Intelligent control method for anti-freezing adjustment of wind power blade

Abstract

The application discloses an intelligent control method for anti-freezing adjustment of a wind power blade, and relates to the technical fields of wind power operation and maintenance and blade anti-freezing; the application synchronously collects four types of data, i.e., environmental meteorological data, blade surface state data, heating system operation data and unit operation data, through a multi-source sensing module, outputs a four-grade icing risk level through an icing probability model evaluation based on multi-parameter weighted fusion, matches a corresponding graded heating strategy to output differentiated power of an air circulation heating system in a blade inner cavity, simultaneously combines blade load and power generation power double coefficient dynamic correction heating output, finally relies on intelligent inspection verification deicing effect and feeds back optimization model parameters, realizes full-process intelligent management and control of pre-identification of icing hazards, on-demand precise heating, unit safety linkage and strategy continuous iteration.

Description

Technical Field

[0001] This invention relates to the field of wind power operation and maintenance and blade anti-icing technology, specifically to an intelligent control method for wind turbine blade anti-icing adjustment. Background Technology

[0002] With the rapid development of my country's new energy industry and the continuous growth of wind power installed capacity, wind farms located at high altitudes and in some humid areas of northern and southern China generally face the problem of blade icing in low-temperature winter environments. Blade icing alters the original aerodynamic shape, reduces wind energy capture efficiency, and increases the structural load on the blades and the vibration of the turbine. In severe cases, it can lead to blade damage, turbine shutdown, and other safety accidents, directly affecting the operational reliability and power generation revenue of the wind farm. Blade antifreezing and de-icing have become a key technical aspect of the operation and maintenance of low-temperature wind farms.

[0003] Existing blade antifreeze solutions mostly employ a fixed-power heating mode, triggering heating start-stop solely based on a single temperature threshold. This fails to dynamically adjust output power according to real-time icing risk, resulting in high energy consumption and delayed de-icing response. Many solutions are designed independently of the unit's main control system, without coordinating adjustments based on real-time blade loads and unit power generation status. This can easily lead to load superposition during de-icing, increasing structural safety risks. Furthermore, they lack closed-loop verification of de-icing effectiveness and strategy self-optimization capabilities, exhibiting poor adaptability to different wind farm micro-meteorological conditions, making it difficult to simultaneously meet the multiple requirements of de-icing effectiveness, operational economy, and unit safety.

[0004] Overall, existing blade antifreeze technologies suffer from drawbacks such as crude control methods, high energy consumption, weak coordination with turbine operation, and insufficient adaptive optimization capabilities. They cannot meet the operation and maintenance needs of wind farms in complex low-temperature environments. Therefore, it is particularly important to develop an intelligent control method for wind turbine blade antifreeze regulation that features intelligent prediction, precise regulation, safe linkage, and closed-loop optimization. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent control method for wind turbine blade antifreeze regulation. It can simultaneously collect four types of data through multi-source sensing modules: environmental meteorology, blade surface condition, heating system operation, and unit operation. After evaluation and output of a four-level icing risk level by a multi-parameter weighted fusion icing probability model, it matches the corresponding graded heating strategy to output differentiated power to the blade internal air circulation heating system. At the same time, it dynamically corrects the heating output by combining the blade load and power generation dual coefficients. Finally, it relies on intelligent inspection to verify the de-icing effect and feed back to optimize the model parameters, realizing full-process intelligent management and control of icing hazard identification, on-demand precise heating, unit safety linkage, and continuous strategy iteration.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent control method for antifreeze adjustment of wind turbine blades, the method comprising the following components:

[0007] Step 1: Multi-source sensing data acquisition, real-time acquisition of four types of data: environmental meteorology, blade surface condition, heating system operation, and unit operation status, and synchronization to the tower base control unit;

[0008] Step 2: Icing risk level assessment. Based on the collected multi-source data, the icing prediction model outputs a four-level risk level to complete the preliminary identification of icing hazards.

[0009] Step 3: Graded heating control, matching the heating power strategy to the corresponding risk level, and outputting differentiated heating power to the air circulation heating system inside the blade cavity;

[0010] Step 4: Operation status linkage correction. Dynamically adjust the heating output power by combining the real-time operating conditions of the unit and the blade load data to balance the de-icing effect and the safe operation of the unit.

[0011] Step 5: Closed-loop feedback of effects, real-time verification of de-icing effect and feedback to optimize prediction model parameters and control strategies, to achieve continuous iteration of control logic.

[0012] Furthermore, in step one, four types of data are collected at differentiated intervals. Environmental meteorological data is collected by the wind farm's anemometer tower and the nacelle meteorological station, with a collection period of 60-120 seconds, covering parameters such as ambient temperature, relative humidity, wind speed, wind direction, and atmospheric pressure. Blade surface temperature data is collected by the infrared thermal imaging module on the top of the nacelle, with a collection period of 27 minutes, covering parameters such as blade surface temperature distribution, icing area percentage, and average icing thickness. Heating system operation data is acquired by the hub-end acquisition module, with a collection period of 5-15 seconds, covering parameters such as real-time power of the PTC heating element, circulating fan speed, and temperatures at the front, middle, and rear points of the blade's inner cavity. Unit operating status data is collected by the unit... The SCADA system acquires data synchronously, with a collection cycle of 1-2 seconds. Parameters cover unit output power, pitch angle, impeller speed, and blade flapping and oscillation directional loads. All data are processed by the edge gateway using the sliding 3σ criterion to remove outliers. The calculation window consists of 20 consecutive sampling points, and jump verification is performed synchronously. When the relative deviation between two adjacent sampling values ​​exceeds 5%, it is determined to be a jump and replaced with the previous valid value. The relative deviation is calculated based on the previous sampling value. The processed data is uploaded to the tower base control unit for unified storage and retrieval through the wind farm ring network. The above collection cycle is configured according to the actual micro-meteorological conditions of the wind farm and the unit response speed. During periods of frequent icing, the cycle can be appropriately shortened to improve the timeliness of early warning.

[0013] Furthermore, in step two, a multi-parameter weighted fusion formula for calculating the icing probability is used, which is as follows: ,in This represents the probability of leaf icing within the next 3 hours, with a value ranging from 0 to 1. This refers to the ambient temperature, expressed in °C. The standard freezing critical temperature is set at 0℃. The relative humidity is expressed as a percentage. This is real-time wind speed, in m / s. The rated reference wind speed is taken as 12 m / s. The average surface temperature of the blade is expressed in °C. , , , These are the weighting coefficients of each influencing factor. The nonlinear adjustment coefficients are obtained by fitting 132 sets of icing event samples from three consecutive icing periods of the target wind field using the Levenberg-Marquardt algorithm. The fitting objective is to maximize the icing prediction accuracy. This is an example, not a limitation, in a preferred embodiment of a certain wind field. The value is 0.42. The value is 0.28. The value is 0.18. The value is 0.12. The value is set to 1.3. All temperature difference terms in the formula are truncated to positive values. When the calculated value is less than 0, it is counted as 0 in the weighted sum to ensure that the formula has a valid real solution in the entire temperature range. The formula output is directly used to classify and determine the level of icing risk.

[0014] Furthermore, in step two, the risk is divided into four levels based on both icing probability and icing thickness. The no-risk level corresponds to an icing probability of less than 0.1 and no measured icing. The low-risk level corresponds to an icing probability of 0.1-0.31 and an icing thickness of less than 1.1mm. The medium-risk level corresponds to an icing probability of 0.31-0.69 and an icing thickness of 1.1mm-4.7mm. The high-risk level corresponds to an icing probability of not less than 0.69 or an icing thickness of not less than 4.7mm. The dual-dimensional judgment adopts the principle of choosing the higher level. When the levels corresponding to icing probability and icing thickness are inconsistent, the higher level is taken as the final risk level. The boundary critical state refers to the situation where the parameter value is equal to the level threshold, and it is uniformly classified into the lower level. A hysteresis protection mechanism is set for risk level switching. Upgrading the level requires two consecutive data collections that meet the target level conditions, and downgrading the level requires three consecutive data collections that meet the lower level conditions. A single data fluctuation does not trigger level switching, avoiding equipment damage and energy waste caused by frequent switching of heating levels.

[0015] Furthermore, in step three, a three-level control cabinet is used to distribute and execute heating control. The tower base control cabinet issues the global heating strategy, the nacelle control cabinet monitors the overall heating system operation status, and the hub control cabinet is responsible for the independent heating control of the three blades. The blade cavity is divided into three heating zones along the blade length: the root section accounts for the first proportion of the blade length, the middle section accounts for the second proportion, and the tip section accounts for the third proportion. This proportion can be optimized according to the blade structure. As an example, the root section accounts for 34%, the middle section accounts for 36%, and the tip section accounts for 30%. Under no-risk conditions, the heating system enters standby mode, with only the data acquisition module and the communication module remaining operational. At low risk levels, a low-power preheating mode is activated, with the PTC heating element outputting a low percentage of its rated power. The circulating fan operates at a low speed, and only the blade root section heating area is activated. At medium risk levels, a medium-power de-icing mode is activated, with the PTC heating element outputting a moderate percentage of its rated power. The circulating fan operates at a moderate speed, and the blade root and middle sections heating areas are activated. At high risk levels, a full-power de-icing mode is activated, with the PTC heating element outputting full power. The circulating fan operates at full speed, and all three sections of the blade heating area are activated. Each blade independently adjusts its heating power according to its own icing status, without the need for all three blades to operate simultaneously.

[0016] Furthermore, the three-level control cabinet adopts a distributed control architecture. The tower base control cabinet is responsible for data interaction with the wind farm's main control system and operation and maintenance management platform, executing global power scheduling and control strategy distribution, and using the Modbus TCP protocol for communication. The nacelle control cabinet is responsible for monitoring the overall operating status of the heating system and fault alarms, and uploading operating data to the tower base control cabinet in real time. The hub control cabinet is responsible for precise heating control of individual blades and local data acquisition. Data is transmitted between the nacelle and the hub through a dedicated Ethernet channel of the hub's conductive slip ring, using the Profinet protocol for communication. Both the hub control cabinet and the nacelle control cabinet are equipped with LoRa wireless backup communication links. The failure of a single communication link does not affect data transmission. When a single blade heating system fails, the corresponding branch is automatically isolated, without affecting the normal operation of the other two blades. When the hub control cabinet is offline, the nacelle control cabinet automatically takes over the heating control and operates at 73% of the base power corresponding to the current risk level. The base power is defined as the benchmark value of the rated heating power of a single blade under the current risk level, ensuring basic anti-freezing capability.

[0017] Furthermore, the blade heating system uses PTC positive temperature coefficient heating elements. The Curie temperature of the element corresponds to a maximum steady-state surface temperature of no more than 82℃, possessing constant temperature self-limiting characteristics and preventing local overheating. The heating elements are arranged along the leeward side of the main beam inside the blade cavity and are installed as a whole inside the blade cavity without altering the original lightning arrester and lightning protection current guiding structure of the blade, and without adding additional aerodynamic protrusions on the windward side of the blade. All electrical components of the heating system are double-insulated, with a grounding resistance of no more than 3.7Ω. The system has built-in triple protection for over-temperature, over-current, and short circuit. The over-temperature protection action threshold is set at 87℃ as a secondary protection under abnormal operating conditions. The over-current protection is triggered when the current reaches 1.2 times the rated value. When any protection is triggered, the corresponding blade heating power supply is immediately cut off. When the wind farm lightning monitoring system issues a thunderstorm warning signal within a 30km range, all blade heating systems automatically shut down and disconnect the main power supply. Operation is gradually restored after the warning is lifted, ensuring the safety of the equipment and unit throughout the process.

[0018] Furthermore, in step four, the actual output power is calculated using a dynamic correction formula for heating power. The formula is as follows: ,in This refers to the actual heating output power of a single blade, expressed in kW. This refers to the base heating power for the corresponding risk level, expressed in kW. This is the blade load correction factor, with a value ranging from 0.5 to 1. This is a correction factor for the unit's power generation, with a value ranging from 0.8 to 1. The load ratio is calculated from the ratio of the real-time blade flapping direction load to the design safety threshold. When the load ratio is less than the first load threshold... The load ratio is always 1 when it reaches the first load threshold. The value is 0.8, and the load ratio reaches the second load threshold. When the value is 0.5, and the load ratio is greater than the second load threshold. The constant value is 0.5. As an example, the first load threshold can be 0.77, and the second load threshold can be 0.93. Calculated by the ratio of the unit's current generating power to its rated power, when the generating power ratio is lower than the first power threshold. When the value is 0.8, the proportion of power generation is higher than the second power threshold. The value is set to 1, and intermediate values ​​are calculated using linear interpolation. For example, the first power threshold can be 0.21, and the second power threshold can be 0.53, assuming the unit is shut down. The value is always 1. This formula serves as the benchmark for continuous power adjustment under normal operating conditions. When the load exceeds the limit and triggers the protection, the gear-type downshifting action is executed first. The formula output is used for real-time power adjustment of the heating system.

[0019] Furthermore, in step four, the blade load data of the unit's SCADA system is read in real time. Using the blade flapping direction load as the core criterion, the design safety threshold is directly taken from the blade's factory design documents. When the load ratio reaches the first load threshold, a first-level power reduction is triggered, lowering the heating power to 80% of the currently calculated value. Simultaneously, the unit's main control system adjusts the pitch angle in the feathering direction by a safe step to reduce the impeller aerodynamic load. When the load ratio reaches the second load threshold, a second-level power reduction is triggered, lowering the heating power to 50% of the currently calculated value. The pitch angle is then adjusted in the same phase length towards the feathering direction. When the load ratio falls below the recovery threshold, the speed is gradually restored to the calculated power by increasing one gear at certain time intervals. The above step length and recovery rate can be calibrated according to the pitch response capability of the unit and the blade thermal time constant. As an example, the step length is 1.7° and the recovery interval is 13 minutes. When the unit is in a shutdown state, the power generation correction limit is lifted, and the heating system can operate at full power. When the unit is in a grid-connected state, the total heating power consumption does not exceed 5% of the current power generation of the unit, ensuring the stability of the power output at the grid connection point.

[0020] Furthermore, in step five, the de-icing effect is verified using an intelligent inspection system. This system employs a combined visual recognition and infrared temperature measurement inspection method, with an inspection cycle of 30-60 minutes. The system determines the progress of ice melting, locates the remaining ice-covered area and its corresponding heating section, and increases the output of that section by 10%-15% based on its current power. When the ice on all blades is completely melted and the risk of icing drops below low risk, the heating power is gradually reduced in a gradient from high power to medium power to low power to standby. Each level is maintained for 20-30 minutes before switching to the next level to avoid a sudden drop in temperature that could cause secondary icing. Meteorological parameters, heating strategies, de-icing duration, total energy consumption, and de-icing effect for each complete de-icing process are stored in a historical database. Every 31 days, the weighting coefficients of the icing probability formula are incrementally fitted and optimized, with the icing prediction accuracy as the optimization objective. The iterative convergence threshold is set at a maximum relative change in the weighting coefficients of less than 1×10⁻⁻⁻⁶. 4 Meanwhile, the base power value of the graded heating strategy is adaptively adjusted based on the cumulative de-icing effect to improve the localization accuracy of the model.

[0021] Compared with existing technologies, this intelligent control method for wind turbine blade antifreeze regulation has the following advantages:

[0022] I. This method combines a multi-parameter weighted fusion icing probability prediction model with four-level graded heating control to identify icing risks in advance and match corresponding heating power. Combined with gradient heating and cooling control, it avoids ineffective energy consumption and secondary icing, effectively reducing the overall operating energy consumption of the blade antifreeze system, shortening the downtime of the unit due to icing, and improving the winter power generation efficiency of the wind farm.

[0023] Second, this method introduces a dual-dimensional linkage correction mechanism between blade load and power generation, which can dynamically avoid the risk of structural load exceeding limits during de-icing. At the same time, combined with the closed-loop verification of de-icing effect and the incremental optimization mechanism of the model, it can continuously improve the localized control accuracy with the accumulation of operational experience, taking into account both the safety of unit operation and the long-term adaptability of control strategy.

[0024] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0026] Figure 1 The main flowchart is for an intelligent control method used for antifreeze regulation of wind turbine blades;

[0027] Figure 2 A flowchart of multi-source sensing data acquisition and preprocessing for an intelligent control method for wind turbine blade antifreeze regulation;

[0028] Figure 3 This is a flowchart illustrating the dynamic correction of heating power and load linkage in an intelligent control method for wind turbine blade antifreeze regulation. Detailed Implementation

[0029] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0030] In a winter anti-freezing operation and maintenance scenario for a 1.5MW double-fed wind turbine at a mountainous wind farm in northern Shanxi Province, the wind farm, located at an altitude of approximately 1800m, experiences continuous icing from November to March of the following year, with extreme minimum temperatures reaching -22℃. Freezing rain, wet snow, and excessively cold rime ice are common in winter, easily leading to uneven icing on the blade leading edges. Traditional fixed-power heating solutions suffer from high energy consumption and delayed de-icing response. Furthermore, the heating system operates independently of the turbine's main control system, which can easily cause blade load fluctuations and grid-connected power deviations. The intelligent control method for wind turbine blade anti-freezing regulation, as described in this invention, is used for blade anti-freezing management. The specific implementation process is as follows: Figure 1As shown, the main process of this method includes five stages: multi-source sensing data acquisition, icing risk level assessment, graded heating control, operational status linkage correction, and effect closed-loop feedback.

[0031] like Figure 2 As shown, the multi-source sensing data acquisition and preprocessing process is as follows: In this embodiment, the four types of data adopt differentiated acquisition cycles, and the cycle setting is determined based on the change rate of the corresponding parameters and the control response requirements. Environmental meteorological data is jointly acquired by the wind farm anemometer tower and the meteorological station on the top of the nacelle, with an acquisition cycle of 65 seconds. The acquired parameters include five items: ambient temperature, relative humidity, wind speed, wind direction, and atmospheric pressure. The temperature and humidity sensors are installed inside the ventilation hood on the top of the nacelle to avoid measurement deviations caused by direct sunlight. Blade surface condition data is acquired by the infrared thermal imaging module on the meteorological rack in the nacelle, with an acquisition cycle of 27 minutes. The parameters cover the blade surface temperature distribution, the proportion of icing area, and the average thickness of icing. The thermal imaging module is preset with the cruising angle of the three blades, and the temperature imaging of the leeward side of the three blades is completed sequentially in each acquisition. Heating system operation data is acquired by the hub-end distributed acquisition module, with an acquisition cycle of 11 seconds. The parameters cover the real-time power of the PTC heating element, the speed of the circulating fan, and the temperature at the front, middle, and rear points of the blade cavity. Each blade corresponds to an independent acquisition branch. The unit's operating status data is directly obtained synchronously from the unit's SCADA system, with an acquisition cycle of 1 second. The parameters cover the unit's output power, pitch angle, impeller speed, and blade flapping and oscillation directional loads. The unit's existing sensing system is utilized without the need for additional sensors. The above cycle is a specific value in this embodiment. In actual applications, it can be adaptively adjusted according to wind field conditions and response requirements. For example, when icing changes rapidly, the infrared acquisition cycle can be shortened to less than 30 seconds.

[0032] All collected data is first preprocessed by the edge gateway on the nacelle side. Outlier removal is performed using the sliding 3σ criterion. The calculation window consists of 20 consecutive sampling points. For each new sampling point, the mean and standard deviation within the window are updated. Sampling points exceeding the mean plus or minus three times the standard deviation are considered outliers and directly replaced with the valid values ​​from the previous time step. Jump checks are performed synchronously, using the relative deviation between two adjacent sampling values ​​as the criterion. The relative deviation is calculated based on the previous sampling value. If the relative deviation exceeds 5%, it is considered a data jump, and the data is also replaced with the previous valid value to avoid control malfunctions caused by momentary sensor failures. After preprocessing, the data is uploaded to the tower base control unit via the wind farm industrial ring network and stored in timestamp alignment for subsequent control logic retrieval.

[0033] The tower base control unit receives the aligned multi-source data and calls the icing prediction model to calculate the probability of blade icing within the next 3 hours. The calculation uses a multi-parameter weighted fusion formula for the icing probability; the formula is: ,in This represents the probability of blade icing within the next 3 hours. For ambient temperature, The standard freezing critical temperature, For ambient relative humidity, For real-time wind speed, The rated reference wind speed, The average temperature of the blade surface. , , , These are the weighting coefficients of each influencing factor. The formula structure, based on the thermodynamic and kinetic mechanisms of icing formation, is a nonlinear adjustment coefficient. All parameters in the formula are first normalized to eliminate dimensional differences. The temperature term is truncated to a positive value; when the ambient temperature or blade surface temperature is above 0°C, the corresponding term is set to 0 to avoid calculation errors due to negative non-integer powers. The weighting coefficients are set according to the degree of influence of each factor on the icing process. Ambient temperature determines the thermodynamic conditions for supercooled water freezing and is the core influencing factor for icing; therefore, the weights are... The value is 0.42; relative humidity determines the amount of supercooled water droplets in the air, which is a necessary condition for ice formation, and its weight is... The value is set to 0.28; wind speed affects the rate and frequency of supercooled water droplets impacting the blade surface, with a weighting of... The value is set to 0.18; the blade surface temperature reflects the state of cold accumulation in the blade matrix, with a weight of 0.18. The value is set to 0.12. The nonlinear adjustment coefficient b is set to 1.3, which is used to fit the nonlinear characteristics of the icing probability increasing rapidly with the deterioration of environmental conditions. The coefficient value is obtained by fitting the historical icing events of the target wind field for three consecutive icing periods.

[0034] Based on the calculated icing probability and the measured icing thickness, a four-level risk classification is established according to the principle of choosing the higher of the two dimensions. The classification references the general grading standards for wind power blade icing and incorporates on-site operation and maintenance strategies. The no-risk level corresponds to an icing probability below 0.1 and no measured icing, requiring no heating intervention. The low-risk level corresponds to an icing probability between 0.1 and 0.31 and an icing thickness less than 1.1 mm, representing the icing initiation stage, requiring preheating for prevention. The medium-risk level corresponds to an icing probability between 0.31 and 0.69 and an icing thickness between 1.1 mm and 4.7 mm, representing the light icing stage, requiring proactive de-icing. The high-risk level corresponds to an icing probability of at least 0.69 or an icing thickness of at least 4.7 mm, representing moderate to severe icing, requiring intensive de-icing. The two-dimensional judgment follows the principle of choosing the higher value; if the two parameters correspond to different levels, the higher level is taken as the final judgment result, with boundary thresholds assigned to the higher level.

[0035] A hysteresis protection mechanism is implemented for risk level switching. Upgrading a risk level requires two consecutive data collections that meet the target level's conditions, while downgrading requires three consecutive data collections that meet the conditions for the next lower level. A single data fluctuation will not trigger a level switch. This mechanism avoids frequent switching of heating levels due to random data fluctuations near the threshold, reduces the number of electrical contactor operations, and extends the lifespan of the heating system hardware.

[0036] In this embodiment, the blade heating system adopts a three-level control cabinet distributed architecture to perform heating control. The blade cavity is divided into three independent heating zones along the blade length: 0-34% of the blade length is the blade root zone, 34%-70% of the blade length is the blade middle zone, and 70%-100% of the blade length is the blade tip zone. The zone division is determined based on the blade structural thickness distribution and icing distribution pattern. The blade root zone has a large structural thickness, slow heat conduction, and poor ventilation conditions near the hub area, making it a zone prone to icing. The blade middle zone is the main aerodynamic working surface of the blade, and icing has the greatest impact on power generation. The blade tip zone has a high linear velocity, rapid icing growth, but the overall thickness is relatively thin.

[0037] The four risk levels correspond to four heating strategies. At no-risk levels, the heating system enters standby mode, with only the data acquisition and communication modules operating to minimize standby power consumption. At low-risk levels, a low-power preheating mode is activated, with the PTC heating element outputting 21%-29% of its rated power and the circulating fan running at 32% of its rated speed. Only the blade root section heating area is activated to maintain the blade root substrate temperature and prevent icing. At medium-risk levels, a medium-power de-icing mode is activated, with the PTC heating element outputting 54%-67% of its rated power and the circulating fan running at 62% of its rated speed. The blade root and middle sections heating areas are activated simultaneously to melt existing thin ice and inhibit further icing. At high-risk levels, a full-power de-icing mode is activated, with the PTC heating element outputting full power and the circulating fan running at full speed. All three sections of the blade heating area are activated to quickly melt thicker ice and restore the unit's aerodynamic performance as soon as possible. The heating power of the three blades can be adjusted independently without synchronous operation, adapting to actual conditions where icing is uneven on individual blades.

[0038] The three-tiered control cabinets collaboratively execute control logic. The tower base control cabinet is responsible for data interaction with the wind farm's main control system and operation and maintenance management platform, executing global power scheduling and distributing control strategies. Communication with the upper-level system uses the Modbus TCP protocol. The nacelle control cabinet is responsible for monitoring the overall operating status of the heating system and issuing fault alarms, uploading operating data to the tower base control cabinet in real time, and acting as a control relay link between the upper and lower levels. The hub control cabinet is responsible for precise heating control of individual blades and local data acquisition. Data is transmitted between the nacelle and the hub via a dedicated Ethernet channel of the hub's conductive slip ring, and communication uses the Profinet protocol to ensure real-time control. Both the hub control cabinet and the nacelle control cabinet are equipped with LoRa wireless backup communication links. When the wired link packet loss rate exceeds 5% for 10 consecutive seconds, the system automatically switches to the backup link. The failure of a single communication link does not affect the overall data transmission. When a single blade heating system malfunctions, the corresponding branch is automatically isolated, without affecting the normal operation of the other two blades. When the hub control cabinet is offline, the nacelle control cabinet automatically takes over the heating control and operates at 73% of the base power corresponding to the current risk level. The base power is defined as the benchmark value of the rated heating power of a single blade under the current risk level, ensuring that the basic antifreeze capability is not interrupted.

[0039] The blade heating system uses PTC positive temperature coefficient heating elements. The Curie temperature of the element corresponds to a maximum steady-state surface temperature of no more than 82℃, possessing constant temperature self-limiting characteristics and preventing localized overheating under normal operation. The heating elements are arranged along the leeward side of the main beam inside the blade cavity, without damaging the main blade structure. They are installed integrally within the blade cavity without altering the original lightning arrester and lightning protection flow guiding structure, and without adding additional aerodynamic protrusions to the windward side of the blade. All electrical components of the heating system are double-insulated, and the system grounding resistance is no greater than 3.7Ω. It has built-in over-temperature, over-current, and short-circuit protection. The over-temperature protection threshold is set at 87℃ as a secondary protection under abnormal operating conditions to cope with extreme fault scenarios such as the circulating fan stopping. The over-current protection is triggered when the current reaches 1.2 times the rated value. The corresponding blade heating power supply is immediately cut off when any protection is triggered. When the wind farm lightning monitoring system issues a thunderstorm warning signal within a 30km range, all blade heating systems automatically shut down and disconnect the main power supply. Operation is gradually restored after the warning is lifted, ensuring the lightning protection safety of the equipment and unit throughout the process.

[0040] like Figure 3 As shown, the dynamic correction and load linkage process for heating power is as follows: Before outputting heating power, dynamic correction needs to be performed based on the real-time operating conditions of the unit. The correction calculation uses the dynamic correction formula for heating power; the formula is: ,in This represents the actual heating output power of a single blade. This is the base heating power for the corresponding risk level. This is the blade load correction factor. This is the correction factor for the unit's power generation. The formula uses the base heating power under the corresponding risk level as a benchmark, multiplying it by the blade load correction factor and the unit's power generation correction factor respectively to obtain the final actual output power, taking into account both structural safety and grid-connected power quality requirements.

[0041] The blade load correction factor is based on the load in the blade flapping direction, and the design safety threshold is directly taken from the ultimate flapping load value in the blade's factory design documents. When the load ratio is less than 0.77, the load correction factor is always 1, and there is no restriction on heating power; when the load ratio reaches 0.77, the correction factor is 0.8, corresponding to the first level of load reduction; when the load ratio reaches 0.93, the correction factor is 0.5, corresponding to the second level of load reduction; when the load ratio is greater than 0.93, the correction factor is always 0.5, maintaining the minimum protection level. Linear interpolation is used to calculate values ​​within the range to achieve a smooth power transition. The 0.77 threshold corresponds to 77% of the design load, reserving sufficient safety margin to cope with sudden gusts; the 0.93 threshold is close to the ultimate load warning line, and mandatory load reduction must be implemented to avoid structural damage.

[0042] It should be noted that the above load correction factor It is used for continuous power regulation under normal operating conditions. When the load ratio reaches the protection threshold, the system will independently trigger a downshift action, and the action will not be repeated. To avoid excessive power reduction, the following measures are implemented: When the load ratio reaches 0.77, a first-level power reduction is triggered, lowering the heating power to 80% of the calculated value. Simultaneously, the main control system adjusts the pitch angle 1.7° towards the feather direction, reducing aerodynamic lift by decreasing the blade angle of attack, thereby reducing the flapping load. When the load ratio reaches 0.93, a second-level power reduction is triggered, lowering the heating power to 50% of the calculated value. The pitch angle is then adjusted another 1.7° towards the feather direction to further reduce the aerodynamic load. When the load ratio falls below 0.71, the power is gradually restored to the calculated power level at a rate of one level every 13 minutes to avoid frequent power fluctuations.

[0043] The generator power correction factor is adjusted according to the generator's operating status. When the generator is shut down, the correction factor is always 1, and power restrictions are lifted to melt ice as quickly as possible. When the generator is connected to the grid, the correction factor is 0.8 when the generator power ratio is below 0.21, and 1 when the generator power ratio is above 0.53, with linear interpolation within the range. Under grid-connected conditions, the total heating power consumption is controlled to not exceed 4.7% of the generator's current generator power to avoid sudden changes in generator output power caused by heating power fluctuations, meeting the grid connection guidelines for active power fluctuations. Under normal operating conditions, the adjustment benchmark is the value calculated using the continuous formula. When the load exceeds the limit and triggers the protection, the step-down action is executed first to ensure the highest safety priority.

[0044] During the de-icing process, an intelligent inspection system is used to verify the de-icing effect. A combined visual recognition and infrared thermography inspection method is employed, with an inspection cycle of 60 minutes. Each inspection sequentially collects visible light images and infrared thermal imaging data from three blades. Image recognition algorithms extract the icing area and thickness distribution to determine the overall de-icing progress. Simultaneously, residual icing areas are mapped to their respective heating sections. If residual icing remains in a specific heating section, the output of that section is increased by 14% from its current power, specifically enhancing the local de-icing effect without wasting energy by increasing the overall power.

[0045] When it is detected that the ice on all blades has completely melted and the risk of icing has dropped below low risk, the heating system is not shut down directly. Instead, the heating power is gradually reduced in a gradient from high power to medium power to low power to standby, maintaining each level for 27 minutes before switching to the next level. Gradual cooling allows the blade surface temperature to drop slowly, avoiding a sudden drop in temperature that could cause water vapor in the air to condense again on the low-temperature blade surface, thus preventing secondary icing.

[0046] Meteorological parameters, heating strategies, de-icing duration, total energy consumption, and de-icing effect for each complete de-icing process are stored in a historical database. Every 31 days, the weighting coefficients of the icing probability formula are incrementally fitted and optimized. The optimization objective function is the accuracy of icing prediction. All icing event data within the period are extracted and added to the sample database. The Levenberg-Marquardt algorithm is used for iterative fitting. The optimization is successful when the maximum relative change of all weighting coefficients is less than 1×10⁻. 4 The system determines when the iteration converges and updates the model weight parameters. Simultaneously, based on historical de-icing results, the system automatically evaluates the energy consumption and timeliness of the current tiered heating strategy. If the average de-icing time or unit energy consumption under a specific risk level significantly deviates from the historical optimal value, the system adaptively fine-tunes the base heating power value for that level, achieving continuous iteration of the control strategy. With the accumulation of operational data, the model's adaptation accuracy to local wind field micro-meteorological conditions continuously improves, and the accuracy of icing prediction gradually increases.

[0047] In summary, this embodiment addresses the actual operation and maintenance scenario of blade icing in northern mountain wind farms during winter. It constructs a complete control chain from multi-source data acquisition, risk assessment, tiered heating to coordinated correction and closed-loop optimization. Nonlinear icing probability prediction enables proactive identification of icing hazards, while segmented and tiered heating strategies reduce ineffective energy consumption. Dual-dimensional corrections based on load and power generation ensure safe unit operation and grid connection stability. Closed-loop iteration continuously improves localization adaptability. The entire method can be implemented using existing wind farm hardware without large-scale hardware modifications. It effectively reduces the energy consumption of the anti-freezing system while ensuring de-icing effectiveness, minimizes turbine downtime due to icing, and improves the reliability and overall power generation benefits of low-temperature wind farms during winter.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An intelligent control method for antifreeze adjustment of wind turbine blades, characterized in that, The method comprises the following components: Step 1: Multi-source sensing data acquisition, real-time acquisition of four types of data: environmental meteorology, blade surface condition, heating system operation, and unit operation status, and synchronization to the tower base control unit; Step 2: Icing risk level assessment. Based on the collected multi-source data, the icing prediction model outputs a four-level risk level to complete the preliminary identification of icing hazards. Step 3: Graded heating control, matching the heating power strategy to the corresponding risk level, and outputting differentiated heating power to the air circulation heating system inside the blade cavity; Step 4: Operation status linkage correction. Dynamically adjust the heating output power by combining the real-time operating conditions of the unit and the blade load data to balance the de-icing effect and the safe operation of the unit. Step 5: Closed-loop feedback of effects, real-time verification of de-icing effect and feedback to optimize prediction model parameters and control strategies.

2. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, In step one, four types of data are collected according to differentiated cycles. Environmental meteorological data are collected by the wind farm's anemometer tower and the nacelle meteorological station, covering parameters such as ambient temperature, relative humidity, wind speed, wind direction, and atmospheric pressure. Blade surface temperature data are collected by the infrared thermal imaging module on the top of the nacelle, covering parameters such as blade surface temperature distribution, icing area ratio, and average icing thickness. Heating system operation data are acquired by the hub-end acquisition module, covering parameters such as real-time power of the PTC heating element, circulating fan speed, and temperature at the front, middle, and rear points of the blade cavity. Unit operation status data are synchronously acquired by the unit's SCADA system, covering parameters such as unit output power, pitch angle, impeller speed, and blade flapping and oscillation direction loads. All data are processed by the edge gateway using the sliding 3σ criterion to remove outliers. The calculation window consists of 20 consecutive sampling points, and jump verification is performed synchronously. When the relative deviation between two adjacent sampling values ​​exceeds 5%, it is determined to be a jump and replaced with the previous valid value. The relative deviation is calculated based on the previous sampling value. The processed data is uploaded to the tower base control unit for unified storage and retrieval through the wind farm ring network.

3. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, The icing probability calculation formula used in step two is a multi-parameter weighted fusion method, and the formula is as follows: ,in This represents the probability of blade icing within the next 3 hours. For ambient temperature, The standard freezing critical temperature, For ambient relative humidity, For real-time wind speed, The rated reference wind speed, The average temperature of the blade surface. , , , These are the weighting coefficients of each influencing factor. This is a non-linear adjustment coefficient.

4. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, In step two, the risk is divided into four levels based on both the probability of icing and the thickness of icing. The two-dimensional judgment adopts the principle of choosing the higher level. When the levels corresponding to the probability of icing and the thickness of icing are inconsistent, the higher level is taken as the final risk level. The boundary critical state refers to the situation where the parameter value is equal to the level threshold. It is uniformly classified into the lower level of the two adjacent levels of the threshold. A hysteresis protection mechanism is set for risk level switching. Upgrading the level requires two consecutive data collections that meet the conditions of the target level. Downgrading the level requires three consecutive data collections that meet the conditions of the lower level. A single data fluctuation does not trigger the level switching.

5. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, In step three, a three-level control cabinet is used to distribute the heating control. The tower base control cabinet issues the global heating strategy, the nacelle control cabinet monitors the overall heating system operation status, and the hub control cabinet is responsible for the independent heating control of the three blades. The blade cavity is divided into three heating zones along the blade length. Under no-risk conditions, the heating system enters standby mode, with only the data acquisition module and communication module running. Under low-risk conditions, a low-power preheating mode is activated, with the PTC heating element outputting 21% to 29% of its rated power and the circulating fan running at 32% of its rated speed, activating only the blade root heating zone. Under medium-risk conditions, a medium-power de-icing mode is activated, with the PTC heating element outputting a medium percentage of its rated power and the circulating fan running at a medium speed, activating the blade root and middle heating zones. Under high-risk conditions, a full-power de-icing mode is activated, with the PTC heating element outputting full power and the circulating fan running at full speed, activating all three heating zones of the blade.

6. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 5, characterized in that, The three-level control cabinet adopts a distributed control architecture. The tower base control cabinet is responsible for data interaction with the wind farm's main control system and operation and maintenance management platform, executing global power scheduling and control strategy distribution. Communication uses the Modbus TCP protocol. The nacelle control cabinet is responsible for monitoring the overall operating status of the heating system and fault alarms, and uploading operating data to the tower base control cabinet in real time. The hub control cabinet is responsible for precise heating control of individual blades and local data acquisition. Data is transmitted between the nacelle and the hub through a dedicated Ethernet channel of the hub conductive slip ring. Communication uses the Profinet protocol. Both the hub control cabinet and the nacelle control cabinet are equipped with LoRa wireless backup communication links. The failure of a single communication link does not affect data transmission. When a single blade heating system fails, the corresponding branch is automatically isolated. When the hub control cabinet is offline, the nacelle control cabinet automatically takes over the heating control and operates at 73% of the base power corresponding to the current risk level. The base power is defined as the benchmark value of the rated heating power of a single blade under the current risk level.

7. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 5, characterized in that, The blade heating system uses PTC positive temperature coefficient heating elements. The Curie temperature of the element corresponds to a maximum steady-state surface temperature of no more than 82℃, possessing constant temperature self-limiting characteristics and preventing local overheating. The heating elements are arranged along the leeward side of the main beam inside the blade cavity and are installed as a whole inside the blade cavity without altering the original lightning arrester and lightning protection current guiding structure of the blade, and without adding additional aerodynamic protrusions on the windward side of the blade. All electrical components of the heating system are double-insulated, and the grounding resistance is no greater than 3.7Ω. The system has built-in triple protection for over-temperature, over-current, and short circuit. The over-temperature protection action threshold is set at 87℃ as a secondary protection under abnormal operating conditions. The over-current protection is triggered when the current reaches 1.2 times the rated value. When any protection is triggered, the corresponding blade heating power supply is immediately cut off. When the wind farm lightning monitoring system issues a thunderstorm warning signal within a 30km range, all blade heating systems automatically shut down and disconnect the main power supply, and gradually resume operation after the warning is lifted.

8. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, In step four, the actual output power is calculated using a dynamic correction formula for heating power. The formula is as follows: ,in This represents the actual heating output power of a single blade. This is the base heating power for the corresponding risk level. This is the blade load correction factor. This is the correction factor for the unit's power generation.

9. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 8, characterized in that, In step four, the blade load data of the unit's SCADA system is read in real time. The load in the blade flapping direction is used as the core judgment basis. The design safety threshold is directly taken from the blade's factory design document. When the load ratio reaches the first load threshold, the first-level power reduction is triggered, and the heating power is reduced to 80% of the current formula calculation value. At the same time, the unit's main control system is linked to adjust the pitch angle in the feathering direction by a safe step to reduce the aerodynamic load on the impeller. When the load ratio reaches the second load threshold, the second-level power reduction is triggered, and the heating power is reduced to 50% of the current formula calculation value. The pitch angle is then adjusted in the feathering direction by the same step. When the load ratio falls back below the recovery threshold, it is gradually restored to the formula calculation power at a speed of one level at time intervals. When the unit is in a shutdown state, the power generation correction limit is lifted, and the heating system can operate at full power. When the unit is in a grid-connected state, the total heating power consumption does not exceed 5% of the unit's current power generation.

10. The intelligent control method for antifreeze adjustment of wind turbine blades according to claim 1, characterized in that, In step five, the de-icing effect is verified by relying on the intelligent inspection system. The inspection method combines visual recognition and infrared temperature measurement to determine the progress of ice melting, locate the residual ice area and correspond it to the heating section. The output of the heating section is increased by 10%-15% based on the current power. When it is detected that the ice on all blades has completely melted and the risk of icing has dropped below low risk, the heating power is gradually reduced in the gradient of high power, medium power, low power and standby. Each level is maintained for 20-30 minutes and then switched to the next level. The meteorological parameters, heating strategy, de-icing time, total energy consumption and de-icing effect of each complete de-icing process are stored in the historical database.

Images