anti-freeze and ice-melting system is installed at the blade tip, energy consumption can be reduced.

[0078] 5. Blade gas-heat deicing solution

[0079] The basic principle of the blade air-heat deicing scheme is to achieve the purpose of deicing by heating the air inside the blade and circulating it inside the blade, and then transferring the heat to the outer surface of the blade. When heating, there will be airflow in the blade cavity at the tip of the blade, and the airflow will also pass through the air duct formed by the blade tip and the belly plate to form a gas circulation route. The air circulation heating in the cavity can avoid the problem of icing, ensure the heating treatment of the trailing edge and the root of the blade, and have a good anti-icing and deicing effect, which can significantly improve the heat conduction efficiency and reduce the heating power consumption. When the temperature of the outer surface of the blade rises, a water film will appear between the ice layer and the blade, which is to achieve deicing through the centrifugal force of the blade rotation. During deicing, the hot air output temperature is 60-80℃, and the blade surface temperature can be controlled at 10-0℃. Since the blade form will not change, the deicing system will have no effect on the aerodynamic performance and lightning protection performance. Heating deicing can be applied to mild climates and icing occurs in an environment around 0℃. Because deicing must ensure a reasonable blade temperature, the energy consumption is only 50% of anti-icing. After a comprehensive analysis of the reliability and applicability of wind turbine blade deicing technology at home and abroad, the following suggestions are made for wind turbine blade ice detection methods and deicing technology solutions: There are a large number of application cases abroad using vibration sensors installed on blades and analyzing their vibration data to evaluate the ice condition. It has been proven to be a reliable and effective detection method. Using a highly sensitive sensor system and a special algorithm, the measurement resolution of the ice thickness within the millimeter range can be achieved; by analyzing the operation data of wind farms during the icing period, the ice detection based on power curve comparison is usually used by major equipment manufacturers and is currently the most commonly used ice alarm shutdown method. Installing a hot air heating system inside the blade for deicing is a simple and reliable solution with high deicing efficiency, but the process of heating and deicing the blades requires a large amount of self-used electricity. There are three ways to improve the deicing efficiency of the hot air heating system and reduce the deicing energy consumption: heating and deicing after the unit blades are shut down due to ice, and not using the method of maintaining the high temperature of the blade body for anti-icing; using a high-precision ice detection system to control the precise start and stop of the deicing heating system; using a super-hydrophobic coating on the leading edge of the blade to prevent the melted ice from condensing repeatedly. It should be noted that the hot air heating deicing system has a power bottleneck when deicing blades over 60 meters.

[0080] See also Figure 1 This embodiment provides a method for predicting the ice coverage state of wind turbine blades in a wind farm, including:

[0081] S1, obtaining historical operation data and grid-connected power data of the wind farm, performing data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminating invalid data;

[0082] S2, calculating and determining the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrating the two sets of data by taking the intersection of time as the historical data for predicting the icing status of wind turbine blades in the wind farm;

[0083] S3, obtaining the historical records of ice coverage of wind turbine blades in the on-site wind farm;

[0084] S4, establishing an icing prediction model based on a reference wind turbine power curve and a historical record of icing conditions on wind turbine blades in the on-site wind farm;

[0085] S5, according to the icing prediction model, obtaining the time when icing occurs on the wind turbine blades of the wind farm, and recording the status; if icing occurs, it is marked as 1, otherwise it is marked as zero.

[0086] As a preferred implementation, the removing of invalid data includes:

[0087] (1) Eliminate data from the fan scheduled maintenance period;

[0088] (2) Eliminate shutdown data caused by various factors of the fan, including shutdown due to faults, manual shutdown, etc.;

[0089] (3) Eliminate data corruption caused by abnormal factors, including communication interruption, SCADA system failure, etc.;

[0090] (4) Eliminate the power reduction operation data caused by power rationing of the power grid.

[0091] As a preferred implementation, the first time period is 15 minutes. Of course, those skilled in the art can also select a suitable time period according to the specific conditions of the wind farm, which is within the protection scope of the present invention.

[0092] As a preferred implementation, the S4 establishes an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm, including:

[0093] S41, determining multiple index factors that affect blade icing;

[0094] S42, quantitatively calculating the values ​​of the correlation between the multiple index factors and the degree of icing on the wind turbine blades according to the grey correlation analysis, determining the importance of the factors affecting the icing on the wind turbine blades, arranging the multiple index factors affecting the icing on the wind turbine blades in descending order of influence through the grey correlation analysis, and selecting the values ​​of the correlation to establish the parameters used in the blade icing prediction model;

[0095] S43, collecting weather forecast data;

[0096] S44, determining key element data in the weather forecast data according to the value of the correlation degree;

[0097] S45, obtaining an icing state sequence according to a time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind measuring tower in a first time period and a time series of grid-connected power data;

[0098] S46, based on the key element data in the weather forecast data and the icing state sequence, a statistical method and a probability method are used to obtain critical conditions of key meteorological elements when the wind farm predicts icing.

[0099] As a preferred embodiment, the step 41 determines that the multiple indicator factors affecting blade icing are freezing meteorological conditions and meteorological elements related to freezing weather, the freezing meteorological conditions include rime, rime and mixed rime of rain and fog, and the meteorological elements related to freezing weather include air temperature, relative humidity and wind speed.

[0100] The severity of the icing situation in the wind farm will directly reflect the actual output of the wind turbines. Therefore, as a preferred embodiment, the method further includes:

[0101] Step 6, when predicting the occurrence of icing on the fan blades, the icing degree is specifically divided based on the actual output data of the fan blades in the generated time period, and the influence coefficient of different degrees of icing on the actual output of the fan blades is calculated.

[0102] This impact coefficient will be introduced into the wind power short-term prediction model later to improve the prediction accuracy and reduce the assessment.

[0103] As a preferred embodiment, the specific classification of the ice coverage degree includes predicting the degree of freezing and the power correction coefficient of the blade output corresponding to the freezing degree through fuzzy cluster analysis and particle swarm optimization methods.

[0104] Impact of blade icing on wind turbines: Blade icing will change the aerodynamic performance of wind turbine blades, resulting in reduced power generation. At the same time, uneven distribution of ice mass will cause unbalanced loads, affecting the safe and stable operation of wind turbines. When icing is severe, the wind turbine will automatically shut down. On-site maintenance personnel without automatic control will also manually shut down when the output is lower than about 50% of the power curve.

[0105] This embodiment also provides a wind farm wind turbine blade icing state prediction system, including:

[0106] A wind farm data acquisition and preprocessing module is used to acquire historical operation data and grid-connected power data of the wind farm, perform data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminate invalid data;

[0107] The icing state prediction data calculation module is used to calculate and determine the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrate the two sets of data according to the time intersection as the historical data for the prediction of icing state of wind turbine blades in the wind farm;

[0108] An ice coverage history acquisition module is used to acquire the history records of ice coverage of wind turbine blades in the on-site wind farm;

[0109] An icing prediction model establishment module is used to establish an icing prediction model based on a reference wind turbine power curve and in combination with the historical records of icing conditions on wind turbine blades in the on-site wind farm;

[0110] A blade icing state prediction module is used to obtain the time when the wind turbine blades in the wind farm are covered with ice according to the icing prediction model, and record the state; if icing occurs, it is marked as 1, otherwise it is marked as zero;

[0111] The output correction module is used to specifically classify the degree of icing when predicting the occurrence of icing on the fan blades, combined with the actual output data of the fan blades within the generated time period, and calculate the influence coefficient of different degrees of icing on the actual output of the fan blades.

[0112] This embodiment takes Hunan T wind farm, Hunan H wind farm, Yunnan M wind farm, Sichuan Z wind farm, Jiangxi Y wind farm as examples, and conducts a preliminary study on blade icing through on-site observation of meteorological data and wind turbine freezing and shutdown phenomena in the winter of 2016.

[0113] 1. Description of data required for testing

[0114] 1. Test site

[0115] In the southern region prone to freezing weather disasters, by analyzing the actual data provided on site and combining it with the freezing records provided on site, we finally selected Hunan T Wind Farm, Hunan H Wind Farm, Yunnan M Wind Farm, Sichuan Z Wind Farm, and Jiangxi Y Wind Farm for freezing test.

[0116] 2. Test data

[0117] For Hunan T wind farm, Hunan H wind farm, Sichuan Z wind farm and Jiangxi Y wind farm, the actual power data, short-term predicted power data, weather forecast information (temperature, humidity, etc.) of each meteorological source and actual freezing sample data analyzed in the early stage were selected from the time period of 00:00 on November 1, 2016 to 23:45 on February 28, 2017.

[0118] For the Yunnan M wind farm, the actual power data from 00:00 on November 1, 2016 to 23:45 on March 31, 2017, the short-term predicted power data, the weather forecast information (temperature, humidity, etc.) of each meteorological source, and the actual freezing sample data analyzed in the early stage were selected.

[0119] 3. Statistics on the accuracy of wind farm freezing prediction

[0120] (1) Make predictions based on weather forecast information

[0121] Through the analysis of the actual data in the early stage, it is concluded that the main factors affecting the freezing of the on-site fans are temperature and humidity, and when the temperature and humidity reach different thresholds, the electric field will freeze to varying degrees. Considering the deviation between the weather forecast information and the actual weather information, different temperature thresholds (-5-6 degrees, with a change step of 0.5 degrees) and different humidity thresholds (80-92, with a change step of 2) are set. It is assumed that when the predicted temperature at a certain moment is less than the temperature threshold and the predicted humidity is greater than the humidity threshold, it is predicted that freezing will occur at that moment, otherwise it is predicted that freezing will not occur at that moment. Finally, the overall prediction accuracy, the prediction accuracy under the actual freezing condition, and the prediction accuracy under the actual non-freezing condition of each meteorological source under different temperature thresholds and humidity thresholds are statistically calculated. Then, according to the temperature threshold and humidity threshold corresponding to the highest overall prediction accuracy, the prediction accuracy under the actual freezing condition and the prediction accuracy under the actual non-freezing condition under the corresponding conditions are found. The statistical results are shown in Table 1.

[0122] Table 1 Statistical results of point-by-point prediction based on weather forecast information

[0123]

[0124]

[0125] (2) Make predictions on a daily basis based on weather forecast information

[0126] Through the analysis of the actual data in the early stage, it is concluded that the main factors affecting the freezing of the on-site fans are temperature and humidity, and when the temperature and humidity reach different thresholds, the electric field will freeze to varying degrees. Considering the deviation between the weather forecast information and the actual weather information, different temperature thresholds (-5-6 degrees, the change step is 0.5 degrees) and different humidity thresholds (80-92, the change step is 2) are set. According to the weather forecast information, the forecast is made on a daily basis, that is, the average humidity value of the lowest temperature value in a day is obtained in units of days. Assuming that when the minimum temperature in a day is less than the temperature threshold and the average humidity value of a day is greater than the humidity threshold, it is predicted that all 96 points will freeze on that day, otherwise it is predicted that none of the 96 points will freeze on that day. Finally, the overall forecast accuracy, the forecast accuracy under actual freezing conditions, and the forecast accuracy under actual non-freezing conditions of each meteorological source under different temperature thresholds and humidity thresholds are calculated respectively. Then, according to the temperature threshold and humidity threshold corresponding to the highest overall prediction accuracy, the prediction accuracy under the corresponding conditions where freezing actually occurs and the prediction accuracy under the conditions where freezing does not actually occur are found. The statistical results are shown in Table 2.

[0127] Table 2 Statistical results of daily forecasts based on weather forecast information

[0128]

[0129]

[0130] (3) Comparison of forecast results by point and day based on weather forecast information

[0131] From Table 2, we can get:

[0132] Hunan T Wind Farm, when using Meteorological 2 for daily forecasts, had the highest overall forecast accuracy, forecast accuracy under actual freezing conditions, and forecast accuracy under actual non-freezing conditions, which were 88.33%, 65.52%, and 95.60%, respectively.

[0133] Hunan H Wind Farm, when using Meteorological 2 for daily forecasting, has the highest overall forecast accuracy and forecast accuracy under actual freezing conditions. The overall forecast accuracy, forecast accuracy under actual freezing conditions, and forecast accuracy under actual non-freezing conditions are 95.00%, 88.89%, and 95.50%, respectively.

[0134] For the Yunnan M wind farm, when using Meteorological 3 for daily forecasts, the overall forecast accuracy, the forecast accuracy under actual freezing conditions, and the forecast accuracy under actual non-freezing conditions were the highest, which were 96.69%, 16.67%, and 100%, respectively.

[0135] For Sichuan Z wind farm, when using weather forecast 1 for daily forecast, the forecast accuracy under actual freezing conditions is the highest. The overall forecast accuracy, the forecast accuracy under actual freezing conditions, and the forecast accuracy under actual non-freezing conditions are 95.00%, 80.00%, and 95.65%, respectively.

[0136] For Jiangxi Y Wind Farm, when using Meteorological 2 for daily forecasts, the overall forecast accuracy and the forecast accuracy under actual freezing conditions were the highest. The overall forecast accuracy, the forecast accuracy under actual freezing conditions, and the forecast accuracy under actual non-freezing conditions were 97.50%, 85.71%, and 98.23%, respectively.

[0137] Since the prediction accuracy rate of the actual freezing conditions of the Yunnan M wind farm is only 16.67%, that is, only one day was accurately predicted in the six-day freezing sample, in this case, the subsequent research on the freezing correction coefficient will be of little significance. In order to study the effect of the freezing test on the site, the research direction is adjusted according to the following ideas:

[0138] (1) First, the minimum limit of the overall prediction accuracy is set to 85%, and the prediction accuracy under actual freezing conditions corresponding to the overall prediction accuracy being no less than 85% is found;

[0139] (2) Secondly, according to the prediction accuracy under the actual freezing conditions selected in (1), find the maximum value, and then find the prediction accuracy under the actual non-freezing conditions under the corresponding conditions;

[0140] (3) Finally, based on the prediction accuracy under the actual non-freezing conditions selected in (2), find the temperature threshold and humidity threshold corresponding to the maximum value.

[0141] According to the above ideas, the final result is: the meteorological source is meteorological 3, the prediction method is daily prediction, the temperature threshold is 1, and the humidity threshold is 80. Under this condition, the overall prediction accuracy is 88.08%, the prediction accuracy under actual freezing conditions is 50.00%, and the prediction accuracy under actual non-freezing conditions is 89.66% (that is, when there is no freezing in reality, the probability of false alarm as freezing is 10.34%).

[0142] (4) Study on correction coefficients under different freezing degrees

[0143] In order to study the correction coefficient under different degrees of freezing conditions, different correction coefficient values ​​need to be set according to different temperature ranges and humidity ranges. When the lowest predicted temperature and the average humidity of the day are within the corresponding range, the corresponding coefficient values ​​are assigned to the 96 points of the day, and then the short-term predicted power is multiplied by the coefficient to obtain the revised predicted power value of the day. The percentage point of average improvement in prediction accuracy after revision can be calculated based on the actual power, predicted power and revised predicted power, and this is used as the optimization target of the particle swarm algorithm to obtain the corresponding revision coefficient value when the average improvement in prediction accuracy is the largest.

[0144] Since there is a certain error between the predicted temperature and humidity and the measured temperature and humidity, there is a possibility that freezing weather will be misjudged when freezing weather does not actually occur, which will affect the effect of power revision to a certain extent. In order to minimize this effect, the temperature and humidity ranges are set according to the following ideas:

[0145] (1) If freezing is predicted on that day, the lowest predicted temperature and average humidity of that day will be counted;

[0146] (2) Analyze the weather forecast information (the lowest predicted temperature in a day, the average humidity in a day) on a daily basis when freezing is predicted, and try to set the weather information of the false alarm day and the weather information of the accurate forecast day in different temperature and humidity ranges.

[0147] According to the above ideas and methods, the accuracy before and after revision is compared, and the comparison results are shown in Table 3. The accuracy results given in the table are the average accuracy of 4 months.

[0148] Table 3 Accuracy comparison results

[0149]

[0150] It can be seen from Table 3 that after correcting the short-term prediction power for predicting freezing days, the monthly average accuracy has been improved to a certain extent.

[0151] This embodiment also provides a memory storing a plurality of instructions, wherein the instructions are used to implement the method as described in the embodiment.

[0152] like Figure 2 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301, the memory 302 stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute the method as in the embodiment.

[0153] This embodiment:

[0154] 1) According to weather forecast data, before conducting icing prediction analysis on wind farms, it is necessary to preprocess the historical operation data of wind farms and eliminate invalid data to make the prediction more accurate.

[0155] 2) Grey correlation analysis can quantitatively calculate the influence of various factors on the icing degree of wind turbine blades, determine the importance of factors affecting blade icing, and lay the foundation for establishing an icing prediction model.

[0156] 3) According to the results of grey correlation analysis, the key element data in the numerical weather forecast data are sorted out, and combined with the icing state sequence data, statistical methods and probability methods are used to obtain the critical conditions of the key meteorological elements when predicting icing in wind farms.

[0157] 4) When predicting the occurrence of icing, it is necessary to combine the actual output data of the corresponding time period, give a specific classification of the degree of icing, and give the influence coefficient of different degrees of icing on the actual output. Then, the icing influence coefficient is introduced into the short-term wind power prediction model. Through fuzzy clustering analysis, particle swarm optimization and other methods, the degree of freezing and the power correction coefficient corresponding to the freezing degree are predicted, which improves the prediction accuracy and reduces the assessment.

[0158] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting the ice coverage status of wind turbine blades in a wind farm, characterized in that include: Step 1, obtaining historical operation data and grid-connected power data of the wind farm, performing data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminating invalid data; Step 2: Calculate and determine the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrate the two sets of data according to the time intersection as the historical data for predicting the icing status of wind turbine blades in the wind farm; Step 3, obtaining the historical records of ice coverage of wind turbine blades in the on-site wind farm; Step 4, establishing an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm; The step 4 establishes an icing prediction model based on the reference wind turbine power curve and the historical records of icing conditions of wind turbine blades in the on-site wind farm, including: Step 41, determining multiple index factors that affect blade icing; Step 42, according to the grey correlation analysis, quantitatively calculate the values ​​of the correlation between the multiple index factors and the degree of icing on the wind turbine blades, determine the importance of the factors affecting the icing on the blades, and arrange the multiple index factors affecting the icing on the wind turbine blades in descending order of influence through the grey correlation analysis, and select the values ​​of the correlation to establish the parameters used in the blade icing prediction model; Step 43, collecting weather forecast data; Step 44, determining key element data in the weather forecast data according to the value of the correlation degree; Step 45, obtaining an icing state sequence according to the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind measuring tower in the first time period and the time series of grid-connected power data; Step 46, based on the key element data in the weather forecast data and the icing state sequence, using statistical methods and probability methods to obtain critical conditions of key meteorological elements when the wind farm predicts icing; Step 5, according to the icing prediction model, obtain the time when the wind turbine blades in the wind farm are covered with ice, and record the status; if icing occurs, it is marked as 1, otherwise it is marked as zero; Step 6, when predicting the occurrence of icing on the fan blades, the degree of icing is specifically divided according to the actual output data of the fan blades within the time period of the occurrence, and the influence coefficient of different degrees of icing on the actual output of the fan blades is calculated; the specific division of the degree of icing includes predicting the degree of freezing and the power correction coefficient of the blade output under the corresponding degree of freezing through fuzzy clustering analysis and particle swarm optimization methods.

2. A method for predicting the icing status of wind turbine blades in a wind farm according to claim 1, characterized in that The elimination of invalid data includes: (1) Eliminate data from the fan scheduled maintenance period; (2) Eliminate shutdown data caused by various factors of the fan, including shutdown due to faults and manual shutdown; (3) Eliminate data corruption caused by abnormal factors, including communication interruption and SCADA system failure; (4) Eliminate the power reduction operation data caused by power rationing of the power grid.

3. A method for predicting the icing status of wind turbine blades in a wind farm according to claim 1, characterized in that The first time period is 15 minutes.

4. A method for predicting the icing status of wind turbine blades in a wind farm according to claim 3, characterized in that The step 41 determines that the multiple indicator factors that affect blade icing are freezing meteorological conditions and meteorological elements related to freezing weather. The freezing meteorological conditions include rime, rime and mixed rime of rain and fog. The meteorological elements related to freezing weather include air temperature, relative humidity and wind speed.

5. A wind farm wind turbine blade icing state prediction system implementing the method as claimed in claim 1, characterized in that include: A wind farm data acquisition and preprocessing module is used to acquire historical operation data and grid-connected power data of the wind farm, perform data preprocessing on the historical operation data of the wind farm and the grid-connected power data, and eliminate invalid data; The icing state prediction data calculation module is used to calculate and determine the time series of wind speed, wind direction, temperature, humidity, and pressure data of the wind tower in the first time period, and the time series of grid-connected power data, and integrate the two sets of data according to the time intersection as the historical data for the prediction of icing state of wind turbine blades in the wind farm; An ice coverage history acquisition module is used to acquire the history records of ice coverage of wind turbine blades in the on-site wind farm; An icing prediction model establishment module is used to establish an icing prediction model based on a reference wind turbine power curve and in combination with the historical records of icing conditions on wind turbine blades in the on-site wind farm; A blade icing state prediction module, used to obtain the time when the wind turbine blades in the wind farm are iced according to the icing prediction model, and record the state; If icing occurs, it is marked as 1, otherwise it is marked as zero; The output correction module is used to specifically classify the degree of icing when predicting the occurrence of icing on the fan blades, combined with the actual output data of the fan blades within the generated time period, and calculate the influence coefficient of different degrees of icing on the actual output of the fan blades.

6. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is used to read the instructions and execute the method according to claim 1.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, which can be read by a processor to execute the method as claimed in claim 1 .

Images