A multi-band fusion wind farm communication network system based on mobile ad hoc network

Abstract

This application discloses a multi-band converged wind farm communication network system based on a mobile ad hoc network, comprising: multiple wind turbine nodes, each deployed at wind turbines arranged along a strip area within the wind farm; each wind turbine node includes the following modules: a multi-band communication module; an energy sensing and acquisition module; a strip ad hoc network routing module; a dynamic rate control module; and a central scheduling module. This application integrates low- and high-frequency dual-band communication modules at each wind turbine node. The low-frequency band ensures highly reliable long-distance transmission of critical data such as control commands, while the high-frequency band supports the bandwidth requirements of high-capacity services such as video inspection, effectively avoiding dynamic shading caused by wind turbine blade rotation. The dual-frequency collaborative transmission mechanism further realizes load sharing and redundancy backup, ensuring overall communication continuity and stability. Through multi-band convergence and adaptive switching technology, the communication reliability in the complex environment of the wind farm is improved.

Description

Technical Field

[0001] This application belongs to the field of wireless communication network technology, specifically relating to a multi-band fusion wind farm communication network system based on a mobile ad hoc network. Background Technology

[0002] With the continuous expansion of wind farm scale and the increasing demand for intelligent operation and maintenance, services such as wind turbine condition monitoring, video inspection, and fault early warning place higher demands on the coverage, transmission rate, and reliability of communication networks. Traditional wind farm communication mainly uses fiber optic ring networks or single-band wireless networks (such as Wi-Fi or data radios). Although fiber optic communication has high reliability, it suffers from high laying costs, difficult maintenance, and susceptibility to physical damage caused by wind turbine blade rotation and tower sway. Single-band wireless networks struggle to balance coverage distance and transmission rate: low-frequency bands (such as Sub-1GHz) have strong diffraction capabilities and long coverage, but limited bandwidth, failing to meet the high-bandwidth service requirements such as video backhaul; high-frequency bands (such as 2.4GHz / 5.8GHz) have high speeds, but are susceptible to obstruction and multipath effects in the complex electromagnetic environment of densely arranged wind turbines and continuously rotating blades, resulting in poor link stability.

[0003] Furthermore, wind farms are mostly located in remote areas with difficult access to municipal power, and wind turbine nodes typically rely on battery power, resulting in limited energy resources. Existing wireless communication systems are designed with little consideration for node energy harvesting capabilities and energy management, leading to nodes going offline due to power depletion and affecting network continuity. Simultaneously, the strip-like physical layout of wind turbines along roads or transmission lines causes traditional routing protocols suitable for mesh networks (such as AODV and OLSR) to incur significant redundant control overhead during route discovery, and they fail to effectively incorporate node energy status for path optimization. When network load fluctuates, the lack of an effective congestion control mechanism can lead to the loss or delay of critical control commands due to network congestion.

[0004] In summary, existing wind farm communication networks have significant shortcomings in terms of frequency band adaptability, energy support capability, routing efficiency, and congestion control, making it difficult to meet the comprehensive requirements of future intelligent wind farms for high reliability, high bandwidth, and low power consumption communication. Summary of the Invention

[0005] This application provides a multi-band converged wind farm communication network system based on mobile ad hoc networks, which aims to solve the problems in the existing single-band wireless network that is difficult to balance coverage distance and transmission rate, and the fact that wind turbine nodes usually rely on battery power and have limited energy.

[0006] A multi-band converged wind farm communication network system based on mobile ad hoc network includes: multiple wind turbine nodes, each wind turbine node being deployed at wind turbines arranged along a strip area within the wind farm, and each wind turbine node including the following modules: a multi-band communication module, providing at least two independently operating radio frequency transceiver units for each wind turbine node, corresponding to a low-frequency band communication submodule and a high-frequency band communication submodule respectively, and adaptively switching frequency bands based on link quality;

[0007] The energy sensing and acquisition module collects and converts environmental energy into electrical energy for storage, and monitors and controls node energy consumption;

[0008] The strip self-organizing network routing module utilizes the strip distribution characteristics of wind turbines to construct and maintain multi-hop wireless communication paths;

[0009] The dynamic rate control module monitors network congestion in real time and dynamically adjusts the data transmission rate.

[0010] And a central dispatch module, deployed in the wind farm control center, communicates with each wind turbine node via a wireless link to perform centralized monitoring, routing optimization and energy dispatch of all network nodes.

[0011] Optionally, the multi-band communication module includes a baseband processing unit, which is used to perform frequency band adaptive switching, including: establishing a node state vector containing the node's remaining energy, data queue length, real-time wind speed, and link quality indication values ​​for each frequency band;

[0012] Based on the node state vector, calculate the low-frequency band optimization factor and the high-frequency band optimization factor respectively;

[0013] In each decision cycle, the low-frequency band preference factor is compared with the high-frequency band preference factor. If the difference exceeds a preset threshold, a switch from the current frequency band to the target frequency band is executed.

[0014] Optionally, the multi-band communication module is also used to perform dual-band cooperative transmission, always transmitting control signaling through the low-frequency band, and dynamically allocating data load to the high-frequency band or low-frequency band for parallel transmission or redundant transmission according to service requirements or link quality.

[0015] Optionally, the energy sensing and acquisition module includes:

[0016] The hybrid energy harvesting unit, comprising a wind power generation unit, a photovoltaic power generation unit, and a piezoelectric energy harvesting subunit, is used to harvest energy from wind, light, and vibration.

[0017] The energy management and storage unit is used to collect, convert, and store the collected energy in the energy storage battery pack.

[0018] The energy consumption monitoring and control unit is used to monitor the energy consumption of each component in the node in real time and to manage the power supply of each component independently.

[0019] The wind turbine node also includes an energy prediction unit, which is used to generate a predicted energy harvesting power curve for future periods based on historical data and environmental parameters using a combined prediction model.

[0020] Optionally, the wind turbine node further includes a dynamic scheduling unit, used to: prioritize tasks according to their communication task types;

[0021] Based on the current remaining power and the energy harvesting power prediction curve, assess the energy availability index within the future time window;

[0022] The energy security level of a node is determined based on the energy availability index.

[0023] Based on the energy security level and task priority, the access, transmission parameters and execution timing of communication tasks are dynamically scheduled.

[0024] Optionally, the strip ad hoc network routing module is used to perform: restricted flooding along the axial direction of the strip subnet to suppress the broadcast diffusion of routing request packets; and to select a path using a comprehensive routing metric method, which integrates the node's remaining energy factor, the expected number of link transmissions, and the number of hops.

[0025] During the route discovery process, a threshold of the node's remaining energy is used to determine whether the node is allowed to participate in forwarding route request packets.

[0026] Optionally, the strip ad hoc network routing module is further configured to: during the route discovery process, the destination node selects the path with the smallest metric value as the primary route based on the multiple route request packets received, and the path with the second smallest metric value that does not intersect with the primary route node as the backup route.

[0027] By monitoring link quality, triggering updates based on node energy changes, or conducting periodic assessments, the primary and backup routes are dynamically maintained and switched.

[0028] Optionally, the dynamic rate control module includes a congestion prediction unit, used to: divide the network congestion level into multiple discrete states and establish a hidden Markov model;

[0029] Using queue length, collision rate, and channel busy / idle status as observation vectors, the probability of entering a congestion state at future times is predicted using the hidden Markov model.

[0030] When the probability exceeds a preset threshold, a congestion warning is triggered.

[0031] Optionally, the dynamic rate control module further includes a rate decision and execution unit, which is used to execute one or more rate adjustment strategies, including application layer transmit rate adjustment, MAC layer contention window dynamic adjustment, frame aggregation degree adjustment and transmit power adjustment, according to priority when a congestion warning is triggered.

[0032] Optionally, the central scheduling module is used to: collect the status information of each wind turbine node at a fixed period or in an event-triggered manner, and construct and maintain a network topology map;

[0033] Based on the overall network topology, node remaining energy, node load, and service priority, calculate and distribute the main route and backup route across subnets to each wind turbine node;

[0034] Based on the energy prediction curves uploaded by each node, low-energy nodes are identified, and traffic is guided to bypass the low-energy nodes in the global routing calculation, or a cross-node energy coordination mechanism is initiated when there is regional energy shortage.

[0035] Real-time detection of node or link failures and triggering corresponding self-healing operations, including route switching and gateway reselection.

[0036] Compared with the prior art, this application has at least the following beneficial effects:

[0037] This application integrates low- and high-frequency dual-band communication modules at each wind turbine node. The low-frequency band ensures highly reliable long-distance transmission of critical data such as control commands, while the high-frequency band supports the bandwidth requirements of high-capacity services such as video inspection. Based on a multi-dimensional state-aware adaptive frequency band switching method, the optimal frequency band can be dynamically selected according to link quality, node energy, and service type, effectively avoiding dynamic blockage caused by wind turbine blade rotation. The dual-frequency collaborative transmission mechanism further realizes load sharing and redundancy backup, ensuring overall communication continuity and stability. Through multi-band fusion and adaptive switching technology, the communication reliability in the complex environment of wind farms is improved.

[0038] This application integrates three energy harvesting methods—wind, solar, and vibration—to fully utilize the abundant environmental energy resources of wind farms. It proposes an energy prediction model based on ARIMA-LSTM-physical mechanism fusion, which can accurately predict available energy for the next few hours. A dynamic scheduling algorithm built upon this model performs admission control, transmission parameter optimization, and transmission timing selection based on task priority and energy availability. It prioritizes critical services during energy shortages and fully utilizes channel resources when energy is abundant, extending the operational lifespan of nodes and the overall network. Through hybrid energy harvesting and intelligent energy scheduling technologies, it enhances the energy support capabilities of nodes and network lifetime.

[0039] This application leverages the physical characteristic of wind turbines being distributed in a strip along roads or power transmission lines to design a strip-shaped restricted flooding mechanism. This mechanism confines routing requests to the effective propagation direction, significantly reducing redundant broadcasts. A comprehensive routing metric method integrates node remaining energy, link quality, and hop count to guide data flow through energy-sufficient and reliable paths. Combined with backup routing and local repair mechanisms, rapid network fault recovery is achieved. An energy-aware forwarding suppression mechanism effectively protects low-energy nodes, preventing premature failure and extending the overall network lifetime. Through a dedicated routing protocol for strip topologies, routing overhead is reduced and energy-balanced transmission is achieved. Attached Figure Description

[0040] Figure 1 This application provides a schematic diagram of the overall architecture of a multi-band converged wind farm communication network system based on a mobile ad hoc network.

[0041] Figure 2 This is a schematic diagram of the internal module connection of a single wind turbine node in a multi-band converged wind farm communication network system based on a mobile ad hoc network, provided as an embodiment of this application. Detailed Implementation

[0042] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-2 The present application will be further described in detail with reference to the embodiments.

[0043] This application provides a multi-band converged wind farm communication network system based on a mobile ad hoc network, specifically including: multiple wind turbine nodes, each wind turbine node being deployed at wind turbines arranged along a strip area within the wind farm, and the wind turbine nodes including:

[0044] Multi-band communication module, energy sensing and acquisition module, strip self-organizing network routing module, dynamic rate control module, central scheduling module;

[0045] The multi-band communication module integrates at least two independently operating radio frequency transceiver units for each wind turbine node, corresponding to a low-frequency band communication submodule and a high-frequency band communication submodule, respectively. Specific configurations include:

[0046] The low-frequency communication submodule, employing the Sub-1GHz band (preferably 433MHz or 915MHz), is based on IEEE 802.15.4 or LoRa modulation technology. It possesses strong diffraction capability and resistance to multipath fading, and is used in low-data-rate, high-reliability transmission scenarios such as command issuance and node status reporting for wind turbine control systems. This submodule has an adjustable transmit power range of 10dBm to 30dBm, a receive sensitivity of no less than -120dBm, and supports long-distance communication (typical coverage radius of no less than 1km).

[0047] The high-frequency communication submodule uses the 2.4GHz or 5.8GHz frequency band, is based on the IEEE 802.11g / n or 802.11ac standard, and supports a physical layer transmission rate of up to 300Mbps. It is used for high-capacity data services such as wind turbine inspection video transmission, batch firmware upgrades, and fault waveform data upload. This submodule employs a MIMO antenna array (2×2 or 4×4) with beamforming capabilities, enabling it to maintain link stability in dynamic environments caused by wind turbine rotation.

[0048] The baseband processing unit adopts an FPGA+ARM heterogeneous architecture and is responsible for tasks such as protocol stack processing, data splitting and fusion, and frequency band switching decisions for the two frequency band sub-modules. The baseband unit has a built-in channel state monitoring module that collects link quality parameters such as RSSI, LQI, SNR, and bit error rate in real time.

[0049] Each frequency band submodule is connected to an independent antenna port, and the antenna layout adopts a spatial diversity approach to avoid co-channel interference. The low-frequency antenna is mounted on the top of the nacelle in the form of a whip or dipole; the high-frequency antenna is embedded in the blade root or nacelle sidewall in the form of a patch array to reduce wind resistance.

[0050] To address the insufficient adaptability of a single frequency band in the complex environment of wind farms, a frequency band adaptive switching method based on multi-dimensional state awareness is provided, comprising the following steps:

[0051] Step 1: Establish the node state vector. ;

[0052] in, Remaining energy at the node (unit: J). The length of the data queue to be sent (unit: packet). Real-time wind speed (unit: m / s), used to predict the shading period of wind turbine blades. This is the current transmission power. and These are the link quality indication values ​​for low- and high-frequency links, respectively.

[0053] Step 2: Calculate the frequency band optimization factor, and define the low-frequency band optimization factor separately. and high-frequency band optimization factor as follows:

[0054]

[0055]

[0056] in, This is the link quality threshold. At maximum transmission power, Where D is the maximum communication distance and D is the actual distance. For the bandwidth required by the business, For the available bandwidth of the high-frequency band, The current link transmission delay, Let be the weighting coefficient, satisfying It can be dynamically adjusted according to the type of business (e.g., increasing the efficiency of control-related businesses). Video services improved );

[0057] Step 3: Switch between decision and execution. The node operates in each decision cycle (configurable as follows). (seconds) calculation and And execute the following judgment logic:

[0058] If the current frequency band is low and If so, then switch to the higher frequency band;

[0059] If currently using high frequency band and If so, it will switch to the low frequency band;

[0060] like If so, the current frequency band will be maintained;

[0061] The switching execution process includes:

[0062] The baseband unit sends a frequency band switching request frame (including the target frequency band and switching time) to the peer node.

[0063] After confirmation from the peer node, both parties synchronously switch radio frequency channels;

[0064] After the switch is completed, a test frame is sent to verify the link. If it fails three times in a row, it will fall back to the original frequency band.

[0065] To further improve spectrum resource utilization, a dual-frequency cooperative transmission method is also provided, specifically including:

[0066] The communication content is divided into control signaling (such as route updates and heartbeat packets) and data payload (such as sensor data and video streams). Control signaling is always transmitted through low-frequency bands to ensure high reliability and low latency; data payloads are dynamically allocated to high-frequency or low-frequency bands according to service requirements.

[0067] When the high-frequency link quality is good and the queue length exceeds the threshold, the baseband unit starts the load balancing mode, diverting some data to the low-frequency band for parallel transmission to reduce end-to-end latency. The diversion ratio is determined by the current link rate and the queue backlog.

[0068] If a link interruption occurs during high-frequency transmission, unacknowledged data packets are automatically retransmitted to the low-frequency band to avoid service interruption. Meanwhile, critical data transmitted in the low-frequency band is redundantly transmitted in the high-frequency band to improve the data delivery success rate.

[0069] Furthermore, to address potential packet loss and latency jitter issues during frequency band switching, the following optimization measures were introduced:

[0070] Switching the prediction mechanism, based on the rate of change of wind speed and the turbine rotation cycle, to predict the future The probability of blade blockage within seconds is used to trigger frequency band switching in advance, avoiding switching during periods of rapid link deterioration.

[0071] In the soft handover process, the two frequency bands maintain connection simultaneously during the handover. The old frequency band resources are released only after the link of the new frequency band is stable, thus achieving seamless handover.

[0072] The switching history learning process involves nodes recording the triggering conditions for each switch and the link quality after the switch, and using online learning algorithms (such as Q-learning) to dynamically optimize the switching threshold. With weighting coefficients This enables the switching strategy to adapt to changes in the wind farm environment during different seasons and time periods.

[0073] The energy sensing and acquisition module includes a hybrid energy acquisition unit, an energy management and storage unit, and an energy consumption monitoring and control unit.

[0074] The hybrid energy harvesting unit includes a wind power generation unit, which uses a miniature vertical axis wind turbine generator, installed on the top of the wind turbine nacelle or the side of the tower, with a rated power of 5W. 20W, with a starting wind speed of not less than 2.5m / s and an output voltage of 12V / 24V at rated wind speed. This sub-unit includes a maximum power point tracking (MPPT) controller, which uses the disturbance observation method to adjust the load impedance in real time so that the fan operates at the optimal power curve.

[0075] The photovoltaic power generation unit uses flexible thin-film solar cells, which are attached to the sun-facing side of the nacelle or tower, with a total peak power of 10W. 30W, the surface of the solar panel is coated with a hydrophobic and dustproof coating to adapt to the high dust environment of wind farms. The photovoltaic sub-unit is also equipped with an independent MPPT controller and uses the conductivity incremental method to achieve maximum power point tracking.

[0076] The piezoelectric energy harvesting subunit embeds piezoelectric ceramic plates at the root of the wind turbine blades or at the vibration nodes of the tower to harvest the mechanical vibration energy generated during the operation of the wind turbine. After rectification and voltage boosting, the energy is output to the energy storage unit as a supplementary energy source.

[0077] The energy management and storage unit includes a multi-source energy combiner controller, which converts the outputs of the three acquisition units mentioned above into DC power after DC / DC conversion and combines them into a DC bus. The bus voltage is stabilized at 24V±5%. The combiner controller has input overvoltage protection, reverse connection protection and backflow prevention functions.

[0078] The energy storage battery pack uses lithium iron phosphate batteries with a nominal voltage of 24V and a capacity of not less than 20Ah. It supports deep charge and discharge cycles (≥2000 times). The battery pack has a built-in battery management system (BMS) that monitors the voltage, temperature, and charge and discharge current of individual cells in real time. It has overcharge, over-discharge, and over-temperature protection as well as equalization management functions.

[0079] The energy monitoring unit, based on Hall sensors and precision shunt resistors, collects the instantaneous power, cumulative power, and state of charge (SOC) of each collection branch in real time. The SOC estimation adopts the ampere-hour integration method and the open-circuit voltage method for fusion correction, and the error is controlled within ±3%.

[0080] The energy consumption monitoring and control unit includes a node energy consumption monitoring module, which sets a high-precision current detection chip (such as INA219) at the power supply input of the node main control board to monitor the instantaneous power consumption and cumulative energy consumption of the communication module, sensor module, and data processing module in real time, and statistically analyzes the energy consumption distribution by device dimension.

[0081] Programmable power management unit: It adopts a multi-channel independently controlled DC / DC converter to power the low-frequency RF module, high-frequency RF module, main control chip and sensor group respectively. Each channel can be independently turned off or the output voltage can be adjusted to achieve sub-millisecond power consumption switching.

[0082] To address the challenges of highly volatile energy harvesting and difficulties in planning communication tasks in wind farm environments, an energy prediction method based on hybrid time series analysis and machine learning is provided, specifically including the following steps:

[0083] 1. Historical data preprocessing: Nodes record the following historical data sequences at fixed intervals (e.g., 15 minutes): Wind power generation sequence Photovoltaic power generation sequence Wind speed sequence Light intensity sequence Temperature sequence Historical SOC sequences ;

[0084] Use 3 Outliers are removed in principle, and missing data is filled in using linear interpolation.

[0085] 2. Multi-timescale feature extraction and construction of feature vectors ;

[0086] in, Hourly codes for the day (0 twenty three), For seasonal coding (spring, summer, autumn, winter), n is the length of the historical time window, with a value of 4. 8 (corresponding to 1) (2-hour historical data)

[0087] 3. The fusion prediction model is constructed using a combined prediction strategy, comprising the following three sub-models:

[0088] ARIMA time series model: A differential autoregressive moving average model is established on the stationary power series to capture short-term cyclical patterns and predict future trends. Power variation trend over 6 hours;

[0089] LSTM neural network model: based on feature vectors Using this as input, a network structure is constructed containing two LSTM layers (64 neurons per layer) and one fully connected output layer to learn the nonlinear mapping relationship between factors such as wind speed, sunlight, and temperature and power generation, outputting the future 1 6-hour power forecast;

[0090] Physical mechanism model: Based on the formula of wind turbine power curve With photovoltaic physical model Construct a mechanism model, in which air density, For the area swept by the wind turbine, The wind energy utilization coefficient, For the tip speed ratio, The pitch angle is the propeller angle. Where S is the photovoltaic conversion efficiency and S is the area of ​​the photovoltaic panel;

[0091] 4. An adaptive weighted fusion algorithm is adopted to dynamically adjust the weights based on the recent prediction errors of each sub-model:

[0092]

[0093]

[0094] in, Let be the variance of the prediction error of the i-th sub-model over the past 24 hours. To prevent division by zero for extremely small positive numbers, the fused prediction results are updated every 15 minutes, outputting the future 1. 6-hour energy harvesting power prediction curve;

[0095] 5. Prediction error correction: Introducing a real-time correction step based on Kalman filtering: When the deviation between the actual collected power and the predicted value exceeds a set threshold (e.g., 20%), the correction algorithm is activated to update the model parameters or switch the dominant sub-model using the latest observations, so that the prediction quickly converges to the true trend.

[0096] Furthermore, a dynamic scheduling algorithm that comprehensively considers the current energy state, energy prediction results, and communication task priorities is proposed, specifically including the following process:

[0097] 1) Task classification and priority definition

[0098] Node communication tasks are divided into the following three categories, and each is assigned a basic priority:

[0099] Critical control categories include: wind turbine start / stop commands, emergency alarms, and routing control frames. These have high basic priority (P=3) and must respond immediately without delay.

[0100] Routine monitoring includes periodic reporting of sensor data such as wind speed, power, and temperature. In the basic priority (P=2), a certain delay is allowed, but data integrity must be guaranteed.

[0101] Non-real-time data includes: video inspection feedback, log file upload, and firmware upgrade. These have a low basic priority (P=1) and are allowed to have significant delays or intermittent transmission.

[0102] 2) Energy availability assessment, defining the energy availability index of a node within a future time window [t, t+T]:

[0103]

[0104] in, This is the current remaining battery level. The average power for the k-th prediction period. To predict the step size (take 15 minutes). The minimum energy consumption required to maintain basic node operation (including sensor sampling, main control standby, etc.) during the future TT time period;

[0105] 3) Dynamic scheduling decision model: In each scheduling cycle (e.g., 30 minutes), the node executes the following scheduling algorithm:

[0106] Step 1: Energy safety level classification, based on The nodes are divided into three energy levels:

[0107] Sufficient energy state: It allows all types of tasks to be executed;

[0108] Energy balance state: Only tasks with a priority of ≥2 are allowed to be executed, and non-real-time data tasks are restricted.

[0109] Energy stress state: Only tasks with priority=3 are allowed to execute, all non-critical communications are suspended, and the system enters power-saving mode.

[0110] in and For configurable thresholds, a value is typically set to... , ;

[0111] Step 2: Queue Management and Task Admission, maintaining three priority queues. , , These are used to store high, medium, and low priority tasks to be sent. At each scheduling moment, the following admission control is executed based on the current energy level:

[0112] If energy is sufficient: allow all queued tasks to participate in scheduling;

[0113] If energy balance is achieved: only allow and Task participation in scheduling, Task suspended;

[0114] If energy is tight: only allow The task participates in scheduling, the remaining queues are suspended, and the power-saving mode is triggered;

[0115] Step 3: Transmission parameter optimization. For tasks allowed to participate in scheduling, further optimize their transmission parameters to match the current energy situation:

[0116] Transmit power adjustment: Dynamically adjust the RF transmit power based on the current link quality and energy level. When energy is scarce, reduce the transmit power to the lowest feasible value while ensuring communication success rate; when energy is sufficient, higher power can be used to increase the transmission rate.

[0117] Data compression and aggregation: When energy is scarce, for Task data can be lightweight compressed (e.g., differential coding, run-length coding), or multiple small data packets can be aggregated into a single large packet for transmission, reducing the number of transmissions.

[0118] Transmission Timing Selection: Based on energy forecast results, non-urgent tasks are postponed to the predicted peak energy harvesting period. For example, if increased sunlight is predicted within the next two hours, video inspection tasks will be postponed to that time period.

[0119] Step 4: Cross-node collaborative scheduling. When a node enters a state of energy shortage and critical tasks cannot be guaranteed, the cross-node collaborative mechanism is activated:

[0120] Send assistance requests to nearby nodes with sufficient energy, and delegate some monitoring tasks to them for uploading;

[0121] It accepts temporary route adjustments from the central scheduling module, avoids undertaking data forwarding tasks, and only operates as a leaf node.

[0122] The strip-shaped self-organizing network routing module, deployed within each wind turbine node, works in conjunction with the multi-band communication module and the energy sensing and acquisition module to construct and maintain multi-hop wireless communication paths within the wind farm environment. Considering the physical layout of wind turbine nodes along roads or transmission lines, which are linearly or nearly linearly distributed, this module employs a dedicated routing protocol for strip topologies. This ensures communication reliability while minimizing control overhead, as detailed below:

[0123] Network topology modeling and node role assignment: Wind turbine clusters within a wind farm are divided into several strip subnets based on their geographical distribution characteristics. Each strip subnet consists of turbines arranged along the same wind duct or transmission line. The distance between adjacent nodes is typically between 100 and 300 meters, and the node communication radius is designed to cover 2 to 3 adjacent nodes, thus ensuring connectivity while avoiding excessive overlap. Each strip subnet uses the same communication channel, while adjacent subnets are allocated different channel resources to prevent co-channel interference. Nodes located at subnet boundaries are designated as gateway nodes. These nodes must operate on two channels simultaneously and can achieve cross-subnet data forwarding through dual-RF modules or time-division multiplexing.

[0124] Based on the functions of nodes in the routing protocol, each wind turbine node can be dynamically assigned to one or a combination of the following roles: Ordinary nodes are only responsible for data acquisition and transmission within their own subnet, participating in intra-subnet routing but not undertaking cross-subnet forwarding tasks; relay nodes, in addition to their own services, also need to forward data from other nodes within the same subnet, and are elected by the routing algorithm based on the node's remaining energy, link quality, and topological location; gateway nodes are located at the subnet boundary, equipped with dual-frequency communication capabilities, responsible for connecting adjacent subnets and forwarding cross-subnet data; boundary nodes are nodes within the subnet closest to the gateway, undertaking data aggregation functions, aggregating data from their own subnet destined for other subnets and submitting it to the gateway. Node roles are not fixed but can be dynamically adjusted according to network topology changes, such as node failures or additions, to adapt to dynamic changes during wind farm operation and maintenance.

[0125] To adapt to the unique characteristics of wind farm environments where node energy is limited and link quality is easily affected by weather conditions, this invention designs a comprehensive routing metric method that considers three dimensions: remaining node energy, link quality, and hop count. The smaller the metric value, the better the path, thus guiding route selection.

[0126] The node energy factor is defined as the ratio of the current remaining energy to the initial energy, i.e. This value ranges between 0 and 1, with a higher value indicating more abundant node energy. Link quality is measured by the expected number of transmissions, and its calculation formula is as follows: ,in Let be the forward delivery success rate from node i to node j. The reverse delivery success rate is obtained by periodically exchanging probe packets between nodes. The smaller the ETX value, the better the link quality.

[0127] Based on this, the comprehensive path metric from the source node to the current node j is defined as follows:

[0128]

[0129] in, The energy factor of the previous hop sending node. This represents the number of hops from the source node to the current node. , , The weighting coefficients are satisfied. The weighting coefficients can be dynamically adjusted according to the network status; for example, they can be appropriately increased when the average energy level of the entire network is low. The value of is chosen to enhance the guiding role of energy factors. This metric is accumulated hop by hop during the route discovery process. The destination node ultimately selects the path with the smallest metric as the primary route and the path with the second smallest metric as the backup route to achieve rapid failover.

[0130] Furthermore, to reduce the control overhead in the route discovery process, this invention utilizes the characteristics of strip topology to design a strip-restricted flooding mechanism, which ensures that route requests propagate only in the strip direction, effectively suppressing broadcast diffusion to useless directions.

[0131] Each node is pre-configured with the axis direction vector of its subnet. This vector is determined during the wind farm planning phase based on the road or transmission line direction and is fixed in the node configuration. Nodes know their own geographical location through a built-in GPS module or preset coordinates and can calculate the position vectors of any neighboring nodes relative to themselves. When a node receives a routing request packet, it first calculates the angle between the neighbor's position vector and the subnet axis direction vector. If this angle is less than a preset strip angle threshold (usually 45°), the neighbor is considered to be within the strip direction, and forwarding is allowed; otherwise, the routing request packet is discarded and no further processing is performed.

[0132] To protect low-energy nodes from premature energy depletion due to frequent participation in routing forwarding, this module introduces an energy-aware forwarding suppression mechanism. When a node's remaining energy falls below a preset threshold (e.g., 20% of its initial energy), the node will not participate in forwarding routing request packets, unless it is a destination node or a gateway node. This effectively extends the lifespan of low-energy nodes while preventing frequent route breaks caused by node energy depletion.

[0133] The route request packet adds several fields to the standard AODV protocol format, including the source node's current energy factor, the cumulative aggregate metric value from the source to the current node, the minimum energy factor of all nodes on the path for energy balance judgment, and an optional striped direction vector. When the source node initiates route discovery, it initializes the cumulative metric to zero and the minimum path energy value to its own energy factor, then broadcasts the route request packet to all neighbors. Neighboring nodes sequentially perform striped direction judgment and energy threshold check, and then update the cumulative metric and minimum path energy value based on the energy factor of the previous hop node and the link ETX. If the metric value provided by the route request packet is better than the existing record, the routing table is updated and forwarding continues; otherwise, it is discarded. The destination node may receive multiple route request packets, selects several paths with the smallest metric values ​​as candidate routes, and transmits them back to the source node along the reverse path via a route reply packet.

[0134] Furthermore, the route reply packet generated by the destination node is unicast back to the source node along the reverse path established by the route request packet. The reply packet carries the cumulative metric value of the path, the minimum path energy, and the backup path identifier. After receiving the route reply packet, the intermediate node establishes a forward route to the destination node and continues to forward it towards the source node. The source node may receive multiple route reply packets, selecting the path with the smallest metric value as the primary route and the path with the second smallest metric value that does not intersect with the primary route node or has no intersecting links as the backup route. The establishment of backup routes allows for rapid switching when the primary route fails, avoiding the latency and overhead of re-initiating route discovery.

[0135] Each node maintains a routing table entry containing the destination node identifier, next-hop node identifier, hop count, path metric, minimum path energy, backup next-hop node identifier, route validity period, and a route flag (indicating whether the entry is a primary route, a backup route, or has expired). The routing table provides fast lookup support during data forwarding.

[0136] Furthermore, route maintenance and updates are accomplished collaboratively through mechanisms such as link quality monitoring, energy-triggered updates, local route repair, and periodic optimization.

[0137] Nodes periodically send Hello packets to their neighbors and calculate the ETX value of their links with each neighbor. When the ETX value exceeds a preset threshold (e.g., 3.0), the link quality is considered to have severely degraded, triggering a local route repair process. When a node's energy factor changes by more than a certain amount (e.g., a decrease of 10%), it proactively broadcasts an energy update announcement to its neighbors. Nodes receiving the announcement recalculate the routing metrics related to that node. If the current route's metric is worse than the backup route, it proactively switches to the backup route and may choose to initiate a new route discovery to update the backup route.

[0138] When the next-hop link of the primary route is interrupted or fails due to node energy depletion, the node attempts to repair the local route by broadcasting a limited-hop-count local route request packet to its neighbors to find an alternative path to the original destination node. If the local repair is successful, the routing table is updated; otherwise, a routing error packet is sent to the upstream node, triggering the source node to switch to the backup route or re-initiate a full network route discovery.

[0139] To balance network load and adapt to dynamically changing energy and link conditions, nodes evaluate the metrics of the current primary route and backup route at regular intervals (e.g., every 10 minutes). If the metric of the backup route is significantly better than that of the primary route (e.g., the metric is more than 20% smaller), the backup route is actively promoted to the primary route, and a new backup route is selected to achieve a balanced distribution of load.

[0140] Furthermore, cross-subnet communication relies on gateway nodes. Gateway nodes periodically broadcast gateway advertisements within their subnet, containing their own address, subnet identifier, and metric information to neighboring subnets. Ordinary nodes identify the nearest gateway node by listening to these advertisements. When a source node needs to send data to a node in another subnet, it first generates a route to its own subnet's gateway, sends the data to the gateway, and the gateway forwards it to neighboring subnets. Gateway nodes maintain routes to neighboring subnet gateways, using a striped routing protocol similar to that within the subnet, but the striped directional constraints can be relaxed to ensure cross-subnet connectivity.

[0141] Gateway nodes need to coordinate their operation on two channels. After receiving data on a channel within the subnet, if forwarding to an adjacent subnet is required, the node switches to the adjacent subnet's channel for transmission. The timing of channel switching is coordinated by the MAC layer to ensure that no data packet loss occurs during the switching process.

[0142] Furthermore, a standardized interface is established between the strip ad hoc network routing module and the energy sensing and acquisition module. This interface periodically acquires the following information: the remaining energy factor of the current node, the energy level classification results, and the energy prediction curve for future periods. The routing module dynamically adjusts the weighting coefficients in the comprehensive metric based on the energy level, appropriately increasing the link quality weight when energy is sufficient. Prioritize communication performance and increase the energy factor weight during energy shortages. Prioritizing extending node lifetime and correspondingly reducing the node's enthusiasm for participating in routing and forwarding. Through this cross-layer collaborative design, routing decisions can respond promptly to changes in node energy status, maximizing the overall network lifetime.

[0143] The dynamic rate control module includes the following functional units:

[0144] The queue status monitoring unit monitors in real time the length Q(t) of the network layer output queue (i.e., the MAC layer sending queue), the average arrival rate λin(t), and the average service rate μ(t) of the queue. The queue length is counted using a counter to determine the current number of buffered packets; the arrival rate is calculated using an exponentially weighted moving average (EWMA).

[0145]

[0146] in Sampling period The number of new data packets arriving within 100ms (e.g.) This is a smoothing factor (usually taken as 0.8). 0.9). Service rate is obtained by measuring the time interval between successful packet transmissions;

[0147] The channel state monitoring unit, based on statistical information provided by the 802.11 MAC layer, collects the following indicators:

[0148] Channel busy / idle The percentage of time during which physical carrier sensing is "busy" within the sampling period;

[0149] Collision rate The percentage of failed transmissions (due to collisions) within the sampling period out of the total number of transmission attempts;

[0150] Average retreat time : The average backoff time of a node before each transmission within the sampling period;

[0151] Retransmission count distribution: Record different retransmission counts (1 7) The frequency of occurrence is used to assess link stability;

[0152] The congestion prediction unit is based on historical state sequences and uses a combination of Hidden Markov Model (HMM) and linear regression to make short-term predictions of network congestion trends.

[0153] Specifically, the short-term congestion prediction method based on the combination of Hidden Markov Models (HMM) and linear regression includes the following steps:

[0154] Step 1: Define the congestion state by dividing the network congestion level into K discrete states (e.g., K=3):

[0155] state (Low load): Queue length Collision rate Channel busy / idleness ;

[0156] state (Medium load): ,or ,or ;

[0157] state (congestion): ,or ,or ;

[0158] threshold , , , , , It can be statically configured or dynamically adaptively adjusted based on network size and service type (e.g., based on the 95th percentile of historical statistics).

[0159] Step 2: Hidden Markov Model Training, with sampling period Using units of 1 second (e.g., 1 second), a multidimensional observation vector is observed at each time t. Assume there exists a hidden sequence of congestion states. Control the probability distribution of observed values;

[0160] The specific model parameters are as follows:

[0161] Initial state distribution ;

[0162] State transition probability matrix ,in ;

[0163] Observational probability distribution Assume that the components of the observation vector are independent and follow a Gaussian distribution:

[0164]

[0165] The model parameters are continuously updated using the online EM algorithm (Baum-Welch algorithm) to adapt the model to dynamic changes in the network. The update window can take the past N samples (e.g., N=100), and each... (For example, every 10 seconds) Re-estimate the parameters;

[0166] Step 3: Congestion Trend Prediction. Given the current time t and the previous observation sequence O(1:t), use the Viterbi algorithm to estimate the most likely hidden state. Then, based on the state transition matrix A, the state probability distribution for the next h steps (e.g., h=5 steps, corresponding to 5 seconds) is predicted:

[0167]

[0168] in Given the current state distribution vector (obtainable from a forward-backward algorithm), define the state that will enter congestion within h steps in the future. The probability of () is:

[0169]

[0170] like Exceeding the threshold If the value is 0.6, it is determined that "congestion is about to occur" and a rate adjustment is triggered.

[0171] Step 4: Linear Regression Correction. To enhance response to sudden business disruptions, a linear regression model is used to make short-term predictions of queue length trends.

[0172]

[0173] in This is a correction coefficient (<1), used for smoothing predictions. If If it does, a congestion warning will be triggered immediately, regardless of the HMM prediction period.

[0174] The two prediction results are logically ORed to form the final congestion warning signal.

[0175] Furthermore, when a congestion warning is triggered or the current congestion state has entered... When this occurs, the rate decision and execution unit initiates the following adjustment strategies, executing them in priority order:

[0176] Application-layer transmission rate adjustment dynamically adjusts the data rate injected into the network by upper-layer applications based on the level of congestion.

[0177]

[0178] in This is the descent step size factor (e.g., 0.2). This is the normalization factor (taken as 1.0). If the congestion state is... Then set directly (A pre-configured minimum rate, e.g., 1 packet / s). Once congestion is relieved, the rate recovers exponentially: after each congestion-free period, , where η is the growth factor (e.g., 0.1);

[0179] The MAC layer contention window is dynamically adjusted based on the collision rate. Adjusting the minimum competition window To reduce the probability of collision:

[0180]

[0181] in For adjustment factors (e.g., 2.0), and It must not exceed the maximum value specified in the standard (e.g., 1023). When After reduction, Gradually restore to the initial value;

[0182] Frame aggregation degree adjustment: If the node supports 802.11n / ac frame aggregation (A-MSDU, A-MPDU), the aggregation degree can be dynamically adjusted according to the congestion level.

[0183]

[0184] The more severe the congestion, the lower the aggregation degree, in order to reduce the retransmission cost of a single transmission failure. Aggregation degree can be increased under low load to improve channel efficiency.

[0185] Transmit power adjustment, combined with the energy level of the energy sensing module, can be appropriately reduced during congestion to reduce interference, but basic connectivity must be maintained. Power adjustment follows these rules:

[0186]

[0187] The specific power value is predetermined based on the link budget (e.g.) ).

[0188] Multi-priority queue support: To ensure the quality of service for critical business operations (such as control commands), nodes maintain multiple priority queues (high / medium / low). During rate adjustment:

[0189] High-priority queues are not subject to application layer rate limits;

[0190] The medium priority queue is adjusted as described above using rapprapp;

[0191] Low-priority queues can be completely blocked from sending during congestion until the congestion is relieved.

[0192] Furthermore, it also has a coordination mechanism with the strip routing module, as detailed below:

[0193] Congestion status announcements are made by nodes periodically (e.g., every 5 beacon cycles) broadcasting their local congestion status within the strip subnet. After receiving the data, the neighboring node incorporates a congestion penalty factor into the routing metric calculation.

[0194]

[0195] in A penalty value (e.g., 5.0) is used to cause the routing algorithm to actively avoid congested nodes;

[0196] Routing switch is triggered when a node itself enters a congested state. When congestion occurs, the node proactively sends a "congestion notification" to the upstream node, suggesting that the upstream node switch to a backup route (if one exists). At the same time, the node can temporarily reduce its relay participation (e.g., stop forwarding low-priority data) until the congestion is relieved.

[0197] Rate negotiation in local route repair involves the node initiating the repair carrying the maximum allowed sending rate. Nodes on the new path must ensure that they can support this rate; otherwise, they will refuse to join the path.

[0198] The central dispatch module, deployed at the wind farm's booster station or control center, serves as the core of the entire communication network management. It is responsible for centralized monitoring, routing optimization, and energy dispatching of all wind turbine nodes in the network. It maintains real-time connection with each wind turbine node through wireless communication links, collects node status information, analyzes the network operation status, and issues control commands to achieve global optimization of network resources.

[0199] The central scheduling module adopts an industrial-grade redundant architecture at the hardware level to ensure high reliability even in harsh industrial environments. The system is configured with two high-performance servers (one primary and one backup) as core computing units. Each server uses dual Xeon processors, 128GB ECC memory, and a 2TB RAID5 disk array, running a real-time optimized operating system. The communication front-end is responsible for establishing physical links with wind turbine nodes, equipped with a multi-channel 4G / 5G wireless module as the primary communication method and a satellite communication module as a backup link. A fiber optic Ethernet interface is also reserved for interfacing with the wind farm's existing monitoring system. The communication front-end supports multiple industrial protocol conversions and is compatible with underlying equipment from different manufacturers. The data storage array adopts a distributed architecture to store historical monitoring data, routing table snapshots, energy scheduling logs, and alarm records, with a data retention period of no less than three years. The clock synchronization unit deploys a GPS / BeiDou dual-mode time receiver, providing microsecond-level time synchronization to all network nodes via the NTP protocol, laying the foundation for time-series data analysis.

[0200] The software architecture adopts a layered and modular design, consisting of a data acquisition layer, a status monitoring layer, a decision control layer, a command issuance layer, and a human-machine interaction layer, from bottom to top. Each layer interacts through standardized interfaces, reducing module coupling and facilitating functional expansion and maintenance. The data acquisition layer is responsible for periodic communication with wind turbine nodes, collecting node status, link quality, energy data, and alarm events, and supports breakpoint resume and data compression to handle temporary communication interruptions. The status monitoring layer cleans, aggregates, and stores the collected data, and calculates the overall network health indicators in real time, providing data support for visualized monitoring. The decision control layer includes a routing calculation engine, an energy scheduling engine, and a fault diagnosis engine, making optimization decisions based on global information. The command issuance layer converts the decision results into specific control commands, issues them to the corresponding nodes, and tracks the execution results. The human-machine interaction layer provides web and mobile access interfaces, supporting multi-user permission management, historical data query, report generation, and remote operation and maintenance functions.

[0201] The central scheduling module sends status request frames to each wind turbine node at a fixed period. The polling period can be configured between 1 and 10 seconds depending on network size and service requirements. Node responses cover multiple dimensions, including identity information, energy status, communication status, routing status, sensor data, and alarm events. Identity information includes node identifier, subnet identifier, node role type, and latitude and longitude coordinates, used to uniquely determine the node's location and function within the network. Energy status includes the remaining percentage of power, instantaneous power of various energy harvesting units, energy level classification results, and an energy prediction curve for the next 6 hours. The prediction curve is generated by the node's local model and uploaded, reflecting the expected changes in the node's energy supply. Communication status includes the current operating frequency band, transmit power, neighbor node list and expected transmission counts for corresponding links, queue length, collision rate, and channel busy / idle status, comprehensively characterizing the node's communication environment and load. Routing status includes the number of routing table entries, the next-hop node of the primary and backup routes, the number of forwarded data packets, and the number of route discovery attempts, used to evaluate the effectiveness of the routing protocol. Sensor data can optionally include wind speed, wind direction, temperature, vibration, and other information for correlation analysis with communication status. Alarm events are reported proactively; nodes report immediately when they detect low energy, link interruption, hardware failure, or congestion warnings, without being limited by polling cycles.

[0202] The construction and maintenance of the network topology are achieved through a combination of initial discovery and incremental updates. Upon system startup, a topology request is broadcast to all known nodes. Nodes return their own neighbor lists, and the central scheduling module iteratively processes these neighbor relationships to construct an initial network topology map. During operation, nodes proactively report topology update events when they detect changes in neighbor relationships. The central scheduling module then updates the topology map accordingly and triggers local route recalculation. Topology visualization uses a GIS map as a background, displaying real-time node locations, link status, node roles, and energy levels. Link status is indicated by color to represent the expected number of transmissions. Zooming, filtering, and historical playback functions are supported, providing operations and maintenance personnel with an intuitive means of network situational awareness.

[0203] The overall network health assessment uses a weighted composite index, which weights and merges four sub-indicators: online node ratio, average energy level, congested node ratio, and average packet delivery rate, to obtain a health score ranging from 0 to 100. When the health score falls below a preset threshold, the system automatically pops up an alarm and prompts maintenance personnel to intervene and investigate. This index can be used to quantitatively assess the overall network operational quality and provide a reference for maintenance decisions.

[0204] Although each wind turbine node possesses local route discovery capabilities, the central scheduling module can calculate optimal cross-subnet routes using a global view and optimize the overall network routing layout through periodic distribution. The route calculation cycle is set to 15 minutes, and real-time recalculation can be triggered when network topology changes exceed a certain threshold. The input to the routing algorithm includes the overall network topology map, the remaining energy factor of each node, the current load of the nodes, and service priority requirements. The output is the primary and backup routes between each pair of source and destination nodes.

[0205] The algorithm aims to comprehensively consider multiple optimization dimensions: minimizing the sum of expected transmissions along the path to ensure communication efficiency; maximizing the minimum energy factor of all nodes along the path to avoid premature path failure due to individual low-energy nodes; balancing network load to prevent critical nodes from prematurely exhausting their energy due to excessive forwarding tasks; and satisfying service priority constraints, such as limiting the number of hops for control-related services to no more than three to ensure low latency. The algorithm implementation employs an improved multi-constraint shortest path method. First, it integrates the energy factor and expected transmission count into a link weight, calculates the top three shortest paths from the source node to the destination node, and then selects paths that satisfy energy constraints and have non-intersecting nodes or links as the primary and backup routes, respectively.

[0206] Routing is distributed using a differentiated strategy to adapt to different scenarios. When a node first connects to the network or its routing table is cleared, the central scheduling module performs a full distribution, sending the complete routing table to the node. During normal operation, incremental distribution is used, distributing only changed routing entries to reduce communication overhead. For network-wide routing information such as gateway addresses, broadcast distribution can be used to send the information to all nodes at once. After receiving the routes distributed by the central module, the nodes merge them with their locally discovered routing tables. The merging rules follow the principle of central route priority, meaning that centrally distributed routes are used as primary routes, and locally discovered routes can be used as supplementary or backup routes. If a conflict occurs between central routes and local routes, the central route takes precedence, and a conflict log is recorded for analysis by the central scheduling module.

[0207] A routing quality monitoring and feedback mechanism ensures continuous optimization of routing decisions. Nodes periodically report the end-to-end latency and packet loss rate of the primary and backup routes to the center. The central scheduling module collects feedback and evaluates the actual performance of the routes. If a route is found to be significantly worse than expected, such as a packet loss rate consistently exceeding 5%, a local route optimization process or a global route recalculation is triggered to ensure that the network always operates in a better state.

[0208] The central scheduling module collects energy prediction curves uploaded by each node to form a network-wide energy prediction distribution map. The node prediction curves cover the changes in energy collection power over the next 6 hours in 15-minute increments. The central scheduling module combines meteorological forecast data to correct the prediction curves, improving prediction accuracy. Based on the current remaining power percentage, all network nodes are divided into three energy levels: high-energy nodes (those with at least 70% remaining power) can serve as primary relays or gateway candidates; medium-energy nodes (those with between 30% and 70% remaining power) participate normally in routing and forwarding; and low-energy nodes (those with less than 30% remaining power) must have their forwarding tasks restricted to prioritize their own data reporting.

[0209] Energy-balanced routing is one of the core methods of energy scheduling. The central scheduling module actively guides traffic to bypass low-energy nodes and prioritize high-energy nodes during global route calculation. Specific implementation methods include introducing an energy penalty factor into link weights to increase the link weights issued by low-energy nodes; setting a forwarding cap for low-energy nodes to limit the number of routes passing through them; and when a low-energy node is about to run out of energy, the central scheduling module can forcibly downgrade it to a leaf node, ceasing its relay tasks and issuing route updates to its neighboring nodes.

[0210] When multiple nodes in a region simultaneously enter a low-energy state, the central scheduling module activates a cross-node energy coordination mechanism. The task migration strategy temporarily relocates the monitoring tasks of low-energy nodes to nearby high-energy nodes, with data uploaded through the high-energy nodes, allowing the low-energy nodes to hibernate or reduce their workload. If a node is equipped with a wireless charging receiver, the central scheduling module can dispatch drones or mobile charging vehicles to the low-energy node for recharging. The hibernation rotation strategy, while ensuring network coverage, arranges for some low-energy nodes to enter deep hibernation, with neighboring nodes covering their monitoring area, until they are awakened again after energy recovery.

[0211] Joint optimization of energy and communication further improves resource utilization efficiency. When the central scheduling module predicts a decrease in wind speed or photovoltaic output, it proactively reduces the communication rate of non-critical services, extending node operating time. During periods of sufficient energy, it actively increases data acquisition frequency or initiates energy-intensive tasks such as firmware upgrades. In conjunction with wind turbine maintenance plans, it performs data backup and routing adjustments for nodes facing planned power outages in advance, preventing data loss or communication interruptions due to planned power outages.

[0212] The central scheduling module detects node or link failures in real time using multiple methods. Heartbeat timeout detection is the most basic fault detection method; nodes send heartbeat packets to the center every 30 seconds, and if no heartbeat is received after three consecutive attempts, the node is considered offline. Link quality degradation detection is based on the expected number of transmissions reported by the node; when this value consistently exceeds a preset threshold, a link failure is determined. Energy depletion detection determines that a node is about to fail when it reports remaining power below 5% and shows no signs of charging. Data anomaly detection is used to identify sensor failures; when a node reports logical errors in its data, a corresponding alarm is triggered.

[0213] For complex faults, the central scheduling module uses topology relationships and historical data for fault location. For example, if all nodes in a subnet are unable to communicate with the central system, but nodes within the subnet can still communicate with each other, then the fault can be determined to be the gateway node of that subnet. The location results can provide precise targets for subsequent self-healing operations;

[0214] Based on the fault type, the central scheduling module automatically executes corresponding self-healing operations. When a node goes offline, the system updates the topology map, recalculates the routes in the affected area, and notifies relevant nodes to switch to backup paths; if a node is determined to be permanently failed, it is marked and removed from the routing table. When a link fails, the system notifies nodes at both ends of the link to try switching operating frequency bands or increasing transmission power; if recovery is still not possible, routes bypassing the link are recalculated. When a gateway fails, the system selects a new gateway node within the subnet, typically a node with sufficient energy, a suitable location, and dual-frequency communication capabilities, issues a gateway switching command, and updates the entire network route. Energy warnings trigger load adjustments according to the energy scheduling strategy. When a sensor fails, the system marks the sensor data as unavailable, uses data from neighboring nodes for interpolation, or notifies maintenance personnel for on-site repair.

[0215] All fault events are recorded in the database for subsequent analysis. The central scheduling module regularly generates fault statistics reports, calculates indicators such as mean time between failures and mean time to repair, identifies frequently failing areas or nodes, and provides data support for operation and maintenance decisions.

[0216] The monitoring dashboard provides a panoramic visualization of the wind farm's communication network. The topology view uses a GIS map as its base, dynamically drawing node locations, link connections, and subnet boundaries, and supports clicking to view detailed node information. The energy view displays the entire network's energy distribution in heatmap form, with colors indicating remaining power levels, and supports scrolling through a timeline to view historical energy changes. The performance view displays real-time curves showing key performance indicators such as network throughput, average latency, packet loss rate, and route discovery count. The alarm view scrolls through the current alarm list, categorized by urgency, importance, and warning levels, and supports alarm confirmation and disabling.

[0217] The operations and maintenance console provides remote configuration, command issuance, historical query, and report generation functions. Remote configuration supports batch modification of node parameters, such as transmit power, sampling period, and routing weight coefficients, and supports importing and exporting configuration templates. Command issuance can manually trigger operations such as route recalculation, energy scheduling, node restart, and firmware upgrade. Historical queries support retrieving historical data by time, node, and event type and exporting charts. Report generation automatically outputs daily, weekly, and monthly reports, covering content such as node online rate, energy consumption, communication quality, and fault statistics.

[0218] The mobile application is compatible with smartphones and tablets, making it convenient for maintenance personnel to view key information, receive alarm push notifications, and perform simple operations on-site, such as confirming alarms or restarting nodes.

[0219] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A multi-band converged wind farm communication network system based on a mobile ad hoc network, characterized in that, It includes: multiple wind turbine nodes, each of which is deployed at wind turbines arranged along a strip within the wind farm, and each wind turbine node includes the following modules: The multi-band communication module provides at least two independently operating radio frequency transceiver units for each wind turbine node, corresponding to the low-frequency band communication submodule and the high-frequency band communication submodule respectively, and performs adaptive frequency band switching based on link quality; The energy sensing and acquisition module collects and converts environmental energy into electrical energy for storage, and monitors and controls node energy consumption; The strip self-organizing network routing module utilizes the strip distribution characteristics of wind turbines to construct and maintain multi-hop wireless communication paths; The dynamic rate control module monitors network congestion in real time and dynamically adjusts the data transmission rate. And a central dispatch module, deployed in the wind farm control center, communicates with each wind turbine node via a wireless link to perform centralized monitoring, routing optimization and energy dispatch of all network nodes.

2. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 1, characterized in that, The multi-band communication module includes a baseband processing unit, which is used to perform frequency band adaptive switching, including: establishing a node state vector containing node remaining energy, data queue length, real-time wind speed and link quality indicator values ​​for each frequency band; Based on the node state vector, calculate the low-frequency band optimization factor and the high-frequency band optimization factor respectively; In each decision cycle, the low-frequency band preference factor is compared with the high-frequency band preference factor. If the difference exceeds a preset threshold, a switch from the current frequency band to the target frequency band is executed.

3. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 2, characterized in that, The multi-band communication module is also used to perform dual-band cooperative transmission, always transmitting control signaling through the low-frequency band, and dynamically allocating data load to the high-frequency band or low-frequency band for parallel transmission or redundant transmission according to service requirements or link quality.

4. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 1, characterized in that, The energy sensing and acquisition module includes: The hybrid energy harvesting unit, comprising a wind power generation unit, a photovoltaic power generation unit, and a piezoelectric energy harvesting subunit, is used to harvest energy from wind, light, and vibration. The energy management and storage unit is used to collect, convert, and store the collected energy in the energy storage battery pack. The energy consumption monitoring and control unit is used to monitor the energy consumption of each component in the node in real time and to manage the power supply of each component independently. The wind turbine node also includes an energy prediction unit, which is used to generate a predicted energy harvesting power curve for future periods based on historical data and environmental parameters using a combined prediction model.

5. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 4, characterized in that, The wind turbine node also includes a dynamic scheduling unit, used to: prioritize communication tasks according to their types; Based on the current remaining power and the energy harvesting power prediction curve, assess the energy availability index within the future time window; The energy security level of a node is determined based on the energy availability index. Based on the energy security level and task priority, the access, transmission parameters and execution timing of communication tasks are dynamically scheduled.

6. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 1, characterized in that, The strip ad hoc network routing module is used to perform: restricted flooding along the axis of the strip subnet to suppress the broadcast spread of routing request packets; and to select a path using a comprehensive routing metric method that integrates the node's remaining energy factor, the expected number of link transmissions, and the number of hops. During the route discovery process, a threshold of the node's remaining energy is used to determine whether the node is allowed to participate in forwarding route request packets.

7. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 6, characterized in that, The strip self-organizing network routing module is also used for: during the route discovery process, the destination node selects the path with the smallest metric value as the primary route based on the multiple route request packets received, and the path with the second smallest metric value that does not intersect with the primary route node as the backup route. By monitoring link quality, triggering updates based on node energy changes, or conducting periodic assessments, the primary and backup routes are dynamically maintained and switched.

8. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 1, characterized in that, The dynamic rate control module includes a congestion prediction unit, used to: divide the network congestion level into multiple discrete states and establish a hidden Markov model; Using queue length, collision rate, and channel busy / idle status as observation vectors, the probability of entering a congestion state at future times is predicted using the hidden Markov model. When the probability exceeds a preset threshold, a congestion warning is triggered.

9. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 8, characterized in that, The dynamic rate control module also includes a rate decision and execution unit, which is used to execute one or more rate adjustment strategies in priority, such as application layer transmit rate adjustment, MAC layer contention window dynamic adjustment, frame aggregation degree adjustment and transmit power adjustment, when a congestion warning is triggered.

10. The multi-band converged wind farm communication network system based on mobile ad hoc network according to claim 1, characterized in that, The central scheduling module is used to: collect the status information of each wind turbine node at a fixed period or in an event-triggered manner, and to construct and maintain the network topology map. Based on the overall network topology, node remaining energy, node load, and service priority, calculate and distribute the main route and backup route across subnets to each wind turbine node; Based on the energy prediction curves uploaded by each node, low-energy nodes are identified, and traffic is guided to bypass the low-energy nodes in the global routing calculation, or a cross-node energy coordination mechanism is initiated when there is regional energy shortage. Real-time detection of node or link failures and triggering corresponding self-healing operations, including route switching and gateway reselection.

Images