A Vertical Axis Wind Turbine Power Prediction Method Based on Multidimensional Environmental Data

Abstract

This invention provides a method for predicting the power of a vertical axis wind turbine based on multidimensional environmental data. It relates to the field of vertical axis wind turbine power technology using multidimensional environmental data. The invention utilizes a sensor network to collect multidimensional data in real time, including environmental parameters such as wind speed, turbulence intensity, temperature, and humidity; operating states such as rotor speed and blade pitch angle; and structural responses such as blade strain and vibration acceleration. Next, empirical mode decomposition is performed on the vibration acceleration signal to construct a physical information neural network. Using environmental data and the frequency domain energy spectrum of aerodynamic loads as input, the real-time wind energy utilization coefficient and aerodynamic coefficients are output. Finally, based on the time domain signal of the aerodynamic loads, online rainflow counting with dynamic window length is used to extract the load cycle spectrum. Combined with a humidity-corrected fatigue damage model, the cumulative damage value is calculated. When the remaining lifetime is below a threshold, the load peak suppression and power generation efficiency are balanced through the coordinated control of linear decay of magnetoresistive torque and fuzzy PID control of the flaps.

Description

Technical Field

[0001] This invention relates to the field of vertical axis wind turbine power technology based on multidimensional environmental data, specifically a method for predicting vertical axis wind turbine power based on multidimensional environmental data. Background Technology

[0002] Vertical axis wind turbines are increasingly used in distributed power generation and urban environments due to their compact structure and good wind direction adaptability. However, the unsteady characteristics of their aerodynamic loads, such as dynamic stall and turbulent transients, pose significant challenges to fatigue life prediction and operation control. In traditional methods, aerodynamic load monitoring relies on simplified mechanistic models or pure data-driven black-box predictions. The former is inaccurate because it ignores dynamic stall effects and three-dimensional flow disturbances, while the latter lacks physical constraints and is prone to overfitting. Neither of these methods can achieve real-time high-precision monitoring under complex operating conditions. At the same time, existing fatigue life assessment methods are mostly based on static load spectra or simplified cycle counting, without considering the transient correction of dynamic angle of attack by turbulence intensity and the coupling effect of humidity on material crack propagation. This results in a remaining life prediction deviation of more than 30%, causing the active load reduction strategy to respond with lag or be overly conservative, which seriously affects the reliability and power generation efficiency of the unit.

[0003] Existing technologies mainly address this problem through two approaches: First, based on the modified leaf element momentum theory combined with the NREL air density correction formula, but without considering the nonlinear influence of turbulent transients on dynamic stall; Second, by directly fitting SCADA data using time-series neural networks such as LSTM, although the error can be reduced to 15% under steady-state conditions, the generalization ability drops sharply under extreme turbulence due to the lack of physical constraints, and the error rebounds to over 35%.

[0004] The shortcomings of existing technologies are mainly reflected in the following aspects: 1) The aerodynamic model and data-driven method are separated, and it is impossible to take into account the coupling of dynamic stall effect and real-time environmental parameters, such as the correlation between humidity and fatigue life; 2) Fatigue life assessment relies on static rain flow count and standard material parameters, ignoring the humidity-induced crack propagation acceleration effect. Experiments show that increasing humidity from 30% to 90% can shorten fatigue life by 40%; 3) The load reduction strategy adopts a fixed threshold trigger and is not dynamically coupled with real-time life indicators, resulting in the coexistence of power generation efficiency loss and mechanical overload risk.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for predicting the power of a vertical axis wind turbine based on multidimensional environmental data, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for predicting the power of a vertical axis wind turbine based on multidimensional environmental data, comprising the following steps:

[0009] Step 1: Acquire real-time multidimensional data of the vertical axis wind turbine through a sensor network, including environmental data, operational data and structural response data. The environmental data includes wind speed, turbulence intensity, air temperature and humidity. The operational data includes rotor speed and blade pitch angle. The structural response data includes blade root strain and vibration acceleration.

[0010] Step 2: Perform mode decomposition on the vibration acceleration signal to extract aerodynamic load components. Based on historical multidimensional data and aerodynamic load components as input, use the corresponding instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient as labels to construct and train a physical information neural network model. Constrain the neural network output by the vertical axis aerodynamic equation residual and dynamic stall hysteresis effect deviation in the loss function.

[0011] Step 3: Based on the real-time output of the physical information neural network model after training, the normal force coefficient and tangential force coefficient are combined with the real-time air density, wind speed, blade azimuth angle and wind turbine swept area to calculate the aerodynamic load time domain signal through the aerodynamic load formula.

[0012] Step 4: Perform online rainflow counting on the aerodynamic load time domain signal, calculate the fatigue damage value in combination with the Miner cumulative damage criterion, and output the remaining life index. When the output remaining life index is less than the set threshold, trigger the load reduction strategy, including adjusting the magnetoresistive torque and flap deflection angle.

[0013] The formula for the loss function is:

[0014] L = ||P pred -P meas ||2+α·||R BEM (C p C L C D )||2+β·||R stall (α eff )||2

[0015] Among them, P pred P is the theoretical power value predicted by the physical information neural network based on the input multidimensional environmental data. meas This represents the real-time output power of the vertical axis wind turbine, directly measured by sensors during actual operation. α and β are weighting coefficients, and R... BEM The BEM equation residuals for the corrected dual-disc multi-flow tube model of a vertical axis wind turbine, R stallFor the residuals of the Leishman-Beddoes dynamic stall model, α eff C is the dynamic angle of attack correction factor. p C L C D These are the instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient, respectively.

[0016] Furthermore, the wind speed is set to V, the turbulence intensity to TI, and the temperature to T in the environmental data. air The humidity is H, the rotor speed in the operating data is ω, the blade pitch angle is θ, the blade root strain in the structural response data is ε, and the vibration acceleration is a. g ;

[0017] According to temperature T air The formula used to calculate real-time air density ρ based on humidity H is:

[0018]

[0019] Among them, P atm For atmospheric pressure, R air is the air gas constant, and e is the partial pressure of water vapor.

[0020] Furthermore, the logic for performing mode decomposition on the vibration acceleration signal to extract the aerodynamic load components is as follows:

[0021] For vibration acceleration a g Empirical mode decomposition was performed to obtain several intrinsic mode functions. Based on the wind turbine rotation frequency, frequency domain energy concentrated in the aerodynamic load-dominated frequency band was selected.

[0022] The formula for calculating the wind turbine rotation frequency is as follows:

[0023]

[0024] In the formula, ω is the rotor speed;

[0025] The selection criteria for intrinsic mode functions are as follows:

[0026] f aero ∈[1f rot ,2f rot ,3f rot ]

[0027] The selected intrinsic mode functions are subjected to fast Fourier transform, and their frequency domain energy spectrum E(f) is calculated as a frequency domain characterization of the aerodynamic load.

[0028] The intrinsic mode function selection criteria for the empirical mode decomposition are: the energy concentration is set to... The resolution of the frequency domain energy spectrum E(f) is less than or equal to 0.5 Hz, and the frequency band f aero The boundary error is less than or equal to 5%.

[0029] Furthermore, based on historical multidimensional data and aerodynamic load components, a physical information neural network model is constructed and trained. The logic for using the corresponding instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient as labels is as follows:

[0030] The historical multidimensional data includes wind speed, turbulence intensity, rotor speed, blade pitch angle, frequency domain energy spectrum, and blade root strain.

[0031] The corresponding instantaneous wind energy utilization coefficient is obtained by back-calculation through measured power. The normal force coefficient and tangential force coefficient are calculated based on aerodynamic load decomposition, and correspond to the components of the aerodynamic force on the blade in the direction perpendicular to the blade's motion and along the motion direction, respectively.

[0032] A physical information neural network architecture is adopted, with historical multidimensional data and aerodynamic load components after modal decomposition as inputs, and the corresponding instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient as outputs. The mean square error between the predicted value and the label is minimized, and the aerodynamic equation and dynamic stall model are forced to be satisfied through a loss function. The predicted instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient are substituted into the BEM equation modified by the dual-disc multi-flow tube model to calculate the theoretical residuals. Based on the Leishman-Beddoes dynamic stall model, the residuals between the aerodynamic coefficient hysteresis effect and the measured values ​​are calculated.

[0033] The predicted power P in the loss function pred The formula is:

[0034]

[0035] Where ρ is the real-time air density, A is the swept area of ​​the wind turbine, V is the real-time wind speed, and C is the real-time air density. p Instantaneous wind energy utilization coefficient;

[0036] Angle of attack corrected based on turbulence intensity TI Where ζ is the turbulence sensitivity coefficient calibrated through wind tunnel experiments, and α geo The geometric angle of attack is calculated from the blade azimuth angle φ and the direction of the incoming flow, where U is the relative wind speed and c is the blade chord length.

[0037] The formula for the turbulence sensitivity coefficient is ζ = 0.25·(TI / 0.15) 0.6 The dynamic angle of attack correction term The relative wind speed of the blades of a vertical axis wind turbine should be calculated based on the azimuth angle: Where R is the radius of the wind turbine;

[0038] The residual term R of the BEM equation corrected by the dual-disc multi-flow pipe model BEM The formula is:

[0039]

[0040] in, Let φ be the blade azimuth angle, and φ be the cumulative angle obtained by integrating the real-time angular velocity ω from the initial time to the current time t. Where N is the real number, c is the blade chord length, R is the rotor radius, a is the axial induction factor, and C is the rotor radius. L C D These are the normal force coefficients and tangential force coefficients output by the physical information neural network;

[0041] The dynamic stall residual term R stall The formula is:

[0042]

[0043] Among them, C L,invi For inviscid lift, ΔC L,vort ΔC represents the lift increment caused by vortex generation. L,sep The lift loss is caused by flow separation. τ1 and τ2 are the time constants of vortex convection and reattachment, which are related to the blade chord length c and the relative wind speed U.

[0044] Furthermore, the normal force coefficient C output by the trained physical information neural network model... L and tangential force coefficient C D Using real-time air density ρ, wind speed V, blade azimuth angle φ, and rotor swept area A, the aerodynamic load time-domain signal F is calculated through the aerodynamic load formula. aero The formula used is:

[0045]

[0046] Among them, F aero As a time-domain signal input to the online rainflow counter;

[0047] For the time-domain signal F of the aerodynamic load aero Perform Kalman filtering with a cutoff frequency set to 3f. rot It retains the dominant aerodynamic frequency band and filters out high-frequency noise.

[0048] Furthermore, the logic for calculating fatigue damage values ​​by performing online rainflow counting on the aerodynamic load time-domain signal and combining it with the Miner cumulative damage criterion is as follows:

[0049] From the time domain signal F aeroExtracting peak-valley sequences, if a segment consisting of four consecutive peak-valley points satisfies the following constraints, then the segment is considered a closed loop, thereby extracting all closed loops in the time-domain signal.

[0050] The constraint is: within this segment, if the value of the previous peak point is not less than the value of the next peak point and the value of the previous valley point is not greater than the value of the next valley point, or if the value of the previous peak point is not greater than the value of the next peak point and the value of the previous valley point is not less than the value of the next valley point, then this region is defined as a closed loop.

[0051] The online rainflow counting is based on a sliding time window, which detects the time-domain signal F in real time. aero The closed loops are selected and the stress amplitude of each closed loop is calculated using the following formula:

[0052]

[0053] In the formula, Δσ i P represents the stress amplitude corresponding to the i-th closed loop. i V represents the highest peak value of the i-th closed loop. u Let F be the lowest valley value of the i-th closed loop, where i is the index of the closed loop and i∈[1,N], and N represents the time-domain signal F. aero The number of closed loops in the middle;

[0054] The fatigue life is calculated by combining the stress amplitude and humidity correction term, using the following formula:

[0055]

[0056] Where, N f,i Let σ be the fatigue life of the i-th closed cycle. fat ,m is the material fatigue parameter, γ H Humidity influence coefficient;

[0057] The formula for the fatigue damage value D is:

[0058]

[0059] Where N is the total number of closed loops.

[0060] Furthermore, the formula used to output the remaining lifetime index is as follows:

[0061] L remain =L design ·(1-D)

[0062] Among them, L design This is the design life, which is the rated operating time;

[0063] When L remain <0.3Ldesign At this time, a load reduction strategy is triggered, including adjusting the reluctance torque. Where k em Here are the reluctance gain coefficients and the flap deflection angle δ = K. p ·D;

[0064] The magnetoresistive gain coefficient k em The dynamic adjustment formula is:

[0065]

[0066] in, This is the decay rate coefficient;

[0067] The control of the flap deflection angle δ adopts a fuzzy PID algorithm, with a proportional coefficient... Where σ crit For the material's yield strength, It is the rated reluctance torque, A ref This represents the cross-sectional area of ​​the blade.

[0068] Compared with the prior art, the beneficial effects of the present invention are:

[0069] This invention proposes a power-lifetime collaborative monitoring and control method for vertical axis wind turbines based on physical-data fusion. This method systematically solves three major technical challenges of inaccurate power prediction, distorted lifetime assessment, and lagging load reduction control in complex environments. First, to address the problem that traditional aerodynamic models ignore the coupling between dynamic stall and multi-dimensional environment, a BEM residual constraint corrected by embedding a physical information neural network into a dual-disc multi-flow tube model is used. Combined with real-time turbulence intensity and humidity data, the angle of attack is dynamically corrected, which reduces the prediction error of the wind energy utilization coefficient under extreme turbulent conditions. At the same time, the accuracy of capturing transient fluctuations in the normal force coefficient is significantly improved, and the stability of power output is significantly optimized.

[0070] This invention addresses the shortcomings of existing life assessments that neglect humidity-accelerated crack propagation. By employing the Miner criterion of online rainflow counting and humidity correction, it calculates the cyclic damage of aerodynamic load time-domain signals in real time. Combined with the humidity influence factor calibrated by salt spray experiments, it reduces fatigue life prediction bias and achieves accurate early warning of life decay. Finally, to address the shortcomings of fixed threshold load reduction strategies, it utilizes multi-stage load reduction control triggered by remaining life to dynamically adjust magnetoresistive torque and flap deflection angle. The closed-loop mechanism of real-time multi-dimensional data acquisition by sensor network → aerodynamic load prediction by physical constraint neural network → adaptive load reduction triggered by dynamic life assessment overcomes the industry challenges of model mismatch, life assessment distortion, and control response lag under environmental coupling. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0072] Figure 2 This is a schematic diagram of the wind speed-theoretical power curve of the present invention;

[0073] Figure 3 This is a schematic diagram of the fatigue damage-remaining life curve of the present invention;

[0074] Figure 4 This is a fitting curve of the remaining lifespan of the vertical axis wind turbine of this invention;

[0075] Figure 5 This is a schematic diagram of the fatigue damage-theoretical power curve of the present invention. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0077] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0078] Example:

[0079] Please see Figures 1-5 The present invention provides a technical solution:

[0080] A method for predicting the power of a vertical axis wind turbine based on multidimensional environmental data, comprising the following steps:

[0081] Step 1: Acquire real-time multidimensional data of the vertical axis wind turbine through a sensor network, including environmental data, operational data and structural response data. The environmental data includes wind speed, turbulence intensity, air temperature and humidity. The operational data includes rotor speed and blade pitch angle. The structural response data includes blade root strain and vibration acceleration.

[0082] The performance and lifespan of vertical axis wind turbines are comprehensively affected by the environment, operating conditions, and structural response. Environmental data quantifies wind resource characteristics and atmospheric conditions to provide input for aerodynamic models. Operating data monitors the real-time control status of the wind turbine and optimizes power output. Structural response data assesses blade fatigue damage, predicts remaining lifespan, and triggers protection strategies.

[0083] The environmental data is set as follows: wind speed V, turbulence intensity TI, and temperature T. air The humidity is set to H, the rotor speed in the operating data is set to ω, the blade pitch angle is set to θ, the blade root strain in the structural response data is set to ε, and the vibration acceleration is set to a. g ;

[0084] ω is the angular velocity of the wind turbine rotation, which determines the tip speed ratio. An increase in rotational speed leads to an increase in the tip speed ratio. The pitch angle θ is the angle between the blade and the plane of rotation, which is used to adjust aerodynamic load and power output. An increase in the pitch angle leads to a decrease in the lift coefficient, which in turn leads to a decrease in power output.

[0085] The strain ε at the blade root is the amount of material deformation at the blade root, directly reflecting aerodynamic loads and structural stresses. Increased strain leads to an increase in the amplitude of alternating stress, which in turn accelerates the accumulation of fatigue damage; vibration acceleration a g It is the blade vibration intensity, used to identify the aerodynamic load components and the structural resonance frequency. An increase in acceleration amplitude leads to an increased risk of dynamic stall or mechanical failure, which requires triggering load reduction control.

[0086] According to temperature T air The formula used to calculate real-time air density ρ based on humidity H is:

[0087]

[0088] Among them, P atm For atmospheric pressure, R air Let be the gas constant for air, and e be the partial pressure of water vapor.

[0089] ρ reflects the mass of air per unit volume and directly affects wind energy conversion efficiency. The higher the density, the greater the kinetic energy captured at the same wind speed. ρ also varies with temperature T. air As humidity increases, the temperature decreases, meaning the gas expands. The temperature decreases slightly as humidity H increases because water vapor displaces dry air.

[0090] Atmospheric pressure P atm This is static atmospheric pressure, which is related to altitude. The standard sea level value is 101.325 kPa. It can be set as a constant in a specific geographical location or measured in real time by a barometric pressure sensor; the air gas constant R... air These are physical properties of dry air, which are only related to the air composition and do not change with the environment;

[0091] Temperature Tair Ambient temperature determines the speed of thermal motion of gas molecules. Increased temperature leads to increased intermolecular distance, which in turn leads to decreased density. Humidity H is the ratio of water vapor content to saturation content in the air, ranging from [0,1]. Increased humidity leads to increased water vapor partial pressure e, which in turn leads to a slight decrease in air density ρ.

[0092] Water vapor partial pressure e is the pressure exerted by water vapor in the air, reflecting the effect of humidity on air density, and is determined by temperature T. air It is determined by both temperature and humidity (H), and the increase in temperature leads to an increase in the saturated vapor pressure exponent, while the increase in humidity causes e to increase linearly.

[0093] This formula converts temperature and humidity into air density, correcting for environmental factors in wind power calculations. The larger the ρ, the denser the air, and the greater the kinetic energy that the wind turbine can capture at the same wind speed, resulting in a higher theoretical power output. The smaller the ρ, the thinner the air, requiring a higher wind speed to reach the rated power, which is common in high-temperature, high-altitude, or high-humidity areas.

[0094] Temperature T air For every 10°C increase in temperature, density decreases by about 3%, significantly affecting power output in low-wind-speed areas; as humidity H increases from 0% to 100%, density decreases by a maximum of 2%, a relatively minor impact that needs to be considered in high-precision models.

[0095] Step 2: Perform mode decomposition on the vibration acceleration signal to extract aerodynamic load components. Based on historical multidimensional data and aerodynamic load components as input, use the corresponding instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient as labels to construct and train a physical information neural network model. Constrain the neural network output by the vertical axis aerodynamic equation residual and dynamic stall hysteresis effect deviation in the loss function.

[0096] The logic for performing modal decomposition on the vibration acceleration signal to extract the aerodynamic load components is as follows:

[0097] For vibration acceleration a g Empirical mode decomposition was performed to obtain several intrinsic mode functions. Based on the wind turbine rotation frequency, frequency domain energy concentrated in the aerodynamic load-dominated frequency band was selected.

[0098] The formula for calculating the wind turbine rotation frequency is as follows:

[0099]

[0100] In the formula, ω is the rotor speed;

[0101] The selection criteria for intrinsic mode functions are as follows:

[0102] f aero ∈[1f rot ,2frot ,3f rot ]

[0103] The selected intrinsic mode functions are subjected to fast Fourier transform, and their frequency domain energy spectrum E(f) is calculated as a frequency domain characterization of the aerodynamic load.

[0104] Vibration acceleration a g It reflects the dynamic response of the blade under aerodynamic loads, mechanical vibrations and noise. It is a single-component signal obtained after IMF (Inherent Mode Function) decomposition. Each IMF represents a vibration mode at a different time scale. The IMF frequencies are arranged from low to high. High-frequency IMFs correspond to noise, and low-frequency IMFs correspond to mechanical or aerodynamic loads.

[0105] f rot Directly determined by the rotor speed ω, it reflects how fast the wind turbine rotates. As the speed ω increases, the rotational frequency f... rot linearly increasing; f aero The aerodynamic load energy is concentrated at the fundamental frequency, second harmonic, and third harmonic of the rotational frequency, and as f... rot Real-time updates, dynamically adapting to current operating conditions; the purpose of this formula is to lock the characteristic frequency of the aerodynamic load and avoid interference from mechanical vibrations, such as bearing failure frequencies. If the energy of a certain IMF is concentrated at f... aero This indicates that it is dominated by aerodynamic loads;

[0106] The purpose of empirical mode decomposition (IMF) on vibration acceleration signals is to separate aerodynamic load components from other disturbances, such as mechanical resonance and noise. The larger the amplitude of the IMF, the greater the contribution of that frequency component to the vibration.

[0107] The intrinsic mode function selection criteria for the empirical mode decomposition are as follows: the energy concentration is defined as... The resolution of the frequency domain energy spectrum E(f) is less than or equal to 0.5 Hz, and the frequency band f aero The boundary error is less than or equal to 5%.

[0108] E(f) characterizes the energy distribution of a signal at different frequencies, representing the energy density corresponding to frequency f. Higher energy indicates a more significant frequency component. E(f) represents the energy density at frequency f. aero The peak value appears at this point, indicating significant aerodynamic load.

[0109] η is the proportion of aerodynamic load frequency band energy to total energy. The larger η is, the higher the proportion of aerodynamic load component in the IMF, ensuring that the screened IMF is dominated by aerodynamic load, i.e., the proportion is greater than or equal to 70%, and eliminating noise interference; if η<0.7, the IMF may contain too much noise or mechanical vibration component and should be discarded; resolution of 0.5Hz distinguishes adjacent harmonics; boundary error of 5% ensures that the frequency band covers the real aerodynamic frequency;

[0110] Based on historical multidimensional data and aerodynamic load components, a physical information neural network model is constructed and trained. The logic for using the corresponding instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient as labels is as follows:

[0111] The historical multidimensional data includes wind speed, turbulence intensity, rotor speed, blade pitch angle, frequency domain energy spectrum, and blade root strain.

[0112] The corresponding instantaneous wind energy utilization coefficient is obtained by back-calculation through measured power, and the calculation formula is as follows: Where ρ is the real-time air density, A is the swept area of ​​the wind turbine, and V is the real-time wind speed;

[0113] The normal force coefficient and tangential force coefficient are calculated based on aerodynamic load decomposition, and correspond to the components of the aerodynamic force on the blade perpendicular to the blade's motion direction and along the motion direction, respectively.

[0114] A physical information neural network architecture is adopted, with historical multidimensional data and aerodynamic load components after modal decomposition as inputs, and the corresponding instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient as outputs. The mean square error between the predicted value and the label is minimized, and the aerodynamic equation and dynamic stall model are forced to be satisfied through a loss function. The predicted instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient are substituted into the BEM equation modified by the dual-disc multi-flow tube model to calculate the theoretical residuals. Based on the Leishman-Beddoes dynamic stall model, the residuals between the aerodynamic coefficient hysteresis effect and the measured values ​​are calculated.

[0115] The formula for the loss function is:

[0116] L = ||P pred -P meas ||2+α·||R BEM (C p C L C D )||2+β·||R stall (α eff )||2

[0117] Among them, P pred P is the theoretical power value predicted by the physical information neural network based on the input multidimensional environmental data. meas This refers to the real-time output power of a vertical axis wind turbine, directly measured by sensors during actual operation. The power is obtained from a wind turbine output power sensor. α and β are weighting coefficients used to balance the importance of data error terms and physical residual terms. R BEM The BEM equation residuals for the corrected dual-disc multi-flow tube model of a vertical axis wind turbine, R stall For the residuals of the Leishman-Beddoes dynamic stall model, α effC is the dynamic angle-of-attack correction factor, reflecting the effect of turbulent transients on aerodynamic loads. p C L C D These are the instantaneous wind energy utilization coefficient, normal force coefficient, and tangential force coefficient, respectively.

[0118] P meas This refers to the real-time output power of a vertical axis wind turbine, directly measured by sensors during actual operation. These sensors are typically installed at the generator output or grid connection point to directly record the instantaneous value of the generated power. pred It is the theoretical power value predicted by the physical information neural network based on the input multidimensional environmental data, such as wind speed, turbulence intensity, and rotational speed.

[0119] Wind energy utilization coefficient C p C reflects the efficiency of converting wind energy into mechanical energy. p The larger the value, the higher the aerodynamic efficiency, and the closer the power output is to the theoretical maximum value; C L C D These are the normal and tangential force coefficients, characterizing the aerodynamic load distribution and affecting blade stress and fatigue life. C... L Increased lift may lead to dynamic stall; C D Increased resistance leads to increased resistance loss.

[0120] R BEM The smaller the residual, the better the neural network prediction conforms to the physical laws; R is used to measure the deviation between the neural network output and the theoretical aerodynamic model. stall The difference between the actual lift and the stall model is quantified; the smaller the residual, the more accurately the dynamic stall effect is modeled.

[0121] This formula improves C by combining data-driven approaches with physical constraints. p C L C D The reliability of the prediction, the data error term P in the loss function pred -P meas The physical residual term R reflects the model's fitting ability. BEM R stall Ensure the prediction conforms to aerodynamic principles;

[0122] The predicted power P in the loss function pred The formula is:

[0123]

[0124] Where ρ is the real-time air density, A is the swept area of ​​the wind turbine, V is the real-time wind speed, and C is the real-time air density. p Instantaneous wind energy utilization coefficient;

[0125] P predThis reflects the theoretical output power of the wind turbine under current operating conditions; the power varies with ρ and V. 3 C p The density increases significantly with increasing density. ρ, calculated from temperature and humidity, affects wind energy capture efficiency; higher density results in a linear increase in power, while power decreases in low-temperature and high-altitude areas. A is the effective windward area of ​​the vertical axis wind turbine, a geometric parameter; a larger area results in a linear increase in power. V is the primary energy source, and its effect on power is cubic; increased wind speed significantly increases power, for example, doubling the wind speed results in an eightfold increase in power. C p It is an aerodynamic efficiency index, with a maximum theoretical value of 0.593 (Bates limit). It is predicted by a neural network and constrained by the aerodynamic model. C p The closer it gets to the Bates limit, the higher the efficiency.

[0126] This formula, based on Bates theory, quantifies the theoretical power output of a vertical axis wind turbine, embedding real-time environmental parameters ρ and V, and aerodynamic efficiency C. p This provides a physical benchmark for power prediction and guides the optimization of wind turbine control strategies;

[0127] Angle of attack corrected based on turbulence intensity TI Where ζ is the turbulence sensitivity coefficient calibrated through wind tunnel experiments, and α geo The geometric angle of attack is calculated from the blade azimuth angle φ and the direction of the incoming flow, where U is the relative wind speed and c is the blade chord length.

[0128] α eff This is the corrected blade angle of attack, reflecting the transient effects of turbulence on aerodynamic loads. An increase in the angle of attack indicates the presence of a stall critical value, and the lift coefficient C... L First increase, then decrease; it is the theoretical angle of attack calculated from the blade azimuth angle φ and the incoming flow direction.

[0129] ζ is the correction magnitude of the calibrated turbulence intensity to the angle of attack; the stronger the turbulence, the more significant the correction. TI is the ratio of the standard deviation of wind speed fluctuation to the mean wind speed, reflecting the degree of turbulence in the flow field. As TI increases, α... eff Increased speed leads to a higher risk of dynamic stall; U is the vector result of the combined incoming airflow velocity and blade velocity. The higher the rotational speed ω, the larger U becomes, resulting in increased α. eff The formula is used to correct for angle-of-attack fluctuations caused by turbulence, thereby improving the prediction accuracy of dynamic stall models.

[0130] The formula for the turbulence sensitivity coefficient is ζ = 0.25·(TI / 0.15) 0.6 The dynamic angle of attack correction term The relative wind speed of the blades of a vertical axis wind turbine should be calculated based on the azimuth angle: Where R is the radius of the wind turbine;

[0131] The vertical axis aerodynamic model is an aerodynamic model specifically designed for the asymmetric flow field characteristics of vertical axis wind turbines. Its core is the blade element momentum theory modified by the dual-disc multi-tube model. Unlike the traditional horizontal axis BEM, the DMS model divides the flow field into upper and lower disk regions and calculates the blade aerodynamic loads at different azimuth angles to more accurately reflect the periodic aerodynamic characteristics of the vertical axis wind turbine.

[0132] The rotor's rotation plane is divided into an upper half (blades moving against the wind) and a lower half (blades moving with the wind), and the induction factor a is calculated for each half. up and a down ;

[0133] The residual term R of the BEM equation corrected by the dual-disc multi-flow pipe model BEM The formula is:

[0134]

[0135] in, Let φ be the blade azimuth angle, and φ be the cumulative angle obtained by integrating the real-time angular velocity ω from the initial time to the current time t. Let N be the real number, c be the blade chord length, R be the rotor radius, and a be the axial induction factor. Through iterative solutions, the momentum conservation and blade force balance in the upper and lower disk regions are ensured. L C D These are the normal force coefficients and tangential force coefficients output by the physical information neural network;

[0136] In the formula, t represents the moment when the data is collected in real time, that is, the time span from the initial moment, such as the system startup moment t=0, to the current data point. In real-time monitoring, t is the timestamp of the sensor data, and the integration interval [0,1] represents the cumulative time from the initial moment to the current moment.

[0137] R BEM It is the C that measures the predictions of neural networks. L C D The smaller the residual deviation from the aerodynamic model, the more the neural network output conforms to the momentum-blade element theory, and the higher the model's reliability. σ is the ratio of the total blade area to the swept area; the higher the realism, the stronger the aerodynamic interference and the larger the axial induction factor b. a is the speed reduction ratio of the airflow passing through the impeller; the larger a is, the lower the flow velocity behind the impeller and the greater the thrust. φ is the rotation angle of the blade relative to the incoming flow direction, which changes periodically with the rotational speed ω, affecting the direction of lift and drag. This formula corrects the neural network output through physical constraints, namely momentum conservation and blade element force balance, to ensure that the prediction conforms to aerodynamic principles.

[0138] Dynamic stall constraints are used to describe the transient aerodynamic effects of airflow separation caused by abrupt changes in angle of attack during the periodic rotation of a vertical axis wind turbine blade. They are constructed using the Leishman-Beddoes dynamic stall model to build residual terms, ensuring the accuracy of the lift coefficient C predicted by the neural network. L Includes the effects of dynamic stall;

[0139] The dynamic stall residual term R stall The formula is:

[0140]

[0141] Among them, C L,invi As inviscid lift, ΔC is calculated using potential flow theory. L,vort ΔC represents the lift increment caused by vortex generation. L,sep The lift loss is caused by flow separation, and τ1 and τ2 are the time constants of vortex convection and reattachment, which are related to the blade chord length c and the wind speed U.

[0142] R stall The smaller the residual, the more accurately the dynamic stall effect is modeled, and the more accurate the C is. L,invi It is the ideal lift calculated using potential flow theory, neglecting viscous effects. It is calculated using the surface element method or vortex lattice method, and depends on α. eff ;ΔC L,vort It is the enhanced lift brought about by the leading-edge vortex during dynamic stall, which is positively correlated with the vortex intensity as measured by wind tunnel transient experiments.

[0143] ΔC L,sep The decrease in lift is caused by airflow separation. The larger the separation area, the more significant the loss. τ1 and τ2 represent the delay effects of vortex convection and reattachment, respectively. τ1 = c / κ1U and τ2 = c / κ2U. κ1 and κ2 are empirical coefficients that distinguish between vortex generation and reattachment processes. This formula captures the unsteady effects of dynamic stall and improves the accuracy of transient load prediction.

[0144] Step 3: Based on the real-time output of the physical information neural network model after training, the normal force coefficient and tangential force coefficient are combined with the real-time air density, wind speed, blade azimuth angle and wind turbine swept area to calculate the aerodynamic load time domain signal through the aerodynamic load formula.

[0145] The normal force coefficient C output by the trained physical information neural network model. L and tangential force coefficient C D Using real-time air density ρ, wind speed V, blade azimuth angle φ, and rotor swept area A, the aerodynamic load time-domain signal F is calculated through the aerodynamic load formula. aero The formula used is:

[0146]

[0147] Among them, F aero As a time-domain signal input to the online rainflow counter;

[0148] For the time-domain signal F of the aerodynamic load aero Perform Kalman filtering with a cutoff frequency set to 3f. rot Preserve the dominant aerodynamic frequency band and filter out high-frequency noise;

[0149] F aero The aerodynamic load reflects the aerodynamic forces experienced by the blades during rotation, directly affecting structural stress and fatigue life. It dynamically changes with air density, wind speed, lift and drag coefficients, and azimuth angle. The larger the density ρ, the linearly the aerodynamic load increases, while the load decreases in low-temperature and high-altitude areas. The larger the area A, the linearly the aerodynamic load increases. When the wind speed V doubles, the aerodynamic load increases to 4 times.

[0150] C L C is the lift coefficient, predicted by a neural network, characterizing the aerodynamic efficiency of the blade in the normal direction (perpendicular to the incoming flow direction). L Increased lift becomes dominant, especially with increased aerodynamic loads when φ≈90°; C D C is the drag coefficient, predicted by a neural network, characterizing the aerodynamic drag of the blade in the tangential (parallel to the incoming flow direction). D As the azimuth angle φ increases, drag offsets lift, especially when φ≈0°, the aerodynamic load decreases; changes in azimuth angle φ cause sinφ and cosφ to periodically modulate the contribution ratio of lift to drag.

[0151] C represents the flow pressure, which is the energy source of aerodynamic forces. L sinφ is the component of lift perpendicular to the plane of the wind turbine, C D cosφ is the component of the drag parallel to the wind turbine plane; F aero It is the vector sum of the normal and tangential forces, reflecting the periodic alternating loads experienced by the blade during rotation;

[0152] When φ = 90°, the blades are perpendicular to the incoming flow, sinφ = 1, lift is dominant, F aero Significantly increased; when φ=0°, the blades are parallel to the incoming flow, cosφ=1, drag counteracts lift, F aero Decrease; for each rotation of the blades of a vertical axis wind turbine, φ changes periodically from 0 to 2π, F aero It exhibits sinusoidal fluctuations with a frequency equal to the wind turbine's rotational frequency f. rot And its harmonics.

[0153] Step 4: Perform online rainflow counting on the aerodynamic load time domain signal, calculate the fatigue damage value in combination with the Miner cumulative damage criterion, and output the remaining life index. When the output remaining life index is less than the set threshold, trigger the load reduction strategy, including adjusting the magnetoresistive torque and flap deflection angle.

[0154] The logic for calculating fatigue damage values ​​by performing online rainflow counting on the time-domain signal of the aerodynamic load components and combining it with the Miner cumulative damage criterion is as follows:

[0155] The logic for calculating fatigue damage values ​​by performing online rainflow counting on the aerodynamic load time-domain signal and combining it with the Miner cumulative damage criterion is as follows:

[0156] From the time domain signal F aero Extracting peak-valley sequences, if a segment consisting of four consecutive peak-valley points satisfies the following constraints, then the segment is considered a closed loop, thereby extracting all closed loops in the time-domain signal.

[0157] The constraint is: within this segment, if the value of the previous peak point is not less than the value of the next peak point and the value of the previous valley point is not greater than the value of the next valley point, or if the value of the previous peak point is not greater than the value of the next peak point and the value of the previous valley point is not less than the value of the next valley point, then this region is defined as a closed loop.

[0158] The online rainflow counting is based on a sliding time window, which detects the time-domain signal F in real time. aero The closed loops are selected and the stress amplitude of each closed loop is calculated using the following formula:

[0159]

[0160] In the formula, Δσ i P represents the stress amplitude corresponding to the i-th closed loop. i V represents the highest peak value of the i-th closed loop. i Let F be the lowest valley value of the i-th closed loop, where i is the index of the closed loop and i∈[1,N], and N represents the time-domain signal F. aero The number of closed loops in the middle;

[0161] The fatigue life is calculated by combining the stress amplitude and humidity correction term, using the following formula:

[0162]

[0163] Where, σ fat m is the material fatigue parameter, SN curve parameter, and N f,i For the fatigue life of the i-th closed cycle, γ H For humidity influence coefficient; N f,i The smaller the value, the higher the damage rate of that stress amplitude;

[0164] Δσ i It is the stress variation amplitude caused by alternating load, Δσ i The larger N is f,i The smaller the value, the higher the damage rate; σ fat This is the reference stress value of the SN curve, corresponding to N. f,i =1, the stress amplitude is determined by material fatigue experiments and is independent of humidity; m is the slope of the SN curve, which reflects the sensitivity of stress amplitude to lifespan. The larger m is, the more significant the effect of stress amplitude on lifespan.

[0165] γ H The accelerating effect of humidity on fatigue life was quantified; the higher the humidity (H), the greater the effect of humidity (N). f,i The greater the reduction, the more accurate the measurement. This is determined through salt spray testing. The experimental method includes: applying alternating loads within a humidity range of [30%, 90%] and recording the crack propagation rate; and calibrating γ through experiments. H At that time, ensure γ H ≤1 / H max , such as H max =0.9, γ H ≤1.11;

[0166] The formula for the fatigue damage value D is:

[0167]

[0168] Where N is the total number of closed loops;

[0169] D reflects the degree of fatigue damage accumulation caused by alternating stress in the material. D∈[0,1], the closer it is to 1, the shorter the remaining life. D=1 indicates failure.

[0170] The formula used to output the remaining lifetime index is:

[0171] L remain =L design ·(1-D)

[0172] Among them, L design Design life, i.e., rated operating time;

[0173] L remain It is the remaining safe operating time predicted based on the current damage value. The larger D is, the longer L is. remain The smaller the value, the more urgent the lifespan warning; L design It is the expected service life under rated operating conditions, such as 20 years, which is determined by the manufacturer based on the materials and operating conditions.

[0174] When L remain <0.3L design At this time, a load reduction strategy is triggered, including adjusting the reluctance torque. Where k em Here are the reluctance gain coefficients and the flap deflection angle δ = K. p ·D;

[0175] The magnetoresistive gain coefficient k em The dynamic adjustment formula is:

[0176]

[0177] in, The decay rate coefficient is optimized using historical fault data; ζ controls the sensitivity of the control gain to changes in remaining lifetime, also optimized using historical fault data. The larger the value, the more sensitive the gain is to changes in lifetime; 0.2 is used to adjust the critical point for triggering load shedding. remain / L design When the gain is less than 0.3, the gain decreases rapidly.

[0178] The control of the flap deflection angle δ adopts a fuzzy PID algorithm, with a proportional coefficient... Where σ crit For the material's yield strength, It is the rated reluctance torque, A ref This refers to the cross-sectional area of ​​the blade.

[0179] σ crit This refers to the ultimate tensile / compressive strength of the blade material, which is determined by material testing. The higher the strength, the greater the allowable flap deflection. It is the maximum output torque designed for the generator. The larger the torque, the higher the proportional coefficient K. p The smaller the size, the smoother the control; A ref It is the blade cross-sectional area, used as a dimensionless scaling factor to ensure K p Dimensionless to avoid unit conflicts.

[0180] Table 1 Statistical Table of Factors Affecting Power

[0181]

[0182]

[0183] During the operation of wind turbines, acquiring environmental data is fundamental. Table 1 shows 30 sets of data under different operating conditions, with wind speeds ranging from 5 m / s to 34 m / s and turbulence intensities ranging from 0.1 to 0.2. These variations reflect the working environment of wind turbines under different meteorological conditions. For example, in scenario 1, the wind speed is 5 m / s and the turbulence intensity is 0.1, while in scenario 30, the wind speed reaches 34 m / s and the turbulence intensity is 0.2. This diverse environmental data provides a basis for model training and validation, ensuring its adaptability and accuracy under various operating conditions.

[0184] Calculating air density is a crucial step in the model. Table 1 lists the air density for each data set, ranging from 1.171 kg / m³. 3 Up to 1.204 kg / m 3 These data were calculated based on temperature and humidity, and meet the requirements of the technical solution. Air density directly affects the prediction of theoretical power. In each scenario, the theoretical power calculation is based on current environmental data and the wind turbine's design parameters. For example, in scenario 1, the wind speed is 5 m / s and the air density is 1.204 kg / m³. 3 Assuming the wind turbine area is 10m² 2 and instantaneous wind energy utilization coefficient C p With a value of 0.3, the theoretical power can be derived, ensuring the accuracy of the model under different wind speeds and environmental conditions. Correspondingly, as the wind speed increases, the theoretical power reaches 67.5kW in scenario 30, showing a clear upward trend, which demonstrates the direct impact of wind speed on power output.

[0185] The actual power in Table 1 serves as the validation standard for the model output. By comparing the theoretical power and the actual power, the predictive ability of the model can be evaluated. For example, in scenario 2, the theoretical power is 2.0 kW, and the actual power is 1.85 kW. The relative closeness between the two indicates that the model performs well in this condition. As wind speed and turbulence intensity change, the actual power also changes, demonstrating the model's effectiveness in dynamic environments.

[0186] Fatigue damage and remaining life are important indicators for the safe operation of wind turbines. Fatigue damage is calculated based on the stress amplitude under actual operating conditions and evaluated in conjunction with the design life. For example, in scenario 10, the fatigue damage is 0.23, and the remaining life is 830 hours. This data can help maintenance personnel take timely measures to avoid failures caused by excessive equipment fatigue.

[0187] The data provided in Table 1, through the acquisition of multi-dimensional environmental parameters, calculation of air density, comparison of theoretical and actual power, and assessment of fatigue damage and remaining life, comprehensively demonstrate the practicality and effectiveness of this technical solution. This data not only provides a reliable basis for model training and validation but also promotes the performance optimization and safe operation of wind turbines, ensuring the sustainable utilization of wind energy.

[0188] Table 2. Statistical Table of Theoretical Power and Remaining Life

[0189]

[0190]

[0191] In each scenario in Table 2, the comparison between theoretical and actual power reveals the performance of the wind turbine in actual operation. For example, in scenario 1, the theoretical power is 1.5kW, while the actual power is 1.4kW, a small difference, indicating that the model's prediction effect is good under this condition. As wind speed and rotational speed increase, the theoretical power also rises. In scenario 30, the theoretical power reaches 67.5kW, while the actual power is 64.5kW, a difference still within an acceptable range (only 3kW), demonstrating the model's rationality and accuracy under high loads. This accuracy reflects the model's effective predictive ability for wind turbine performance, ensuring the effective utilization of wind energy under different environmental conditions. This result is consistent with the concept of power prediction through multi-dimensional environmental data emphasized in the technical solution.

[0192] Fatigue damage value is a key indicator for assessing the operational reliability of wind turbines. The table shows that the fatigue damage value gradually increases from 0.1 in scenario 1 to 0.64 in scenario 30. This indicates that as power output increases, the workload of the wind turbine also increases, leading to progressively more severe fatigue damage. In scenario 10, the fatigue damage value reaches 0.23, which may indicate the need for inspection and maintenance to avoid potential failures.

[0193] Remaining life is a crucial indicator for assessing the usable time of a wind turbine under current operating conditions. In the table, the remaining life gradually decreases from 990 hours to 430 hours, a process closely related to increased fatigue damage. Lower fatigue damage results in a relatively longer remaining life, indicating good equipment operation. For example, scenario 12 shows a remaining life of 790 hours, demonstrating ample usability under these environmental conditions. Conversely, in scenario 30, the remaining life drops to 430 hours, suggesting more frequent maintenance and attention. This approach not only reduces the risk of equipment failure but also effectively extends the overall service life of the wind turbine, meeting the reliability and maintenance optimization requirements of the technical solution.

[0194] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0195] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0196] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0197] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for power prediction of a vertical axis wind turbine based on multi-dimensional environmental data, characterized in that, The specific steps include: Step 1: Obtain real-time multi-dimensional data of the vertical axis wind turbine through the sensor network, including environmental data, operating data and structural response data, the environmental data including wind speed, turbulence intensity, air temperature and humidity, the operating data including rotor speed and pitch angle, and the structural response data including blade root strain and vibration acceleration; Step 2: Perform modal decomposition on the vibration acceleration signal to extract the aerodynamic load component, based on historical multi-dimensional data and the aerodynamic load component as input, corresponding instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient as label, construct and train a physical information neural network model, and constrain the neural network output through the residual error of the vertical axis aerodynamic equation and the dynamic stall lag effect bias in the loss function; Step 3: Output real-time normal force coefficient and tangential force coefficient based on the trained physical information neural network model, combine real-time air density, wind speed, blade azimuth angle and wind wheel swept area, and calculate the aerodynamic load time domain signal through the aerodynamic load formula; Step 4: Perform online rainflow counting on the aerodynamic load time domain signal, calculate the fatigue damage value based on the Miner cumulative damage criterion, and output the remaining life index, when the output remaining life index is less than the set threshold, trigger the load shedding strategy, including adjusting the magnetic resistance torque and the flap deflection angle; The formula of the loss function is: wherein, is the theoretical power value predicted by the physical information neural network according to the input multi-dimensional environment data, is the real-time output power directly measured by the sensor in the actual operation of the vertical axis wind turbine, are weight coefficients, is the BEM equation residual term corrected by the double-disk multi-stream tube model of the vertical axis wind turbine, is the Leishman-Beddoes dynamic stall model residual term, is the dynamic angle of attack correction factor, are the instantaneous wind energy utilization coefficient, the normal force coefficient and the tangential force coefficient, respectively; The normal force coefficient output by the trained physical information neural network model and tangential force coefficient With real-time air density Wind speed Blade azimuth angle Wind turbine sweep area The time-domain signal of the aerodynamic load is calculated using the aerodynamic load formula. The formula used is: wherein, as a time domain signal to an online rainflow counter; Time domain signal of aerodynamic load Kalman filter is used with cutoff frequency set as , which retains the dominant frequency band of aerodynamic force and filters out high frequency noise.

2. A method for power prediction of a vertical axis wind turbine based on multi-dimensional environmental data according to claim 1, characterized in that: The wind speed in the setting environment data is set as , the turbulence intensity is , the air temperature is , the humidity is , the rotor rotating speed in the running data is , the pitch angle is , the blade root strain in the structural response data is , and the vibration acceleration is ; According to temperature and humidity Calculate real-time air density The formula used is: wherein is the atmospheric pressure, is the air gas constant, is the water vapor partial pressure.

3. A method of power prediction for a vertical axis wind turbine based on multi-dimensional environmental data according to claim 2, characterized in that: The logic of modal decomposition on the vibration acceleration signal to extract the aerodynamic load component is: Vibration acceleration The empirical mode decomposition is performed to obtain a plurality of intrinsic mode functions, and based on a rotating frequency of the wind wheel, a frequency domain energy concentrated in a dominant frequency band of the aerodynamic load is screened out. The calculation formula of the wind wheel rotation frequency is: In the formula, is the rotor speed; The selection rule of the inherent modal function is: Perform a Fast Fourier Transform on the selected intrinsic mode functions and calculate their frequency domain energy spectrum. , as a frequency domain characterization of aerodynamic load; The intrinsic mode function screening condition of the empirical mode decomposition is that the energy concentration degree is set as , the resolution of the frequency domain energy spectrum is less than or equal to 0.5 Hz, and the boundary error of the frequency band is less than or equal to 5%.

4. A method of power prediction for a vertical axis wind turbine based on multi-dimensional environmental data according to claim 3, characterized in that: The logic of constructing and training a physical information neural network model based on historical multi-dimensional data and aerodynamic load components, and taking corresponding instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient as label is: The historical multi-dimensional data includes wind speed, turbulence intensity, rotor speed, pitch angle, frequency energy spectrum and blade root strain; The corresponding instantaneous wind energy utilization coefficient is obtained by back calculation of the measured power, and the normal force coefficient and the tangential force coefficient are calculated based on the aerodynamic load decomposition, corresponding to the components of the aerodynamic force on the blade perpendicular to the blade movement direction and along the movement direction respectively; The physical information neural network architecture is adopted, the input is historical multi-dimensional data and modal decomposed aerodynamic load component, and the output is corresponding instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient; Minimize the mean square error of the predicted value and the label, and force the aerodynamic equation and the dynamic stall model to satisfy through the loss function, substitute the predicted instantaneous wind energy utilization coefficient, normal force coefficient and tangential force coefficient into the BEM equation corrected by the double disc multi-flow pipe model, calculate the theoretical residual error, and calculate the residual error of the aerodynamic coefficient lag effect based on the Leishman-Beddoes dynamic stall model; the loss function in which the predicted power is given by the formula: wherein, is the real-time air density, is the wind wheel swept area, is the real-time wind speed, is the instantaneous wind energy utilization coefficient; based on the turbulence intensity corrected angle of attack wherein is a turbulence sensitivity factor calibrated by wind tunnel experiments, is the geometric angle of attack, which is calculated from the blade azimuth angle and the incoming flow direction, is the relative wind speed, is the blade chord length; The formula of the turbulence sensitivity coefficient is ; the dynamic angle of attack correction term The relative wind speed of the vertical axis wind turbine blade should be calculated based on the azimuth angle: , wherein is the wind wheel radius; The double disk multi-stream tube model modified BEM equation residual term The formula is: wherein, is a blade azimuth angle, is a cumulative angle obtained by integrating the real-time angular velocity from the initial moment to the current moment is a cumulative angle obtained by integrating the real-time angular velocity is a real degree, is a number of blades, is a blade chord length, is a wind wheel radius, is an axial induction factor, is a normal force coefficient and a tangential force coefficient output by the physical information neural network; The dynamic stall residual term The formula is: wherein, is the no-stick lift, is the lift increment due to vortex generation, is the lift loss due to flow separation, is the vortex convection and reattachment time constant, related to the blade chord length and the relative wind speed .

5. A method of power prediction for a vertical axis wind turbine based on multi-dimensional environmental data according to claim 4, characterized in that: The logic of performing online rainflow counting on the aerodynamic load time domain signal, and calculating the fatigue damage value based on the Miner cumulative damage criterion is: The peak-valley sequence is extracted from the time domain signal If a section composed of four continuous peak-valley points satisfies the following constraint condition, the section is regarded as a closed cycle, and all the closed cycles in the time domain signal are extracted in this way. The constraint condition is that, in the section, if the previous peak point value is not less than the next peak point value, and the previous valley point value is not greater than the next valley point value, or the previous peak point value is not greater than the next peak point value, and the previous valley point value is not less than the next valley point value, the region is defined as a closed loop; The online rainflow counting is based on a sliding time window, and detects a time domain signal in real time , and a stress amplitude of each closed cycle is calculated according to the following formula: wherein is the stress amplitude of the nth closed cycle, is the highest peak value of the nth closed cycle, is the lowest valley value of the nth closed cycle, is the lowest valley value of the nth closed cycle, is the lowest valley value of the nth closed cycle, is the lowest valley value of the nth closed cycle, is the index of the closed cycle, and , denotes the number of closed cycles in the time domain signal . The fatigue life is calculated in combination with the stress amplitude and the humidity correction term, and the formula is: wherein, is the fatigue life of the ith closed cycle, is the material fatigue parameter, is the humidity influence coefficient; The fatigue damage value The formula is: wherein is the total number of closed cycles.

6. A method of power prediction for a vertical axis wind turbine based on multi-dimensional environmental data according to claim 5, characterized in that: The formula for outputting the residual life index is: wherein is the design life, i.e. the rated operating time; When a load shedding strategy is triggered, including adjusting the reluctance torque wherein is a reluctance gain coefficient, and the flap deflection angle ; The magnetic reluctance gain coefficient The dynamic adjustment formula is: wherein is the decay rate coefficient; The flap deflection angle The control uses a fuzzy PID algorithm, the proportional coefficient wherein is the material yield strength, is the rated reluctance torque, is the blade cross-sectional area.

Images