Low-wind-speed fan and energy storage collaborative power supply system and method

Abstract

The application discloses a low-wind-speed fan and energy storage collaborative power supply system and method, and belongs to the technical field of wind power generation and energy storage. The system comprises a pole-tower co-construction integrated architecture, a wind power generation module, an energy storage module, a power conversion module, an automatic switching module and an edge intelligent control terminal. The edge intelligent control terminal cooperates with a cloud platform to execute normal power supply mode, energy supplement power supply mode, emergency power supply mode and fault protection mode. The edge intelligent control terminal adopts a bus voltage nonlinear control strategy based on Lyapunov stability, an extended Kalman filter state of charge online estimation and an adaptive step maximum power point tracking algorithm. The cloud platform runs a long short-term memory network wind speed prediction model and dynamically adjusts the lower limit of the reserved state of charge of the energy storage module. The application realizes efficient capture of wind energy in a low-wind-speed section and effectively improves the power supply reliability of key loads in remote areas.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation and energy storage technology, specifically relating to a low-wind-speed wind turbine and energy storage coordinated power supply system and method. Background Technology

[0002] As wind power development gradually extends from high-wind-speed resource areas to low-wind-speed and ultra-low-wind-speed areas, how to efficiently utilize wind energy resources in low-wind-speed areas and ensure the reliability of power supply to end loads has become a research hotspot in the field of new energy power systems.

[0003] Patent CN121307814A proposes an integrated wind-solar-storage DC microgrid device. By directly integrating a vertical-axis wind turbine with a DC bus, it effectively avoids energy losses caused by multiple AC / DC conversions and, combined with an energy storage module, achieves off-grid power supply for agricultural distributed applications. Within conventional wind speed ranges, this solution effectively utilizes wind energy for peak-valley arbitrage and load regulation, demonstrating the technological advantages of DC microgrids in reducing energy link losses. Patent CN121124130A targets high-power applications such as offshore wind power for hydrogen production. By introducing a grid-type wind turbine and energy storage system in synergy, supplemented by a diesel generator set as the final underlying guarantee, it constructs a highly reliable power supply architecture with multiple redundancy characteristics. These solutions demonstrate excellent system anti-disturbance capabilities in dealing with high rated load fluctuations and extreme sea conditions. Through the physical coupling of traditional and new energy sources, they significantly improve the system's operational continuity.

[0004] Most existing control logics are based on conventional wind speed distribution or high-power industrial load models. When faced with the dynamic response characteristics of wind turbines frequently operating at the edge of cutting-off wind speeds in low-wind-speed environments, they often lack sufficient strategy precision. Under extremely low wind speed conditions, the output power of wind turbines is not only weak but also exhibits extremely strong nonlinear fluctuations. At this point, the coupling relationship between the wind turbine and the energy storage system is no longer a simple power smoothing process, but involves electromagnetic support during turbine startup, optimal energy capture during weak power periods, and refined matching with constant load demand. Existing energy storage scheduling technologies are mostly "passive response" types, intervening only when power is insufficient. This can easily lead to deep discharge of the energy storage battery or even system failure when low wind speeds prevent the wind turbine from outputting effective power for extended periods.

[0005] For applications like communication base stations with constant load characteristics and limited space, existing split-type equipment deployments and extensive energy dispatching models have shown serious inadequacy. While the reliance on diesel generators in traditional solutions solves emergency power supply problems, it brings high operation and maintenance costs and carbon emission pressures, contradicting the original intention of green development. Furthermore, in a pure wind-storage coupling framework, how to achieve millisecond-level seamless switching between wind, storage, load, and even surplus grid power through edge intelligent control without external stable power support, and how to possess autonomous path optimization capabilities in the event of a fault, are challenges that current technical solutions cannot provide complete logical support on an integrated hardware platform. The utilization rate of low-wind-speed energy remains low, and the system's power supply stability struggles to meet telecom-grade standards. Summary of the Invention

[0006] The purpose of this invention is to provide a low-wind-speed wind turbine and energy storage coordinated power supply system and method, which solves the technical problems of low energy capture efficiency, insufficient system redundancy, and poor power supply stability of unattended sites in remote areas under low wind speed conditions.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for coordinating power supply from low-wind-speed wind turbines and energy storage, the method being executed by an edge intelligent control terminal, includes the following steps: Step 1: Collect data on the power generation of the wind power generation module, the state of charge of the energy storage module, the real-time power of the load, and the voltage and frequency of the mains power. Step 2: Determine the mains power status based on the data collected in Step 1. When the mains power is normal and the power generation of the wind power generation module is greater than or equal to the load power, control the wind power generation module to supply power to the load and charge the excess power into the energy storage module. Step 3: When the mains power is normal and the power output of the wind power generation module is less than the load power, control the wind power generation module to operate in maximum power point tracking mode. The insufficient load power is supplemented by the mains power, while the energy storage module is kept in float charging state. Step 4: When the mains power is abnormal, the load will be switched to an independent power supply path consisting of the wind power generation module and the energy storage module within 20 ms. Step 5: When an electrical fault is detected, isolate the fault source and switch to the backup power supply path within 2 ms.

[0008] Furthermore, in step two, the excess electrical energy is charged into the energy storage module using a constant current-constant voltage segmented strategy, with the charging cutoff condition being that the state of charge reaches the upper limit threshold of 0.90 to 0.95.

[0009] Furthermore, the mains power abnormality mentioned in step four is defined as any of the following situations: (a) the mains voltage is lower than 70% of the rated value and the duration of this voltage state exceeds 200ms; (b) the mains voltage is higher than 120% of the rated value and the duration of this voltage state exceeds 200ms; (c) the mains frequency deviation exceeds ±2Hz.

[0010] Furthermore, in step four, when the state of charge drops to the warning threshold of 0.20, non-critical loads are cut off step by step according to preset priority. The non-critical loads include heaters, monitoring auxiliary equipment, and secondary frequency bands of radio frequency units.

[0011] Furthermore, the maximum power point tracking employs an adaptive step-size perturbation observation method, with an adaptive step size... Determined by the following formula: ; in This represents the proportionality coefficient, with a value ranging from 0.5 to 2.0. This represents the change in power, expressed in watts (W). This indicates the rated power of the wind power generation module, in watts (W). This represents the baseline duty cycle step size, ranging from 0.005 to 0.02, and is dimensionless.

[0012] Furthermore, the state of charge is estimated online using an extended Kalman filter algorithm based on a first-order resistor-capacitor equivalent circuit, and the state equation is: ; The observation equation is: ; in: Indicates time The state of charge; Indicates time The polarization voltage, in V; This indicates the sampling period, expressed in seconds, with a value ranging from 0.01 s to 0.1 s. This represents the polarization time constant, in seconds. Indicates the charge / discharge efficiency coefficient; This indicates the battery's nominal capacity, measured in Ah. Indicates time The battery circuit current, in amperes (A); This represents polarization resistance, measured in Ω. Represents the process noise vector; Indicates time The battery terminal voltage, in V; A nonlinear function representing the relationship between open-circuit voltage and state of charge; This represents the internal resistance in ohms, measured in Ω. Indicates time The battery circuit current, in amperes (A); Indicates measurement noise; The process noise corresponding to the charged state dimension at time k-1 is... The process noise corresponds to the polarization voltage dimension at time k-1; the two together form a two-dimensional process noise vector. The whole follows a Gaussian distribution with zero mean.

[0013] Furthermore, the electrical faults mentioned in step five include overvoltage, overcurrent, short circuit, and insulation degradation. These are detected by a fault diagnosis module based on Fast Fourier Transform. When the total harmonic distortion rate is greater than 10% and the amplitude of the third harmonic exceeds 5% of the fundamental amplitude, and the amplitude of the fifth harmonic exceeds 3% of the fundamental amplitude, it is determined to be an insulation micro-arc flash fault. When the amplitude of high-frequency components above 1 kHz suddenly increases to more than 10 times the normal value, it is determined to be a switch short circuit fault.

[0014] Furthermore, the present invention also includes a bus voltage control step based on Lyapunov stability: calculating the bus voltage error. ,in This is the reference value for bus voltage. The real-time bus voltage is used as the basis for calculating the reference current of the bidirectional converter. ,in For load current, Inject bus current into the wind power generation module Bus capacitance, in F. The convergence coefficient is expressed in s. - ¹ Finally, adjust the duty cycle of the bidirectional converter to make the actual current track the signal. .

[0015] Furthermore, the method also includes a dynamic load sensing and instantaneous power compensation step: real-time monitoring of the instantaneous load power. and its rate of change When detected At W / ms, the output compensation power is within 50μs. ,in This indicates the load power deviation, in watts (W). This represents the proportionality coefficient, ranging from 0.8 to 1.0. This represents the differential coefficient, ranging from 0.01 to 0.05.

[0016] In addition, this invention also discloses a low-wind-speed wind turbine and energy storage coordinated power supply system, comprising: The integrated architecture for co-construction of poles and towers adopts precast integrated poles and towers made of ultra-high performance concrete. The compressive strength of the ultra-high performance concrete is not less than 120 MPa, and it incorporates steel fibers with a volume percentage of 2.0% to 3.5%. A wind power generation module is installed on the top of the integrated tower, and the cut-in wind speed of the wind power generation module is 2.0 m / s to 2.5 m / s; The energy storage module, composed of lithium iron phosphate battery cells, is installed at the bottom of the integrated tower or in its internal mounting compartment. The power conversion module includes a three-phase pulse width modulation rectifier, a bidirectional DC / DC converter, and a bidirectional DC / AC inverter; The automatic switching module uses a high-speed electromagnetic drive automatic transfer switch, with a mechanical switching time of less than or equal to 20 ms; An edge intelligent control terminal is integrated into the control compartment of the integrated tower. The edge intelligent control terminal stores and runs program instructions, which are used to execute the low wind speed wind turbine and energy storage coordinated power supply method according to any one of claims 1 to 9.

[0017] Furthermore, the system also includes a cloud platform that runs a Long Short-Term Memory (LSTM) network wind speed prediction model. The model's inputs are the wind speed sequence, air pressure, ambient temperature, and relative humidity from the past 6 hours, and its output is the wind speed sequence for the next 24 hours, with a time resolution of 1 hour. The edge intelligent control terminal dynamically adjusts the reserved charge state limit of the energy storage module based on the predicted wind speed. The adjustment rules are as follows: If the expected cumulative power generation Greater than or equal to total load demand ,but ; like ,but ; like ,but .

[0018] Furthermore, the control chamber is equipped with a hybrid passive and active thermal management system. The outer shell of the control chamber is made of corrugated aluminum alloy heat sinks and filled with thermally conductive phase change material with a melting point of 45 ℃. When the internal temperature of the control chamber exceeds 45 ℃, the edge intelligent control terminal starts the axial flow fan, and the fan speed is adjusted by the proportional-integral-derivative controller according to the temperature difference.

[0019] Furthermore, in the fault protection mode, when switching to the backup power supply path after isolating the fault source, an active damping smooth switching method is adopted: first, the difference between the output voltage of the backup power conversion module and the bus voltage is adjusted to be less than 1 V, and then the inrush current is suppressed by a virtual resistor. The initial value of the virtual resistor is 0.5 Ω, which linearly decreases to 0 Ω within 20 ms.

[0020] Furthermore, the edge intelligent control terminal also runs a thermal management reinforcement learning algorithm based on a deep Q-network, in a state space... and reward function Optimize the cooling fan speed, among which Indicates the temperature inside the tower. Indicates the temperature outside the tower. Indicates the average battery temperature. Indicates the battery's heating power. Indicates the fan speed. and These are the weighting coefficients. This indicates the real-time electrical power of the fan. This indicates the rated power of the fan.

[0021] Furthermore, the rated power of the wind power generation module is configured with redundancy of 1.5 to 3.0 times the rated power of the load, and the total energy capacity of the energy storage module is configured to maintain the critical load for no less than 24 hours under windless and grid power-free conditions. The critical load refers to the core communication equipment of the communication base station and the necessary sensors of the environmental monitoring station.

[0022] Furthermore, the cloud platform also runs a quantile regression forest model, outputting the 10th, 50th, and 90th quantile prediction curves for wind speed over the next 24 hours; the edge intelligent control terminal calculates the reserved charge state limit based on the quantile prediction curves. Specifically: First calculate the expected cumulative power generation corresponding to the 10th percentile. Then calculate the required energy storage reserve capacity. Then let ,in This represents the total demand for critical loads over 24 hours, in kWh. This indicates the total energy of the battery, expressed in kWh.

[0023] Furthermore, the edge intelligent control terminal incorporates a built-in charge-discharge optimization strategy for sensing energy storage lifespan degradation, based on an aging model: ; The equivalent cyclic aging rate is calculated, and rolling optimization is performed under emergency power supply mode with the objective of minimizing the weighted sum of load shedding power and aging cost. This indicates the current battery capacity, in Ah. Indicates time, in seconds; This represents the cyclic aging rate constant under reference operating conditions, in seconds. - ¹; This represents the absolute value of the battery circuit current, in amperes (A). This represents the reference current, expressed in amperes (A). This represents the current acceleration factor, which is dimensionless. The activation energy is expressed in J / mol. Let J represent the ideal gas constant, taken as 8.314 J / (mol·K); This indicates the reference temperature, expressed in Kelvin (K). This indicates the actual temperature of the battery, in Kelvin (K). It represents the magnitude of the change in state of charge during one charge-discharge cycle, and is dimensionless. Indicates reference depth, dimensionless; This represents the depth acceleration factor, which is dimensionless.

[0024] Compared with the prior art, the present invention has the following beneficial effects: This invention integrates wind power generation modules, energy storage modules, power conversion modules, and edge intelligent control terminals into the internal or external mounting points of ultra-high performance concrete precast towers through a unified tower co-construction architecture. This significantly reduces the system's footprint and cable loss, improves structural strength and wind vibration resistance, and utilizes hollow cavities for natural heat dissipation and cable routing, significantly enhancing the equipment protection level and operational reliability in remote, unattended scenarios.

[0025] A two-tiered control architecture, combining edge intelligent control terminals and a cloud platform, is adopted. The edge intelligent control terminals independently execute millisecond-level real-time control tasks, while the cloud platform is responsible for long-cycle wind speed prediction and global optimization scheduling. This ensures both the speed of local control and the dynamic state-of-charge adjustment based on wind speed prediction using a long short-term memory network. This allows the energy storage module to reserve sufficient capacity in advance during low-wind-speed periods, effectively preventing power outages caused by over-discharge of stored energy. The system is designed with four operating modes: normal power supply, supplementary power supply, emergency power supply, and fault protection. Combined with a high-speed electromagnetic automatic switching module, it achieves seamless switching in less than 20 milliseconds when the mains power is abnormal. When an electrical fault is detected, it isolates the fault source within 2 milliseconds and automatically switches to the backup power path, greatly improving the system's power supply continuity and meeting the telecom-grade reliability requirements of critical loads such as communication base stations.

[0026] Simultaneously, a nonlinear control strategy for the bus voltage based on Lyapunov stability is introduced, combined with online state-of-charge estimation using extended Kalman filtering and an adaptive step-size maximum power point tracking algorithm. This achieves rapid, overshoot-free convergence of the DC bus voltage under low wind speed fluctuation conditions, improving the capture efficiency of weak wind energy while suppressing voltage drops caused by load abrupt changes. An integrated fault diagnosis module based on Fast Fourier Transform, a dynamic load sensing and compensation module, and a hybrid thermal management system are implemented to accurately identify and rapidly isolate early faults such as insulation degradation and switch short circuits. Microsecond-level power compensation is performed for pulsed loads, and reinforcement learning is used to optimize the energy consumption of the cooling fan, comprehensively enhancing the system's autonomous operation capability and energy storage cycle life in harsh environments. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 This is an overall flowchart of the method described in this invention.

[0029] Figure 2 This is a flowchart of the maximum power point tracking method of the adaptive step-size perturbation observation method of the present invention.

[0030] Figure 3 This is a flowchart of the bus voltage control based on Lyapunov stability according to the present invention. Detailed Implementation

[0031] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0032] The following is in conjunction with the appendix Figures 1-3 The embodiments of the present invention will be described in detail below.

[0033] This invention provides a low-wind-speed wind turbine and energy storage collaborative power supply system and method. Through ultra-high integration tower architecture, multi-level intelligent control strategy, cloud-edge collaborative prediction, and bus voltage control method based on Lyapunov stability, it achieves efficient capture of weak wind energy and highly reliable uninterrupted power supply.

[0034] The system includes: an integrated architecture for joint construction of poles and towers, a wind power generation module, an energy storage module, a power conversion module, an automatic switching module, and an edge intelligent control terminal.

[0035] The integrated pole and tower co-construction architecture adopts precast integrated poles and towers made of ultra-high performance concrete. The ultra-high performance concrete has a compressive strength of not less than 120 MPa and a flexural strength of not less than 18 MPa, and incorporates 2.0% to 3.5% copper-plated steel fiber by volume. The integrated pole and tower has a hollow pipe pile structure, with a total height set at 12 m to 25 m based on local wind resources and load power. The wall thickness is designed according to the bending moment distribution gradient, with a bottom wall thickness of 150 mm to 200 mm and a top wall thickness of 80 mm to 120 mm. The longitudinal cavity inside the integrated pole and tower serves as a cable channel and natural heat dissipation duct, while the exterior is equipped with multi-layer mounting flanges for installing auxiliary equipment.

[0036] The wind power generation module uses a permanent magnet synchronous generator, with a rated power redundancy of 1.5 to 3.0 times the load's rated power. The blades of the wind power generation module adopt a high lift-to-drag ratio aerodynamic shape based on the NACA 63-4xx series airfoil, with a root chord length increased by 12% to 18% compared to conventional designs, and swept-back blade tips. The cut-in wind speed of the wind power generation module is set to 2.0 m / s to 2.5 m / s, and the cut-out wind speed is set to 20 m / s to 25 m / s.

[0037] The energy storage module is composed of lithium iron phosphate cells connected in series and parallel. Its total energy capacity is configured to maintain critical loads for at least 24 hours under windless and mains power-free conditions. The critical loads refer to the core communication equipment of the communication base station and the necessary sensors of the environmental monitoring station. The energy storage module integrates a battery management system, which has functions for acquiring individual cell voltage, current, and temperature data, as well as passive balancing.

[0038] The power conversion module includes a three-phase pulse-width modulation rectifier, a bidirectional DC / DC converter, and a bidirectional DC / AC inverter. The three-phase pulse-width modulation rectifier adopts a three-phase full-bridge topology. The bidirectional DC / DC converter connects the energy storage module to the DC bus, and the bidirectional DC / AC inverter supplies power to the AC load. The power devices are silicon carbide metal-oxide-semiconductor field-effect transistors with a switching frequency of 50 kHz to 100 kHz.

[0039] The automatic switching module employs a high-speed electromagnetically driven automatic transfer switch with a mechanical switching time of less than or equal to 20 ms. The contact material is a silver-nickel alloy, and an arc-extinguishing chamber is internally installed. The automatic transfer switch connects to the mains input terminal, the inverter output terminal, and the load terminal.

[0040] The edge intelligent control terminal is based on an ARM Cortex-M7 architecture microcontroller with a main frequency of no less than 400 MHz and a built-in floating-point unit. It is equipped with multiple 16-bit analog-to-digital converters, a controller area network bus interface, an RS485 interface, an Ethernet interface, and a 4G / 5G communication module. The edge intelligent control terminal internally integrates energy management system software, executing all local control algorithms.

[0041] The system adopts a two-tier control architecture that coordinates edge intelligent control terminals and a cloud platform. The edge intelligent control terminal, with a control cycle of 10 ms to 100 ms, collects real-time data on wind turbine speed, DC bus voltage, battery state of charge, ambient wind speed, and load power parameters, and independently executes tasks such as mode switching, maximum power point tracking, state of charge estimation, and real-time fault protection. The cloud platform interacts bidirectionally with the edge intelligent control terminal via a wireless communication link, and is responsible for long-cycle data storage, wind speed prediction model training, global optimization scheduling command issuance, and remote firmware upgrades.

[0042] The edge intelligent control terminal calculates the power balance equation of the system in real time and drives the system to automatically switch between the following four modes: Normal power supply mode: When the mains power is normal and the wind power generation module's output is greater than or equal to the load demand, the wind power generation module's output is directly supplied to the load after rectification. Excess electrical energy is charged into the energy storage module via a bidirectional DC / DC converter using a constant current-constant voltage segmented strategy. The charging cutoff condition is when the state of charge reaches the upper limit threshold, which is set to 0.90 to 0.95. If the wind power generation module's output is much greater than the load demand and the state of charge has reached the upper limit threshold, the rectifier is adjusted to enter an overmodulation state or the wind turbine's pitch angle is adjusted away from the optimal position to limit the captured power.

[0043] Supplemental power supply mode: When the mains power is normal and the wind power generation module's output power is less than the load demand, the wind power generation module operates in maximum power point tracking mode to output the current maximum power, and the insufficient load power is automatically supplemented by the mains power. The energy storage module is in float charging state, and the edge intelligent control terminal controls the bidirectional DC / DC converter to maintain the voltage at the energy storage module terminal within a preset range, which is 3.35 V to 3.40 V for lithium iron phosphate cells.

[0044] Emergency Power Supply Mode: When the mains power is abnormal (defined as the mains voltage being lower than 70% or higher than 120% of the rated value for more than 200 ms, or the mains frequency deviation exceeding ±2 Hz), the automatic switching module will switch the load to the independent wind-storage power supply path within 20 ms. If the wind power generation module's output is greater than or equal to the load's required power, the wind power generation module will supply power independently and can charge the energy storage module; if the wind power generation module's output is greater than 0 and less than the load's required power, the wind power generation module and the energy storage module will supply power together; if the wind power generation module's output is equal to 0, the energy storage module will supply power independently. When the state of charge drops to the warning threshold (set to 0.20), the edge intelligent control terminal will cut off non-critical loads according to preset priorities and send an alarm signal to the cloud platform.

[0045] Fault Protection Mode: The edge intelligent control terminal performs real-time diagnosis of overvoltage, overcurrent, short circuit, and insulation degradation faults on the wind power generation module side, energy storage module side, load side, and mains power side. Upon detecting a fault, the fault source is isolated within 2 ms via a solid-state circuit breaker or by blocking the pulse width modulation drive pulse, and automatically switches to the backup power supply path according to the system health topology. The backup power supply path includes a backup power conversion module or a mains direct-connection circuit. Fault information and waveform data from 200 ms prior to the fault are uploaded to the cloud platform.

[0046] The system satisfies the power balance equation at every instant: ; in: Indicates the wind power generation module at time Output active power, in watts (W). Indicates the energy storage module at time The discharge power, in W, is positive when discharging and 0 when not discharging; Indicates the time of mains power. The active power injected into the system is measured in W. The actual value is taken when the mains power is normal and the system is in operation. The value is 0 when the mains power is abnormal or the system is not in operation. Indicates the load at time. Total active power consumed, in watts (W). Indicates the energy storage module at time The charging power is expressed in watts (W). It is a positive value when charging and 0 when not charging. This represents the total power loss of the power conversion module, circuitry, and thermal management equipment, expressed in watts (W), calibrated through offline experiments. , , Nonlinear functions.

[0047] The control objective of the edge intelligent control terminal is to ensure that the DC bus voltage... The power is stabilized at the set value, and power balance is achieved by adjusting the duty cycle of the bidirectional converter.

[0048] To address the transient stability problem of bus voltage under load surges or wind energy fluctuations, this invention provides a nonlinear voltage control strategy based on the Lyapunov direct method. The bus voltage error is defined as follows: ,in This is the reference value for the bus voltage. The Lyapunov function is selected as... To ensure the system is asymptotically stable, the following conditions must be met: The design of the control law makes ,in The convergence coefficient is used. From this, the current command for the bidirectional converter is derived as: ; in: This is the reference current for the bidirectional converter, measured in A. A positive value indicates current injected from the energy storage module side into the bus, while a negative value indicates current drawn from the bus side into the energy storage module side. The load current is measured in amperes (A) and is measured in real time using a Hall sensor. This refers to the current injected into the busbar after rectification of the wind power generation module, expressed in amperes (A). This refers to the DC bus capacitance, measured in F, with values ​​ranging from 0.01 F to 0.1 F. The convergence coefficient is expressed in s. - ¹, with values ​​ranging from 100 to 500; , represents the bus voltage error, in units of V.

[0049] The edge intelligent control terminal calculates in each control cycle. The duty cycle of the bidirectional converter is adjusted by a current inner-loop proportional-integral controller to enable the actual current to track the circuit. This control method ensures no overshoot convergence of the bus voltage during load step changes, and the convergence time is reduced from... The decision time is typically 5 ms to 20 ms.

[0050] This invention employs an extended Kalman filter algorithm based on a first-order resistor-capacitor equivalent circuit to estimate the state of charge online. The discretized state-space model is as follows: Equations of state: ; Observation equation: ; in: Indicates time The state of charge is dimensionless and ranges from 0 to 1; Indicates time The polarization voltage, in V; This indicates the sampling period, expressed in seconds, with a value ranging from 0.01 s to 0.1 s. This represents the polarization time constant, measured in seconds, and is obtained through battery characteristic testing. This represents the charge / discharge efficiency coefficient, which is taken as 0.95 to 0.98 during charging and 0.98 to 1.0 during discharging. This indicates the battery's nominal capacity, measured in Ah. Indicates time The battery circuit current, in amperes (A), is positive when discharging and negative when charging; This represents polarization resistance, measured in Ω. This represents polarization capacitance, measured in F. The process noise corresponding to the charged state dimension at time k-1 is... The process noise corresponds to the polarization voltage dimension at time k-1; the two together form a two-dimensional process noise vector. The whole follows a Gaussian distribution with zero mean; Indicates time The battery terminal voltage, in V; The nonlinear function representing the relationship between open-circuit voltage and state of charge is obtained through offline low-current pulse discharge calibration; This represents the internal resistance in ohms, measured in Ω. Indicates time The battery circuit current, in amperes (A); The measurement noise is represented by a zero-mean Gaussian distribution with variance . .

[0051] The edge intelligent control terminal sequentially executes the time update step and the measurement update step in each sampling period to recursively obtain the optimal estimate of the state of charge, and uses the optimal estimate of the state of charge for charge and discharge boundary control and reserved capacity scheduling.

[0052] This invention employs a composite maximum power point tracking algorithm that integrates adaptive step-size perturbation observation with wind speed prediction feedforward. The steps are as follows: Step 1: Track control cycle at each maximum power point Inside, the edge intelligent control terminal collects the DC voltage after rectification from the wind power generation module. and DC current Calculate instantaneous power The maximum power point tracking control period is between 20 ms and 50 ms.

[0053] Step 2: Calculate the power change and duty cycle change Adaptive perturbation step size Determined by the following formula: ; in: This represents the proportionality constant, ranging from 0.5 to 2.0, and is dimensionless. This represents the change in power, expressed in W. This indicates the rated power of the wind power generation module, in watts (W). This represents the baseline duty cycle step size, ranging from 0.005 to 0.02, and is dimensionless.

[0054] like ,in To preset the power variation threshold, take 0.1% of the rated power. .

[0055] Step 3: The decision logic for the perturbation direction is as follows: like and ,or and If the disturbance direction is maintained, then the disturbance direction is maintained. ; Otherwise, reverse the direction of the disturbance, that is... .

[0056] Step 4: Feedforward Compensation: When the wind speed sensor detects a sudden change in wind speed... At m / s, the edge intelligent control terminal directly sets the initial value of the duty cycle using the offline-stored wind speed-optimal duty cycle mapping table. Then immediately switch back to the disturbance observation method in steps 2 to 3; the wind speed-optimal duty cycle mapping table is obtained through wind tunnel experiments, with an interpolation accuracy of 0.1 m / s.

[0057] The cloud platform runs a long short-term memory network wind speed prediction model. The input features of the model include wind speed sequence, air pressure, ambient temperature, and relative humidity over the past 6 hours. The sampling interval for the wind speed sequence is 10 minutes. The Long Short-Term Memory (LSTM) network has the following structure: an input layer with a feature dimension of 4, two hidden layers each containing 64 LSM units, and an output layer containing 24 neurons, outputting the average wind speed for the next 1 to 24 hours. The LSM network is trained using the Adam optimizer with root mean square error as the loss function. The training dataset consists of historical data from weather stations and measured data from this system. The cloud platform updates the prediction results every 6 hours and sends the wind speed curve for the next 24 hours to the edge intelligent control terminal.

[0058] The edge intelligent control terminal predicts wind speed. ( , indicating the future Calculate the expected cumulative power generation over the next 24 hours using the average wind speed over h. : ; in Based on predicted wind speed The power curve of the wind power generation module is obtained by interpolation, and the power curve is obtained through offline calibration. The total load demand is also calculated. h. Dynamically adjust the reserved state of charge limit of the energy storage module. : like ,but ; like ,but ; like ,but It also instructed that the energy storage module be recharged in advance when mains power is available.

[0059] The edge intelligent control terminal integrates a fault diagnosis module based on Fast Fourier Transform (FFT). Using the DC bus current signal as the analysis object, 2048 points are collected every 10 ms at a sampling frequency of 200 kHz. After applying a Hanning window, a FFT is performed to extract the amplitudes of the fundamental frequency and the 2nd to 20th harmonics. The total harmonic distortion (THD) is calculated. ; in: Represents the total harmonic distortion rate, which is dimensionless; This represents the effective value of the fundamental current, in amperes (A). Indicates the first The effective value of the second harmonic current, in amperes (A); This represents the highest harmonic order, which is 20.

[0060] The fault determination rules are as follows: when Furthermore, if the amplitude of the third harmonic exceeds 5% of the fundamental amplitude, and the amplitude of the fifth harmonic exceeds 3% of the fundamental amplitude, it is determined to be a micro-arc flash fault caused by insulation aging. When the amplitude of high-frequency components above 1 kHz suddenly increases to more than 10 times the normal value, it is determined to be a short circuit fault in the power semiconductor switch.

[0061] After confirming the fault, the edge intelligent control terminal triggers the solid-state circuit breaker to isolate the faulty branch within 2 ms and records the fault waveform.

[0062] The edge intelligent control terminal possesses high-speed dynamic load sensing capabilities to address the pulsed loads present in communication base stations. The edge intelligent control terminal monitors the DC bus voltage in real time via an analog-to-digital converter with a 1 MHz sampling rate. and load-side current Calculate the instantaneous power of the load. and its rate of change When detected At W / ms, the edge intelligent control terminal initiates feedforward compensation control within 50 μs, controlling the bidirectional converter to draw or absorb instantaneous compensation power from the energy storage module side. : ; in: , This is a reference value for load power, in watts (W). This represents the proportionality constant, ranging from 0.8 to 1.0, and is dimensionless. This represents the differential coefficient, ranging from 0.01 to 0.05, and is dimensionless. Indicates time, in seconds (s).

[0063] After compensation, the bus voltage fluctuation is limited to within ±3%.

[0064] The control chamber integrates a hybrid passive and active thermal management system. The outer shell of the control chamber uses corrugated aluminum alloy heat sinks with a heat dissipation area of ​​no less than 2 m², and is filled with a thermally conductive phase change material. This phase change material has a melting point of 45 ℃ and a latent heat of phase change of no less than 200 kJ / kg. When the internal temperature is below 45 ℃, only passive cooling is used; when the internal temperature exceeds 45 ℃, the edge intelligent control terminal activates the built-in axial flow fan. The fan speed is controlled by a proportional-integral-derivative controller based on the temperature difference. adjust: ; in: This indicates the fan speed, expressed in r / min. This represents the proportionality constant, which ranges from 10 to 50 and is dimensionless. This represents the integral coefficient, with values ​​ranging from 0.1 to 1.0, and the unit is s. - ¹; This represents the differential coefficient, with values ​​ranging from 0.01 to 0.1, and the unit is seconds (s). This indicates the temperature inside the control chamber, expressed in °C. This indicates the temperature difference, expressed in degrees Celsius (°C). This represents the integral variable, with units of seconds (s).

[0065] To facilitate a better understanding of the present invention by those skilled in the art, the present invention will be further described below with reference to specific embodiments.

[0066] Example 1: This example describes the deployment of a communication base station in an inland area. The base station is located in a remote area with low power grid reliability and an average annual wind speed of 3.5 m / s, which is a typical low wind speed area.

[0067] The communication base station equipment has a rated power of 1.5 kW and exhibits constant load characteristics. The core communication load is 1.1 kW (including the baseband processing unit, main control board, and optical transmission equipment), while the non-critical loads include 0.1 kW for environmental monitoring equipment and 0.3 kW for the secondary frequency band of the radio frequency unit. The core load requires uninterrupted power supply 24 / 7, with an allowable interruption time of less than 50 ms.

[0068] The tower is constructed using ultra-high performance precast concrete, standing 15 meters tall with a bottom wall thickness of 180 mm and a top wall thickness of 100 mm, and a steel fiber content of 2.5%. The tower is hollow internally, with an equipment compartment at the bottom for installing energy storage and power conversion modules, and a wind power generation module at the top. Ladders and maintenance platforms are provided on the tower surface, and the internal longitudinal cavity serves as both a cable channel and a natural cooling duct.

[0069] The wind turbine module has a rated power of 3.0 kW (2.0 times the load's rated power redundancy), a cut-in wind speed of 2.2 m / s, a cut-out wind speed of 22 m / s, and a rotor diameter of 4.2 m. The blades utilize an optimized NACA 63-418 airfoil design, with a 15% increase in root chord length compared to conventional designs, and swept-back blade tips to reduce aerodynamic noise. The generator is a permanent magnet synchronous generator with a 48-pole stator design, a rated speed of 300 r / min, and outputs three-phase AC power which is rectified and connected to the DC bus.

[0070] The energy storage module uses a lithium iron phosphate battery pack with a total energy of 38.4 kWh and a rated voltage of 51.2 V. The battery pack consists of 16 series-connected lithium iron phosphate modules (each with a nominal voltage of 3.2 V), and each module consists of 15 50 Ah cells connected in parallel (750 Ah per string), for a total capacity of 750 Ah. The battery management system is integrated inside the battery pack, with sampling accuracy of voltage ±2 mV and current ±0.5%, and features single-cell voltage, current, and temperature acquisition, as well as passive equalization functions. The battery pack is installed in the equipment compartment at the bottom of the tower, and the compartment is equipped with insulation and ventilation openings.

[0071] Power conversion module: A three-phase rectifier with silicon carbide metal-oxide-semiconductor field-effect transistors, a bidirectional DC / DC converter, and a DC / AC inverter, with a rated conversion efficiency of 96.5%. The three-phase rectifier converts the frequency-converted AC power output from the fan into 51.2V DC power; the bidirectional DC / DC converter connects the battery pack to the DC bus, enabling bidirectional charging and discharging control; the DC / AC inverter converts the DC power into 220V / 50Hz AC power to supply the load. The switching frequency of the power devices is set to 80kHz.

[0072] Automatic switching module: High-speed electromagnetically driven automatic transfer switch with a mechanical switching time of 18 ms. The contact material is silver-nickel alloy, and an internal arc-extinguishing chamber is incorporated. The switch has two input terminals: one for mains input and one for inverter output; the output terminal is connected to the load. The switching logic is controlled by an edge intelligent control terminal, supporting manual / automatic mode selection.

[0073] Edge intelligent control terminal: Employs an STM32H743 microcontroller with a 400 MHz clock speed and a built-in floating-point unit. It is equipped with an 8-channel 16-bit analog-to-digital converter, a 2-channel controller area network (CAN) bus interface, a 2-channel RS485 interface, a 1-channel Ethernet interface, and a 5G communication module. The control terminal is installed in a control compartment in the middle of the tower, with an IP65 protection rating.

[0074] Cloud platform: Deployed on Alibaba Cloud servers, it runs a long short-term memory network wind speed prediction model, outputs the wind speed sequence for the next 24 hours every 6 hours (step size 1 hour), and distributes it to edge intelligent control terminals via 5G network.

[0075] The control parameters are set as follows: Edge intelligent control terminal control cycle: 100 ms.

[0076] Maximum power point tracking period: 20 ms, reference duty cycle step size proportionality coefficient .

[0077] Upper limit of state of charge threshold State of charge limit warning threshold The value is dynamically adjusted based on wind speed forecast (baseline value is 0.20).

[0078] Extended Kalman filter sampling period s, Battery model parameters: Ohmic internal resistance Ω, polarization resistance Ω, polarized capacitor F, polarization time constant s, nominal capacity Ah, charge / discharge efficiency coefficient .

[0079] Lyapunov Bus Voltage Control: Bus Voltage Reference Value V, bus capacitance F, convergence coefficient .

[0080] Long Short-Term Memory Network Prediction: The cloud platform sends out the wind speed curve for the next 24 hours every 6 hours, with a step size of 1 hour.

[0081] The detailed process is as follows: After the system is powered on, the edge intelligent control terminal first executes an initialization self-test program, sequentially checking whether the three-phase voltage and current of the wind power generation module are within the normal range, whether the battery management system communication of the energy storage module is normal, whether the drive circuits of each power conversion module are normal, whether the contact status of the automatic switching module is normal, and whether the 5G communication link is unobstructed. After the self-test passes, the control terminal enters the main control loop and executes the following control tasks with a period of 100 ms.

[0082] Operating during the morning period (wind speed 4.0 m / s to 6.0 m / s): At this time, wind resources are relatively abundant, and the power output of the wind turbine modules is between 1.8 kW and 2.5 kW, which exceeds the load demand of 1.5 kW. The system automatically enters the normal power supply mode. The control process is as follows: The edge intelligent control terminal first collects mains power parameters through voltage and current transformers to confirm that the mains power is normal (voltage 220 V ± 10%, frequency 50 Hz ± 0.5 Hz). Then, it reads the DC bus voltage and current to calculate the wind turbine output power. Since the wind turbine power exceeds the load demand, the control terminal sends a charging command to the bidirectional DC / DC converter, setting the charging current to 0.2 times the nominal capacity (150 A). After rectification by the three-phase rectifier, a portion of the wind turbine output power (1.5 kW) is directly supplied to the load, while the remaining power is charged into the energy storage module via the bidirectional DC / DC converter in a constant current manner. Simultaneously, the control terminal continuously monitors the battery pack voltage. When the battery voltage reaches 56.8 V (corresponding to 3.55 V per cell, i.e., a total voltage of 56.8 V), it automatically switches to constant voltage charging mode, and the charging current gradually decreases. When the charging current drops to 2 A, the control terminal determines that the battery is fully charged and stops charging. At this point, the state of charge (SOC) increases from the initial 0.70 to 0.92.

[0083] To prevent overcharging, if the fan power is still greater than the load demand after the state of charge reaches 0.92, the control terminal will adjust the modulation ratio of the three-phase rectifier to enter the over-modulation state, limiting the fan capture power, or adjust the fan pitch angle away from the optimal position so that the excess power is dissipated as heat.

[0084] Afternoon operation (wind speed 2.0 m / s to 3.5 m / s): As wind speed decreases, the power output of the wind turbine module drops to 0.6 kW to 1.2 kW, which is less than the load demand of 1.5 kW. The system automatically switches to supplemental power supply mode.

[0085] The edge intelligent control terminal controls the three-phase rectifier to operate in maximum power point tracking mode. It uses an adaptive step-size perturbation observation method to track the optimal operating point of the wind turbine in real time, ensuring the turbine outputs the maximum power at the current wind speed (e.g., 0.9kW). The insufficient load power (0.6kW) is automatically supplemented by the mains power through the mains direct-connection loop of the automatic switching module. At this time, the energy storage module is in float charging mode. The control terminal controls the bidirectional DC / DC converter to maintain the battery pack terminal voltage between 53.5V and 53.8V (corresponding to 3.344V to 3.3625V for a single cell). This voltage range is the optimal float charging voltage range for lithium iron phosphate batteries, maximizing battery life.

[0086] Nighttime operation: The cloud platform predicted 6 hours in advance that the wind speed would remain below 2.0 m / s for the next 6 hours (no effective power generation), automatically adjusting the reserved state of charge (SBC) limit to 0.50 and instructing the edge intelligent control terminal to replenish the battery in advance when the mains power was normal. At an SBC of 0.55, a simulated mains power outage occurred. The edge intelligent control terminal, through its mains voltage detection circuit, detected the mains voltage drop (voltage below 154 V, i.e., 70% of the rated value) within 100 μs and immediately issued a switching command to the automatic switching module. The automatic switching module completed the mechanical switching within 18 ms, switching the load from the mains circuit to the inverter output circuit. Because the switching time was less than the maximum allowable interruption time (50 ms) of the communication base station, the load equipment did not reset.

[0087] After the switchover is complete, the system enters emergency power supply mode. Since the wind speed is lower than the cut-in wind speed, the fan output power is 0, and all load power is provided by the energy storage module. The battery pack discharges at 1.5 kW. The state of charge (SOC) begins to decrease from 0.55%. The control terminal monitors the SOC in real time. When the state of charge drops to 0.20, the control terminal first disconnects the environmental monitoring load (0.1kW) according to the preset priority. When the state of charge drops to 0.15, the secondary frequency band of the radio frequency unit (0.3 kW) is further cut off, leaving only the core communication load of 1.1 kW.

[0088] At this point, the remaining available capacity is 38.4 kWh × 0.15 = 5.76 kWh, with a support time of approximately 5.76 / 1.1 ≈ 5.24 h. If the charge level starts at 0.55 and discharges at 1.5 kW to 0.15, the theoretical support time is 38.4 × (0.55 - 0.15) / 1.5 = 10.24 h. The actual support time meets the core communication requirements of the base station at night.

[0089] The fault simulation is as follows: A short-circuit fault was artificially created in the silicon carbide metal-oxide-semiconductor field-effect transistor (MOSFET) switch inside the power conversion module. The edge intelligent control terminal analyzed the DC bus current signal in real time using a Fast Fourier Transform (FFT) fault diagnosis module. After the fault occurred, the high-frequency harmonic amplitude suddenly increased by 15 times. The control terminal detected the anomaly within 1.8 ms, immediately blocked all pulse-width modulation (PWM) drive pulses, and triggered a solid-state circuit breaker to isolate the faulty branch. Simultaneously, the control terminal uploaded the fault information (including fault type, timestamp, and waveform data from 200 ms prior to the fault) to the cloud platform via the 5G network and sent an alarm SMS to maintenance personnel. The backup power conversion module automatically activated within 10 ms, the load voltage fluctuation was less than 5%, and the load equipment did not detect the power outage.

[0090] During a year-long field test, the system's operational data are as follows: annual effective power generation time was 4820 hours (the actual number of hours the wind turbines generated out of a total of 8760 hours); the system's mean time between failures (MTBF) was 13800 hours (approximately 1.58 years); the energy capture efficiency in the low wind speed range (2.5 m / s to 4.5 m / s) was 31.2% (the ratio of measured power generation to theoretical wind energy); the total outage time for core loads was 0 minutes (the energy storage system provided full power during mains power outages, excluding brief outages of non-critical loads); the energy storage cycle life (capacity decayed to 80% under 0.5C charge / discharge conditions) was 4800 cycles; and the tower top vibration amplitude (at a wind speed of 10 m / s) was 1.3 mm. These data were obtained through statistical analysis of the operational logs.

[0091] Example 2: The system configuration is the same as in Example 1, but the following advanced control modules are additionally enabled in the edge intelligent control terminal: Bus voltage control based on Lyapunov stability (replacing traditional proportional-integral control); Dynamic allocation of energy storage discharge power based on model predictive control (for emergency power supply mode). A smooth switching method for active damping control (for fault protection mode).

[0092] Model predictive control discharge power allocation method: In emergency power supply mode, the edge intelligent control terminal solves the following optimization problem with a period of 100 ms, aiming to extend the support time of the energy storage system as much as possible while meeting the power supply reliability constraints: ; Constraints: ; ; ; in: Indicates the first Battery discharge power within each control cycle, in kW; This indicates the prediction time domain and is dimensionless. , This represents the weighting coefficient, which is dimensionless. This represents a reference value for the state of charge and is dimensionless. It represents the limit of the state of charge and is dimensionless. It represents the upper limit of the state of charge and is dimensionless. Indicates the first Load power within each control cycle, in kW; kWh represents the total energy of the battery; s, needs to be converted to hours: h; kW represents the maximum discharge power; Indicates the first The state of charge for each control cycle is dimensionless.

[0093] An explicit model predictive control method is adopted, and the optimal solution under different combinations of state of charge and load power is pre-calculated as a lookup table. The sampling step size of the lookup table is: 1% step for state of charge from 20% to 90%, and 0.1 kW step for load power from 0.5 kW to 1.5 kW. The edge intelligent control terminal obtains the optimal discharge power in real time through bilinear interpolation, and the calculation time for a single calculation is less than 1 ms.

[0094] In a scenario simulating a windless period of 8 hours (initial SOC=0.90, load 1.5 kW), conventional proportional-integral control can support approximately 16 hours (limited by the lower voltage limit), while model predictive control extends the support time to 19.2 hours.

[0095] Active damping smooth switching method: When the main power conversion module fails and a switch to the backup power conversion module is required, the edge intelligent control terminal first reads the output voltage of the backup module. and bus voltage Calculate the voltage difference .like V, the controller adjusts the synchronous rectifier of the backup module so that its output voltage gradually approaches the bus voltage through a first-order inertial element: ; in: This indicates the reference value of the standby module's output voltage, in volts (V). This indicates the bus voltage, measured in volts (V). This represents the voltage difference, measured in volts (V). Indicates time, in seconds; It is a natural constant.

[0096] The adjustment process lasts 10 ms, reducing the voltage difference to within 1 V. A switching command is then issued, and a virtual resistor is connected in series in the switching loop. Ω, virtual inductance The virtual resistance is reduced to 0 μH to suppress inrush current. Within 20 ms after switching, the virtual resistance is linearly reduced to 0 Ω and the virtual inductance is reduced to 0 μH.

[0097] In a scenario simulating an 8-hour windless period, model predictive control extended the support time for a 1.5 kW load from 16 hours with conventional proportional-integral control to 19.2 hours. When the load jumped from 1.5 kW to 2.5 kW (rise time 100 μs), Lyapunov control reduced the bus voltage drop from 12% with conventional proportional-integral control to 3.5%. In 100 consecutive simulated fault switching operations, the voltage overshoot was less than 5%, the peak current surge was less than 1.2 times the rated current, and the load equipment did not reset or suffer damage.

[0098] Example 3: The system configuration is the same as in Example 1, but a reinforcement learning-based thermal management energy consumption optimization module is deployed in the edge intelligent control terminal.

[0099] Reinforcement learning agent design: The state space includes six variables: tower temperature. (°C), outside temperature of the tower (°C), Average battery temperature (°C), Battery heating power (W) State of charge (Dimensionless) Current fan speed (r / min).

[0100] The action range is the fan speed adjustment amount. The value range is -200 r / min to +200 r / min, discretized into 9 actions: -200, -150, -100, -50, 0, 50, 100, 150, 200 r / min. The reward function is defined as: ; in: Indicates time The reward value is dimensionless; This represents the temperature penalty weight, which is dimensionless. This represents the energy consumption penalty weight, which is dimensionless. This indicates the average battery temperature, in °C. This indicates the real-time electrical power of the fan, in watts (W). W represents the rated power of the fan; This indicates a penalty for the battery temperature deviating from the optimal range by 25±5 ℃, in ℃.

[0101] The deep Q-network structure consists of 6 nodes in the input layer, 64 nodes in the hidden layer (using linear rectified activation functions), and 9 nodes in the output layer (using linear activation functions). The network was pre-trained on a cloud platform using 5000 hours of thermal operation data collected in the lab and 20000 hours of virtual data generated from a thermal simulation model. The Adam optimizer was used with a learning rate of 0.001 and an experience replay pool capacity of 10000. After training, the network weights were quantized from 32-bit floating-point to 8-bit integers and deployed to an edge intelligent control terminal. A single inference operation took less than 5 ms.

[0102] During the online operation phase, the edge intelligent control terminal performs reinforcement learning inference every 10 seconds, outputs the optimal fan speed adjustment, and randomly explores with a 5% probability (i.e., randomly selects actions to explore new strategies 10% of the time). Every 24 hours, the edge intelligent control terminal compresses and uploads the operation trajectory (state-action-reward sequence) to the cloud platform for periodic retraining to adapt to the thermal characteristics differences caused by seasonal changes.

[0103] During continuous operation for 7 days on hot summer days (maximum temperature 38 ℃), the average daily energy consumption of the cooling fan was 112 Wh, a 38% reduction compared to the traditional proportional-integral-derivative control (average 180 Wh per day). The battery pack's highest temperature was 35.2 ℃, and the lowest was 33.8 ℃, with an internal temperature difference of 1.4 ℃, meeting the optimal operating temperature range (15 ℃ to 35 ℃) for lithium iron phosphate batteries. Accelerated cycle life test data for the battery pack shows that, under the same operating conditions, the capacity decay rate of the battery pack using this thermal management method is 25% lower than that of the traditional method.

[0104] Example 4: Same as Example 1, but with the following advanced modules additionally deployed in the cloud platform and edge intelligent control terminal: Probabilistic wind speed prediction model based on quantile regression forest; A dynamic optimizer for energy storage reserved capacity based on value at risk; A charging and discharging strategy for sensing energy storage lifespan degradation.

[0105] The cloud platform runs a quantile regression forest model to output the 10th, 50th, and 90th quantile prediction curves for wind speed over the next 24 hours. The model input features include: wind speed sequences from the past 6 hours (sampling interval 10 minutes), air pressure, ambient temperature, relative humidity, and forecast wind speeds obtained from numerical weather prediction. The quantile regression forest uses a random forest framework, generating 200 decision trees through bootstrap sampling. The node splitting of each tree is based on minimizing the weighted quantile loss function. ; in: Indicates the quantile level, taken as 0.10, 0.50, or 0.90; The quantile loss function is dimensionless. Indicates the first The actual wind speed of each sample is in m / s; Indicates the first The predicted wind speed for each sample is in m / s. This represents the number of training samples; it is dimensionless. This is the quantile test function. For indicator functions (when) If the value is 1, then the value is 0.

[0106] The model is retrained every 6 hours, using historical data from the last 30 days. After training, the cloud platform outputs three quantile prediction curves for the next 24 hours every hour and distributes them to the edge intelligent control terminal.

[0107] After receiving the predicted wind speed quantile curve, the edge intelligent control terminal calculates the corresponding power generation quantile curve. ( , indicating the future h), obtained by interpolating the power curve of the wind power generation module.

[0108] Define power supply reliability requirements: Under extreme conditions of no mains power and no wind (actual wind speed lower than the cut-in wind speed), the system should guarantee power supply to critical loads with a probability of no less than [percentage missing]. Based on probabilistic wind speed prediction, the edge intelligent control terminal calculates the cumulative power generation for the next 24 hours. The three quantiles: ; in Define the risk level. Take the closest one The expected power generation corresponding to the quantile, i.e. (The 10th percentile indicates a 90% probability that the actual power generation will be higher than this value, but a power supply reliability of no less than 99.5% is required, so a more conservative 10th percentile is used.) Then calculate the required energy storage reserve capacity. : ; in h represents the total critical load demand, in kWh. The reserved state-of-charge limit is then set. for: ; in kWh represents the total battery energy. To prevent excessive reserve, The upper limit is 0.50, and the lower limit is 0.10. If... ,but .

[0109] The edge intelligent control terminal incorporates an accelerated aging model for lithium iron phosphate batteries, quantifying the impact of charge / discharge depth, charge / discharge rate, and operating temperature on lifespan into equivalent cycle counts. Battery capacity decay rate. Described by the following formula: ; in: This indicates the current battery capacity, in Ah. Indicates time, in seconds; This represents the cyclic aging rate constant under reference operating conditions, in seconds. - ¹, take ; This represents the absolute value of the battery circuit current, in amperes (A). This represents the reference current, in A, and is set to 375 A (corresponding to a 0.5C rate, based on a 750 Ah capacity). This represents the current acceleration factor, which is dimensionless and is taken as 1.2. The activation energy is expressed in J / mol. ; Let J represent the ideal gas constant, taken as 8.314 J / (mol·K); This indicates the reference temperature, in K, taken as 298.15 K (25 ℃). This indicates the actual temperature of the battery, in Kelvin (K). It represents the magnitude of the change in state of charge during one charge-discharge cycle, and is dimensionless. Indicates the reference depth, dimensionless, taken as 0.8 (80% depth of discharge); This represents the depth acceleration factor, which is dimensionless and is set to 1.5.

[0110] The edge intelligent control terminal, within each control cycle (100 ms), according to the current... , Based on the predicted charge / discharge depth distribution over the next 24 hours, calculate the expected aging cost over the next 24 hours. In emergency power supply mode, the optimization objective changes from simply extending the support time to minimizing the weighted sum: ; in: This indicates the critical load power that was cut off due to insufficient energy storage capacity, in kW; This indicates the rated power of the critical load, in kW, taken as 1.1 kW; This represents the predicted aging cost over the next 24 hours, expressed in equivalent cycle counts. This represents the reference aging cost, which is dimensionless and takes 0.1 equivalent cycle count. , This represents the weighting coefficient, which is dimensionless.

[0111] The edge intelligent control terminal, through online rolling optimization, prioritizes the use of lower power ratios while meeting power supply reliability constraints. The system actively maintains the energy storage state of charge in the middle range (0.4 to 0.6) when sufficient wind speed is predicted for the next 24 hours, in order to extend cycle life.

[0112] The system of Example 4 was deployed at an inland base station with an average annual wind speed of 3.5 m / s and a wind speed fluctuation coefficient (standard deviation / mean) of 0.6, and operated continuously for 6 months. Statistics show that the actual power supply reliability (percentage of uninterrupted power supply time to critical loads) was 99.72%, close to the target value of 99.5%; the daily average equivalent charge / discharge depth of the energy storage system was 42%, a decrease of 13 percentage points compared to 55% in Example 1; the battery pack temperature fluctuation range was controlled within 2.1 ℃; and the number of critical load power outages due to insufficient energy storage was 0. This example extends the effective support time and predicts battery life for the same energy storage capacity.

[0113] Comparative Example 1: The system configuration adopts a traditional split-type wind power supply system, without using the integrated tower architecture and control method of this invention.

[0114] Tower: Ordinary concrete tower (strength grade C40, compressive strength 40 MPa), 15 m high, 200 mm thick, with no internal integrated design. The wind power generation module is independently installed at the top of the tower, while the energy storage module, power conversion module, and control equipment are distributed and installed in ground cabinets (protection grade IP54). All equipment is connected via outdoor armored cables.

[0115] Wind power generation module: Conventional permanent magnet synchronous generator, rated power 3.0 kW, cut-in wind speed 3.0 m / s, cut-out wind speed 20 m / s, rotor diameter 3.8 m, blades are conventional airfoil (no low wind speed optimization). No maximum power point tracking control; the rectified wind turbine is directly connected to the DC bus, and the bus voltage fluctuates with wind speed.

[0116] Energy storage module: Valve-regulated lead-acid battery pack with a total energy of 20 kWh and a rated voltage of 48 V, consisting of 24 2 V 500 Ah cells connected in series. The battery management system only has overvoltage protection, overcurrent protection, and undervoltage alarm functions, and lacks accurate state-of-charge estimation (it only roughly judges based on the terminal voltage).

[0117] Power conversion module: conventional silicon-based insulated gate bipolar transistor rectifier and inverter, rated conversion efficiency 92%.

[0118] Automatic switching module: Manual transfer switch, no automatic switching function. After a mains power outage, on-site maintenance personnel must manually close the switch to switch to energy storage power supply.

[0119] Control logic: Voltage threshold control. When the DC bus voltage is below 44V (corresponding to battery undervoltage), the circuit is manually closed for energy storage and discharge; when it is above 56V, the charging circuit is manually disconnected. No edge-controlled intelligent terminal, no cloud platform.

[0120] Thermal management: The ground-level cabinet uses natural ventilation and has no active heat dissipation. In summer, the internal temperature of the cabinet can reach above 55 ℃.

[0121] The results are as follows: Operating at the same location (annual average wind speed 3.5 m / s) for one year. Annual effective power generation time is 3120 h (the wind turbine cannot generate power in the 2.5 m / s to 3.0 m / s wind speed range due to high cut-in wind speed). The system's mean time between failures (MTBF) is 5100 h. The energy capture efficiency in low wind speed ranges is 16.8%. The total duration of core load power outages is 1340 min, mainly due to: the need for manual switching to energy storage power after a mains power outage, a switching time of approximately 5 to 10 minutes, causing the communication base station to lose power during this period; inaccurate estimation of the energy storage's uncharged state, resulting in multiple over-discharges that damaged the batteries (replacing the batteries twice within one year); lack of predictive capabilities, leading to base station power outages after the energy storage's energy is depleted during consecutive windless periods. The energy storage cycle life (capacity decay to 80%) is approximately 1100 cycles. The tower top vibration amplitude (10 m / s wind speed) is 5.2 mm. Poor heat dissipation caused the temperature inside the control cabinet to exceed 55°C in summer, increasing the failure rate of power modules and requiring the inverter to be replaced three times within a year.

[0122] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0123] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for coordinating power supply of a low-wind-speed wind turbine and energy storage, characterized in that, The method is executed by an edge intelligent control terminal and includes the following steps: Step 1: Collect data on the power generation of the wind power generation module, the state of charge of the energy storage module, the real-time power of the load, and the voltage and frequency of the mains power. Step 2: Determine the mains power status based on the data collected in Step 1. When the mains power is normal and the power generation of the wind power generation module is greater than or equal to the load power, control the wind power generation module to supply power to the load and charge the excess power into the energy storage module. Step 3: When the mains power is normal and the power output of the wind power generation module is less than the load power, control the wind power generation module to operate in maximum power point tracking mode. The insufficient load power is supplemented by the mains power, while the energy storage module is kept in float charging state. Step 4: When the mains power is abnormal, switch the load to an independent power supply path composed of the wind power generation module and the energy storage module within 20 ms. Step 5: When an electrical fault is detected, isolate the fault source and switch to the backup power supply path within 2 ms.

2. The method for coordinated power supply of low-wind-speed wind turbines and energy storage according to claim 1, characterized in that, In step two, the excess electrical energy is charged into the energy storage module using a constant current-constant voltage segmented strategy. The charging cutoff condition is when the state of charge reaches the upper limit threshold of 0.90 to 0.

95.

3. The method for coordinated power supply of low-wind-speed wind turbines and energy storage according to claim 1, characterized in that, The mains power abnormality mentioned in step four is defined as any of the following situations: (a) the mains voltage is lower than 70% of the rated value and the duration of this voltage state exceeds 200ms; (b) the mains voltage is higher than 120% of the rated value and the duration of this voltage state exceeds 200ms; (c) the mains frequency deviation exceeds ±2Hz.

4. The method for coordinated power supply of low-wind-speed wind turbines and energy storage according to claim 1, characterized in that, In step four, when the state of charge drops to the warning threshold of 0.20, non-critical loads are cut off step by step according to the preset priority. The non-critical loads include heaters, monitoring auxiliary equipment, and secondary frequency bands of radio frequency units.

5. The method for coordinated power supply of low-wind-speed wind turbines and energy storage according to claim 1, characterized in that, The maximum power point tracking employs an adaptive step-size perturbation-observation method, with an adaptive step size... Determined by the following formula: ； in This represents the proportionality coefficient, with a value ranging from 0.5 to 2.

0. This represents the change in power, expressed in watts (W). This indicates the rated power of the wind power generation module, in watts (W). This represents the baseline duty cycle step size, ranging from 0.005 to 0.02, and is dimensionless.

6. The method for coordinated power supply of low-wind-speed wind turbines and energy storage according to claim 1, characterized in that, The state of charge is estimated online using an extended Kalman filter algorithm based on a first-order resistor-capacitor equivalent circuit, and the state equation is: The observation equation is: ； in: Indicates time The state of charge; Indicates time The polarization voltage, in V; This indicates the sampling period, expressed in seconds, with a value ranging from 0.01 s to 0.1 s. This represents the polarization time constant, in seconds. Indicates the charge / discharge efficiency coefficient; This indicates the battery's nominal capacity, measured in Ah. Indicates time The battery circuit current, in amperes (A); This represents polarization resistance, measured in Ω. The process noise corresponding to the charged state dimension at time k-1 is... The process noise corresponds to the polarization voltage dimension at time k-1; together, they constitute a two-dimensional process noise vector. The whole follows a Gaussian distribution with zero mean; Indicates time The battery terminal voltage, in V; A nonlinear function representing the relationship between open-circuit voltage and state of charge; This represents the internal resistance in ohms, measured in Ω. Indicates time The battery circuit current, in amperes (A); This indicates measurement noise.

7. A method for co-powering a low-wind-speed wind turbine and energy storage according to claim 1, characterized in that, The electrical faults mentioned in step five include overvoltage, overcurrent, short circuit, and insulation degradation. They are detected by a fault diagnosis module based on fast Fourier transform. When the total harmonic distortion rate is greater than 10% and the amplitude of the third harmonic exceeds 5% of the fundamental amplitude and the amplitude of the fifth harmonic exceeds 3% of the fundamental amplitude, it is determined to be an insulation micro-arc flash fault. When the amplitude of high-frequency components above 1 kHz suddenly increases to more than 10 times the normal value, it is determined to be a switch short circuit fault.

8. A method for co-powering a low-wind-speed wind turbine and energy storage according to claim 1, characterized in that, It also includes a bus voltage control step based on Lyapunov stability: calculating the bus voltage error. ,in This is the reference value for bus voltage. The real-time bus voltage is used as the basis for calculating the reference current of the bidirectional converter. ,in For load current, Inject bus current into the wind power generation module Bus capacitance, in F. The convergence coefficient is expressed in s. - ¹ Finally, adjust the duty cycle of the bidirectional converter to make the actual current track the signal. .

9. A method for co-powering a low-wind-speed wind turbine and energy storage according to claim 1, characterized in that, It also includes dynamic load sensing and instantaneous power compensation steps: real-time monitoring of load instantaneous power. and its rate of change When detected At W / ms, the output compensation power is achieved within 50 μs. ,in This indicates the load power deviation, in watts (W). This represents the proportionality coefficient, ranging from 0.8 to 1.

0. This represents the differential coefficient, ranging from 0.01 to 0.

05.

10. A low-wind-speed wind turbine and energy storage coordinated power supply system, characterized in that, include: The integrated architecture for co-construction of poles and towers adopts precast integrated poles and towers made of ultra-high performance concrete. The compressive strength of the ultra-high performance concrete is not less than 120 MPa, and it incorporates steel fibers with a volume percentage of 2.0% to 3.5%. A wind power generation module is installed on the top of the integrated tower, and the cut-in wind speed of the wind power generation module is 2.0 m / s to 2.5 m / s; The energy storage module, composed of lithium iron phosphate battery cells, is installed at the bottom of the integrated tower or in its internal mounting compartment. The power conversion module includes a three-phase pulse width modulation rectifier, a bidirectional DC / DC converter, and a bidirectional DC / AC inverter; The automatic switching module uses a high-speed electromagnetic drive automatic transfer switch, with a mechanical switching time of less than or equal to 20 ms; The edge intelligent control terminal collects real-time operating data from the wind power generation module, energy storage module, load, and mains power, and coordinates with the power conversion module and automatic switching module to execute the low wind speed wind turbine and energy storage collaborative power supply method as described in any one of claims 1 to 9.

Images