A self-powered wireless ad hoc network terminal based on perovskite photovoltaic
By employing perovskite photovoltaic modules and energy management modules in distributed IoT terminals in the field, and dynamically adjusting communication parameters, combined with supercapacitors and anti-tamper modules, the problems of terminal power supply stability and data security were solved, enabling long-term stable operation of the terminal under low light conditions and network self-healing capability.
Patent Information
- Application Number
- CN202611172631.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-04
- Publication Date
- 2026-11-17
AI Technical Summary
Existing power supply systems for distributed IoT terminals in the wild have limited battery life, are prone to voltage drops on power rails due to sudden communication loads, lack dynamic energy status adjustment and anti-tampering mechanisms, and cause network interruptions due to premature power depletion of wireless self-organizing network relay nodes, resulting in insufficient data security.
Perovskite photovoltaic modules are used as the power source, combined with an energy management module, a built-in energy storage unit and a communication module. The radio frequency transmission power and sleep cycle are dynamically adjusted. The anti-tamper module monitors the physical boundary state of the equipment and executes a safe destruction process when the power supply drops. A supercapacitor is introduced to store energy in low light conditions. A multi-level dynamic power consumption state machine and an energy-aware routing algorithm are used to optimize the network topology.
It enables long-term stable operation of terminals in a passive environment, reduces the risk of unauthorized disassembly, improves data security and network self-healing capabilities, optimizes energy utilization efficiency, and avoids network interruption caused by nodes running out of power.
Smart Images

Figure CN123124137A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of wireless communication and new energy technology, specifically to a self-powered wireless self-organizing network terminal based on perovskite photovoltaics. Background Technology
[0002] The long-term operation of distributed IoT terminals in the field is highly dependent on ambient energy. Existing devices mostly use traditional batteries or silicon-based solar modules. Traditional batteries have limited battery life and high maintenance costs; silicon-based solar modules are rigid and heavy, making it difficult to conformally fit into miniaturized terminal casings, and their power generation capacity is poor under low light conditions, making it difficult to maintain system charging. In addition, existing devices lack a power consumption scheduling mechanism based on remaining power, and in low-power states, sudden communication loads can easily cause voltage drops in the power supply rail, leading to frequent system power outages and resets.
[0003] In multi-hop wireless ad hoc network communication, existing routing algorithms often rely solely on signal strength as a criterion, neglecting the remaining energy status of relay nodes. This leads to core nodes in critical topology locations prematurely exhausting their power due to frequent data forwarding, causing local network disruptions. Furthermore, existing solutions simply overlay power supply and communication functions without coordinated optimization, failing to dynamically adjust RF transmit power and sleep cycles based on energy status.
[0004] Furthermore, unattended terminals face a high risk of unauthorized dismantling and data theft. Existing anti-tampering mechanisms mostly rely on single mechanical switches, which are prone to false alarms under outdoor temperature differences and vibrations. When the terminal is damaged or the main power is cut off, the lack of backup transient energy storage design means the processor cannot obtain enough power to execute a complete data erasure procedure, leading to the potential leakage of sensitive data such as network keys. Simultaneously, existing circuit designs lack irreversible blocking mechanisms at the chip's underlying structure, allowing unauthorized personnel to directly read core asset data through the motherboard's physical debugging interface. Summary of the Invention
[0005] To address the above problems, the present invention provides the following technical solution: The first aspect of the present invention provides a self-powered wireless self-organizing network terminal based on perovskite photovoltaics. The terminal includes a shell, on the surface of the shell of the self-powered wireless self-organizing network terminal are perovskite photovoltaic modules, and inside the shell are an energy management module, a built-in energy storage unit, a processing module and a communication module. The power output terminal of the perovskite photovoltaic module is connected to the power input terminal of the energy management module, converting the received ambient light into DC power and inputting it into the energy management module. The energy management module is connected to the built-in energy storage unit, the processing module, and the communication module, respectively, converting the DC power into a standard charging voltage to charge the built-in energy storage unit, and supplying power to the processing module and the communication module. The energy management module calculates the remaining power value by collecting the instantaneous terminal voltage parameters and charging / discharging current of the built-in energy storage unit and sends it to the processing module. The processing module reads the remaining power value and generates or caches the data payload to be sent based on locally collected data, terminal status data, or data to be forwarded received as a relay node. When the remaining power value meets the communication recovery conditions, the processing module issues a network initialization command to the communication module. The processing module also generates an RF configuration command based on the remaining power value, enabling the communication module to set the RF transmission power and receiver sleep / wake-up cycle. The communication module starts the MESH self-organizing network protocol stack and establishes a multi-hop routing link according to the network initialization command, and selects the next-hop node to forward the data payload to be sent based on the link quality parameters of adjacent nodes.
[0006] Using the above scheme, the terminal collects ambient solar energy using perovskite photovoltaic modules and monitors and distributes power through an energy management module. The processing module dynamically controls the network establishment and routing scheduling parameters of the communication module based on the remaining power, achieving dynamic matching between system power consumption and available energy, and ensuring long-term operation of the node in a passive environment.
[0007] Furthermore, the processing module is also connected to a sensing module and an anti-tamper module; the sensing module collects environmental sensor data and RFID tag information and sends them to the processing module to generate the data payload to be sent; the anti-tamper module obtains the status parameters of the shell connection part, and sends a hardware interrupt level signal to the processing module when the status parameters exceed the preset normal threshold range, so that the processing module generates a physical damage alarm payload and enters a safety protection preparation state; when the physical damage alarm payload is reported, the reporting reaches the preset timeout time, or the main power supply rail voltage is detected to drop below the preset power supply drop threshold, the processing module enters the safety destruction process.
[0008] Using the above scheme, the anti-tamper module and the sensing module monitor the physical boundary status of the equipment. Combined with the power supply drop detection logic, the terminal can lock the abnormal state and send back alarm data through a hardware interrupt trigger mechanism before it is damaged and shuts down or loses power.
[0009] Furthermore, the perovskite photovoltaic module adopts a flexible thin-film laminated structure conformally bonded to the surface of the outer shell; the perovskite photovoltaic module is sequentially bonded from the inside out with a flexible polymer substrate, a transparent electrode layer, a bottom charge transport layer, a perovskite light absorption layer with a thickness between 300nm and 500nm, a top charge transport layer, a metal back electrode layer, and a flexible multilayer encapsulation structure; the flexible multilayer encapsulation structure is composed of a high barrier layer, an ultraviolet filter layer, a scratch-resistant layer, and a hydrophobic self-cleaning layer, and together with the terminal outer shell sealing structure, the terminal achieves an IP68 protection level.
[0010] Using the above scheme, the perovskite absorber layer can achieve broadband energy harvesting. A multi-layered composite encapsulation structure resists external moisture penetration and UV degradation. Simultaneously, the conformal bonding design adapts to the curved surface of the outer shell, increasing the light-receiving area, and the synergistic sealing enhances the equipment's outdoor protection level.
[0011] Furthermore, the energy management module includes a maximum power point tracking unit and a low-light charging management unit, and is connected in parallel to a supercapacitor for low-light pulse energy storage; the energy management module is connected to an ambient light sensor and reads the light intensity data collected by the ambient light sensor; when the light intensity is between 200 lux and 500 lux, the energy management module enters a pulse charging mode and guides the DC power to the built-in nanowatt-level low-power boost converter circuit for voltage pumping to the supercapacitor for energy storage; when the terminal voltage of the supercapacitor reaches the upper limit discharge threshold, the energy management module disconnects the input-side charging switch and closes the electronic switch with active current limiting function to pulse charge the built-in energy storage unit.
[0012] Using the above scheme, the system accumulates a small current using a supercapacitor under low-light conditions. When the terminal voltage reaches a set threshold, a controlled pulse is injected into the energy storage unit by closing an electronic switch, thereby maintaining the charging capability in low-light environments.
[0013] Furthermore, the processing module is configured with a multi-level dynamic power consumption state machine; the multi-level dynamic power consumption state machine switches between full-speed mode, energy-saving mode, low-power mode, and emergency sleep mode according to the remaining power value and task priority; the processing module is preset with multiple decreasing power determination thresholds and hysteresis compensation constants; when the remaining power value drops below the corresponding power determination threshold, the processing module triggers a downgrade switch; when the remaining power value rises and exceeds the sum of the corresponding power determination threshold and the hysteresis compensation constant, the processing module triggers an upgrade switch.
[0014] Using the above scheme, hysteresis comparison logic is introduced into the power state machine. The set hysteresis compensation constant covers the voltage rebound range caused by internal resistance after the energy storage unit disconnects the load, thus avoiding the problem of frequent mode switching at the power determination critical point of the system.
[0015] Furthermore, the sensing module integrates an edge computing unit; the environmental sensor data includes high-frequency sampled vibration acceleration parameters and low-frequency sampled temperature, humidity, and water level parameters; the RFID tag information includes personnel location identification codes and asset identification codes obtained by the RFID reader unit; for the vibration acceleration parameters, the edge computing unit subtracts the arithmetic mean point by point to remove the DC component of gravity and performs amplitude normalization processing before inputting it into a one-dimensional convolutional neural network model; the one-dimensional convolutional neural network model outputs the probabilities of normal vibration state, illegal damage state, and geological anomaly state; when the output probability of illegal damage state or geological anomaly state is greater than a preset classification threshold and the same type of output is maintained for at least two consecutive judgment time windows, the edge computing unit generates an anomaly state tag and generates the data payload to be sent by combining the statistical characteristics of the low-frequency sampled temperature, humidity, and water level parameters.
[0016] Using the above scheme, the edge computing unit performs filtering and neural network classification to convert the raw high-frequency vibration sampling data into abnormal state tags. Combined with a time window judgment mechanism to eliminate transient mechanical interference, the RF power consumption of the communication module transmitting redundant data is reduced. Simultaneously, the fusion of RFID information enhances the terminal's perception of assets and personnel.
[0017] Further, the communication module obtains the received signal strength and signal-to-interference-plus-noise ratio of neighboring nodes through a broadcast discovery process to generate link quality parameters; the communication module receives the current remaining battery power parameters of the neighboring nodes; the energy-aware routing unit in the processing module adds a zero-bias constant to the normalized link quality parameters, takes the reciprocal, and multiplies it by the link quality weight coefficient to obtain the first value; it adds the zero-bias constant to the normalized neighboring node's feedback power parameters, takes the reciprocal, and multiplies it by the energy state weight coefficient to obtain the second value; the energy-aware routing unit adds the first value and the second value to obtain the routing cost. The system selects the neighboring node with the lowest routing cost as the next-hop node; multiple terminals form a decentralized network through the MESH self-organizing network protocol; when any relay node suffers physical damage, self-destruction due to anti-tampering, or power depletion and shutdown, the neighboring node generates a link error message and broadcasts it within the keep-alive message timeout window, causing the data source node to call the suboptimal path in the local backup routing table to take over the transmission task; when the keep-alive message period is no more than 100ms, the media access control layer single response timeout time is no more than 50ms, and there is at least one suboptimal path in the local backup routing table, the network completes the transmission link switch and resumes data communication within 0.5s.
[0018] The above scheme comprehensively considers both RF signal quality and the remaining energy of adjacent nodes in the routing calculation. Setting a zero-bias constant prevents floating-point overflow when the microcontroller performs reciprocal operations. This mechanism guides the network to prioritize nodes with sufficient energy and stable signals for data relay, balancing the overall energy consumption of the network topology. Furthermore, a rapid self-healing mechanism ensures that the failure of a single node does not cause network-wide paralysis.
[0019] Furthermore, when entering low-power mode, the processing module calculates the target sleep period based on the remaining power value and the instantaneous power generation of the perovskite photovoltaic module, limits the target sleep period by boundary, and sends it to the communication module, so that the communication module updates the receiver sleep-wake cycle to the target sleep period; the processing module also reduces the sampling frequency of environmental parameters such as temperature, humidity and water level, and sets the vibration sensor to interrupt threshold standby mode.
[0020] By adopting the above scheme, the system extends the basic sleep period due to reduced power consumption while using the instantaneous power generation of photovoltaic modules as a time compensation factor. The sleep interval is appropriately shortened according to the actual sunlight conditions, maintaining the network activity of nodes under the premise of energy balance; at the same time, the dynamic adjustment of sensor sampling frequency and standby mode further reduces the underlying static power consumption of the system.
[0021] Furthermore, the anti-tamper module includes a housing anti-tamper unit; the housing anti-tamper unit includes a micro switch and a light sensor; the micro switch monitors the opening and closing state of the main housing closure joint of the housing; the light sensor is arranged in a sealed dark chamber area inside the housing and monitors the external ambient light entering after the sealed housing breaks; the processing module combines the real-time temperature data collected by the onboard temperature sensor to perform temperature compensation on the base voltage constant, and calculates the percentage of light leakage state parameters based on the temperature-compensated base voltage constant; the processing module performs a logical OR operation on the micro switch open flag and the light leakage flag of the light sensor to generate a physical damage alarm load.
[0022] Using the above scheme, the system combines the mechanical state of the microswitch with the photosensitive state of the photosensitive sensor for tamper detection. Real-time temperature data is used to compensate for the substrate voltage of the photosensitive sensor, eliminating substrate deviation caused by device temperature drift and improving the accuracy of tamper detection in outdoor temperature variation scenarios.
[0023] Furthermore, in the secure destruction process, when the main power supply rail voltage drops below the preset power supply drop threshold, the energy management module switches the system power supply to the backup energy storage network; the processing module writes a tamper-erase instruction and pulls down the physical reset control line of the backup domain storage unit to clear the high-density data; and when the transient available energy of the backup energy storage network is greater than the physical fuse energy threshold, electronic fuse programming is performed through the on-chip electronic fuse programming controller, or the MOS transistor is controlled to break down through the chip fuse unit to perform a physical fuse operation.
[0024] Using the above scheme, when the main power supply rail drops, the backup energy storage network provides transient power. The processing module preempts control through a non-maskable interrupt, clearing the backup domain data; simultaneously, it establishes a high-voltage power domain to physically blow electronic fuses or MOSFETs, blocking the path to read core data through the debug interface at the hardware level.
[0025] This invention provides a self-powered wireless self-organizing network terminal based on perovskite photovoltaics. It has the following beneficial effects: 1. This invention utilizes a supercapacitor and a boost circuit to accumulate a small current under low light conditions, charging the built-in energy storage unit in pulse form to maintain charging capability in low-light environments. Simultaneously, a multi-level dynamic power consumption state machine introduces a hysteresis compensation constant. This constant covers the voltage rebound range after the energy storage unit is disconnected from the load, preventing the system from frequently switching power consumption modes at critical power levels and improving the long-term power supply stability of the terminal.
[0026] 2. This invention uses an edge computing unit to perform local classification and filtering of high-frequency sampled data, reducing the power consumption generated by the communication module sending redundant data. Furthermore, the energy-aware routing unit integrates the link quality and remaining power of adjacent nodes, and introduces an anti-zero bias constant to avoid computational overflow. It prioritizes nodes with stable signals and sufficient energy as relays, balancing the overall energy consumption of the network topology and preventing nodes from failing due to power depletion.
[0027] 3. This invention combines a microswitch and a photosensitive sensor to monitor physical boundaries and uses temperature data to compensate for the substrate voltage error of the photosensitive device, improving the accuracy of tamper detection under outdoor temperature difference environments. When safety destruction is triggered and the main power supply drops, the system switches to the backup energy storage network. The processing module clears the backup domain data through a non-maskable interrupt and uses a charge pump to boost the voltage and physically blow the electronic fuse, blocking the data leakage path at the hardware level. Attached Figure Description
[0028] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is an overview flowchart of the self-powered wireless self-organizing network method of the present invention; Figure 3This is a flowchart of the energy management and low-light adaptive charging state machine of the present invention; Figure 4 This is a flowchart of the data-aware sampling and edge computing feature extraction process of the present invention; Figure 5 This is a flowchart of the multi-level dynamic power consumption state machine and energy-sensing routing addressing of the present invention; Figure 6 This is a comparison curve of the terminal state of charge under continuous low light shading environment according to the present invention. Figure 7 This is a bar chart comparing the energy consumption of direct pass-through and edge computing in a single burst event according to the present invention. Figure 8 This is a two-dimensional broken line graph showing the remaining energy distribution of network nodes according to the present invention. Detailed Implementation
[0029] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0030] Reference Appendix Figure 1 The present invention provides a self-powered wireless self-organizing network terminal based on perovskite photovoltaics, which may include: perovskite photovoltaic modules, energy management modules, processing modules, communication modules, sensing modules, and anti-tampering modules.
[0031] The perovskite photovoltaic module is mounted on the surface of the self-powered wireless self-organizing network terminal's casing. The output terminal of the perovskite photovoltaic module is connected to the power input terminal of the energy management module.
[0032] The power output terminal of the energy management module is connected to the power input terminals of the processing module, communication module, sensing module, and anti-tamper module, respectively, to provide them with operating voltage.
[0033] In an optional embodiment, the data communication interface of the processing module is connected to the communication interfaces of the energy management module, the communication module, the sensing module, and the anti-tamper module, respectively, to realize data interaction and command control.
[0034] Reference Appendix Figure 2 This invention provides a self-powered wireless ad hoc network method, comprising the following steps: S10, the perovskite photovoltaic module receives ambient light and converts it into DC power. The energy management module adjusts the voltage conversion circuit to convert the DC power into a standard charging voltage to charge the built-in energy storage unit. It monitors the charging and discharging current and calculates the remaining power value and sends it to the processing module. S20, the processing module reads the remaining power value and sends a network initialization command to the communication module. The communication module starts the MESH self-organizing network protocol stack, sends broadcast messages to detect adjacent nodes within the communication range and establishes multi-hop routing links. S30, the sensing module reads environmental sensor data and RFID tag information according to the preset sampling sequence and sends them to the processing module. The processing module performs preprocessing operations on the environmental sensor data and RFID tag information to generate the data payload to be sent. S40, the processing module executes power consumption scheduling logic based on the remaining power value and sends RF configuration instructions to the communication module. The communication module sets the RF transmission power and receiver sleep-wake cycle based on the RF configuration instructions, extracts the link quality parameters of adjacent nodes and the power parameters fed back by adjacent nodes, calculates the routing cost, selects the next hop node, and forwards the data payload through the multi-hop routing link according to the set RF transmission power. S50, the anti-tamper module continuously acquires the mechanical and electrical status parameters of the shell connection parts. When the shell status parameters are detected to exceed the preset normal threshold range, a hardware interrupt level signal is sent to the processing module. The processing module first generates the highest priority physical damage alarm data and preemptively reports it. When the alarm reporting is completed, or the alarm reporting reaches the preset timeout time, or the main power rail voltage is simultaneously detected to drop below the preset power drop threshold, the processing module enters the safe destruction process. If the main power rail voltage has dropped, the processing module uses the backup capacitor energy to maintain the safe erasure unit operation. If the fuse energy verification condition is met, the processing module further activates the internal charge pump and electronic fuse programming circuit to complete the irreversible erasure of sensitive data and the physical fuse operation of the debugging interface.
[0035] The following will provide a detailed description of the specific technical solution of this invention in conjunction with the hardware structure and software algorithm logic of each module.
[0036] Reference Appendix Figure 3The perovskite photovoltaic module provided in this embodiment of the invention adopts a flexible thin-film laminated structure conformally bonded to the outer shell surface of a self-powered wireless self-organizing network terminal. The perovskite photovoltaic module physically comprises a flexible polymer substrate, a transparent electrode layer, a bottom charge transport layer, a perovskite light absorption layer, a top charge transport layer, a metal back electrode layer, and a flexible multilayer encapsulation structure. In an optional embodiment, the flexible multilayer encapsulation structure is specifically composed of a high-barrier layer, an ultraviolet filter layer, a scratch-resistant layer, and a hydrophobic self-cleaning layer to provide conditions that block water and oxygen penetration. The edge area of the flexible multilayer encapsulation structure is bonded to the terminal shell surface with weather-resistant sealant, and the electrode leads are waterproofed with potting compound or sealing terminals, limiting the minimum bending radius during module bonding to prevent cracks in the encapsulation layer on the non-planar shell. The above-mentioned flexible multilayer encapsulation structure and the terminal shell sealing structure together constitute the overall waterproof and dustproof structure, enabling the terminal to meet the IP68 protection requirement after sealing testing. The specific layer construction and manufacturing process of the perovskite photovoltaic module includes the following steps: S101, a polyimide (PI) film or a polyethylene naphthalate (PEN) film with high heat resistance is selected as the flexible polymer substrate to withstand the high-temperature environment required for subsequent perovskite crystallization. Plasma activation treatment is performed on the surface of this flexible polymer substrate, and a silica buffer layer is coated onto its surface. Coating with a silica buffer layer increases the interfacial adhesion of the subsequently deposited film. For the configuration of operating parameters of the plasma cleaning equipment and routine degreasing operations, those skilled in the art can select a standard cleaning process route based on the substrate material. The surface cleaning process is well-known in the art and will not be described in detail here.
[0037] S102, a transparent conductive material is deposited on the surface of the silica buffer layer to form a transparent electrode layer. In this embodiment, the transparent conductive material is selected from indium tin oxide (ITO) or silver nanowires. When ITO is selected as the transparent conductive material, a transparent electrode layer is formed on the substrate surface using low-temperature magnetron sputtering, low-temperature evaporation, or coating-type ITO nano-slurry film deposition processes. When silver nanowires are selected as the transparent conductive material, a silver nanowire dispersion is attached to the substrate using spraying, slot coating, or spin coating processes, and a transparent conductive network is formed by low-temperature curing or UV-assisted curing. After the above treatment, a transparent electrode layer with a sheet resistance of less than 15 Ω / sq is formed. This sheet resistance threshold is predetermined by the trade-off between electrode transmittance and conductivity.
[0038] S103, Prepare a bottom charge transport layer on the surface of the transparent electrode layer. Depending on the selected heterojunction structure, spin-coat a tin dioxide precursor solution over the transparent electrode layer, and form an electron transport layer after annealing and curing; or spin-coat a relevant hole material (e.g., Spiro-OMeTAD, PTAA, or PEDOT:PSS) to form a hole transport layer.
[0039] S104. A perovskite light-absorbing layer is prepared above the underlying charge transport layer. A mixed cationic perovskite precursor solution containing formamidinium and methylamine is prepared according to a stoichiometric ratio. The precursor solution is uniformly coated onto the surface of the underlying charge transport layer using a slot coating process or inkjet printing process. After coating, the semi-finished component is placed in a heating device and annealed at a temperature range of 100°C to 150°C (when using polyethylene naphthalate film, the annealing temperature is controlled below its glass transition temperature, for example, 100°C to 120°C), for a duration of 10 min to 30 min. The above annealing temperature and time range are determined through preliminary experiments based on the crystallization kinetics of mixed cationic crystals to promote grain growth while avoiding thermal degradation of the material. After annealing, a perovskite film with a thickness between 300 nm and 500 nm is formed. This thickness range ensures sufficient absorption of photon energy while maintaining a low internal carrier recombination rate. This step controls the thickness of the core light absorption layer to below 1μm to meet the bonding requirements of the non-planar deformation of the terminal shell.
[0040] S105, a top charge transport layer is prepared above the perovskite light-absorbing layer. In accordance with the polarity of the bottom charge transport layer in step S103, a charge transport material solution with the opposite polarity is spin-coated at this point and dried and cured to form the top charge transport layer (if the bottom layer is an electron transport layer, the top layer is a hole transport layer such as Spiro-OMeTAD or PTAA; conversely, the bottom layer is the opposite). This top charge transport layer, together with the bottom charge transport layer, encapsulates the perovskite light-absorbing layer, forming a complete charge-separated structure.
[0041] S106, a metal conductive material is deposited above the top charge transport layer to form a metal back electrode layer. Elemental gold or silver is deposited onto the component surface using vacuum evaporation or magnetron sputtering equipment to form a metal thin film with a thickness of 80 nm to 100 nm. For the vacuum degree adjustment and target arc initiation control of the vacuum evaporation equipment, those skilled in the art can consult standard process manuals based on the preset film thickness. Metal thin film sputtering is a well-known technique in the field and will not be elaborated further here.
[0042] S107 features a flexible multilayer encapsulation structure covering the outer side of the metal back electrode layer and the edge area of the module. An alumina film is deposited on the outer side of the module using atomic layer deposition (ALD) as a high-barrier layer. A polymer film containing UV absorbers is coated onto the surface of the high-barrier layer as a UV filter layer. A fluoropolymer film is then attached to the outside of the UV filter layer to form a scratch-resistant layer. A nano-scale silane coupling agent is sprayed onto the surface of the scratch-resistant layer to create a hydrophobic self-cleaning layer.
[0043] After the above structure is fabricated, when incident light irradiates the perovskite photovoltaic module, the perovskite light-absorbing layer absorbs photons with energy greater than its band gap and excites electron-hole pairs. Under the influence of the electric field built into the device, the electrons and holes separate and drift to the electrodes on both sides through the electron transport layer and hole transport layer, respectively, thereby forming a direct current in the external circuit. The photoelectric conversion efficiency of this module follows the following physical relationship: ; In the formula, Represents the photoelectric conversion efficiency of perovskite photovoltaic modules. This represents the voltage across the component when it is in an open-circuit state. This represents the loop current of the component under short-circuit conditions. The fill factor represents the degree of carrier recombination loss within the device. This represents the incident light power density incident on the component surface. This represents the physical surface area of the module that receives light. Thanks to the aforementioned thin light-absorbing layer structure and broad spectral absorption characteristics, even in low-light environments such as dense forests or cloudy days with illuminance as low as 200 lux, the module can still generate sufficient transition carriers to output micro-power electrical energy that can be collected by the energy management module. This micro-power electrical energy is stored by a nanowatt-level low-power boost converter and a supercapacitor, and then replenishes the built-in energy storage unit in a pulsed manner. This allows the terminal to operate intermittently in a low-power state, rather than requiring the photovoltaic module to directly and continuously drive all subsequent circuits under such low-light conditions. In one specific embodiment, the light-receiving area of the perovskite photovoltaic module is 50 cm². 2 Up to 300cm 2 The built-in energy storage unit has a capacity of 500mAh to 6000mAh, with a preferred capacity of 5000mAh. The equivalent capacitance of the backup energy storage network ranges from several microfarads to several farads. Under standard test conditions, the photoelectric conversion efficiency of the perovskite photovoltaic module is greater than 25%. The communication module operates at different currents in transmit, receive, idle listening, and deep sleep states. The processing module establishes an energy budget table based on the standard inspection cycle, energy-saving sampling cycle, single alarm transmission energy consumption, security erase energy consumption, and electronic fuse programming energy consumption, and determines the power consumption state machine threshold, security destruction trigger conditions, and backup energy storage network capacity accordingly. In this embodiment of the invention, the energy management module internally incorporates a multi-source power supply access network and a battery management and protection circuit to comprehensively manage the input power of the perovskite photovoltaic modules and external backup DC power supply, and dynamically assess the remaining power of the built-in energy storage unit. The hardware control and power assessment logic of this part includes the following steps: S108, the energy management module's input is equipped with a multi-source selection circuit. This circuit is simultaneously connected to the output of the perovskite photovoltaic module and the external backup DC charging interface. To achieve hot-swapping and seamless switching of the power supply links, the multi-source selection circuit utilizes metal-oxide-semiconductor field-effect transistors (MOSFETs) to construct an ideal diode network. Compared to traditional PN junction diodes with a fixed forward voltage drop, using MOSFETs for conduction reduces heat loss during power transmission. When the external backup DC charging interface is connected, the comparator inside the multi-source selection circuit obtains the voltage difference between the two inputs, conducts the power supply link with the higher voltage difference, and shuts off the other link to prevent reverse current from damaging the photovoltaic device or the external power supply. For the construction of the MOSFET-based anti-reverse current circuit and the suppression of transient voltage during hot-swapping, those skilled in the art can select and configure it according to the application manuals of conventional power management chips. The specific circuit structure and discrete component arrangement are well-known technologies in the field and will not be elaborated here.
[0044] S109, the electrical energy entering the energy management module is processed by the voltage conversion circuit and then sent to the built-in battery management unit. In addition, the energy management module also has a supercapacitor connected in parallel for low-light pulse energy storage and a backup tantalum capacitor network for tamper-proof power-off protection. The battery management unit monitors the terminal voltage, loop current, and surface temperature of the built-in energy storage unit in real time through an analog front-end chip. In a preferred embodiment, the built-in energy storage unit uses a special wide-temperature lithium-ion battery with a set lower limit of 3.0V for discharge voltage, an upper limit of 4.2V for safe charging voltage, and a set safe operating temperature range between -40℃ and +85℃. The battery management unit sets the allowable charging temperature range and allowable discharging temperature range according to the characteristics of the selected wide-temperature lithium-ion battery material. When the temperature is below -20℃, the battery management unit automatically prohibits charging and only allows discharging. The above threshold parameters are determined based on the basic electrochemical characteristics of the selected wide-temperature lithium-ion battery material and thermal runaway safety specifications. When the terminal voltage or surface temperature exceeds the above threshold range, the battery management unit cuts off the charging / discharging circuit switch. Because this hardware-level cutoff mechanism operates independently of the main program's software logic, it can avoid protection failures caused by software crashes and reduce the risk of overcharging, over-discharging, and thermal runaway in the built-in energy storage unit.
[0045] Within the safe operating temperature and voltage range, the S110 energy management module uses a high-precision shunt resistor or Hall current sensor connected in series in the power supply circuit to perform high-frequency sampling of the instantaneous current flowing into and out of the built-in energy storage unit, and simultaneously obtains the instantaneous terminal voltage parameters of the energy storage unit through an analog-to-digital converter. The sampled data is temporarily stored in the module's internal register as the basic input variable for subsequent assessment of the remaining power status.
[0046] The S111 energy management module's internal processing chip calculates the remaining charge of the built-in energy storage unit in real time based on collected instantaneous current and terminal voltage data, using an ampere-hour integration method combined with an open-circuit voltage calibration mechanism. The physical essence of the ampere-hour integration method is to accumulate the amount of charge flowing into and out of the battery over time, reflecting the dynamic changes in the overall state of charge. Because a single ampere-hour integration method can accumulate calculation errors due to zero-point drift of the current sensor after long-term operation, the system extracts the current stable open-circuit terminal voltage and periodically corrects the initial charge value when it detects that the absolute value of the charging / discharging current is less than a preset sleep current threshold for a preset time window. This is done in conjunction with a built-in voltage-capacity mapping table. The sleep current threshold is pre-calibrated based on the static maintenance power consumption of the self-powered wireless self-organizing network terminal; the voltage-capacity mapping table is fitted from discrete data points obtained from charging / discharging tests of the same batch of energy storage units in a standard laboratory environment. The specific logic for calculating the remaining charge value follows the following physical expression: ; In the formula, Representative at The normalized value of the instantaneous remaining power obtained at any time (the value ranges from 0 to 1); Represents the initial moment The calibrated baseline remaining power normalization value is obtained directly by reading the stable open-circuit voltage and querying the aforementioned built-in mapping table when the device is cold-started or woken from a long-term hibernation state. This represents the actual maximum available ampere-hour capacity of the built-in energy storage unit in its current aging state, which is linearly reduced over its lifespan based on the number of charge-discharge cycles. Representative at The instantaneous current flowing through the energy storage unit at all times is defined as positive in the charging direction and negative in the discharging direction. Representing the coulomb efficiency coefficient, its value is dynamically determined by an internal lookup table function, influenced by the current operating temperature and charge / discharge rate. Specifically, in the charging direction (… When 0), the efficiency discount obtained from the lookup table is applied in the discharge direction ( When the value is 0, the coefficient is equivalent to 1. The calculated data will be periodically encapsulated into status data frames and sent to the processing module via the communication bus to provide a basis for decision-making regarding power consumption scheduling and routing selection in the subsequent system.
[0047] The energy management module integrates a low-light charging management unit. This unit adjusts the power conversion topology based on the real-time light intensity during operation. Its low-light adaptive charging state machine control mechanism includes the following steps: S112, the energy management module reads the light intensity data collected by the ambient light sensor via an external communication bus or independent interface. This light intensity data serves as a parameter for determining the charging mode switching. The low-light charging management unit divides the charging mode into five levels according to preset segmented control logic: When the light intensity is greater than or equal to 50,000 lux, it enters MPPT fast charging mode; When the light intensity is between 10,000 and 50,000 lux, it enters MPPT constant current mode; When the light intensity is between 500 and 10000 lux, it enters boost charging mode; When the light intensity is between 200 and 500 lux, it enters pulse charging mode; When the light intensity is less than 200 lux, it enters sleep mode.
[0048] The aforementioned light intensity thresholds are pre-calibrated based on the output power characteristic curves of perovskite photovoltaic modules under different illuminance levels. To avoid frequent mode switching caused by fluctuations in ambient light around the thresholds, hysteresis comparison logic is introduced for these thresholds. Taking the 50,000 lux threshold boundary as an example, when the illuminance rises above 50,000 lux, the system enters MPPT fast charging mode; when the illuminance decreases, it must fall below a preset drop threshold to exit this mode. Corresponding hysteresis intervals are set for other threshold boundaries.
[0049] S113, when entering MPPT fast charging mode, the energy management module activates the built-in maximum power point tracking unit. This unit uses a perturbation-observation method to adjust the converter's equivalent input impedance by fine-tuning the converter's duty cycle. The system continuously samples the photovoltaic output voltage and current and calculates the real-time power, adjusting the operating point to make the rate of change of photovoltaic output power with voltage approach zero. When entering MPPT constant current mode, the system limits the maximum amplitude of the output current while maintaining maximum power point tracking, maintaining a constant current output to the built-in energy storage unit.
[0050] S114, when entering boost charging mode, the maximum power point voltage (Vmpp) across the perovskite photovoltaic module is lower than the lower limit of the charging voltage of the built-in energy storage unit, and the direct conduction circuit cannot generate charging current. The energy management module activates the internal boost chopper circuit. By controlling the on and off of the power switch, the inductor energy storage is superimposed with the photovoltaic input voltage, boosting the output voltage to within the standard charging voltage range. For the selection parameters of the inductor and capacitor in the boost chopper circuit, those skilled in the art can calculate them according to conventional switching power supply design guidelines; its basic topology is well-known in the field and will not be elaborated here.
[0051] S115, when entering pulse charging mode, the continuous current generated by the perovskite photovoltaic module is lower than the static consumption and internal resistance of the charging management chip. The energy management module disconnects the DC charging channel directly connected to the built-in energy storage unit, guiding the current at the photovoltaic input to the built-in nanopower low-power boost converter circuit. After voltage boosting, the current is stored in the supercapacitor connected to the forward bypass. The energy management module uses a low-power voltage comparator to monitor the voltage across the supercapacitor. When the supercapacitor voltage rises and reaches the preset upper discharge threshold, the voltage comparator outputs an enable level. The system first disconnects the input-side charging switch between the photovoltaic module and the supercapacitor, and then closes the main circuit electronic switch with active current limiting function. The charge accumulated inside the supercapacitor is injected into the built-in energy storage unit in a controlled pulse form. This pulse injection process is constrained by the current limiting, voltage limiting, and temperature protection logic of the battery management unit. When the voltage at the built-in energy storage unit approaches the upper limit of the safe charging voltage, the surface temperature exceeds the safe operating temperature range, or the ambient temperature is lower than the prohibited charging threshold, the main circuit electronic switch remains off. When the voltage at the supercapacitor terminals drops to the preset lower stop threshold, the electronic switch in the main circuit opens and reopens the input-side charging switch between the photovoltaic module and the supercapacitor, allowing the system to re-enter the next energy storage cycle. The upper discharge threshold is dynamically determined by the control logic based on the current terminal voltage of the built-in energy storage unit, the on-state voltage drop of the electronic switch, and margin parameters. The lower stop threshold is set higher than the current terminal voltage of the built-in energy storage unit, thereby ensuring positive charge injection and preventing reverse flow of battery energy to the supercapacitor.
[0052] S116, when entering sleep mode, the energy management module shuts down all input switches connected to the perovskite photovoltaic module, physically cutting off the charging circuit to block the path of reverse leakage of electrical energy from the built-in energy storage unit through the photovoltaic module. In this mode, the system maintains periodic sampling by the light sensor, waiting to wake up the charging management function when the light conditions recover.
[0053] In this embodiment of the invention, the communication module supports the MESH self-organizing network protocol and is internally configured with a MESH protocol stack unit. This unit is used to implement a decentralized self-organizing network, multi-hop relay, and route discovery and maintenance. The MESH protocol stack startup and broadcast discovery process includes the following steps: S201, the processing module sends a network initialization command to the communication module. After receiving the command, the communication module activates the MESH protocol stack unit, loads the RF hardware driver, configures the transceiver's operating frequency band and channel bandwidth parameters, and controls the RF front-end to enter the idle listening state.
[0054] S202, the MESH protocol stack unit generates a network probe broadcast message. The data payload of this broadcast message encapsulates the current node's physical layer address, network layer identifier, timestamp, and initial node energy state data. The communication module uses a carrier sense multiple access (CSense) mechanism to detect the current channel state. When the radio frequency hardware determines that the channel has been idle for more than a preset inter-frame interval, the communication module modulates the network probe broadcast message into a radio frequency signal and transmits it to the surrounding space. The backoff algorithm and channel idle assessment process in the CSense multiple access mechanism can be implemented by those skilled in the art with reference to standard wireless LAN protocol stacks; these are well-known technologies in the field and will not be elaborated upon here.
[0055] S203: After neighboring nodes within the communication range detect the radio frequency signal, they demodulate and extract the network probe broadcast message. The neighboring nodes use the received signal strength detection circuit built into the radio frequency front-end or baseband processing chip to collect the radio frequency energy at the receiving end, and output the quantized received signal strength parameter through a hardware register. This received signal strength parameter is affected by spatial propagation distance and environmental fading, forming the basic data for the system to evaluate the physical link quality.
[0056] In step S204, the MESH protocol stack unit of the neighboring node parses the network probe broadcast message. If the received signal strength exceeds a preset link establishment threshold, it activates a built-in pseudo-random number generator to generate a backoff duration. After the corresponding backoff duration, the neighboring node sends a unicast routing response message to the source node to avoid packet collisions in the air channel. This routing response message contains the neighboring node's physical address and node energy status data. Upon receiving the routing response message, the source node extracts the node information and writes it into its neighbor node table maintained in local memory. The neighbor node table dynamically records single-hop reachable nodes that meet the link establishment threshold conditions, completing the neighbor node probe task in the route discovery phase.
[0057] In step S205, after the neighbor node table is established, the MESH protocol stack unit executes a multi-hop relay mechanism to extend network coverage. When a source node needs to send data to a target node outside the single-hop communication range, the source node searches its local routing table. If there is no matching entry for the target node in the routing table, the source node forwards a route request message to all nodes in its neighbor node table. This route request message contains an identifier sequence number and a time-to-live (TTL) field. Upon receiving the route request message, an intermediate relay node checks its local cache for a record with the same sequence number. If a record exists, it discards it to prevent broadcast storms; if not, it records the reverse path, decrements the TTL value by one, and continues broadcasting the message until the route request message reaches the target node, provided the TTL is greater than zero. The target node replies with a route response message along the reverse path, establishing an end-to-end multi-hop routing link, thus achieving network topology construction and route maintenance.
[0058] In this embodiment of the invention, the energy-aware routing unit within the processing module obtains the physical link status and energy reserves of neighboring nodes through the communication module, and performs relay node screening and transmission path maintenance. The process of multi-hop network topology construction and link break reconstruction includes the following steps: S206, the communication module obtains basic status data of neighboring nodes through a pre-broadcast discovery process. The physical layer generates corresponding link quality values based on the received signal strength parameter quantization. Simultaneously, neighboring nodes encapsulate the remaining energy value of the node, as assessed by the battery management unit, in their routing response messages. These link quality and remaining energy parameters are written to local memory as decision variables for multi-hop network topology construction. To prevent division-by-zero errors in subsequent calculations that could lead to system anomalies, the energy-aware routing unit performs performance checks when reading data, directly discarding node entries with link quality or remaining energy values below a preset safety threshold, excluding them from candidate routes.
[0059] S207, when performing route selection, the energy-aware routing unit uses a multi-condition joint metric routing cost function to evaluate the transmission cost through a single relay node. This routing cost function comprehensively considers the link quality between the sending node and its corresponding neighboring nodes, as well as the remaining energy of the neighboring node, and introduces dynamic weighting coefficients to adjust the proportion of link stability and energy balance in the routing decision. The control logic dynamically configures the weight allocation mechanism according to the type of the current communication task: when transmitting data with high real-time requirements, priority is given to ensuring the quality of the communication link; when the overall network power is low or when performing routine inspection tasks, the route is forced to avoid low-power nodes to prevent single-point exhaustion. The specific mathematical model for cost calculation and the logic for preventing division by zero will be elaborated in detail in the routing addressing mechanism section below.
[0060] In step S208, the energy-aware routing unit traverses reachable neighbor nodes and calculates the routing cost value for each neighbor node according to the aforementioned cost function. When constructing a multi-hop path to the target node, the routing request message carries the path cumulative cost value field, the path node list field, and the hop count field during forwarding. Each intermediate relay node, upon receiving a routing request message from the previous hop node, calculates the current single-hop routing cost value based on the received link quality from the previous hop node to this node and the remaining energy of this node, and adds this single-hop routing cost value to the path cumulative cost value field. Subsequently, the intermediate relay node updates the path node list field and the hop count field, and continues to broadcast the routing request message if the lifetime field is greater than zero. Upon receiving multiple routing request messages for different paths, the target node filters based on the path cumulative cost value, hop count, and the node with the lowest remaining energy in the path, and returns a routing response message along the reverse path. The source node selects the path with the lowest cumulative cost value as the final data transmission link based on the received multiple routing response messages. Simultaneously, the system automatically stores the paths ranking second and third in total cost into the local backup routing table. This mechanism, at its underlying logic, prioritizes nodes with good link quality and high remaining energy as relays, preventing critical nodes from running out of power due to frequent data forwarding and extending the overall network lifespan. For the specific operation of the minimum cost path generation algorithm at the routing protocol's underlying level, those skilled in the art can refer to the shortest path algorithm in the relevant theory for programming calculations. Its core optimization logic is well-known in the field and will not be elaborated upon here.
[0061] S209, after the multi-hop network topology is established and enters the data transmission state, the communication module executes the background link failure monitoring program. Adjacent nodes exchange keep-alive messages at a set periodic frequency. When a relay node suffers physical damage, self-destruction due to tamper protection, or power depletion and shutdown, if its adjacent nodes fail to receive the keep-alive message from that node within the set timeout window, or encounter the maximum retransmission failure of the Media Access Control layer when sending data frames to that node, the communication module determines that the single-hop link has been broken.
[0062] In step S210, the node detecting the link failure immediately generates a link error message and broadcasts it internally to the network, notifying the data source node to invalidate the original path entry in its local routing table. Upon receiving the link error message, the source node's energy-aware routing unit searches its local backup routing table and directly invokes the pre-calculated suboptimal path to take over the transmission task. If no backup node is available, a routing request message is resent to find a new relay node. By setting a high-frequency keep-alive message interaction period and a hardware media access control layer response timeout interrupt triggering mechanism, combined with seamless switching of the local backup routing table, this link failure reconstruction mechanism enables the network to complete the transmission link switch and restore data communication within 0.5 seconds when a node fails, provided that the keep-alive message period is no greater than 100ms, the media access control layer's single response timeout is no greater than 50ms, at least one suboptimal path exists in the local backup routing table, and the adjacent nodes participating in the switch are not in emergency sleep mode. If the above conditions are not met, the system enters the re-route discovery process and determines the actual reconstruction time based on the communication duty cycle set by the current power state machine.
[0063] Reference Appendix Figure 4 In this embodiment of the invention, to achieve adjustment of communication transmission power and dynamic scheduling of power consumption of each module in the terminal, the processing module needs to periodically collect the current energy status of the device, external environmental parameters, and underlying radio frequency communication characteristics. The sampling control of external parameters and radio frequency characteristics specifically includes the following steps: S301, the processing module sends a status polling command to the energy management module. The energy management module calls the built-in battery management unit to extract the voltage and charging / discharging current values of the built-in lithium battery. The battery management unit calculates and outputs the remaining battery capacity parameter based on the ampere-hour integration method. To eliminate the cumulative drift error caused by operation without external power supply, when the battery is in a low-rate discharge state, the battery management unit uses the built-in open-circuit voltage and capacity mapping curve to look up a table and correct the initial reference value of the ampere-hour integration. The processing module reads the light intensity value collected by the external ambient light sensor and the photovoltaic module power generation recorded by the energy management module. The obtained remaining battery capacity parameter and power generation constitute the energy state data in the system scheduling logic.
[0064] S302, the processing module initiates a radio frequency (RF) characteristic read request to the communication module. The baseband processing chip of the communication module extracts the received signal strength and signal-to-interference-plus-noise ratio (SNR) of the current communication link from the hardware registers. To avoid evaluation bias caused by environmental fading of a single RF indicator, the processing module normalizes the extracted received signal strength and SNR according to the upper and lower limits of the hardware measurement, and then performs weighted summation according to a preset weighting factor to quantify and generate a comprehensive link quality parameter. Before normalization, the control logic performs a limiting operation on the input parameters, clamping the acquired values that exceed the range to the set upper and lower limit calibration values. This link quality parameter is written to local memory as the data basis for the transmit power control unit and the energy-sensing routing unit to perform decisions.
[0065] S303: The processing module reads the physical state of the surrounding environment through the multi-sensor interface unit within the sensing module and evaluates the task priority based on the type of business being executed. When the RFID reader unit within the sensing module fails to scan a tag, and the temperature, humidity, and vibration sensor data are all within the preset safe range, the processing module sets the current task priority to the routine inspection level. When the multi-sensor interface unit detects that the vibration amplitude exceeds the alarm threshold, or the anti-tamper unit of the anti-tamper module detects an unauthorized disassembly, the processing module responds to a hardware interrupt and adjusts the task priority to the emergency alarm level. For non-destructive abnormal vibration events, when the physical trigger event causing the emergency alarm disappears and the timer reaches the set timeout threshold, the task priority automatically falls back to the routine inspection level; however, for unauthorized disassembly events such as opening the casing or light leakage, the system will treat them as a fatal threat, immediately lock the task priority to the emergency alarm level, suspend routine data acquisition, routing maintenance, and low-power scheduling processes, and enter a safety protection standby state. Upon completion of any one of the following conditions—alarm data reporting, alarm reporting timeout, or detection of a main power supply drop—the processing module initiates an irreversible safety destruction process. When a main power supply rail drop triggers a non-maskable interrupt, the processing module suspends any unfinished routine data acquisition, routing maintenance, or alarm retransmission processes and directly enters the safety destruction process. After the safety destruction process is completed, the system writes a death flag and enters a logical deadlock state, no longer allowing priority fallback.
[0066] S304, the processing module timestamps the collected parameters of remaining battery power, light intensity, link quality, and task priority, and writes them to the local data cache unit. The processing module performs a moving average filtering on the continuously collected numerical parameters, i.e., calculates the arithmetic mean of the historical sampling sequence within a set sliding window length. In the initial stage of system wake-up, if the number of sampled data in the circular buffer is less than the set sliding window length, the processing module reduces the calculation window according to the actual data volume to avoid invalid data participating in the calculation and causing output distortion. The filtered parameter sequence serves as the input parameter for the power dynamic scheduling unit to calculate the working mode. In this embodiment of the invention, the terminal status data includes at least one of remaining battery power, light intensity, link quality, task priority, current power state machine mode, and anti-tamper status flag.
[0067] In this embodiment of the invention, the sensing module integrates an edge computing unit. This unit is used to preprocess the raw data collected by the multi-sensor interface unit and the RFID reader unit locally, and to reduce the amount of data that the communication module needs to transmit back in the multi-hop network through feature extraction and data packaging. The specific process of edge computing feature extraction and payload generation includes the following steps: S305, the edge computing unit reads the raw multi-sensor sampling sequences within a continuous time window from the local data cache unit. These raw multi-sensor sampling sequences are divided into high-frequency sampling of vibration acceleration parameters and low-frequency sampling of temperature, humidity, water level, and tilt parameters. For the temperature, humidity, water level, and tilt parameters, which exhibit relatively stable changes, the edge computing unit performs statistical feature extraction. The control logic traverses the sampling sequences within the current time window, calculates and extracts the maximum, minimum, and arithmetic mean values, and compares each parameter with a preset safety baseline, generating corresponding out-of-limit flags. This preset safety baseline is calibrated by technicians based on the statistical maximum boundary values of the equipment deployment environment under historical normal operating conditions and is stored in a local register. The edge computing unit retains these statistical features and out-of-limit flags while discarding low-frequency raw sampling points, achieving dimensionality reduction of the parameters in the normal environment.
[0068] S306: For high-frequency sampled vibration acceleration parameters, the edge computing unit calls a built-in one-dimensional convolutional neural network model to classify physical states, avoiding the transmission of raw time-domain waveform data back to the wireless communication network. The input data for this one-dimensional convolutional neural network model is a multi-channel one-dimensional time series vector. The specific number of channels in the vector is determined by the physical directions of the sensor's three-axis acceleration, and the vector length is determined by the total number of fixed sampling points within a single decision time window. Before the data is input into the model, the edge computing unit performs preprocessing logic on the input sequence, namely, subtracting the arithmetic mean of the sequence point by point to remove the DC component of gravity, and performing amplitude normalization processing according to the hardware range of the accelerometer.
[0069] S307, this one-dimensional convolutional neural network model internally includes a data input layer, two cascaded one-dimensional convolutional layers, a global average pooling layer, and a classification output layer. ReLU activation function layers are cascaded at the outputs of both one-dimensional convolutional layers to perform nonlinear mapping on the linear feature maps output by the convolutional kernels. In one specific embodiment, a vibration accelerometer collects triaxial acceleration data at a sampling frequency of 100Hz to 800Hz, with each decision time window containing 128 to 512 consecutive sampling points. The model input tensor is a three-channel one-dimensional time series. The first one-dimensional convolutional layer has 8 to 32 convolutional kernels with a kernel length of 3 to 7 and a stride of 1; the second one-dimensional convolutional layer has 16 to 64 convolutional kernels with a kernel length of 3 to 7 and a stride of 1. The classification output layer outputs three probability values: normal vibration state, illegal damage state, and geological anomaly state. When the output probability of illegal damage or geological anomaly states exceeds a preset classification threshold, and the same type of output is maintained for at least two consecutive decision time windows, the edge computing unit generates the corresponding anomaly state label. A one-dimensional convolutional layer extracts local abrupt change features of vibration signals at different time scales using internally parameterized sliding convolution kernels. The feature map, after nonlinear activation, is passed to a global average pooling layer, compressing the spatial dimension into a one-dimensional vector, and then passed to the classification output layer. The results output by this classification output layer correspond to three specific business physical states: normal vibration states caused by environmental wind load disturbances, illegal damage states caused by impacts on the equipment casing, and geological anomaly states caused by tilting or deformation of the deployment base.
[0070] S308, the aforementioned one-dimensional convolutional neural network model adopts an engineering deployment method that separates offline training from terminal inference. During the offline construction phase, technicians collect sensor vibration waveform sequences from a real deployment environment and manually label them with one-hot codes corresponding to the physical events to construct the model training sample set. The training process uses a cross-entropy loss function to evaluate the difference between the network's predicted probability distribution and the actual one-hot code label distribution. The model iteratively updates the network layer weights on the host computing platform using the backpropagation algorithm. After the loss value converges, the network layer floating-point parameters are quantized into low-bit-width integers and burned into the read-only memory of the edge computing unit for local inference. The gradient chain differentiation and optimizer operation logic in the backpropagation algorithm can be configured by those skilled in the art with reference to standard deep learning training frameworks; these are well-known technologies in the field and will not be elaborated upon here.
[0071] S309, the edge computing unit performs memory alignment and data aggregation on the extracted low-frequency statistical features, the physical state tags output by the neural network model, and the personnel location identification codes and asset identification codes obtained by the RFID card reader unit. In a product-level embodiment, the RFID card reader unit can read passive RFID tags within a range of 3m to 5m to meet the needs of forestry inspection personnel for location, asset identification, and short-range passive tag identification. The processing module encapsulates the aggregated information into a protocol stack according to the task priority of the data source, generating a standard format communication data payload. When the remaining battery power reported by the energy management module is lower than the preset energy-saving threshold, the edge computing unit activates the built-in LZW dictionary compression algorithm to perform secondary byte compression on the generated communication data payload. To prevent the high information entropy of the data payload from causing volume expansion after compression, the edge computing unit calculates the volume ratio before and after compression. If the number of bytes in the compressed payload is greater than or equal to the number of bytes in the original payload, the control logic discards the compression result and retains the original communication data payload to execute subsequent processes. The final payload data that passes the verification will be delivered to the communication module to await radio frequency transmission.
[0072] Reference Appendix Figure 5 In this embodiment of the invention, to improve the battery life of self-powered devices, the processing module is configured with proactive power scheduling logic based on hardware state awareness. This logic combines the current remaining battery power with task priority to drive a multi-level dynamic power state machine to perform system-level operating mode switching. The operation process of the multi-level dynamic power state machine specifically includes the following steps: S401: The processing module reads the remaining battery power parameter reported by the energy management module and uses it as the determination condition for the power consumption state machine. The control logic has multiple decreasing power thresholds pre-configured in the local register. When the remaining power is greater than 80%, it enters full-speed mode. When the remaining battery power is between 50% and 80%, it enters energy-saving mode; When the remaining battery power is between 20% and 50%, it enters low power mode; When the remaining battery power is less than or equal to 20% and no emergency alarm event is triggered, it enters emergency sleep mode; In a preferred embodiment, the high battery threshold is set to 80%, the medium battery threshold to 50%, the low battery critical threshold to 20%, and the extreme low battery warning threshold to 8%. These specific numerical points, together with the hysteresis compensation mechanism, constitute the preferred calibration scheme for the power consumption state machine, and the value of the hysteresis compensation constant is greater than the equivalent battery change value corresponding to the voltage rebound of the energy storage unit after load removal. To prevent frequent state machine transitions caused by small fluctuations in the remaining battery charge around a specified threshold, the processing module introduces a hysteresis comparison mechanism in the state switching determination. When the remaining charge drops below a certain threshold, the system triggers a downgrade switch; when the remaining charge recovers, it must exceed the threshold and be supplemented by a preset hysteresis compensation constant before the system triggers an upgrade switch. The aforementioned battery charge determination thresholds and hysteresis compensation constants are pre-calibrated by technicians based on the rated capacity and discharge characteristic curve of the built-in lithium battery. The value of the hysteresis compensation constant is greater than the equivalent charge value of the terminal voltage rebound after the battery load is removed, so as to ensure the physical stability of the state determination and avoid the invalid power loss caused by frequent reconfiguration of the hardware in the critical state.
[0073] The S402 multi-level dynamic power consumption state machine is divided into full-speed mode, energy-saving mode, low-power mode, and emergency sleep mode. The processing module executes state transitions based on the task priority output from the multi-sensor interface unit. When the task priority is at the emergency alarm level, the processing module first compares whether the current remaining battery power is higher than the software-set extreme low battery warning threshold (this warning threshold is calibrated to be slightly higher than the microcontroller's underlying hardware power-loss reset threshold). If the software-set extreme operating power is met, the state machine transitions to full-speed mode, enabling the highest sampling rate of all sensors and keeping the RF receiving link of the communication module always open. If the current power is lower than the extreme low battery warning threshold, the processing module further determines whether the current available energy is higher than the single alarm transmission energy threshold. If it is higher than the single alarm transmission energy threshold, the system downgrades to execute single alarm data packet transmission and then enters sleep mode; if it is lower than the single alarm transmission energy threshold, the system abandons regular RF transmission, prioritizes local sensitive data erasure or writes the alarm payload to the low-power hold buffer, and waits for photovoltaic input to recover or the backup energy storage network to reach the transmission energy threshold before attempting to report again. When the task priority is at the routine inspection level, the state machine performs a smooth transition between power saving mode, low power mode and emergency sleep mode based on the power range determination result with hysteresis comparison mechanism.
[0074] In S403, when the state machine is in power-saving or low-power mode, the processing module controls overall power consumption by adjusting the sleep cycle of the communication module and the sampling frequency of the sensing module. The control logic incorporates a power configuration lookup table, with different clock multiplication factors and timer reload values corresponding to different modes. Upon entering low-power mode, the processing module extends the polling interval of the multi-sensor interface unit based on the lookup table parameters, and simultaneously sends a sleep command to the communication module via the internal bus, increasing its duty cycle in the sleep state at the media access control layer. By extending the wake-up interval, the system reduces dynamic current consumption while meeting basic network keep-alive response requirements.
[0075] S404: When the remaining battery power is below the low-power threshold and no emergency alarm event is triggered, the state machine enters emergency sleep mode. The processing module cuts off the external power supply circuits of the multi-sensor interface unit, RFID card reader unit, and wireless communication RF front-end, but retains the low-power hold-up power domain. The low-power hold-up power domain includes at least the energy management module's status monitoring unit, the low-frequency oscillator of the internal real-time clock, the micro-switch interrupt detection circuit, the low-power comparator of the casing light leakage detection front-end, and the light recovery wake-up detection unit. Thus, in emergency sleep mode, the terminal can still respond to hardware events such as casing opening, casing damage and light leakage, main power supply drop, and photovoltaic input recovery. When photovoltaic input recovers and continues for more than the preset anti-shake delay period, the energy management module sends a hardware wake-up interrupt to the processing module; after the processing module exits emergency sleep mode, it reassesses the remaining power of the energy storage unit, and when the communication recovery conditions are met, it controls the communication module to reload the RF hardware driver, execute neighbor node discovery, time synchronization, neighbor state routing table refresh, and necessary routing request processes; when the reassessment result is still below the communication recovery conditions, the system continues to maintain only the low-power hold-up power domain. The conditions for communication recovery include that the remaining power of the built-in energy storage unit is higher than the preset communication recovery power threshold, the power rail voltage of the communication module is stable, and no safe destruction process is detected.
[0076] In this embodiment of the invention, when the multi-level dynamic power state machine switches to low-power mode, the processing module intervenes in the clock management of the underlying hardware at the micro-timing scale. This is achieved by dynamically reconfiguring the medium access timing of the communication module and the analog-to-digital converter triggering logic of the sensing module, thereby reducing the overall dynamic power consumption of the system. The specific process of RF duty cycle compression and sampling frequency reduction includes the following steps: In step S405, the processing module extracts the remaining battery power and the instantaneous power generation of the photovoltaic modules, and calculates the target sleep period for the communication modules in the next stage based on the joint scheduling model. The control logic uses the set basic network keep-alive time period as a benchmark, calculates the power attenuation ratio based on the ratio of the current remaining battery power to the power threshold, and extends the basic period positively. Simultaneously, it extracts the current instantaneous power generation of the photovoltaic modules, multiplies it by a preset energy gain weight, and negatively deducts from the extended time, thus dynamically deriving the target sleep period. To prevent the calculated sleep period from being too long, causing the node to be judged as offline by the network routing, or from being too short due to a sudden increase in photovoltaic power, rendering the sleep period ineffective for energy saving, the processing module performs boundary limiting operations on the target sleep period after calculation. The control logic forces the calculated value to be between the maximum keep-alive interval and the minimum communication interval allowed by the network medium access control layer protocol.
[0077] In S406, the processing module sends the limited target sleep cycle value to the communication module via the internal communication bus. The baseband processing chip of the communication module then rewrites the wake-up timer register of the internal media access control layer. During sleep, the communication module disables the low-noise amplifier and frequency synthesizer at the front end of the RF receiver, maintaining only the low-power internal resistor-capacitor oscillator to calculate the elapsed time. When the internal wake-up timer overflows, the RF receiver opens a fixed-length receive listening window to capture network beacon messages from neighboring nodes. To eliminate timing drift errors caused by ambient temperature changes in the internal resistor-capacitor oscillator, after successfully capturing a neighboring node's network beacon message, the communication module extracts the global timestamp from the message to calibrate the local clock and dynamically adjusts the advance opening compensation amount of the next receive listening window based on the calibration deviation, ensuring the alignment and overlap of the transmit and receive timings. By extending the time interval between adjacent listening windows, the RF module's receive duty cycle is compressed, reducing RF standby current consumption.
[0078] S407, synchronized with the sleep cycle adjustment on the communication side, implements asymmetric sampling frequency reduction control for the sensing module. For slowly changing environmental parameters such as temperature, humidity, and water level, the processing module extends the trigger cycle of the corresponding analog-to-digital conversion channel by modifying the reload value of the internal hardware timer. For vibration acceleration parameters, directly reducing the sampling rate would lead to missed high-frequency abrupt signal changes. Therefore, the processing module disables the continuous analog-to-digital conversion function of the vibration sensor and configures it to interrupt threshold standby mode via the serial peripheral interface. In this mode, when the selected vibration sensor has an internal comparator or threshold interrupt unit, it continuously monitors the external input level in a low-power state. When physical vibration is converted into an electrical signal and the level exceeds the preset hardware alarm threshold, the analog comparator outputs a hardware interrupt pulse to wake up the processing module. The system then immediately cuts off sleep mode and resumes high-frequency continuous sampling to capture the complete vibration waveform sequence.
[0079] S408: As the sampling frequency decreases and the RF duty cycle is compressed, the edge computing unit synchronously adjusts the local data buffer's delivery logic. Discrete sensor data acquired through frequency reduction is accumulated and appended to the circular queue of the local data buffer unit. When the available storage space in the circular queue falls below the set safety level and the communication module is still in a dormant state, the processing module triggers a priority overwrite arbitration mechanism. The control logic traverses the current queue, forcibly overwriting the newly generated over-limit alarm data and high-frequency characteristic data to the address of the existing regular low-frequency environmental inspection data in the queue. If there is no low-frequency environmental inspection data in the queue, the oldest high-frequency characteristic data in the queue is overwritten based on the timestamp, thus ensuring the local retention of critical emergency information. When the number of data payload bytes accumulated in the queue reaches the maximum transmission unit length set by the communication module, or the data dwell time reaches the set mandatory reporting time tolerance, the processing module activates the communication module's transmission link to execute data packet transmission. For the specific software pointer offsets and circular overwrite operations of data caching and queue memory management, those skilled in the art can configure them according to the standard application programming interface of the embedded operating system, which is a well-known technology in the field and will not be described in detail here.
[0080] In this embodiment of the invention, when the communication module performs multi-hop relay data transmission, it is constrained by the energy-limited characteristics of the self-powered device. Therefore, the processing module is equipped with an energy-aware routing unit. This unit jointly evaluates the spatial radio frequency characteristics and the physical energy state of the node when establishing the communication routing table and selecting the next-hop node. The specific process of the energy-aware routing cost function and node selection addressing includes the following steps: S409, the processing module controls the communication module to periodically listen to the routing beacon messages broadcast by neighboring network nodes. The baseband processing chip extracts the unique identifier of the neighboring node and the current remaining battery power parameter of the neighboring node from the demodulated beacon messages. Simultaneously, the processing module combines the received signal strength and signal-to-interference-plus-noise ratio measured by local RF hardware to quantize and generate the link quality parameters for reaching the neighboring node. To unify the dimensions of different physical parameters and eliminate the interference of numerical magnitude differences on subsequent algebraic operations, the processing module uses a preset full-scale reference value to map the extracted remaining battery power parameter and link quality parameter to dimensionless intervals. The above normalized parameters are integrated into table entries and updated in the neighbor state routing table in local memory.
[0081] S410, when performing neighbor state table updates, route discovery, or route table failure reconstruction, the energy-aware routing unit traverses the neighbor state routing table. It calculates the routing cost of each candidate neighbor node. From the physical principles of routing addressing, the more remaining power and the better the RF link quality of the next-hop node, the lower its corresponding routing cost should be. This routing cost value is calculated using the following formula: ; in, The normalized link quality parameters; This represents the normalized remaining battery power parameter of adjacent nodes. This refers to the link quality weighting coefficient. Energy state weighting coefficient; It is obtained by normalizing the received signal strength and the signal-to-interference-plus-noise ratio. Specifically, the received signal strength and the signal-to-interference-plus-noise ratio are first clamped to their respective hardware measurement ranges, then mapped to the dimensionless interval of 0 to 1, and then summed according to the preset radio frequency weights. The remaining battery percentage reported by candidate neighbor nodes is mapped to a dimensionless interval of 0 to 1. To prevent zero bias constants, which are used to simultaneously affect the denominator of the link quality term and the node energy term, their value is greater than the lower limit of the processor's floating-point precision and less than one-tenth of the normalized resolution of the link quality and node energy.
[0082] Based on the above formula, the specific calculation logic for the routing cost value is as follows: Add a preset anti-zero bias constant to the normalized link quality parameter, take the reciprocal, and multiply by the link quality weight coefficient to obtain the first value; add the preset anti-zero bias constant to the normalized candidate neighbor node's remaining battery power parameter, take the reciprocal, and multiply by the energy state weight coefficient to obtain the second value; finally, add the first and second values to output the routing cost value for the corresponding candidate neighbor node. The anti-zero bias constant is introduced here. To prevent the processor's underlying hardware from causing a division-by-zero anomaly when the denominator approaches zero in actual physical measurements, its value is greater than the lower limit of the processor's floating-point arithmetic precision and less than one-tenth of the normalized resolution of link quality and node energy.
[0083] S411, to prevent nodes with excessive forwarding loads from going offline prematurely due to energy depletion, the processing module dynamically calculates the link quality weight coefficient and energy status weight coefficient based on the task priority of the current communication data load. When transmitting emergency alarm-level tasks, the system must prioritize low latency and high reliability of network transmission, and the control logic sets the link quality weight coefficient to be greater than the energy status weight coefficient. When transmitting routine environmental inspection-level tasks, the system prioritizes network-level energy balance over transmission latency, and the control logic increases the energy status weight coefficient to be greater than the link quality weight coefficient. Through this dynamic weight allocation logic, the routine inspection data stream will adaptively bypass neighboring areas with low energy levels, achieving spatial balance of network-level energy consumption.
[0084] S412, when the terminal, acting as a relay node, receives a data frame sent by the previous hop node, the processing module parses the destination address and path information in the data frame header. If the destination address is not the current node, the data frame is cached as data to be forwarded in the local data cache unit, and the next hop node is determined based on the local routing table or the energy-aware routing unit. When the communication data payload needs to be sent, the processing module first checks whether there is a path entry to the destination node in the local routing table. If a path entry exists, the next hop node is determined based on the path entry, and the physical medium access control address of the next hop node is encapsulated in the data frame header and sent to the radio frequency link for transmission. If there is no path entry in the local routing table, or if an existing path entry fails due to a link error message, keep-alive timeout, or node energy falling below the safety threshold, the processing module initiates a route discovery process based on the cumulative path cost value and updates the local routing table and backup routing table according to the newly received route response message. After initiating the route discovery process, the energy-aware routing unit first eliminates failed nodes with link quality or remaining energy below a preset safety threshold. If the routing cost of all remaining candidate neighbor nodes exceeds the network island tolerance threshold, it is determined that all available relay nodes in the surrounding area are in a low-energy or channel-fading state. The network island tolerance threshold is calculated by technicians based on a weighted average of the extreme link cost corresponding to the lower limit of the RF receiver sensitivity and the extreme energy cost corresponding to the battery discharge cutoff voltage. At this point, the processing module triggers the route fallback protection mechanism, abandons multi-hop addressing forwarding, increases the power register of the RF transmitter unit to the hardware output limit, and attempts to establish a single-hop direct connection with the network aggregation layer gateway node. If the underlying response mechanism reports a single-hop communication timeout failure, the processing module cancels the sending request and falls back to the local data buffer unit. For the data frame payload fragmentation and encapsulation and channel conflict avoidance mechanisms in the network communication protocol stack, those skilled in the art can refer to the standard wireless sensor network protocol for configuration. The basic networking logic is well-known in the field and will not be elaborated here.
[0085] In this embodiment of the invention, to address the risk of physical damage to self-powered devices in unattended outdoor environments, an anti-tamper module integrates a shell anti-tamper unit. This unit combines mechanical and optical dual sensing mechanisms to monitor the physical sealing status of the shell. The specific process of hardware monitoring of shell integrity parameters includes the following steps: The S501 tamper-proof module's housing tamper-proof unit includes a microswitch and a light sensor. The microswitch is installed on the inner edge of the access cover or the main housing's closed joint, its mechanical contacts being in a pressure-closed state when the equipment is closed. The light sensor is located on the main board in a sealed, dark chamber area inside the housing, with its photosensitive surface facing the housing wall, used to monitor external ambient light entering after the sealed housing breaks. The two sensors are physically complementary, covering both cover opening and housing breakage as physical damage modes.
[0086] The S502 microswitch's electrical circuit is connected to the system power rail via a pull-up network, and its signal output is connected to an external interrupt pin with asynchronous wake-up functionality in the processing module. When the casing is fully closed, the contacts are grounded, and the interrupt pin remains low. When the cover is opened, the contact spring releases, and the output level changes. Considering that the self-powered device is in a low-power sleep mode for extended periods, to prevent the device from frequently waking up the main control chip due to short-term glitches caused by mechanical vibration or electromagnetic interference from environmental wind loads, a hardware RC low-pass filter circuit is configured at the front end of the interrupt input channel for preliminary physical debouncing. When the level transition signal passes through the hardware filter channel, it triggers system wake-up. The processing module performs secondary state confirmation at the software level, continuously reading the pin state within the set debouncing time window. If the reading result remains continuously high, the control logic sets the microswitch's on flag.
[0087] In S503, the photosensitive sensor is connected to the analog-to-digital converter (ADC) channel of the processing module via a voltage divider sampling circuit. In one specific embodiment, the photosensitive sensor is connected between the voltage divider sampling node and the system power supply, and a pull-down resistor is connected between the voltage divider sampling node and ground. When the housing structure is intact, no light enters the internal dark chamber, and the ADC collects the static base voltage caused by the dark current of the components. When the housing is damaged by external force, causing cracks and allowing ambient light to leak in, the equivalent impedance of the photosensitive sensor decreases, resulting in an increase in the voltage of the voltage divider sampling node. The processing module calculates the current light leakage state parameter based on the photosensitive mutation judgment logic. Specifically, the control logic extracts the currently read photosensitive sensor sampling voltage value, subtracts the base voltage constant pre-calibrated in the factory fully sealed state of the device, and calculates the relative voltage offset. Subsequently, this relative voltage offset is divided by the full-scale reference voltage of the ADC to obtain the relative percentage value of the light leakage state parameter. Considering that drastic temperature fluctuations inside outdoor equipment can cause dark current drift in photosensitive elements, leading to a shift in the static substrate voltage, the control logic uses real-time temperature data extracted from the onboard temperature sensor to dynamically compensate for the substrate voltage constant. When the percentage of light leakage calculated after temperature compensation exceeds a preset light leakage alarm threshold, the processing module sets the light leakage flag on the photosensitive sensor. This light leakage alarm threshold is determined by technicians based on the light transmittance of the housing material, the extent of internal wiring obstruction, and the expected minimum aperture light leakage.
[0088] In step S504, the processing module performs a logical OR operation on the on flag of the microswitch and the light leakage flag of the photosensitive sensor. When either flag is set, the system determines that the integrity of the casing has been compromised. At this time, the processing module suspends the currently executing routine environmental parameter acquisition or sleep process and generates a physical damage alarm payload containing a damage type identifier and the device's physical address. This payload is transferred to the communication module, triggering the highest priority transmission of the Direct Memory Access (DMA) channel. Under the premise of complying with the media access control layer channel access rules, the communication module places the alarm payload into the highest priority transmission queue and completes the network reporting operation using the minimum contention window, the shortest allowed inter-frame interval, and the maximum allowed retransmission count.
[0089] In this embodiment of the invention, when the integrity of the casing is compromised or the device faces unauthorized intrusion, the anti-tamper module, through linkage with the underlying microcontroller hardware mechanism, performs an emergency destruction operation on local core data to prevent the physical theft of network access credentials and encryption keys. The specific process of hardware interrupt triggering and sensitive data erasure includes the following steps: In the S505, while executing the aforementioned alarm operation, the internal hardware voltage monitoring module continuously monitors the main power rail status. In actual damage scenarios, abnormal disassembly of the casing is often accompanied by a physical disconnection of the main power supply connection. When any of the following conditions are met: alarm reporting is completed, alarm reporting reaches a preset timeout period, or the main power rail voltage drops below a set power drop threshold, the processing module initiates the safe destruction process. If the main power rail voltage has not dropped, the processing module, under main power supply conditions, performs sensitive data clearing, external memory locking, and death flag writing through the safe erasure unit; if the main power rail voltage has dropped, the interrupt controller triggers a non-maskable interrupt request, and the control logic forcibly preempts the current execution pointer of the central processing unit, suspending all running operating system threads. At this time, the energy management module automatically controls the analog switch to seamlessly switch the system power supply to the backup energy storage network composed of onboard tantalum capacitors, providing the transient discharge energy required to maintain the emergency erasure operation for the core logic circuits and storage modules of the microcontroller.
[0090] After entering the secure erasure process, the processing module executes asynchronous, differentiated erasure timing according to the preset data sensitivity level. When the secure erasure process is triggered by a main power supply drop, the erasure timing runs within a non-maskable interrupt service routine. Based on the engineering principle of being constrained by both time and energy, the erasure strategy for different data types must be adapted to the erasure characteristics of their physical carriers. For high-security data such as device network access credentials, communication encryption keys, and local identity certificates, the system has centrally mapped them to the microcontroller's internal battery backup register and real-time clock backup domain during memory architecture partitioning. The processing module does not perform conventional software addressing erasure operations; instead, it directly writes the specified tamper-proof erasure instruction to the internal power control register. This instruction forcibly pulls low the physical reset control line of the backup domain storage unit through internal hardware connections, disrupting the latch state of the static random access memory array and completing the physical zeroing of all high-security data within a finite number of clock cycles. This hardware-based erasure timing avoids the risk of power failure due to excessively long software cyclic erasure times.
[0091] S507, after confirming that high-security data has been cleared, the processing module uses the currently available power supply to intervene in the destruction process of routine business data; when the main power supply rail has fallen, the currently available power supply is provided by the backup energy storage network. For environmental acquisition data and operation logs stored in an external serial non-volatile memory, since the energy and time required to perform a full-page erase far exceed the supply limit of the backup energy storage network, the processing module issues write protection and read protection lock commands to the external memory through the serial peripheral interface. In a preferred embodiment, the external serial non-volatile memory is a secure storage device with one-time programmable security control bits, read protection lock bits, or secure erase commands; the processing module configures the target storage area to an irreversible read protection state by modifying its one-time programmable security area control bits. Environmental acquisition data, operation logs, and network communication cache data are preferably encrypted using a local data encryption master key before being written to the external serial non-volatile memory. For ordinary non-volatile memory that does not have irreversible read protection, the processing module prioritizes the local data encryption master key erasure and access control locking operations to prevent the encrypted business data in the external memory from being recovered and read.
[0092] After completing the zeroing and locking operations of the storage medium (S508), the processing module temporarily stores the system death flag write request and continues to determine whether the electronic fuse programming conditions are met. If the processing module does not have an on-chip electronic fuse array, or the transient available energy of the backup energy storage network is insufficient to complete the electronic fuse programming, the processing module writes the immutable system death flag to the secure boot area of its internal non-volatile memory. If the processing module has an on-chip electronic fuse array and the transient available energy meets the physical fusing requirements, the system first executes the electronic fuse programming process and completes the write or latch operation of the system death flag before triggering a system reset. The system death flag is preferably stored in a one-time programmable memory area, an electronic fuse array, a secure boot control area, or a secure non-volatile memory area protected by hardware read / write protection mechanisms. The bootloader prioritizes reading the system death flag during each power-on reset; when the flag is detected to be set, the bootloader refuses to load the application firmware and keeps the communication interface, debug interface, and firmware upgrade interface closed. After completing the above security lock, the control logic writes a freeze instruction to the independent watchdog register to prevent reset triggering.
[0093] In this embodiment of the invention, to address the risks of invasive micro-probe detection and firmware reverse engineering that self-powered devices may face under extreme security threats, the tamper protection module, after performing data logic erasure, further activates the physical irreversible destruction mechanism of the underlying microcontroller. The specific process of the core chip's physical fuse execution logic includes the following steps: In S509, upon receiving a destruction command from the tamper protection module, the processing module cuts off the peripheral power supply to the general purpose input / output pins and suspends the internal direct memory access controller and non-core peripheral bus to reduce the system's basic static power consumption. When the processing module has an on-chip charge pump boost circuit and an electronic fuse programming controller, the processing module activates the on-chip integrated hardware charge pump boost circuit. When the processing module or the tamper protection module is equipped with a chip fuse unit, the chip fuse unit controls the MOSFET to enter a breakdown state, performing irreversible physical damage to the core chip or debug access path. This charge pump boosts the current system power rail voltage to the high-voltage level required for electronic fuse programming, establishing the high-voltage transient programming power domain required for physical fuse breaking. When the selected processing module does not have an on-chip electronic fuse array, the tamper protection module performs secure erase, debug interface lock, and secure boot lock operations as an alternative implementation. This alternative implementation does not perform the electronic fuse programming pulse and shadow register reload verification steps, but instead puts the terminal into a logical deadlock state through key erase, debug interface lock, secure boot lock, and system death flag writing.
[0094] In S510, to prevent insufficient residual power in the backup energy storage network from maintaining the full programmed pulse width during a power outage, which could result in the electronic fuse being in a partially blown state, the processing module reads the current power rail voltage value via an analog-to-digital converter and calculates the transient available energy using a preset energy assessment model. Physically, the energy released by the energy storage capacitor is determined by the square difference between the voltages across the capacitor and the efficiency of the subsequent power conversion circuit. The calculation expression for this energy assessment model is as follows: ; in, The transient available energy represents the computational output; This represents the physical nominal capacity of the energy storage capacitors in the backup energy storage network. This represents the current power rail voltage value read by the analog-to-digital converter; This represents the minimum input operating threshold voltage at which the internal hardware charge pump can maintain the boost output. This threshold voltage is determined by the physical cutoff characteristics of the underlying hardware circuitry. The conversion efficiency constant represents the charge pump boost circuit, and its value ranges from 0 to 1. It is determined by actual circuit impedance and switching losses through measured measurements. When the calculated efficiency is... When the energy level exceeds the preset physical fuse breaking energy threshold, the processing module grants execution permission to the electronic fuse programmable controller. This physical fuse breaking energy threshold is calibrated by technicians based on the thermal integral required to break the polysilicon link. If the available energy is lower than this threshold, the processing module abandons the hardware fuse breaking operation and instead triggers a low-energy degradation and destruction mechanism, prioritizing the erasure or invalidation of the device's network access credentials, communication encryption keys, local identity certificates, data encryption master keys, and key index tables. For external business data stored in an encrypted manner, the system erases the data encryption master key, making the external business data undecryptable even if it is retained in the storage medium. If the remaining power still meets the writing conditions of the secure storage area, the processing module further writes an invalid key flag and a destruction status flag to the secure storage area, and then allows the system to enter the undervoltage reset state.
[0095] S511: When the processing module has an on-chip electronic fuse array, an electronic fuse programming controller, and an on-chip charge pump boost circuit, and after passing the transient available energy verification, the processing module writes the target fuse address to the address register of the electronic fuse programming controller and sends a programming trigger bit to the control register. If the processing module does not have the hardware conditions for electronic fuse programming, but the anti-tamper module is configured with a chip fuse unit and the transient available energy of the backup energy storage network meets the energy conditions required for MOSFET breakdown, the system performs a MOSFET breakdown-type physical fuse operation; if neither the electronic fuse programming conditions nor the MOSFET breakdown conditions are met, the system skips the physical fuse step and instead executes the debug interface locking, security boot locking, key erasure, and system death flag writing process.
[0096] In S512, when the system executes the electronic fuse programming step, after the programming pulse ends, the processing module sends a reload command to the electronic fuse programming controller, causing the internal shadow register of the chip to re-latch the state of the physical array. The control logic reads the state verification bit of the shadow register to confirm that the control bits related to the debug interface and safety exhibit a high-impedance open-circuit characteristic. After the verification is completed, the processing module triggers the system reset pin of the internal power management unit, causing the chip to enter a physically locked state without a debug interface and without boot privileges.
[0097] When the system skips the electronic fuse programming step because it lacks an on-chip electronic fuse array, electronic fuse programmable controller, on-chip charge pump boost circuit, or because the transient available energy is lower than the physical fusing energy threshold, the processing module does not perform shadow register reload verification, but directly executes the debug interface locking, security boot locking, key erasure, and system death flag writing process.
[0098] Before or after a reset, the bootloader reads the electronic fuse status and system dead flag. When any safety lock flag is detected as set, the bootloader refuses to load the application firmware and keeps the communication interface, debug interface, and firmware upgrade interface closed, causing the terminal to enter a logical deadlock state. Subsequently, the processing module or power management control logic shuts down the system master clock tree, causing the device to enter a deep suspend state and wait for power depletion.
[0099] Specific application examples In one product-level embodiment, the self-powered wireless self-organizing network terminal has dimensions of 200mm × 200mm × 100mm and a weight of less than 1.5kg. It is suitable for manual deployment, vehicle delivery, scattering, or airdrop deployment, with a scattering radius greater than 500m. The perovskite photovoltaic module has a photoelectric conversion efficiency greater than 25%. The built-in energy storage unit capacity ranges from 500mAh to 6000mAh, with a preferred capacity of 5000mAh. The operating temperature range is -40℃ to +85℃. Under pure battery operation without external light supplementation, the terminal has a continuous operating time of 8 to 10 hours; under 10000 lux illumination, the terminal's operating time is greater than 72 hours. The RFID reader unit can read passive RFID tags within a range of 3m to 5m. When the anti-tamper module detects an open casing, casing damage leading to light leakage, or a drop of the main power supply rail, the response time from hardware interruption to the initiation of the safety destruction process is less than 0.1s. To further assist those skilled in the art in understanding the technical details of the present invention, the following will describe in detail the overall operation process, key numerical calculations, experimental verification, and accompanying figures of the present invention in a progressively structured manner, using specific scenarios of monitoring forest fires and illegal logging in dense forests.
[0100] In practical applications of monitoring forest fires and illegal logging in dense forests, this invention deploys self-powered wireless self-organizing network terminals in virgin forests perpetually shaded by tree canopies (with an average daily ambient illuminance of less than 400 lux). These terminals are responsible for environmental sensing, abnormal vibration identification (such as damage from illegal logging), and terminal anti-tampering and self-destruction. Overall, the system successfully ensures positive energy gains in extremely low-light environments by combining perovskite low-light photovoltaic modules with a low-power pulse charging mode. Simultaneously, it utilizes a lightweight 1D-CNN edge computing model to extract features, reducing the burden on radio frequency transmission through a "computing power for communication" strategy, and employs dynamic weighted routing to avoid low-power nodes.
[0101] In a large-scale outdoor comparative experiment involving 50 nodes over 30 days, the proposed solution achieved significant results: the remaining battery power of the terminal was always maintained in a safe range of over 65%, the total power consumption of hardware and software in a single emergency was reduced by about 90.2%, and the network lifespan was extended by 50% compared with the traditional solution, successfully avoiding network paralysis and "communication island" phenomena caused by excessive power consumption of the central node.
[0102] In the actual implementation process, the intelligent scheduling and security of the terminal rely on the evaluation of two key mathematical models.
[0103] The first item is the dynamic addressing mechanism of energy-aware routing, the core of which lies in the multi-condition routing cost function: ; When performing routine temperature and humidity checks (with remaining power weight set) Link quality weight The cost function forces the data flow to bypass low-power nodes; however, when the edge computing module detects an emergency alarm event, the weights are instantly reversed. , Prioritize securing high-quality radio frequency links to ensure extremely low latency.
[0104] The second is a safety destruction mechanism in the event of physical damage. When the main power supply drops due to violent dismantling, the system switches to the backup tantalum capacitor network. Power supply. According to the transient available energy assessment formula: ; Charge pump cutoff threshold ( Charge pump cutoff threshold Boost efficiency The energy released in the emergency can be calculated to be 4.59 mJ. This energy is much greater than the 2.5 mJ thermal integral required to burn out the on-chip electronic fuse, thus ensuring that the core key completes physical-level irreversible melting before a complete power failure.
[0105] The actual operational effects and experimental results of the above technical solutions can be intuitively explained through three two-dimensional basic diagrams.
[0106] Reference Appendix Figure 6 The horizontal axis represents the number of test days, and the vertical axis represents the percentage of remaining battery capacity. The lines in the graph are clear and do not intersect. The legend is placed independently in the blank area in the upper right corner of the graph (not overlapping with the data lines). The measured data trajectory of this invention (solid line) maintains a gentle decline of more than 65%, while the traditional monocrystalline silicon solution (dashed line) shows a steep linear decline until it bottoms out.
[0107] Reference Appendix Figure 7The horizontal axis corresponds to the two processing schemes, and the vertical axis represents the total charge energy consumed (mJ). The legend is placed in the upper left corner. It can be seen intuitively that the column height of the traditional transparent transmission scheme is 18.5mJ, while the column height of the edge computing scheme of this invention is only 1.8mJ, which clearly verifies the high energy efficiency ratio of exchanging computing power for communication.
[0108] Reference Appendix Figure 8 The horizontal axis represents the node numbers arranged from the edge to the center according to the physical topology, and the vertical axis represents the current remaining power percentage of each node. In the figure, the remaining power distribution of each node in the scheme of this invention is relatively flat, indicating that energy-aware routing can reduce the forwarding burden of the central node, improve the energy consumption balance of the network, and extend the network lifetime by 50%.
Claims
1. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics, characterized in that: The terminal includes a housing, a perovskite photovoltaic module disposed on the surface of the housing, an energy management module, a built-in energy storage unit, a processing module, and a communication module disposed inside the housing. The perovskite photovoltaic module converts ambient light into DC power and inputs it into the energy management module. The energy management module converts the DC power to charge the built-in energy storage unit, supplies power to the processing module and the communication module, and obtains the remaining power value of the built-in energy storage unit and sends it to the processing module. The processing module reads the remaining battery level and generates or caches the data payload to be sent based on locally collected data, terminal status data, or data to be forwarded received as a relay node. When the remaining battery level meets the communication recovery conditions, the processing module sends a network initialization command to the communication module. The communication module responds to the network initialization command by starting the MESH self-organizing network protocol stack and establishing a multi-hop routing link, and selects the next-hop node to forward the data payload to be sent based on the link quality parameters of adjacent nodes.
2. The self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The energy management module calculates the remaining power value by collecting the instantaneous terminal voltage parameters and charging / discharging current of the built-in energy storage unit, and sends the remaining power value to the processing module. The processing module generates radio frequency configuration instructions based on the remaining battery level, enabling the communication module to set the radio frequency transmission power and the receiver sleep / wake cycle.
3. The self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The perovskite photovoltaic module adopts a flexible thin film stacked structure conformally bonded to the surface of the outer shell, which includes, from the inside out, a flexible polymer substrate, a transparent electrode layer, a bottom charge transport layer, a perovskite light absorption layer with a thickness of 300nm to 500nm, a top charge transport layer, a metal back electrode layer, and a flexible multilayer encapsulation structure. The flexible multilayer packaging structure includes a high barrier layer, an ultraviolet filter layer, a scratch-resistant layer, and a hydrophobic self-cleaning layer, which, together with the terminal shell sealing structure, enable the terminal to achieve an IP68 protection level.
4. The self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The energy management module includes a maximum power point tracking unit and a low-light charging management unit, and is communicatively or electrically connected to an ambient light sensor. The energy management module contains a supercapacitor connected in parallel for low-light pulse energy storage. When the light intensity is between 200 lux and 500 lux, the energy management module enters the pulse charging mode and stores energy in the supercapacitor through the nanowatt-level low-power boost converter circuit. When the voltage at the supercapacitor terminal reaches the upper limit discharge threshold, the energy management module disconnects the input-side charging switch and closes the electronic switch with active current limiting function to pulse charge the built-in energy storage unit.
5. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The processing module is equipped with a multi-level dynamic power consumption state machine, which switches between full-speed mode, energy-saving mode, low-power mode and emergency sleep mode according to the remaining power value and task priority. The processing module presets multiple decreasing power thresholds and hysteresis compensation constants. When the remaining power value drops below the corresponding power threshold, a downgrade switch is triggered. When the remaining power value rises and exceeds the sum of the corresponding power threshold and the hysteresis compensation constant, an upgrade switch is triggered.
6. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 5, characterized in that: When entering low power mode, the processing module calculates the target sleep period based on the remaining power value and the instantaneous power generation of the perovskite photovoltaic module, and sends the target sleep period to the communication module after boundary limiting, so that the communication module updates the receiver sleep-wake cycle. The processing module also reduces the sampling frequency of environmental parameters such as temperature, humidity and water level, and sets the vibration sensor to an interrupt threshold standby mode.
7. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The communication module obtains the received signal strength and signal-to-interference-plus-noise ratio of neighboring nodes through a broadcast discovery process, generates link quality parameters, and receives the current remaining battery power parameters of neighboring nodes. The energy-aware routing unit in the processing module adds a zero-bias constant to the normalized link quality parameter, takes the reciprocal, and multiplies it by the link quality weight coefficient. It also adds a zero-bias constant to the normalized remaining battery power parameter of the adjacent node, takes the reciprocal, and multiplies it by the energy state weight coefficient. The sum is used to obtain the routing cost value, and the adjacent node with the smallest routing cost value is selected as the next hop node. Multiple terminals form a decentralized network through the MESH self-organizing network protocol; When any relay node suffers physical damage, self-destruction due to tamper protection, or shutdown due to power depletion, the adjacent node generates a link error message and broadcasts it within the keep-alive message timeout window, causing the data source node to call the suboptimal path in the local backup routing table to take over the transmission task. If the keep-alive message period is no more than 100ms, the media access control layer single response timeout is no more than 50ms, and there is at least one suboptimal path in the local backup routing table, the network completes the transmission link switch and restores data communication within 0.5s.
8. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The processing module is connected to the sensing module. The sensing module collects environmental sensor data and RFID tag information and sends them to the processing module to generate a data payload to be sent. The environmental sensor data includes temperature and humidity parameters, water level parameters, vibration acceleration parameters, and tilt parameters; The radio frequency tag information includes the personnel location identification code and asset identification code obtained by the RFID card reader unit scanning; The perception module includes an edge computing unit. The edge computing unit demeans and normalizes the vibration acceleration parameters and inputs them into a one-dimensional convolutional neural network model. It outputs the probabilities of normal vibration state, illegal damage state, and geological anomaly state. When the output probability of illegal damage state or geological anomaly state is greater than a preset classification threshold and the same type of output is maintained for at least two consecutive judgment time windows, an anomaly state label is generated.
9. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 1, characterized in that: The processing module is connected to the anti-tamper module, and the anti-tamper module includes an outer shell anti-tamper unit. The tamper-proof housing unit includes a micro switch and a light sensor; The micro switch monitors the opening and closing state of the closed joint of the main shell, and the light sensor is set in the sealed dark room area inside the shell and monitors the external ambient light that enters after the sealed shell breaks. The processing module combines the real-time temperature data collected by the onboard temperature sensor to perform temperature compensation on the substrate voltage constant, and calculates the percentage of light leakage state parameters based on the temperature-compensated substrate voltage constant; the processing module performs a logical OR operation on the microswitch on flag and the light leakage flag of the photosensitive sensor to generate a physical damage alarm load.
10. A self-powered wireless self-organizing network terminal based on perovskite photovoltaics according to claim 9, characterized in that: When the physical damage alarm load is reported, the reporting time reaches the preset timeout period, or the main power supply rail voltage is lower than the preset power supply drop threshold, the processing module enters the safe destruction process. In the safe destruction process, when the main power supply rail voltage is lower than the preset power supply drop threshold, the energy management module switches the system power supply to the backup energy storage network. The processing module writes tamper-erase instructions and pulls down the physical reset control line of the backup domain storage unit to clear the high-density data. When the transient available energy of the backup energy storage network is greater than the physical fuse energy threshold, it performs electronic fuse programming through the on-chip electronic fuse programming controller, or controls the MOS transistor to break down through the chip fuse unit to perform physical fuse operation.
