Fan blade icing monitoring method based on mechanism analysis

Abstract

The invention discloses a fan blade icing monitoring method based on mechanism analysis, and the method comprises the steps: 1, constructing a dynamic mechanism model of "pneumatic-thermal-quality" three-phase strong coupling, local micrometeorological conditions and water drop impact characteristics of each airfoil section and feedback effects of icing growth on aerodynamic configuration and subsequent icing in the blade rotation process are calculated in real time; 2, performing icing prediction based on the constructed dynamic mechanism model, wherein the icing prediction comprises local microenvironment calculation, dynamic icing growth calculation and pneumatic feedback calculation; step 3, judging an icing state; step 4, outputting an icing state and early warning information; the problems that a data-based driving method needs to depend on a large amount of high-quality annotation data, when the data size is insufficient or data distribution changes, the generalization ability and prediction precision of a prediction model can be greatly reduced, the internal mechanism of icing cannot be explained, and the reliability requirement in engineering application is difficult to meet are solved.

Description

Technical Field

[0001] This invention belongs to the field of wind power equipment condition monitoring technology; in particular, it relates to a method for monitoring wind turbine blade icing based on mechanism analysis. Background Technology

[0002] In cold weather and low winter temperatures, wind turbine blades are prone to icing. Icing on wind turbine blades alters their aerodynamic shape and increases their weight, significantly reducing power generation efficiency and causing increased blade vibration and fatigue damage. In severe cases, it can even lead to blade breakage or the collapse of the entire turbine, resulting in safety accidents.

[0003] Current technologies for monitoring or predicting icing on wind turbine blades mainly include:

[0004] Methods based on direct sensor measurement, such as installing strain gauges, accelerometers, and temperature sensors on the surface of wind turbine blades, can determine whether the blades are icing by monitoring the mechanical properties or environmental parameters of the blades. However, this method suffers from problems such as difficult sensor installation, susceptibility to failure due to harsh environments, high maintenance costs, and inability to accurately predict icing.

[0005] Data-driven approaches collect wind turbine operating data and environmental data, and then use machine learning algorithms to build icing monitoring or prediction models. However, data-driven approaches rely on a large amount of high-quality labeled data. When the amount of data is insufficient or the data distribution changes, the generalization ability and prediction accuracy of the prediction model will drop significantly, and it will be unable to explain the underlying mechanism of icing, making it difficult to meet the reliability requirements of engineering applications. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a mechanism-based method for monitoring icing on wind turbine blades, so as to solve the long-standing technical problem of "how to accurately predict non-uniform icing on dynamically rotating wind turbine blades and its performance impact".

[0007] The technical solution of this invention is:

[0008] A method for monitoring icing on wind turbine blades based on mechanism analysis, the method comprising:

[0009] Step 1: Construct a dynamic mechanism model with strong three-phase coupling of "aerodynamic-thermal-mass" to calculate in real time the local micro-meteorological conditions and water droplet impact characteristics of each airfoil section during blade rotation, as well as the feedback effect of icing growth on aerodynamic shape and subsequent icing.

[0010] Step 2: Perform icing prediction based on the constructed dynamic mechanism model, including local microenvironment calculation, dynamic icing growth calculation, and aerodynamic feedback calculation;

[0011] Step 3: Determine the icing status;

[0012] Step 4: Output icing status and early warning information.

[0013] The dynamic mechanism model includes:

[0014] Step 1.1 Calculation of unsteady airflow field and droplet trajectory in three-dimensional rotating coordinate system: In the three-dimensional unsteady computational fluid dynamics (CFD) solver, a rotating coordinate system and dynamic mesh technology are introduced to simulate the flow field of the blade under real wind shear, yaw error and turbulence conditions.

[0015] Step 1.2: Construct a dynamic phase transition and icing growth model based on leaf element thermodynamic equilibrium;

[0016] Step 1.3, Aerodynamic shape evolution and performance feedback: Couple the icing growth model with the aerodynamic performance lookup table or the reduced-order aerodynamic model in real time.

[0017] In three-dimensional unsteady computational fluid dynamics (CFD) solvers, methods for simulating the flow field of blades under realistic wind shear, yaw error, and turbulent conditions include:

[0018] Calculation of local effective wind speed and angle of attack:

[0019] Local effective wind speed vector:

[0020] ;

[0021] r is the radius of the cross section from the leaf root. This represents the incoming wind speed vector. This represents the tangential velocity vector generated by the rotation of the blade, where It is the impeller rotational angular velocity vector. It is a vector. It is the azimuth angle; This represents the turbulent wind speed vector generated by the turbulence model;

[0022] Calculation of effective angle of attack in local areas:

[0023] ;

[0024] in,

[0025] , is the angle of entry;

[0026] : Distribution of blade twist angle;

[0027] Dynamic pitch angle;

[0028] The calculation of the local water droplet collision coefficient includes:

[0029] Stokes number:

[0030] ;

[0031] Let this be the relaxation time of the water droplet; The density of the water droplets; The median volume diameter is the water droplet size distribution. C(r) represents the aerodynamic viscosity; C(r) represents the chord length at the section at a distance r from the blade root radius.

[0032] Local collision coefficients are based on CFD simulations or empirical correlations:

[0033] ;

[0034] For the corresponding airfoil section and current angle of attack The two-dimensional collision efficiency of the Stokes number St; This is a three-dimensional effect correction factor; K is the rotation enhancement coefficient, and k2 and k1 are empirical coefficients obtained by fitting CFD data.

[0035] The construction of the dynamic phase transition and icing growth model based on leaf element thermodynamic equilibrium includes:

[0036] (1) Leaf element heat balance equation:

[0037] ;

[0038] 1) Input heat include:

[0039] Heating is achieved through the kinetic energy generated by the impact of supercooled water droplets; The latent heat released when water droplets freeze. Latent heat of solidification; Temperature of water droplets Cool to component surface temperature The released sensible heat;

[0040] Heat provided for the anti-icing / de-icing system;

[0041] Liquid water content;

[0042] The impact efficiency coefficient is a dimensionless coefficient with a value ranging from 0 to 1.

[0043] Effective impact speed;

[0044] This refers to a time interval, the period during which heat accumulates or the system operates.

[0045] The latent heat of solidification of water;

[0046] This is the specific heat capacity of water;

[0047] The temperature of the water droplet;

[0048] This refers to the surface temperature of the blade.

[0049] This refers to the heater power.

[0050] Heat output include:

[0051] For convection cooling;

[0052] For heat dissipation during evaporation / sublimation;

[0053] For radiative heat dissipation;

[0054] (2) Dynamic phase change fraction and water film migration:

[0055] Dynamic phase transition fraction: ,

[0056] ;

[0057] Water film mass balance: ;

[0058] The rate of change of water film mass over time;

[0059] The impact water mass flow rate;

[0060] This refers to the mass flow rate of the frozen water, also known as the freezing volume.

[0061] Water loss is caused by airflow shear force;

[0062] This is due to water loss caused by centrifugal force.

[0063] Aerodynamic shape evolution and performance feedback include:

[0064] (1) Real-time aerodynamic coefficient calculate:

[0065] Calculate the ice shape at the current moment and dynamically update the lift coefficient of each leaf element. drag coefficient and torque coefficient The airfoil coordinates are modified in real time by parametrically describing the ice shape; then, the airfoil aerodynamic data after ice shape perturbation is used to obtain values ​​from a lookup table or a reduced-order model. ;

[0066] Torque loss and unbalanced load calculation:

[0067] 1) Blade element torque contribution:

[0068] ;

[0069] 2) Total torque:

[0070] , Number of leaves;

[0071] 3) Torque loss factor:

[0072] ;

[0073] 4) Unbalanced load:

[0074] flapping torque of a single blade:

[0075] ;

[0076] , Let r be the lift and drag at the distance from the blade root radius (r) to the i-th cross section, respectively.

[0077] Total unbalanced swing torque vector:

[0078] ,in Let be the unit vector of the i-th blade in the waving direction.

[0079] The icing status assessment includes:

[0080] Step 3.1: Set the judgment threshold;

[0081] Step 3.2: Comprehensive logical judgment, the judgment logic includes:

[0082] (1) Judgment by direct physical quantity:

[0083] Condition A: Does the ice thickness calculated by the model exceed a preset threshold at any leaf element location?

[0084] Condition B: Does the total icing mass calculated by the model, i.e., the water film mass, exceed the preset threshold?

[0085] (2) Judgment of performance derivatives:

[0086] Condition C: Whether the torque loss factor predicted by the model is lower than the preset threshold;

[0087] Condition D: Does the unbalanced load predicted by the model exceed the safety threshold of the wind turbine structure load?

[0088] (3) Logical fusion decision:

[0089] Determining "whether it is covered with ice": If condition A is true in any position, it is determined that "the blade is covered with ice";

[0090] Assess the degree of icing:

[0091] Slight icing: Condition A meets the first-level threshold, and conditions C and D do not trigger the second-level threshold;

[0092] Moderate icing: Condition A meets the secondary threshold, or condition C triggers the secondary threshold;

[0093] Severe icing: Condition A meets the level 3 threshold, or condition C triggers the level 3 threshold, or condition D triggers the danger threshold.

[0094] The beneficial effects of this invention are:

[0095] This invention, by introducing a dynamic coupling mechanism model of aerodynamics, heat, and mass, not only solves the qualitative problem of "whether icing occurs," but also more precisely addresses the quantitative problems such as "how thick the ice is, where it forms, and the extent of its impact on the wind turbine," providing effective decision support for the safe and efficient operation of wind turbines. Its technical advantages include:

[0096] (1) This invention can quantitatively predict the non-uniform distribution of icing on the three-dimensional surface of a blade, including leading edge icing, severe icing at the blade tip, and differential icing between the suction and pressure surfaces. It overcomes the shortcomings of traditional methods that can only determine "whether there is icing" but cannot tell "where the icing is located and what shape it is".

[0097] (2) This invention is a physics-based prediction model, rather than a post-event monitoring model. It can predict the risk, type and growth rate of icing based on real-time meteorological conditions tens of minutes to several hours in advance, before icing is detected by the naked eye or a camera, leaving valuable response time for operation and maintenance decisions.

[0098] (3) By dynamically calculating the phase change fraction and water film migration, the model of this invention can effectively distinguish different types such as light ice, mixed ice and frost ice, thus providing a key basis for assessing the hazards of icing and formulating de-icing strategies.

[0099] (4) At the same time, this invention uniquely outputs unbalanced load indicators, which can directly quantify the fatigue load and ultimate load of the wind turbine structure caused by asymmetric icing. It provides a direct and quantitative scientific basis for preventing catastrophic accidents such as blade breakage and collapse caused by icing, and realizes the leap from "phenomenon alarm" to "structural safety early warning".

[0100] (5) When the model of this invention is coupled with the thermal de-icing system, it can form an intelligent closed-loop control. The model can not only trigger de-icing, but also optimize the de-icing strategy (such as zoned heating and intermittent de-icing) through real-time simulation. While ensuring the de-icing effect, it can minimize energy consumption and avoid secondary load problems caused by uneven de-icing.

[0101] (6) The model described in this invention directly outputs the torque loss factor, which can assess the power generation loss caused by icing in real time and accurately, and helps wind farm operators to accurately grasp the economic losses caused by icing.

[0102] (7) The model of this invention is based on first principles (mass, momentum and energy conservation). It has good adaptability and predictive ability for different climate zones, different terrains, different blade airfoil designs and different wind turbine operating states, and overcomes the shortcomings of pure data-driven models in scenarios with scarce data.

[0103] This addresses the problems that data-driven methods rely on large amounts of high-quality labeled data, and that when the amount of data is insufficient or the data distribution changes, the generalization ability and prediction accuracy of the prediction model will drop significantly, and the underlying mechanism of icing cannot be explained, making it difficult to meet the reliability requirements in engineering applications. Detailed Implementation

[0104] A method for monitoring icing on wind turbine blades based on mechanism analysis, the method comprising:

[0105] Step 1: Construct a dynamic mechanism model with strong coupling of the three phases of "aerodynamics-thermal-mass".

[0106] This invention abandons the traditional mechanism model that treats wind turbine blades as static conductors and establishes a dynamic mechanism model with strong coupling of three phases: aerodynamics, heat, and mass. The core of this model lies in the real-time calculation of the local micro-meteorological conditions and water droplet impact characteristics of each airfoil section during blade rotation, and considers the feedback effect of icing growth on aerodynamic shape and subsequent icing.

[0107] The construction includes:

[0108] Step 1.1: Calculation of unsteady airflow field and water droplet trajectory in three-dimensional rotating coordinate system.

[0109] This step is fundamental to accurately predicting icing distribution and addresses the unsteady effects caused by rotation and dynamic pitch control. In the three-dimensional unsteady computational fluid dynamics (CFD) solver, a rotating coordinate system and dynamic mesh technology are innovatively introduced to simulate the flow field of the blade under realistic wind shear, yaw error, and turbulent conditions.

[0110] (1) Calculation of local effective wind speed and angle of attack:

[0111] For a section at a distance *r* from the blade root radius, the effective wind speed is a vector synthesis of the incoming wind speed, the blade element rotational tangential velocity, and the turbulent fluctuating wind speed. The local instantaneous angle of attack is calculated based on this. It is dynamic and changes, which directly determines the impact characteristics of water droplets.

[0112] 1) Local effective wind speed vector:

[0113]

[0114] in,

[0115] This represents the incoming wind speed vector (considering wind shear, i.e.) ,in (Wind shear index); This represents the tangential velocity vector generated by the rotation of the blade, where It is the impeller rotational angular velocity vector. It is a vector. It is the azimuth angle; This represents the turbulent wind speed vector generated by a turbulence model (such as the Kaimal or Mann model).

[0116] 2) Local effective angle of attack:

[0117]

[0118] in,

[0119] : is the angle of entry.

[0120] : Distribution of blade twist angle.

[0121] : Dynamic pitch angle.

[0122] (2) Calculation of local water droplet collision coefficient:

[0123] 1) Stokes number:

[0124]

[0125] : Water droplet relaxation time.

[0126] The density of the water droplet.

[0127] : The median volume diameter of the water droplet size distribution.

[0128] Aerodynamic viscosity.

[0129] C(r): The chord length at the point where the cross section is at a distance r from the leaf root radius.

[0130] 2) Local collision coefficient: based on CFD simulation or empirical correlation, for example:

[0131]

[0132] : Corresponding airfoil section and current angle of attack The two-dimensional collision efficiency of the Stokes number St can be obtained through the Finnegan formula or by looking up a table.

[0133] : Three-dimensional effect correction factor, taking into account spanwise flow and wing root / wingtip effects.

[0134] The rotation enhancement coefficient is used to quantify the effect of centrifugal force on local water droplet concentration and impact efficiency. Rotation causes the local water droplet flux on the outside of the blade to be higher than that on the inside. Here, k2 and k1 are empirical coefficients obtained by fitting CFD data.

[0135] Step 1.2: Construct a dynamic phase transition and icing growth model based on leaf element thermodynamic equilibrium:

[0136] The effects of centrifugal force from blade rotation and Coriolis force on water film movement, as well as the coupling of an internal active heating / de-icing system (if present), were considered. A novel transient thermal balance equation for each "blade element" was established, and a centrifugally driven water film migration model was introduced.

[0137] (1) Leaf element heat balance equation:

[0138]

[0139] 1) Input heat :

[0140] : Kinetic energy heating from the impact of supercooled water droplets, characterizing the heat converted from kinetic energy when supercooled water droplets impact high-speed transmission line conductors / tower base components, unit: joule (J) or kilojoule (kJ).

[0141] The latent heat released when a water droplet freezes ( (Latent heat of solidification). The total amount of latent heat released into the environment when supercooled water droplets freeze on the surface of a component, measured in joules (J) or kilojoules (kJ).

[0142] Water droplet temperature Cool to component surface temperature Released sensible heat ( (Specific heat capacity of water). The total amount of sensible heat released by a water droplet as it cools from its own temperature to the surface temperature of a component, measured in joules (J) or kilojoules (kJ).

[0143] Heat provided by the anti-icing / de-icing system. The total heat provided by an active anti-icing system (such as electric heating or ultrasonic de-icing) within a specific time period, measured in joules (J) or kilojoules (kJ).

[0144] Liquid water content: the mass of supercooled water droplets contained in a unit volume of air, reflecting the density of water droplets in clouds and fog, unit: grams per cubic meter (g / m³).

[0145] The impact efficiency coefficient is a dimensionless coefficient that represents the proportion of the actual mass of water droplets impacting the surface of the component to the total mass of incident water droplets. Its value ranges from 0 to 1.

[0146] Effective impact velocity is the actual impact velocity of the supercooled water droplet relative to the surface of the transmission line component, expressed in meters per second (m / s).

[0147] The time interval is the period of heat accumulation or system operation, measured in seconds (s).

[0148] Latent heat of freezing of water is the heat released when a unit mass of water freezes into ice at 0°C. It is a constant and is taken as 334 J / g under standard atmospheric pressure.

[0149] Specific heat capacity of water: the amount of heat absorbed (or released) by a unit mass of water when its temperature rises (or falls) by 1°C. It is a constant and is taken as 4.186 J / (g·°C) at room temperature.

[0150] Water droplet temperature: the initial temperature of a supercooled water droplet, usually below 0°C (unfrozen state), unit: degrees Celsius (°C).

[0151] : Blade surface temperature, usually below 0℃ (meeting the conditions for icing formation), unit: degrees Celsius (℃).

[0152] Heater power (spatial-temporal distribution function), the instantaneous power of the anti-icing and de-icing system, is a function of the cross-sectional distance from the blade root radius r and time t, and its unit is watts (W).

[0153] The heater power (spatial-temporal distribution function) is the instantaneous power of the anti-icing and de-icing system, which is a function of the cross-sectional distance from the blade root radius r and time t, and is measured in watts (W).

[0154] 2) Output heat :

[0155] Convection heat dissipation. Convection heat transfer coefficient. A method based on local Reynolds number can be adopted. Empirical formulas, such as: Nusel number ( (For coefficients).

[0156] : Evaporation / sublimation for heat dissipation. Let (Le) be the latent heat of vaporization, and (Le) be the Lewis number. RH is the saturated vapor pressure, p is the relative humidity, and RH is the atmospheric pressure.

[0157] : Radiative heat dissipation. For emission rate, is the Stefan-Boltzmann constant.

[0158] (2) Dynamic phase change fraction and water film migration:

[0159] Dynamic phase transition fraction (freeze rate): The value range is [0, 1]. The traditional "freezing fraction" is a static parameter; this invention dynamizes it, where n(t) depends on the instantaneous thermal equilibrium state of the current microelement, allowing for partial freezing and partial loss / scattering of the water film, thus more accurately simulating the formation of light ice, mixed ice, and frost ice. .

[0160] Water film mass balance:

[0161] : The rate of change of water film mass over time (i.e., the increase or decrease in water film mass per unit time).

[0162] Impact water mass flow rate.

[0163] Freezing water mass flow rate, also known as freezing volume.

[0164] Water loss due to airflow shear force, empirical formula: ,in For wall shear stress, The thickness of the water film.

[0165] Water loss due to centrifugal force. Centrifugal drive speed ,in The centrifugal migration coefficient describes the tendency of liquid water to migrate along the blade surface (especially towards the blade tip) under centrifugal force, and is related to surface tension, contact angle, and water film thickness. This parameter explains why more severe "ice nodules" or frost often appear in the blade tip region. (Simplified model) (This refers to the dynamic viscosity of water).

[0166] Step 1.3, Aerodynamic Shape Evolution and Performance Feedback:

[0167] Achieving a closed-loop prediction mechanism is crucial for forecasting performance degradation and power loss. This involves coupling the icing growth model with an aerodynamic performance lookup table or a reduced-order aerodynamic model in real time.

[0168] (1) Real-time aerodynamic coefficient ( , , )calculate:

[0169] Based on the ice shape calculated in step 1.2, the lift coefficient of each leaf element is dynamically updated. drag coefficient and torque coefficient Specifically, ice shape is described by parameters (such as angular ice thickness (b), ice nodule length). (etc.), modifying the airfoil coordinates in real time. Then, use a lookup table or reduced-order model (e.g., based on angle of attack) to look up the airfoil aerodynamic data after icing disturbance. Quickly obtain Reynolds number (Re) and ice shape parameters. .

[0170] (2) Torque loss and unbalanced load:

[0171] 1) Blade element torque contribution: ;

[0172] 2) Total torque: , This refers to the number of leaves.

[0173] 3) Torque loss factor:

[0174] 4) Unbalanced load (taking swinging moment as an example):

[0175] flapping torque of a single blade: ; and Let r be the lift and drag at the distance r from the blade root of the i-th cross section, respectively.

[0176] Total unbalanced swinging moment vector (in a fixed coordinate system):

[0177] ,in Let be the unit vector of the i-th blade in the waving direction.

[0178] Blade element torque contribution (torque micro-element): The torque generated per unit length micro-element at time t, at a distance r from the blade root cross section. It is the basis for calculating the total torque, and the unit is Newton-meter (Nm).

[0179] Total aerodynamic torque: The total torque generated by the wind turbine at time t after the blade surface is covered with ice (usually less than in the clean state), unit: Newton-meter (Nm).

[0180] Total torque under icing conditions.

[0181] Total torque in clean condition.

[0182] Torque loss factor. A dimensionless coefficient that characterizes the degree of torque attenuation caused by icing on the blade, with a value ranging from 0 to 1.

[0183] : The flapping torque of the i-th blade. The torque generated by the i-th blade about the flapping axis at time t, in Newton-meters (Nm).

[0184] Total unbalanced flapping moment vector. The resultant vector of all blade flapping moments in a fixed coordinate system, unit: Newton-meter (Nm).

[0185] Air density: The mass density of air flowing through the blades, taken as 1.225 kg / m³ at normal temperature and pressure.

[0186] : Effective wind speed at a distance r from the blade root, the actual incoming wind speed at a distance r from the blade root at time t (considering the effects of wind shear, turbulence, etc.), unit: meters per second (m / s).

[0187] Airfoil chord length at a distance r from the blade root: the airfoil chord length (characteristic length) of the blade at a distance r from the blade root. Unit: meter (m).

[0188] : Lift coefficient at a distance r from the blade root, a dimensionless coefficient that characterizes the ability of the blade element to generate lift, and is related to the angle of attack α and the Reynolds number Re.

[0189] : The drag coefficient at the distance r from the blade root, a dimensionless coefficient that characterizes the magnitude of airflow drag on the blade element and is related to the angle of attack α and the Reynolds number Re.

[0190] : Lift force at the i-th cross section at a distance r from the blade root, the force generated by the blade element in the airflow perpendicular to the direction of the incoming flow, unit: Newton (N).

[0191] : The resistance at the i-th cross section at a distance r from the blade root, the force acting on the blade element in the airflow parallel to the direction of the incoming flow, in Newtons (N).

[0192] Radial infinitesimal length: the infinitesimal length unit of the radial direction of the blade, in meters (m).

[0193] : Radius of the blade root, the radial distance between the blade and the hub, in meters (m).

[0194] The radius of the blade tip is the radial distance from the outermost end of the blade (i.e., the rotor radius), in meters (m).

[0195] The number of blades is the total number of blades in the wind turbine, dimensionless (common values ​​are 2 or 3).

[0196] The angle of attack of the airflow is related to the angle between the blade element chord and the direction of the incoming flow (including the combined effects of the installation angle, angle of attack, etc.), and the unit is degrees (°).

[0197] Let be the azimuth angle of the i-th blade, and be the rotation angle of the i-th blade relative to the fixed coordinate system, in degrees (°).

[0198] : The unit vector of the waving direction of the i-th blade, a dimensionless vector, whose direction is along the normal direction of the waving motion of the blade.

[0199] Leaf serial number, dimensionless, used to distinguish different leaves (values ​​1, 2, ..., N) _b ).

[0200] Step 2: Predict icing based on the dynamic mechanism model constructed in Step 1: This specifically includes:

[0201] Step 2.1: Data Input and Initialization

[0202] Step 2.1.1: Real-time data acquisition

[0203] The implementation of this invention relies on the acquisition and fusion of multi-source data on meteorological environmental parameters, wind turbine operating status parameters, and blade icing status parameters. Key measurement parameters include, but are not limited to: liquid water content (LWC), median droplet volume diameter (MVD), effective local wind speed and angle of attack on the blades, blade root load, and icing thickness and distribution measured directly by vision or sensors. These data collectively form the basis for model input, calibration, and validation.

[0204] Temperature, humidity, wind speed and precipitation sensors are installed on the top of the wind turbine tower. Blade surface temperature sensors are installed on the blade surface and speed sensors are installed on the main shaft of the wind turbine. Air temperature, relative humidity, wind speed and precipitation intensity are collected in real time (the water droplet diameter D, blade surface temperature and blade speed n are obtained indirectly through the precipitation sensor (the blade speed n is used to calculate the blade surface linear velocity, which in turn affects the water droplet impact angle α).

[0205] The collected data includes:

[0206] Meteorology and Environment: wind speed, wind direction, temperature, air pressure, humidity, liquid water content (LWC), and median volume diameter (MVD) of water droplets.

[0207] Wind turbine status and operation: impeller speed, yaw angle, pitch angle, output power, blade root bending moment, and power of the thermal de-icing system.

[0208] Blade condition and verification: video information, direct ice thickness measurement, power curve deviation, vibration spectrum, etc.

[0209] Step 2.1.2: Model Initialization

[0210] (1) Load the geometric parameters of a specific wind turbine blade (torsion distribution, chord length distribution, airfoil aerodynamic data table, etc.).

[0211] (2) Set the initial state: assume that the blade surface is clean and the water film thickness is 0.

[0212] Step 2.2: Calculation of the dynamic mechanism model (executed cyclically)

[0213] (Execute in a loop within each time step (e.g., 1 second)

[0214] Step 2.2.1, Local Microenvironment Calculation

[0215] (1) Calculate the local effective wind speed and angle of attack:

[0216] Calculate the local effective wind speed at the blade root radius based on real-time data. and instantaneous angle of attack .

[0217] (2) Calculate the local collision coefficient:

[0218] based on , Based on the measured MVD, the Stokes number St(r,t) is calculated. This is then combined with the rotational enhancement factor. The local water droplet collision coefficient of each leaf element was calculated. .

[0219] Step 2.2.2: Dynamic icing growth calculation

[0220] (1) Solve the transient heat balance equation:

[0221] For each blade element, a heat balance equation is drawn up, including all terms such as kinetic heating, latent heat of phase change, convective / evaporative heat dissipation, and anti-icing heating. Solving this equation yields the dynamic phase change fraction n(t) and blade surface temperature within that time step. .

[0222] (2) Renew water film and ice formation:

[0223] Perform water film mass balance calculation: New water film mass = Original mass + Impact amount - Freezing amount - Shear loss amount - Centrifugal loss amount.

[0224] Based on the amount of freezing Calculate the ice thickness growth within this time step. This involves updating the three-dimensional geometry of the blade and adding the increased ice thickness to the digital blade model.

[0225] Step 2.2.3, Pneumatic Feedback Calculation

[0226] (1) Update aerodynamic coefficients:

[0227] Based on the new ice shape, query or calculate the updated lift coefficient for each leaf element. and drag coefficient .

[0228] (2) Calculate system-level performance indicators:

[0229] 1) Torque loss factor Integrate the torque contribution of all leaf elements and compare it with the torque under clean conditions to calculate the current percentage of total torque loss.

[0230] 2) Unbalanced load Based on the asymmetrical lift and drag of the three blades, the unbalanced flapping moment and oscillation moment acting on the entire wind turbine are calculated.

[0231] Step 3, Icing status assessment:

[0232] Step 3.1: Set the judgment threshold

[0233] The system has preset multiple threshold levels (which can be adjusted according to the wind farm strategy):

[0234] Level 1 Threshold (Warning Level): Slight icing, with acceptable performance loss.

[0235] Level 2 threshold (alarm level): Moderate icing, significant performance loss, and increased load.

[0236] Level 3 Threshold (Hazard Level): Severe icing, involving structural safety or complete stall.

[0237] Step 3.2: Comprehensive logical judgment

[0238] The model does not rely solely on a single ice thickness data point, but rather employs a comprehensive judgment based on multiple indicators and cross-validation, as follows:

[0239] (1) Judgment by direct physical quantity:

[0240] Condition A: Ice thickness calculated by the model at any leaf element location. Has the preset threshold been exceeded?

[0241] Condition B: Does the total icing mass (i.e., water film mass) calculated by the model exceed the preset threshold?

[0242] (2) Judgment of performance derivatives:

[0243] Condition C: Torque loss factor predicted by the model Is it below a preset threshold (e.g., below 0.85, i.e., power loss exceeds 15%)?

[0244] Condition D: Unbalanced loads predicted by the model Does it exceed the safety threshold for wind turbine structural load?

[0245] (3) Logical fusion decision:

[0246] Determining "whether it is covered with ice": If condition A is true in any position, it is determined that "the blade is covered with ice".

[0247] Assess the degree of icing:

[0248] Slight icing: Condition A meets the first-level threshold, and conditions C and D do not trigger the second-level threshold.

[0249] Moderate icing: Condition A meets the secondary threshold, or condition C triggers the secondary threshold.

[0250] Severe icing: Condition A meets the level 3 threshold, or condition C triggers the level 3 threshold (e.g., power loss > 30%), or condition D triggers the danger threshold.

[0251] Active de-icing system intervention judgment: If the system is equipped with thermal de-icing, when it is determined to be "moderate icing", the de-icing system is triggered to start and calculate in real time. This feedback is then fed back into the thermal balance, forming a closed-loop control until the model calculations show that the ice layer has completely melted (ice thickness returns to zero).

[0252] Step 4: Output icing status and early warning information.

[0253] Step 4.1: Output Results

[0254] (1) Icing status: clean / slight / moderate / severe.

[0255] (2) Ice distribution map: Visual display of three-dimensional ice shape.

[0256] (3) Performance impact: current power loss percentage, unbalanced load size.

[0257] (4) Warning information: Based on the judgment results, issue warnings of the corresponding level.

[0258] Step 4.2: Trigger the action

[0259] Warning level: Highlighted on the monitoring interface to notify maintenance personnel to pay attention.

[0260] Alert Level: It is recommended to shorten the inspection cycle and prepare a de-icing plan.

[0261] Hazard level: Automatically performs a shutdown operation and starts the de-icing system (if equipped), while sending an emergency alarm to maintenance personnel.

[0262] The dynamic mechanism model of the "aerodynamic-thermal-mass" three-phase strong coupling in this invention is constructed using MATLAB software; the data acquisition module uses a high-precision sensor with a sampling frequency of 1Hz; the data processing module uses an STM32F407 microcontroller for data preprocessing; the model icing monitoring is implemented using an industrial control computer, running an algorithm program written in C++; the early warning system includes an audible and visual alarm and a remote communication module, which can send early warning signals to the mobile terminals of maintenance personnel; the human-machine interaction module uses a touch screen to display real-time data, icing status, and early warning information.

Claims

1. A method for monitoring icing on wind turbine blades based on mechanism analysis, characterized in that: The method includes: Step 1: Construct a dynamic mechanism model with strong three-phase coupling of "aerodynamic-thermal-mass" to calculate in real time the local micro-meteorological conditions and water droplet impact characteristics of each airfoil section during blade rotation, as well as the feedback effect of icing growth on aerodynamic shape and subsequent icing. Step 2: Perform icing prediction based on the constructed dynamic mechanism model, including local microenvironment calculation, dynamic icing growth calculation, and aerodynamic feedback calculation; Step 3: Determine the icing status; Step 4: Output icing status and early warning information.

2. The method for monitoring wind turbine blade icing based on mechanism analysis according to claim 1, characterized in that: The dynamic mechanism model includes: Step 1.1 Calculation of unsteady airflow field and droplet trajectory in three-dimensional rotating coordinate system: In the three-dimensional unsteady computational fluid dynamics (CFD) solver, a rotating coordinate system and dynamic mesh technology are introduced to simulate the flow field of the blade under real wind shear, yaw error and turbulence conditions. Step 1.2: Construct a dynamic phase transition and icing growth model based on leaf element thermodynamic equilibrium; Step 1.3, Aerodynamic shape evolution and performance feedback: Couple the icing growth model with the aerodynamic performance lookup table or the reduced-order aerodynamic model in real time.

3. The method for monitoring wind turbine blade icing based on mechanism analysis according to claim 2, characterized in that: In three-dimensional unsteady computational fluid dynamics (CFD) solvers, methods for simulating the flow field of blades under realistic wind shear, yaw error, and turbulent conditions include: Calculation of local effective wind speed and angle of attack: Local effective wind speed vector: ； r is the radius of the cross section from the leaf root. This represents the incoming wind speed vector. This represents the tangential velocity vector generated by the rotation of the blade, where It is the impeller rotational angular velocity vector. It is a vector. It is the azimuth angle; This represents the turbulent wind speed vector generated by the turbulence model; Calculation of effective angle of attack in local areas: ； in, , is the angle of entry; : Distribution of blade twist angle; Dynamic pitch angle; The calculation of the local water droplet collision coefficient includes: Stokes number: ； Let this be the relaxation time of the water droplet; The density of the water droplets; The median volume diameter is the water droplet size distribution. C(r) represents the aerodynamic viscosity; C(r) represents the chord length at the section at a distance r from the blade root radius. Local collision coefficients are based on CFD simulations or empirical correlations: ； For the corresponding airfoil section and current angle of attack The two-dimensional collision efficiency of the Stokes number St; This is a three-dimensional effect correction factor; K is the rotation enhancement coefficient, and k2 and k1 are empirical coefficients obtained by fitting CFD data.

4. The method for monitoring wind turbine blade icing based on mechanism analysis according to claim 2, characterized in that: The construction of the dynamic phase transition and icing growth model based on leaf element thermodynamic equilibrium includes: (1) Leaf element heat balance equation: ； 1) Input heat include: Heating is achieved through the kinetic energy generated by the impact of supercooled water droplets; The latent heat released when water droplets freeze. Latent heat of solidification; Temperature of water droplets Cool to component surface temperature The released sensible heat; Heat provided for the anti-icing / de-icing system; Liquid water content; The impact efficiency coefficient is a dimensionless coefficient with a value ranging from 0 to 1. Effective impact speed; This refers to a time interval, the period during which heat accumulates or the system operates. The latent heat of solidification of water; This is the specific heat capacity of water; The temperature of the water droplet; This refers to the surface temperature of the blade. This refers to the heater power. Heat output include: For convection cooling; For heat dissipation during evaporation / sublimation; For radiative heat dissipation; (2) Dynamic phase change fraction and water film migration: Dynamic phase transition fraction: , ； Water film mass balance: ; The rate of change of water film mass over time; The impact water mass flow rate; This refers to the mass flow rate of the frozen water, also known as the freezing volume. Water loss is caused by airflow shear force; This is due to water loss caused by centrifugal force.

5. The method for monitoring wind turbine blade icing based on mechanism analysis according to claim 2, characterized in that: Aerodynamic shape evolution and performance feedback include: (1) Real-time aerodynamic coefficient calculate: Calculate the ice shape at the current moment and dynamically update the lift coefficient of each leaf element. drag coefficient and torque coefficient The airfoil coordinates are modified in real time by parametrically describing the ice shape; then, the airfoil aerodynamic data after ice shape perturbation is used to obtain values ​​from a lookup table or a reduced-order model. ; Torque loss and unbalanced load calculation: 1) Blade element torque contribution: ； 2) Total torque: , Number of leaves; 3) Torque loss factor: ； 4) Unbalanced load: flapping torque of a single blade: ； , Let r be the lift and drag at the distance from the blade root radius (r) to the i-th cross section, respectively. Total unbalanced swing torque vector: ,in Let be the unit vector of the i-th blade in the waving direction.

6. The method for monitoring wind turbine blade icing based on mechanism analysis according to claim 1, characterized in that: The icing status assessment includes: Step 3.1: Set the judgment threshold; Step 3.2: Comprehensive logical judgment, the judgment logic includes: (1) Judgment by direct physical quantity: Condition A: Does the ice thickness calculated by the model exceed a preset threshold at any leaf element location? Condition B: Does the total icing mass calculated by the model, i.e., the water film mass, exceed the preset threshold? (2) Judgment of performance derivatives: Condition C: Whether the torque loss factor predicted by the model is lower than the preset threshold; Condition D: Does the unbalanced load predicted by the model exceed the safety threshold of the wind turbine structure load? (3) Logical fusion decision: Determining "whether it is covered with ice": If condition A is true in any position, it is determined that "the blade is covered with ice"; Assess the degree of icing: Slight icing: Condition A meets the first-level threshold, and conditions C and D do not trigger the second-level threshold; Moderate icing: Condition A meets the secondary threshold, or condition C triggers the secondary threshold; Severe icing: Condition A meets the level 3 threshold, or condition C triggers the level 3 threshold, or condition D triggers the danger threshold.

Images