Intelligent power supply for unattended station

Abstract

The invention relates to the technical field of intelligent power supplies, in particular to an intelligent power supply for an unattended station. Comprising a multi-source power supply scheduling unit; an intelligent energy storage management unit; a double-link communication unit; the double-link communication unit is used for constructing a multi-redundancy communication link of wired, wireless and power line carriers, and the double-link communication unit is based on link quality data, adopts an automatic switching strategy to guarantee data transmission continuity, preprocesses monitoring data by using an edge side data clustering algorithm, and realizes offline data caching and transmission flow optimization. Through a multi-source energy dynamic scheduling strategy, according to photovoltaic real-time output, commercial power time-of-use electricity price and load electricity demand, power supply paths of photovoltaic, commercial power and a storage battery can be dynamically optimized, the consumption proportion of clean energy is improved, the dependency degree of high-cost commercial power is reduced, a complex energy structure of photovoltaic, commercial power and storage battery multi-energy complementation is effectively adapted, and the energy utilization rate is improved. And power supply flexibility is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent power supply, in particular to an unmanned site intelligent power supply. Background Art

[0002] The power supply systems of unmanned sites (such as earthquake monitoring stations and field monitoring equipment) must achieve multi-energy coordinated power supply, precise battery status management, and highly reliable data transmission without human intervention. This presents technical challenges, particularly in adapting to older batteries and ensuring communication stability in complex environments. Existing solutions often suffer from fixed energy scheduling strategies, extensive battery health monitoring, and insufficient communication link redundancy, making them unable to meet the requirements for long-term reliable operation in extreme environments.

[0003] For example, Chinese patent CN202210819069.1 discloses an unattended power supply system that can be charged autonomously. The system comprises a structural module, a lithium-thermal battery module, a lithium-ion battery module and a control module. The control module monitors the electrical properties of the lithium-ion battery through an internal monitoring program, and the lithium-thermal battery module is controlled by a single-chip microcomputer to start and stop charging the ion battery pack. In the power supply system of the present invention, the lithium-ion battery is the main power supply unit and the lithium-thermal battery is its auxiliary power supply. The two are integrated into one through the control module and the structural module, which can realize the long-term working characteristic of the power supply system; the lithium-thermal battery can continuously emit a certain amount of heat when working, and this heat can provide a certain insulation function for the low-temperature operation of the lithium-ion battery, thereby broadening the operating temperature range of the power supply system and further improving the working performance of the power supply system; the power supply system has its own charging function and can be unattended. It can be used in remote areas where there are no charging conditions as an emergency power supply. Another example is Chinese patent CN201720076417.5, which discloses an intelligent power management system for unmanned station seismic instruments. The system includes a microprocessor, a main power supply, a backup power supply, an AD sampling circuit, and a comparison circuit. The microprocessor is connected to the main power supply and the backup power supply via the comparison circuit, and the microprocessor is connected to the seismograph. The backup power supply is connected to the microprocessor via the AD sampling circuit. The microprocessor circuit is connected to a power display unit, the microprocessor circuit is connected to a text message alarm unit, the text message alarm unit circuit is connected to a communication unit, and the communication unit network is connected to a wireless terminal. The operating status of the main power supply and backup power supply of the seismic instrument is detected in real time, and the operating status of the system is notified by text message according to the workflow of the intelligent power management system.

[0004] Although the above-mentioned existing technical solutions all have their design advantages, the existing technical solutions still have the following technical defects, specifically: First, the multi-source energy scheduling is not intelligent enough: Chinese patent CN202210819069.1 adopts "lithium thermal battery-lithium ion battery" fixed logic switching, and Chinese patent CN201720076417.5 relies on the main / backup power supply voltage threshold to trigger switching; both do not integrate multiple energy inputs such as photovoltaic and city power, nor are they associated with external information such as electricity prices and light forecasts. They can only achieve simple power redundancy and cannot dynamically optimize the power supply path according to energy cost and availability, resulting in low energy utilization efficiency (such as photovoltaic power abandonment and dependence on high-priced city power), and it is difficult to adapt to multi-energy complementary scenarios; second, Chinese patent CN202210819069.1 controls charging and discharging by monitoring electrical properties such as voltage and current, and Chinese patent CN201720076417.5 relies on AD Sampling to obtain battery power; such methods only reflect the real-time electrical signals of the battery and cannot distinguish between instantaneous fluctuations and aging attenuation. It is difficult to accurately estimate the SOC (state of charge) (ignoring the effects of discharge rate and internal resistance on power) and cannot predict the SOH (state of health, such as internal resistance growth and capacity decay). Moreover, there is a lack of charge and discharge balancing strategies for different aging batteries. Old battery packs are prone to capacity inconsistency and shortened lifespan. Third, Chinese patent CN202210819069.1 does not have a data communication module, and Chinese patent CN201720076417.5 only transmits status information via a text message link. A single link is prone to failure due to electromagnetic interference or network interruption, and the monitoring data is not locally cached or edge preprocessed (such as data compression and network interruption retransmission). Not only is the risk of data loss during network disconnection high, but it cannot support the continuous data transmission of multiple devices such as sensors and cameras, making it difficult to meet long-term data collection needs. In view of this, we propose an unmanned site intelligent power supply. Summary of the Invention

[0005] The object of the present invention is to provide an intelligent power supply for an unattended site to solve the problems raised in the above background technology.

[0006] To solve the above technical problems, the present invention aims to provide an unmanned site intelligent power supply, comprising:

[0007] A multi-source power supply dispatching unit is used to realize multi-energy hybrid power supply of photovoltaic, mains, and battery. It unifies the power supply system through adaptive rectification technology, and dynamically calls low-cost energy in combination with the electricity price data prediction model to realize intelligent switching of power supply paths.

[0008] An intelligent energy storage management unit, wherein the multi-source power supply scheduling unit and the intelligent energy storage management unit exchange data in real time via the CAN bus. The intelligent energy storage management unit is used for logical management of battery charging and discharging. Based on data collected by multi-source sensors (battery voltage sensor, current sensor, temperature sensor), the intelligent energy storage management unit estimates the state of charge in combination with the Kalman filter algorithm, adopts a dynamic state of charge balancing strategy, and extracts the internal resistance trend in combination with the EIS electrochemical impedance spectroscopy frequency domain analysis technology to achieve battery life prediction and adaptive correction of aging parameters;

[0009] A dual-link communication unit is used to build multiple redundant communication links for wired, wireless, and power line carriers. Based on link quality data, the dual-link communication unit uses an automatic switching strategy to ensure data transmission continuity, pre-processes monitoring data using an edge-side data clustering algorithm, implements disconnected data caching and transmission flow optimization, and enables synchronous power and data transmission through a POE network port;

[0010] The integrated protection and control unit is used to provide sealed protection and intelligent control functions, adopts a gradient heat dissipation structure to optimize equipment thermal management, based on network status monitoring, uses a fault self-healing strategy to restore communication connections, realizes local control through a waterproof operation panel, and integrates multi-sensor fusion technology to realize comprehensive environmental status assessment.

[0011] As a further improvement of this technical solution, the multi-source power supply scheduling unit includes a main control module, wherein:

[0012] The main control module generates power supply path switching instructions based on real-time energy data, supports two groups of battery charging and discharging modes in turn, integrates electricity price prediction models and adaptive rectifier control algorithms, and realizes mains access status monitoring and battery charging and discharging parameter management. The main control module includes an energy scheduling submodule, a battery management submodule, and a mains monitoring submodule, wherein:

[0013] The energy scheduling submodule generates priority scheduling strategies for photovoltaic, mains, and battery storage based on historical electricity price data, weather forecasts, and light intensity data (collected by light intensity sensors) using an LSTM neural network model. It also uses a transfer learning algorithm to adapt to peak and valley electricity pricing rules in different regional power grids to minimize energy costs.

[0014] The battery management submodule uses a bidirectional DC-DC converter to achieve independent charge and discharge control of the two battery groups, automatically switching the charging path based on the state of charge threshold. It also integrates a battery balancing circuit and uses dynamic impedance matching technology to achieve capacity consistency management of the two battery groups.

[0015] The mains monitoring submodule uses a voltage transformer and a frequency detection chip to collect mains voltage and frequency data in real time. When the voltage is detected to exceed the preset threshold range (such as <176V or >264V), an audible and visual alarm is triggered and an alarm log is generated.

[0016] As a further improvement of this technical solution, the multi-source power supply scheduling unit further includes a communication module, wherein:

[0017] The communication module is connected to the photovoltaic inverter, battery pack and mains monitoring submodule (for collecting mains voltage, frequency and access status) through multiple electrically isolated serial ports, and adopts a dual-link communication architecture to realize data interaction; the communication module includes a data cache submodule, wherein:

[0018] The data cache submodule is used to store energy data when the network is disconnected and automatically resume transmitting energy data after recovery.

[0019] As a further improvement of this technical solution, the intelligent energy storage management unit includes a multi-source data perception module and an EIS detection module, wherein:

[0020] The multi-source data perception module includes a data acquisition submodule and a signal conditioning submodule, wherein:

[0021] The data acquisition submodule integrates a voltage sensor, a current sensor, and a temperature sensor, which are used to collect the battery terminal voltage, charge and discharge current, and surface temperature in real time. It also uses a high-precision analog-to-digital conversion chip to achieve three-channel synchronous sampling, and the sampling frequency meets the requirements of real-time battery status monitoring.

[0022] The signal conditioning submodule is used to filter, reduce noise and adapt the level of the sensor output signal, eliminate signal interference and convert it into a standardized digital signal; it also supports sensor fault self-detection and triggers a hardware interrupt when any sensor data is abnormal and reports it to the main control module;

[0023] The EIS detection module includes a sweep frequency excitation submodule, an impedance decoupling submodule, and an aging parameter generation submodule, wherein:

[0024] The frequency sweeping excitation submodule is used to inject a sinusoidal excitation signal of a preset frequency band into the battery pack to stimulate the electrochemical impedance response of the battery; the signal amplitude is dynamically adjusted according to the current state of the battery to ensure that the excitation intensity is adapted to different states of charge;

[0025] The impedance decoupling submodule is used to synchronously collect the voltage and current responses corresponding to the excitation signal, separate and calculate the battery impedance spectrum through the frequency domain analysis algorithm; and extract the internal resistance value of the characteristic frequency point to form an impedance characteristic vector reflecting the internal state of the battery;

[0026] The aging parameter generation submodule establishes an internal resistance-aging model based on the impedance characteristic vector and temperature data, and outputs battery health status correction parameters (correction parameters include internal resistance growth rate, capacity attenuation coefficient, etc.) for dynamically adjusting the energy storage management strategy.

[0027] As a further improvement of the present technical solution, the intelligent energy storage management unit further includes a dynamic balancing control module, which is used to achieve accurate estimation of battery pack status, environmental impact compensation and energy balancing control. The dynamic balancing control module includes a state of charge estimation submodule, a temperature compensation submodule and an active balancing execution submodule, wherein:

[0028] The state-of-charge estimation submodule integrates voltage, current, and temperature data from multiple sensor modules and iteratively estimates the battery pack's state of charge (SOC) using a Kalman filter algorithm. It also uses the internal resistance trend output by the EIS detection module to provide real-time compensation for the SOC estimation result, reducing long-term estimation errors. It also uses a multi-source data fusion architecture to achieve time synchronization and error calibration of data from different sensors, and outputs a high-precision SOC value to the active balancing execution submodule as a basis for decision-making on the balancing strategy.

[0029] The temperature compensation submodule is used to construct a nonlinear correction model for temperature and state of charge, and adjust the state of charge estimation results based on real-time temperature data to eliminate the impact of temperature drift. The nonlinear correction model parameters are generated through self-learning of historical battery data to adapt to state estimation under different ambient temperatures. The internal resistance value detected by EIS is temperature compensated to establish a temperature-internal resistance mapping relationship, thereby improving the accuracy of aging parameters.

[0030] The active balancing execution submodule uses a bidirectional DC-DC converter to achieve bidirectional energy transfer between battery packs, and dynamically adjusts the balancing current (range 0.5C–2C) based on the difference in the state of charge and internal resistance of the battery packs to achieve adaptation of batteries with different capacities and aging levels. It also constructs a multi-objective optimization function to formulate balancing priorities based on the state of charge deviation, internal resistance difference, and temperature distribution, giving priority to battery packs with significant state differences. It also introduces a time-sharing balancing scheduling mechanism, combined with load forecasting to avoid peak power consumption, reduce the interference of the balancing process on the system power supply, and ensure operational stability.

[0031] As a further improvement of the present technical solution, the dual-link communication unit includes a link management module, and the link management module includes a link monitoring submodule and a switching control submodule, wherein:

[0032] The link monitoring submodule is used to collect signal strength, bit error rate, transmission delay and packet loss rate data of wired links, wireless links and power line carrier links;

[0033] The handover control submodule has a built-in multi-dimensional link quality evaluation algorithm and calculates the link comprehensive quality score Q using the following formula:

[0034] Q = w1 × (1-SNR / SNR max )+w2×BER+w3×(RTT / RTT max )+w4×

[0035] PLR;

[0036] Among them, SNR is the signal-to-noise ratio, SNR max is the maximum acceptable signal-to-noise ratio, BER is the bit error rate, RTT is the round-trip delay, and RTT is the max is the maximum acceptable delay, PLR is the packet loss rate, w1, w2, w3, w4 are weight coefficients and satisfy w1+w2+w3+w4=1;

[0037] The handover control submodule performs trend analysis on the comprehensive link quality scores of M consecutive sampling periods based on a sliding time window algorithm, and calculates the quality change trend factor ΔQ using the following formula:

[0038] ΔQ=(Q m -Q1) / (M×T);

[0039] Among them, Q1 is the link quality score at the start of the window, Q m is the link quality score at the end of the window, T is the sampling period; when the comprehensive quality score Q of a link is lower than the preset threshold Q th When the trend factor ΔQ is less than 0, the link pre-switching mechanism is triggered and the connection establishment process of the backup link is started.

[0040] As a further improvement of the present technical solution, the dual-link communication unit further includes a data processing module and a POE transmission module, wherein:

[0041] The data processing module includes an edge computing submodule and a cache scheduling submodule, wherein:

[0042] The edge computing submodule integrates a spatiotemporal correlation feature extraction algorithm, which calculates the monitoring data point x using the following formula: i The spatiotemporal correlation ST(x i ,t):

[0043] ST(x i ,t)=α×Sim s (x i )+β×Sim t (x i ,t)+γ×Corr(x i ,t);

[0044] Among them, Sim s (x i ) is the spatial similarity function, which calculates the spatial similarity between the current data point and the historical cluster center through the Euclidean distance; Sim t (x i ,t) is the time similarity function, which calculates the similarity of data change trends within the time window through the exponential decay model; Corr(x i ,t) is the spatiotemporal correlation function, which calculates the joint probability distribution of data points in the spatiotemporal dimension; α, β, γ are weight coefficients and α+β+γ=1;

[0045] The cache scheduling submodule builds the following three-level cache queue based on the spatiotemporal correlation ST value:

[0046] When ST≥ST high When , it is stored in the high priority queue;

[0047] When ST low ≤ST <ST high When , store it in the medium priority queue;

[0048] When ST <ST low When , it is stored in the low priority queue;

[0049] Among them, ST high Indicates high correlation threshold, ST low indicates a low association threshold;

[0050] The POE transmission module includes a power line carrier communication submodule and an Ethernet power supply submodule, wherein:

[0051] The power line carrier communication submodule has a built-in OFDM modulator and uses OFDM modulation technology to divide the power line channel into N subcarriers and dynamically allocate the number of bits on the subcarriers through an adaptive bit loading algorithm. i :

[0052]

[0053] Among them, SNR i is the signal-to-noise ratio of the i-th subcarrier, SNR0 is the reference signal-to-noise ratio;

[0054] The OFDM modulator is equipped with an impedance matching interface for connecting to an external impedance matching network to achieve dynamic impedance matching of the power line channel;

[0055] The Ethernet power supply module integrates a power negotiation protocol stack, which dynamically adjusts the output voltage V after parsing the classification request of the powered device. out and current I out , so that the total output power satisfies:

[0056]

[0057] Among them, P base is the basic power supply power of the equipment, n is the number of powered devices, P i is the power requirement of the i-th powered device, w i It is a dynamic weight factor (assigned according to device priority, service type and link status).

[0058] Furthermore, the three-level queue storage specifically includes:

[0059] High priority queue (ST ≥ 0.8): stores real-time alarm data and critical battery status;

[0060] Medium priority queue (0.5≤ST<0.8): stores periodic monitoring data (such as environmental parameters every 10 minutes);

[0061] Low priority queue (ST<0.5): stores unstructured data (such as camera video stream fragments);

[0062] In addition, when the network is disconnected, only high-priority data (about 50KB per hour) is cached to avoid storage overflow; after the network is restored, the data is resumed in the order of "high-medium-low", and the resumption rate is limited to 30% of the link bandwidth (to avoid impacting the main business).

[0063] As a further improvement of this technical solution, the power line carrier communication submodule integrates an adaptive equalization circuit and uses the LMS algorithm to calculate the equalizer tap coefficient h(n). The calculation formula is as follows:

[0064] h(n)=h(n-1)+μ·e(n)·x(n);

[0065] Where μ is the step size factor, e(n) is the error signal, and x(n) is the input signal. The error signal e(n) satisfies the following equation: e(n) = d(n) - y(n), where d(n) is the desired response signal and y(n) is the equalizer output signal.

[0066] The characteristic impedance Z of the impedance matching network c satisfy:

[0067]

[0068] Where Z0 is the inherent characteristic impedance of the power line (obtained through offline calibration); Z L (n) is the real-time load impedance, measured by injecting a 20kHz-500kHz sweep signal. The dynamic correction factor γ(n) is calculated based on the channel estimation result: Wherein, H(n) is the channel frequency response estimation value at the nth moment.

[0069] As a further improvement of this technical solution, the integrated protection and control unit includes a gradient heat dissipation module, a fault self-healing module, a waterproof interaction module and a multi-sensor fusion module, wherein:

[0070] The gradient heat dissipation module uses a layered equidistant fin array to optimize the heat flow path. The spacing between adjacent fin groups starts from the reference value and expands layer by layer with a fixed tolerance. It has a built-in serpentine fluid channel, and the channel size can be linearly changed with the number of fin layers.

[0071] The fault self-healing module identifies communication anomalies by real-time monitoring of link status and executes a three-level recovery process of "diagnosis-switching-calibration" to achieve autonomous detection and recovery of communication faults.

[0072] The waterproof interactive module uses a multi-layer sealing structure to block liquid intrusion. The touch area is divided into multiple levels of operation permissions to achieve access control. When operating in a wet state, the sensing mode is automatically switched to ensure the reliability of human-computer interaction.

[0073] The multi-sensor fusion module is used to build a multi-dimensional environmental monitoring network to collect heterogeneous data, identify environmental anomalies through joint judgment, trigger graded early warning responses and realize comprehensive assessment of environmental status.

[0074] As a further improvement of this technical solution, the fault self-healing module includes a link monitoring subunit, a diagnosis and decision subunit, a redundancy switching subunit and a strategy optimization subunit, wherein:

[0075] The link monitoring subunit deploys a multi-dimensional probe matrix to collect real-time operating parameters of the physical layer (signal strength, bit error rate), link layer (packet loss rate, throughput), and application layer (response time) to build a full link status profile.

[0076] The diagnosis decision subunit uses a fuzzy logic algorithm to calculate the fault confidence level based on the fault signature library and inference engine. When the confidence level exceeds the preset threshold, the corresponding recovery strategy is triggered.

[0077] The redundant switching subunit is used to maintain dynamic link priorities and comprehensively evaluate the multi-dimensional indicators of alternative links (such as bandwidth, latency, stability, etc.) through the hierarchical analysis method to achieve rapid switching of the optimal link;

[0078] The strategy optimization subunit is used to build a fault case knowledge base and adopts a reinforcement learning algorithm to dynamically adjust the decision weights according to historical recovery effects to form a strategy iterative optimization closed loop.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. This invention uses a multi-source dynamic energy scheduling strategy to dynamically optimize the power supply paths of photovoltaic, mains, and batteries based on real-time photovoltaic output, mains time-of-use electricity prices, and load power demand. This increases the proportion of clean energy consumption, reduces dependence on high-cost mains electricity, effectively adapts to the complex energy structure of "photovoltaic + mains + battery" multi-energy complementarity, and enhances power supply flexibility.

[0081] 2. Based on the integrated analysis of electrical performance monitoring and electrochemical impedance characteristics, this invention can accurately identify the differences between instantaneous electrical signal fluctuations and long-term aging attenuation in batteries, achieving precise estimation of state of charge and predicting health trends. Combined with dynamic balancing control strategies for batteries of varying aging levels, this technology can alleviate capacity inconsistencies in older battery packs and extend battery cycle life.

[0082] 3. This invention adopts a multi-redundant communication link architecture of wired, wireless, and power line carriers, combined with edge-side data preprocessing and network disconnection caching mechanisms. When a single link encounters electromagnetic interference or physical failure, it can automatically switch to a backup link, reducing the risk of data transmission interruption and loss. It continuously supports stable data transmission from multiple devices such as sensors and cameras, and adapts to the communication reliability requirements in complex electromagnetic environments.

[0083] 4. Through the coordinated operation of energy scheduling, battery management, and communication redundancy modules, this invention further enhances the overall autonomous operation capability of the unmanned site power system under complex working conditions, reduces the need for manual intervention, and ensures long-term stable operation from energy supply, energy storage health, to data interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a system framework diagram of the present invention;

[0085] The meaning of each number in the figure is:

[0086] 100, multi-source power supply dispatching unit; 110, main control module; 111, energy dispatching submodule; 112, battery management submodule; 113, mains power monitoring submodule; 120, communication module; 121, data cache submodule;

[0087] 200, intelligent energy storage management unit; 210, multi-source data perception module; 211, data acquisition submodule; 212, signal conditioning submodule; 220, EIS detection module; 221, sweep frequency excitation submodule; 222, impedance decoupling submodule; 223, aging parameter generation submodule; 230, dynamic balancing control module; 231, state of charge estimation submodule; 232, temperature compensation submodule; 233, active balancing execution submodule;

[0088] 300, dual-link communication unit; 310, link management module; 311, link monitoring submodule; 312, switching control submodule; 320, data processing module; 321, edge computing submodule; 322, cache scheduling submodule; 330, POE transmission module; 331, power line carrier communication submodule; 332, Ethernet power supply submodule;

[0089] 400, integrated protection and control unit; 410, gradient heat dissipation module; 420, fault self-healing module; 421, link monitoring subunit; 422, diagnosis and decision subunit; 423, redundant switching subunit; 424, strategy optimization subunit; 430, waterproof interaction module; 440, multi-sensor fusion module. DETAILED DESCRIPTION

[0090] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0091] like Figure 1 As shown, this embodiment provides an unattended site intelligent power supply, including:

[0092] The multi-source power supply dispatching unit 100 is used to realize multi-energy hybrid power supply of photovoltaic, mains, and batteries. It unifies the power supply mode through adaptive rectification technology, and dynamically calls low-cost energy in combination with the electricity price data prediction model to realize intelligent switching of power supply paths. The multi-source power supply dispatching unit 100 is used to realize multi-energy hybrid power supply of photovoltaic, mains, and batteries. It unifies the power supply mode through three-phase full-bridge rectification topology (adapting to 380V / 220V mains and using PWM pulse width modulation technology to dynamically adjust the output voltage), and dynamically calls low-cost energy in combination with the LSTM neural network electricity price prediction model to realize intelligent switching of power supply paths.

[0093] In this embodiment, the multi-source power supply scheduling unit 100 includes a main control module 110 (which may use an STM32H743ZI high-performance MCU with a main frequency of 480 MHz to meet the requirements of multi-task parallel computing), wherein:

[0094] The main control module 110 generates power supply path switching instructions based on real-time energy data, supports two battery groups to alternately charge and discharge, integrates electricity price prediction models and adaptive rectifier control algorithms, and implements mains access status monitoring and battery charge and discharge parameter management. The main control module 110 includes an energy scheduling submodule 111, a battery management submodule 112, and a mains monitoring submodule 113, wherein:

[0095] The energy scheduling submodule 111 generates priority scheduling strategies for photovoltaic, mains, and battery power based on historical electricity price data, weather forecasts, and light intensity data (which can be collected by light intensity sensors) from the LSTM neural network model. It also uses a transfer learning algorithm to adapt to peak and valley electricity price rules in different regional power grids to minimize energy costs.

[0096] The battery management submodule 112 uses a bidirectional DC-DC converter to achieve independent charge and discharge control of the two battery groups, automatically switching the charging path based on the state of charge threshold; integrates a battery balancing circuit, and uses dynamic impedance matching technology to achieve capacity consistency management of the two battery groups;

[0097] The mains monitoring submodule 113 uses a voltage transformer and a frequency detection chip (ZMPT107 voltage transformer and PIC16F877A frequency detection chip can be used) to collect mains voltage and frequency data in real time; when it is detected that the voltage exceeds the preset threshold range, such as <176V or >264V, an audible and visual alarm is triggered and an alarm log is generated.

[0098] In this embodiment, the multi-source power supply scheduling unit 100 further includes a communication module 120, wherein:

[0099] The communication module 120 is connected to the photovoltaic inverter, battery pack, and mains monitoring submodule 113 (for collecting mains voltage, frequency, and access status) through multiple electrically isolated serial ports, and adopts a dual-link communication architecture to achieve data interaction; the communication module 120 includes a data cache submodule 121, wherein:

[0100] The data cache submodule 121 is used to store energy data when the network is disconnected and automatically resume transmitting the energy data after the network is restored.

[0101] As a further explanation of this embodiment, the energy scheduling execution process in this embodiment is as follows:

[0102] First, data collection: synchronize the PV inverter power, real-time utility electricity price (using RS485 to read meter data), and battery SOC (feedback from the battery management submodule) every 5 minutes;

[0103] Then, a prediction calculation is performed: the data is input into the LSTM model to predict the photovoltaic power generation power and the peak-valley price difference of the mains electricity in the next hour;

[0104] Then, sort by cost priority: PV (priority consumption) → utility power (off-peak period, 0:00-8:00) → battery (discharge) → utility power (peak period, 8:00-22:00) → battery (emergency backup);

[0105] Finally, the output instruction is: that is, the charge and discharge instruction (such as "photovoltaic → battery group A charging, current 10A") is sent to the battery management submodule 112 through the CAN bus, and the access status detection instruction is sent to the mains monitoring submodule 113.

[0106] As a further explanation of this embodiment, the capacity consistency management mechanism of the two battery groups in this embodiment is as follows:

[0107] First, the internal resistance of the two battery groups is tested every hour. When the internal resistance difference is greater than 10%, active balancing is initiated: the energy of the high-capacity battery is transferred to the low-capacity battery through the DC-DC converter, and the balancing current is set to 5A until the internal resistance difference is less than 5%;

[0108] Second, the charge and discharge switching logic is as follows:

[0109] SOC≤20%: PV charging is prioritized (PV power ≥50W), otherwise it switches to charging at the grid price of mains electricity;

[0110] SOC ≥ 90%: Discharge to the load first, and excess energy can be fed back to the grid (if the grid supports bidirectional metering).

[0111] As a further explanation of this embodiment, the alarm log in this embodiment specifically includes:

[0112] Storage medium: On-chip EEPROM (AT24C128, capacity 128KB) is used to record timestamps (accurate to the moment), voltage values, and frequency values;

[0113] Storage rules: Store up to 100 logs, with new data automatically overwriting the earliest record, and support export analysis via serial port;

[0114] Alarm log management, through standardized data recording and storage mechanisms, provides unattended systems with "black box"-like fault tracing capabilities. It is a crucial component in ensuring power system reliability and maintainability. Its core value lies in leveraging historical data to rapidly locate abnormal events and continuously optimize system operating status.

[0115] As a further explanation of this embodiment, the data cache submodule 121 in this embodiment can use W25Q128SPINORFlash (capacity 128MB) as the storage medium; when the main link signal strength is <-90dBm or three consecutive heartbeat packets are lost, it automatically switches to the LoRa link and starts data caching at the same time; after connecting to the network (the main link signal is ≥-85dBm and stable for 30 seconds), historical cache data is uploaded at a rate of 10 items / second, and then real-time data is synchronized.

[0116] It should be added that the system in this embodiment relies on the city power monitoring, main control, energy dispatch and the system's built-in communication module to build a collaborative mechanism, covering abnormal response and data interaction. The specific contents are as follows:

[0117] Voltage abnormality linkage logic:

[0118] When the mains monitoring submodule 113 detects a voltage anomaly, it triggers the main control module through an interrupt signal; the energy scheduling submodule immediately switches to battery power supply mode (prioritizing battery packs with higher SOC (state of charge)), and the communication module pushes an alarm message to the remote platform.

[0119] Data interaction rules:

[0120] The battery management submodule 112 feeds back the battery SOC and internal resistance data to the energy scheduling submodule 111 every 2 minutes;

[0121] The mains monitoring submodule 113 uploads voltage and frequency data every minute;

[0122] The communication module synchronizes energy dispatch instructions to the photovoltaic inverter and battery pack in real time.

[0123] In addition, this embodiment unifies the power supply system through multi-energy hybrid scheduling of photovoltaic, mains, and battery power, and adaptive rectification technology. Combined with algorithms such as LSTM electricity price prediction and dynamic impedance matching, a closed-loop logic of "data acquisition-algorithm decision-making-control execution-feedback optimization" is constructed to achieve intelligent switching of power supply paths. To ensure stable operation of the algorithm, a hardware-software collaborative optimization solution can be adopted. The specific solution is as follows:

[0124] Algorithm hardware carrier: The LSTM model is deployed on the STM32H743ZI MCU, using the DSP instruction set to accelerate matrix operations. Internal resistance detection relies on the AD9833 waveform generator and the ADS1256 ADC for high-precision sampling, and is implemented with the CMSIS-DSP library to implement fast Fourier transform. Frequency detection reuses the MCU ADC, and zero-crossing events are captured through timer interrupts.

[0125] Parameter debugging interface: supports dual-mode debugging via local serial port (baud rate 115200) and remote 4G link: local commands can be used to adjust the model learning rate or modify the equalization trigger threshold; remote debugging receives cloud commands through an encrypted channel.

[0126] Abnormal degradation mechanism: If the LSTM model crashes, it automatically switches to the "fixed threshold scheduling strategy" (photovoltaic priority - grid price - battery discharge); if the internal resistance detection circuit fails, it downgrades to the "voltage difference balancing mode" (capacity estimation by voltage difference), which is still better than the non-balancing solution.

[0127] Intelligent energy storage management unit 200, the intelligent energy storage management unit 200 is used for the logical management of battery charging and discharging. Based on the data collected by multi-source sensors (battery voltage sensor, current sensor, temperature sensor), combined with the Kalman filter algorithm, it estimates the state of charge, and adopts a dynamic state of charge balancing strategy. Combined with EIS electrochemical impedance spectroscopy frequency domain analysis technology, it extracts the internal resistance trend to achieve battery life prediction and adaptive correction of aging parameters; the intelligent energy storage management unit 200 is used for the logical management of battery charging and discharging. Based on the real-time data of the multi-source sensor group (voltage, current, temperature) and EIS electrochemical impedance spectroscopy analysis, it constructs a "perception-analysis-control-correction" closed loop to achieve accurate state of charge (SOC) estimation, battery aging prediction and dynamic balancing control.

[0128] In this embodiment, the intelligent energy storage management unit 200 includes a multi-source data perception module 210 and an EIS detection module 220, wherein:

[0129] The multi-source data perception module 210 includes a data acquisition submodule 211 and a signal conditioning submodule 212, wherein:

[0130] The data acquisition submodule 211 integrates a voltage sensor, a current sensor, and a temperature sensor, which are used to respectively collect the battery terminal voltage, charge and discharge current, and surface temperature in real time. It also uses a high-precision analog-to-digital conversion chip to achieve three-channel synchronous sampling, and the sampling frequency meets the requirements of real-time battery status monitoring.

[0131] The signal conditioning submodule 212 is used to filter, reduce noise, and adapt the level of the sensor output signal to eliminate signal interference and convert it into a standardized digital signal. It also supports sensor fault self-detection and triggers a hardware interrupt when any sensor data is abnormal and reports it to the main control module 110.

[0132] The EIS detection module 220 includes a sweep frequency excitation submodule 221, an impedance decoupling submodule 222, and an aging parameter generation submodule 223, wherein:

[0133] The frequency sweeping excitation submodule 221 is used to inject a sinusoidal excitation signal of a preset frequency band into the battery pack to stimulate the electrochemical impedance response of the battery; the signal amplitude is dynamically adjusted according to the current state of the battery to ensure that the excitation intensity is adapted to different states of charge;

[0134] The impedance decoupling submodule 222 is used to synchronously collect the voltage and current responses corresponding to the excitation signal, separate and calculate the battery impedance spectrum through the frequency domain analysis algorithm; and extract the internal resistance value of the characteristic frequency point to form an impedance characteristic vector reflecting the internal state of the battery;

[0135] The aging parameter generation submodule 223 establishes an internal resistance-aging model based on the impedance characteristic vector and temperature data, and outputs battery health status correction parameters (correction parameters include internal resistance growth rate, capacity attenuation coefficient, etc.) for dynamically adjusting the energy storage management strategy.

[0136] In this embodiment, the intelligent energy storage management unit 200 further includes a dynamic balancing control module 230, which is used to achieve accurate estimation of battery pack status, environmental impact compensation, and energy balancing control. The dynamic balancing control module 230 includes a state of charge estimation submodule 231, a temperature compensation submodule 232, and an active balancing execution submodule 233, wherein:

[0137] The state-of-charge estimation submodule 231 integrates the voltage, current, and temperature data from multiple sensor modules and iteratively estimates the battery pack's state of charge (SOC) using a Kalman filter algorithm. It uses the internal resistance trend output by the EIS detection module 220 to provide real-time compensation for the SOC estimation result, reducing long-term estimation errors. It also uses a multi-source data fusion architecture to achieve time synchronization and error calibration of data from different sensors, and outputs a high-precision SOC value to the active balancing execution submodule 233 as a basis for decision-making on the balancing strategy.

[0138] The temperature compensation submodule 232 is used to construct a nonlinear correction model for temperature and state of charge, and adjust the state of charge estimation results based on real-time temperature data to eliminate the impact of temperature drift; the nonlinear correction model parameters are generated through self-learning of battery historical data to adapt to state estimation under different ambient temperatures; and the internal resistance value detected by EIS is temperature compensated to establish a temperature-internal resistance mapping relationship to improve the accuracy of aging parameters.

[0139] The active balancing execution submodule 233 uses a bidirectional DC-DC converter to realize bidirectional energy transfer between battery packs, and dynamically adjusts the balancing current range of 0.5C-2C according to the difference in charge state and internal resistance of the battery packs, so as to adapt to batteries with different capacities and aging degrees; and constructs a multi-objective optimization function to formulate balancing priorities based on the charge state deviation, internal resistance difference and temperature distribution, giving priority to battery packs with significant state differences; and introduces a time-sharing balancing scheduling mechanism, combined with load forecasting to avoid peak power consumption, reduce the interference of the balancing process on the system power supply, and ensure operational stability.

[0140] As a further illustration of this embodiment, in the data acquisition submodule 211 of this embodiment, the voltage sensor can be connected in parallel to the two ends of the battery pack using INA219 (I2C interface) to realize battery terminal voltage monitoring; the current sensor can be connected in series to the charge and discharge circuit through ACS712 (Hall effect) to collect the charge and discharge current; the temperature sensor can be attached to the surface of the battery to monitor the surface temperature in real time; the analog-to-digital conversion can realize three-channel synchronous sampling through ADS1115 to meet the real-time monitoring requirements of the battery status; and the above-mentioned multiple hardware can use the DMA controller of the STM32H7MCU to realize data synchronous acquisition to ensure the temporal and spatial consistency of the voltage, current and temperature data.

[0141] As a further illustration of this embodiment, in the signal conditioning submodule 212 of this embodiment, filtering processing can use a second-order Butterworth low-pass filter to suppress charging and discharging pulse interference, and the hardware implementation is constructed using an operational amplifier; level adaptation can use an RC voltage divider circuit to condition the sensor signal to 0-3.3V to match the ADC input range; and sensor data rationality checks (such as voltage, current, and temperature threshold detection) can be performed regularly. If the data exceeds the limit, a hardware interrupt is triggered and reported to the main control module 110, and a fault code is recorded in the EEPROM;

[0142] As a further illustration of this embodiment, in the swept frequency excitation submodule 221 of the EIS detection module 220 of this embodiment, the sinusoidal excitation signal can be generated by using an AD9833 waveform generator to generate a 10 MHz to 10 kHz sinusoidal excitation signal, covering the electrochemical characteristic frequency range of the battery; the signal amplitude can be dynamically adjusted according to the current SOC of the battery: a lower amplitude is used in high or low state of charge, and a higher amplitude is used in normal state, to avoid overexcitation and ensure the signal-to-noise ratio of the impedance response;

[0143] In the impedance decoupling submodule 222 of the EIS detection module 220 of this embodiment, synchronous acquisition can use a synchronous sampling ADC to synchronously acquire the voltage and current responses of the excitation signal, and extract the fundamental component through a digital phase-locked loop (DPLL); frequency domain analysis can use a fast Fourier transform (FFT) to calculate the impedance spectrum, extract the internal resistance values ​​at characteristic frequency points in the low-frequency region and the high-frequency region, and form an impedance characteristic vector;

[0144] In the aging parameter generation submodule 223 of the EIS detection module 220 of this embodiment, the basic architecture of the internal resistance-aging model is as follows: the internal resistance-aging model is established based on least squares fitting, and the internal resistance growth rate and the ambient temperature are combined to output the capacity attenuation coefficient and the internal resistance growth rate for dynamic adjustment of the charge and discharge strategy; at the same time, the internal resistance value detected by EIS can be temperature-corrected to establish a temperature-internal resistance mapping relationship to improve the accuracy of the aging parameters.

[0145] As a further explanation of this embodiment, in the state of charge estimation submodule 231 of the dynamic balancing control module 230 of this embodiment, based on the fused multi-source sensor data, the SOC is iteratively estimated through Kalman filtering, and the internal resistance trend detected by EIS is introduced to compensate the estimation result in real time to reduce long-term errors; at the same time, a multi-source data fusion architecture can be used to realize the time synchronization and error calibration of the sensor data, and output the state of charge value to the active balancing execution submodule 233.

[0146] In the temperature compensation submodule 232 of the dynamic balancing control module 230 of this embodiment, a nonlinear correction model of temperature and state of charge is constructed and the SOC estimation result is adjusted according to the real-time temperature, thereby eliminating the influence of temperature drift; at the same time, the nonlinear correction model parameters are generated through self-learning of battery historical data to adapt to state estimation under different ambient temperatures.

[0147] In the active balancing execution submodule 233 of the dynamic balancing control module 230 of this embodiment, a bidirectional DC-DC converter can be used to realize bidirectional energy transfer between battery packs, and support dynamic adjustment of the balancing current according to the difference in charge state and internal resistance of the battery packs; and a multi-objective optimization function can be constructed to formulate the balancing priority based on the charge state deviation, internal resistance difference and temperature distribution, and introduce a time-sharing balancing scheduling mechanism to avoid peak power consumption and reduce interference with the system power supply.

[0148] It should be added that the multi-module collaborative control process in the intelligent energy storage management unit 200 in this embodiment is as follows:

[0149] First, the main control module 110 sends the initial battery parameters to the intelligent energy storage management unit 200, and the EIS detection module 220 performs the first full-frequency sweep to establish the battery initial impedance characteristic library.

[0150] Next, sensor data is collected in real time to update SOC estimation and temperature compensation, and EIS fast frequency sweep is triggered regularly to monitor the trend of internal resistance changes. When the state of charge or internal resistance difference exceeds the threshold, active balancing is automatically started.

[0151] Finally, full-band EIS scans are performed regularly to update the aging parameter model. When the battery health status (SOH) falls below the preset threshold, warning information is reported through the communication module.

[0152] Furthermore, the multi-source power supply dispatching unit 100 and the intelligent energy storage management unit 200 implement bidirectional communication via a real-time data bus. The interaction process and mechanism between the two are as follows:

[0153] (1) Generation of charging and discharging strategies driven by electricity prices:

[0154] The triggering conditions are: the energy scheduling submodule 111 completes the electricity price forecast and photovoltaic power forecast for the next hour based on the LSTM neural network; the intelligent energy storage management unit 200 provides real-time feedback on the battery pack state of charge (SOC), state of health (SOH) and charge and discharge power limits;

[0155] The interaction logic is:

[0156] First, the energy scheduling submodule 111 generates a power supply path priority function based on the electricity price period (peak / valley / flat), photovoltaic available power, and battery SOC:

[0157] P priority =f(electricity price, PV power, SOC, SOH), where electricity price accounts for 40%, PV power accounts for 30%, SOC accounts for 20%, and SOH accounts for 10%;

[0158] Subsequently, the main control module 110 sends the charge and discharge control parameters to the intelligent energy storage management unit 200 via the CAN bus, including:

[0159] Target SOC interval (e.g., [30%, 80%]);

[0160] Charge and discharge current threshold (related to battery internal resistance and temperature);

[0161] Energy type priority identification (e.g., "Priority for grid electricity valley prices" or "Photovoltaic direct supply");

[0162] When SOH is less than 70%, the charging current is automatically limited to ≤0.8C and the discharging current is ≤1.2C;

[0163] When the battery temperature is greater than 50°C or less than 0°C, fast charging (current ≤ 0.5C) is prohibited.

[0164] (2) Strategy modification under battery status constraints:

[0165] The trigger condition is that the intelligent energy storage management unit 200 detects:

[0166] The SOC difference between battery packs is greater than 10%;

[0167] The internal resistance growth rate of a single battery pack is >5% / month;

[0168] The surface temperature exceeds the range of [-25℃, 65℃];

[0169] The interaction logic is:

[0170] First, the dynamic balancing control module 230 sends the status abnormality code and real-time data (such as SOC difference, internal resistance increment, and temperature value) to the main control module 110 through the RS485 interface;

[0171] Subsequently, when the SOC is unbalanced: the main control module 110 suspends the current charging and discharging tasks, prioritizes active balancing control (bidirectional DC-DC converter regulates energy transfer), and switches to photovoltaic independent power supply mode;

[0172] When the internal resistance exceeds the limit: the faulty battery pack is marked as "standby mode", allowing only low-power operation (current ≤ 0.3C), and the charging and discharging priority of the healthy battery pack is increased;

[0173] When the temperature is abnormal: the intelligent energy storage management unit 200 automatically adjusts the Kalman filter algorithm parameters to compensate for the temperature drift, and the energy scheduling submodule 111 adjusts the power supply path to avoid high-load operation under extreme temperatures;

[0174] (3) Coordinated control of dynamic balance and energy dispatch:

[0175] The trigger conditions are:

[0176] The energy scheduling submodule 111 determines that the current period is a “low energy cost period”, such as a city level period or a photovoltaic power surplus period;

[0177] The intelligent energy storage management unit 200 identifies that the battery pack needs to be managed for capacity consistency through the EIS detection module 220;

[0178] The interaction logic is:

[0179] The energy scheduling submodule 111 calculates the available balanced power based on the electricity price, photovoltaic power and load demand: in, is the photovoltaic surplus power, P is the available mains power during the valley price period, load is the real-time load power;

[0180] Active balancing execution submodule 233 balance Dynamically adjust the balancing current (0.5C–2C) and feed real-time energy consumption data to the main control module 110; the energy scheduling submodule 111 simultaneously updates the power supply path priority to ensure that the balancing process does not affect the power supply of the core load;

[0181] (4) Cross-module linkage response under abnormal conditions:

[0182] The triggering conditions are: the mains monitoring submodule 113 detects a voltage anomaly (<176V or >264V);

[0183] The intelligent energy storage management unit 200 determines that the battery pack SOH is less than 50% (end of life);

[0184] The interaction logic includes:

[0185] When the mains power is abnormal, the following response is made:

[0186] The main control module 110 immediately switches to the full battery pack discharge mode and sends a current limiting instruction (current ≤ 1.5C) to the intelligent energy storage management unit 200;

[0187] If the battery SOC is ≤ 25% and the mains power anomaly lasts for more than 30 minutes, a hierarchical power-off strategy will be initiated (prioritizing power cutoff of non-core devices).

[0188] When the battery reaches end of life:

[0189] The intelligent energy storage management unit 200 reports a "battery replacement warning";

[0190] Energy scheduling submodule 111 adjustment strategy:

[0191] Battery charging and discharging is prohibited (only used as a backup power source);

[0192] Force reliance on mains electricity supply and add an "emergency procurement cost" parameter to the electricity price prediction model.

[0193] The dual-link communication unit 300 is used to build multiple redundant communication links for wired, wireless, and power line carriers. Based on link quality data, the dual-link communication unit 300 adopts an automatic switching strategy to ensure data transmission continuity, uses an edge-side data clustering algorithm to pre-process monitoring data, implements disconnected data caching and transmission flow optimization, and realizes synchronous power and data transmission through the POE network port;

[0194] In this embodiment, the dual-link communication unit 300 includes a link management module 310, which includes a link monitoring submodule 311 and a switching control submodule 312, wherein:

[0195] The link monitoring submodule 311 is used to collect signal strength, bit error rate, transmission delay and packet loss rate data of wired links, wireless links and power line carrier links;

[0196] The handover control submodule 312 has a built-in multi-dimensional link quality evaluation algorithm and calculates the link comprehensive quality score Q using the following formula:

[0197] Q = w1 × (1-SNR / SNR max )+w2×BER+w3×(RTT / RTT max )+w4×

[0198] PLR;

[0199] Among them, SNR is the signal-to-noise ratio, SNR max is the maximum acceptable signal-to-noise ratio, BER is the bit error rate, RTT is the round-trip delay, and RTT is the maxis the maximum acceptable delay, PLR is the packet loss rate, w1, w2, w3, w4 are weight coefficients and satisfy w1+w2+w3+w4=1;

[0200] The handover control submodule 312 performs trend analysis on the link comprehensive quality scores of M consecutive sampling periods based on a sliding time window algorithm, and calculates the quality change trend factor ΔQ using the following formula:

[0201] ΔQ=(Q m -Q1) / (M×T);

[0202] Among them, Q1 is the link quality score at the start of the window, Q m is the link quality score at the end of the window, T is the sampling period; when the comprehensive quality score Q of a link is lower than the preset threshold Q th When the trend factor ΔQ is less than 0, the link pre-switching mechanism is triggered and the connection establishment process of the backup link is started.

[0203] In this embodiment, the dual-link communication unit 300 further includes a data processing module 320 and a POE transmission module 330, wherein:

[0204] The data processing module 320 includes an edge computing submodule 321 and a cache scheduling submodule 322, wherein:

[0205] The edge computing submodule 321 integrates a spatiotemporal correlation feature extraction algorithm, which calculates the monitoring data point x using the following formula: i The spatiotemporal correlation ST(x i ,t):

[0206] ST(x i ,t)=α×Sim s (x i )+β×Sim t (x i ,t)+γ×Corr(x i ,t);

[0207] Among them, Sim s (x i ) is the spatial similarity function, which calculates the spatial similarity between the current data point and the historical cluster center through the Euclidean distance; Sim t (x i ,t) is the time similarity function, which calculates the similarity of data change trends within the time window through the exponential decay model; Corr(x i ,t) is the spatiotemporal correlation function, which calculates the joint probability distribution of data points in the spatiotemporal dimension; α, β, γ are weight coefficients and α+β+γ=1; x i represents the i-th data sample; t represents the current evaluation time;

[0208] The cache scheduling submodule 322 constructs the following three-level cache queue based on the spatiotemporal correlation ST value:

[0209] When ST≥ST high When , it is stored in the high priority queue;

[0210] When ST low ≤ST <ST high When , store it in the medium priority queue;

[0211] When ST <ST low When , it is stored in the low priority queue;

[0212] Among them, ST high Indicates high correlation threshold, ST low indicates a low association threshold;

[0213] The POE transmission module 330 includes a power line carrier communication submodule 331 and an Ethernet power supply submodule 332, wherein:

[0214] The power line carrier communication submodule 331 has a built-in OFDM modulator and uses OFDM modulation technology to divide the power line channel into N subcarriers and dynamically allocate the number of bits b on the subcarriers through an adaptive bit loading algorithm. i :

[0215]

[0216] Among them, SNR i is the signal-to-noise ratio of the i-th subcarrier, SNR0 is the reference signal-to-noise ratio;

[0217] The OFDM modulator is equipped with an impedance matching interface for connecting to an external impedance matching network to achieve dynamic impedance matching of the power line channel;

[0218] The Ethernet power supply submodule 332 integrates a power negotiation protocol stack, and dynamically adjusts the output voltage V after parsing the classification request of the powered device. out and current I out , so that the total output power satisfies:

[0219]

[0220] Among them, P base is the basic power supply power of the equipment, n is the number of powered devices, P i is the power requirement of the i-th powered device, w i It is a dynamic weight factor (assigned according to device priority, service type and link status).

[0221] In this embodiment, the power line carrier communication submodule 331 integrates an adaptive equalization circuit and uses the LMS algorithm to calculate the equalizer tap coefficient h(n). The calculation formula is as follows:

[0222] h(n)=h(n-1)+μ·e(n)·x(n);

[0223] Where μ is the step size factor, e(n) is the error signal, and x(n) is the input signal. The error signal e(n) satisfies the following equation: e(n) = d(n) - y(n), where d(n) is the desired response signal and y(n) is the equalizer output signal.

[0224] Characteristic impedance Z of the impedance matching network c satisfy:

[0225]

[0226] Where Z0 is the inherent characteristic impedance of the power line (obtained through offline calibration); Z L (n) is the real-time load impedance, measured by injecting a 20kHz-500kHz sweep signal. The dynamic correction factor γ(n) is calculated based on the channel estimation result: Wherein, H(n) is the channel frequency response estimation value at the nth moment.

[0227] As a further illustration of this embodiment, in the link monitoring submodule 311 of this embodiment, the sampling strategy is: link parameters are collected once every 1 second, the wireless link (4G / LoRa) and the wired link (Ethernet) are monitored in parallel, and the power line carrier link (PLC) is polled every 5 seconds (to avoid channel occupancy).

[0228] As a further illustration of this embodiment, in the handover control submodule 312 of this embodiment, a Xilinx Spartan-6 FPGA can be used to implement a multi-dimensional link quality assessment algorithm, supporting parallel computing and hardware acceleration to meet real-time requirements. Specifically, when calculating the comprehensive link quality score Q, the following parameter configuration can be used:

[0229] Weight coefficients: w1 = 0.4 (SNR priority, ensuring basic communication), w2 = 0.3 (BER second best, ensuring data integrity), w3 = 0.2 (RTT third best, adapting to real-time services), w4 = 0.1 (PLR minimum, handling burst packet loss);

[0230] Sliding window: M = 10 (number of sampling cycles), T = 1 second (single cycle duration), total window duration 10 seconds;

[0231] Threshold setting: Q th =0.6 (0-1 rating system, below the threshold triggers switching).

[0232] As a further illustration of this embodiment, in the edge computing submodule 321 of this embodiment, an ARM Cortex-A7 processor (such as a Raspberry Pi Zero W) can be used to run the spatiotemporal correlation feature extraction algorithm, supporting Python / C mixed programming; the spatiotemporal correlation feature extraction algorithm is implemented as follows:

[0233] Spatial similarity: Among them, c k is the cluster center of historical data (can be generated offline through K-means algorithm); Sim s (x i ) is the spatial similarity; max(·) is the maximum distance;

[0234] Time similarity: (λ = 0.1 / hour);

[0235] Weight configuration: α = 0.3, β = 0.5, γ = 0.2 (time correlation priority).

[0236] As a further explanation of this embodiment, in the cache scheduling submodule 322 of this embodiment, IS61LV25616SRAM (32MB) can be used as a high-speed cache, and paired with W25Q128NOR Flash (128MB) to implement three-level queue storage:

[0237] High priority queue (ST ≥ 0.8): stores real-time alarm data and critical battery status;

[0238] Medium priority queue (0.5≤ST<0.8): stores periodic monitoring data (such as environmental parameters every 10 minutes);

[0239] Low priority queue (ST<0.5): stores unstructured data (such as camera video stream fragments);

[0240] In addition, when the network is disconnected, only high-priority data (about 50KB per hour) is cached to avoid storage overflow; after the network is restored, the data is resumed in the order of "high-medium-low", and the resumption rate is limited to 30% of the link bandwidth (to avoid impacting the main business).

[0241] As a further illustration of this embodiment, the Ethernet power supply submodule 332 in this embodiment can use a TPS23753 PoE controller, and its power distribution rules are as follows:

[0242] Basic power P base =5W (system power supply);

[0243] Dynamic weight w i :

[0244] Seismic data acquisition equipment: w i =0.5 (highest priority);

[0245] Camera: w i =0.3 (second priority);

[0246] Spare equipment: w i =0.2 (lowest priority);

[0247] Overload protection: When the total power exceeds 30W, the power supply to low-priority devices is reduced according to the weight ratio. It should be added that the multi-module collaborative control process of the dual-link communication unit 300 in this embodiment is as follows:

[0248] Initialization phase:

[0249] The link management module 310 scans available links (Ethernet / LoRa / PLC), creates a link list and assigns priorities (Ethernet>LoRa>PLC);

[0250] The POE transmission module 330 negotiates the power requirements of the powered device and initializes the power allocation table.

[0251] Real-time operation stage:

[0252] The link monitoring submodule 311 updates the link quality data every 1 second, and the switching control submodule 312 dynamically adjusts the link priority;

[0253] The edge computing submodule 321 performs spatiotemporal correlation analysis on the sensor data, and the cache scheduling submodule 322 stores the data according to the ST value classification;

[0254] The POE transmission module 330 dynamically adjusts power distribution according to the link status (for example, when the wireless link load is high, the power supply to the communication equipment is prioritized).

[0255] Failover phase:

[0256] When the Q value of the main link is lower than the threshold for three consecutive windows (30 seconds), the pre-switching mechanism is triggered and the LoRa / PLC link is preheated synchronously;

[0257] After the switching is completed, the edge computing submodule 321 automatically adjusts the data compression ratio (when the wireless link bandwidth is low, the compression ratio is increased to 8:1).

[0258] The integrated protection and control unit 400 is used to provide sealed protection and intelligent control functions, adopt a gradient heat dissipation structure to optimize equipment thermal management, based on network status monitoring, use fault self-healing strategies to restore communication connections, realize local control through a waterproof operation panel, and integrate multi-sensor fusion technology to realize comprehensive environmental status assessment.

[0259] In this embodiment, the integrated protection and control unit 400 includes a gradient heat dissipation module 410, a fault self-healing module 420, a waterproof interaction module 430, and a multi-sensor fusion module 440, wherein:

[0260] The gradient heat dissipation module 410 uses a layered equidistant fin array to optimize the heat flow path. The spacing between adjacent fin groups starts from a reference value and expands layer by layer with a fixed tolerance. It has a built-in serpentine fluid channel, and the channel size can be linearly changed with the number of fin layers.

[0261] The fault self-healing module 420 identifies communication anomalies by real-time monitoring of link status and executes a three-level recovery process of "diagnosis-switching-calibration" to achieve autonomous detection and recovery of communication failures;

[0262] The waterproof interactive module 430 uses a multi-layer sealing structure to block liquid intrusion, and the touch area is divided into multiple levels of operation permissions to achieve access control. When operating in a wet state, the sensing mode is automatically switched to ensure the reliability of human-computer interaction.

[0263] The multi-sensor fusion module 440 is used to build a multi-dimensional environmental monitoring network to collect heterogeneous data, identify environmental anomalies through joint criteria, and trigger graded early warning responses to achieve a comprehensive assessment of the environmental status.

[0264] In this embodiment, the fault self-healing module 420 includes a link monitoring subunit 421, a diagnosis and decision subunit 422, a redundancy switching subunit 423, and a policy optimization subunit 424, wherein:

[0265] The link monitoring subunit 421 deploys a multi-dimensional probe matrix to collect real-time operating parameters of the physical layer (signal strength, bit error rate), link layer (packet loss rate, throughput), and application layer (response time) to build a full link status profile;

[0266] The diagnosis decision subunit 422 uses a fuzzy logic algorithm to calculate the fault confidence based on the fault feature library and the inference engine, and triggers the corresponding recovery strategy when the confidence exceeds a preset threshold;

[0267] The redundancy switching subunit 423 is used to maintain dynamic link priorities and comprehensively evaluate the multi-dimensional indicators of alternative links (such as bandwidth, latency, stability, etc. of alternative links) through the hierarchical analysis method to achieve rapid switching of the optimal link;

[0268] The strategy optimization subunit 424 is used to build a fault case knowledge base and adopt a reinforcement learning algorithm to dynamically adjust the decision weight according to the historical recovery effect to form a strategy iteration optimization closed loop.

[0269] As a further illustration of this embodiment, in the gradient heat dissipation module 410 of this embodiment, 6063 aluminum profiles can be used to process layered equidistant fins, with the base spacing set to 5 mm (taking into account both air circulation and compact structure), and the spacing between adjacent fin groups is expanded layer by layer with a tolerance of 1 mm (such as 5 mm for the first layer, 6 mm for the second layer, 7 mm for the third layer...), to adapt to the diffusion law of heat flow from dense to sparse; at the same time, a polytetrafluoroethylene serpentine tube can be built into the fluid channel, and the inner diameter of the channel changes linearly with the number of fin layers (such as Φ3 mm for the first layer, and +0.5 mm for each additional layer), and a micro diaphragm pump (such as Kamoer KVP8) is used to drive the circulation of the ethylene glycol aqueous solution to enhance heat exchange.

[0270] As a further illustration of this embodiment, in the redundant switching subunit 423 of this embodiment, three indicator weights are set for link evaluation: bandwidth accounted for 0.4, latency accounted for 0.3, and stability accounted for 0.3 (stability is calculated based on the average bit error rate (BER) within 10 minutes). Among them, the alternative link evaluation formula is:

[0271]

[0272] Among them, S link Indicates the comprehensive evaluation score of the candidate link; B i The actual bandwidth of the i-th candidate link; B max represents the maximum bandwidth among all candidate links; T min represents the minimum delay among all candidate links; T i represents the actual delay of the i-th candidate link; BER i represents the bit error rate of the i-th candidate link;

[0273] When the confidence of the main link is <0.3, S is automatically selected. link The maximum backup link has a switching delay of ≤500ms.

[0274] As a further illustration of this embodiment, in the policy optimization subunit 424 of this embodiment, policy iteration optimization is implemented based on a reinforcement learning framework, specifically including:

[0275] Status (S): defines the system status as a fault type, covering three scenarios: link interruption, noise interference, and insufficient bandwidth;

[0276] Action (A): The optional action is set as the switching strategy, including fast switching and pre-connection switching modes;

[0277] Reward (R): The reward function is designed as: (The “recovery time countdown” encourages quick recovery, and the “error penalty” suppresses invalid actions, guiding the direction of strategy optimization).

[0278] Iterative logic: After every 10 accumulated failure events, the Q-learning algorithm is used to update the decision weight (learning rate α = 0.1, discount factor γ = 0.9) to continuously optimize the strategy library.

[0279] As a further illustration of this embodiment, in the waterproof interactive module 430 of this embodiment, the multi-layer seal may use EPDM O-rings to construct a three-level sealing structure (covering the joints between the panel and the housing, the buttons and the panel, and the interface and the module); the touch panel may use waterproof ITO glass (with an integrated hydrophobic coating on the surface), and is divided into the following three levels of operation permissions:

[0280] Administrator (Level 1): Modify system parameters (such as IP addresses, thresholds);

[0281] Maintenance operator (Level 2): ​​performs diagnostic and restart operations;

[0282] Ordinary user (level 3): view device status and perform simple operations;

[0283] At the same time, the surface humidity can be monitored through the built-in capacitance sensor of the panel. When the capacitance change is greater than 10%, it is judged to be wet.

[0284] As a further explanation of this embodiment, in the multi-sensor fusion module 440 of this embodiment,

[0285] By integrating heterogeneous sensor data with multi-dimensional weighted criteria, a comprehensive assessment of environmental anomalies can be achieved:

[0286] S env =0.6·temperature and humidity anomaly+0.3·vibration anomaly+0.1·gas anomaly;

[0287] Among them, the temperature and humidity anomaly, vibration anomaly, and gas anomaly are all normalized values ​​​​between 0 and 1 (calculated based on the threshold interval of historical data statistics. When the threshold is exceeded, the anomaly tends to 1, and vice versa, it tends to 0);

[0288] Based on S env Implement graded early warning, including:

[0289] S env ≥0.8: Trigger an audible and visual alarm (red LED + buzzer) and report it through the communication module;

[0290] 0.5≤S env <0.8: Yellow warning, record the log;

[0291] S env <0.5: normal state.

[0292] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be accomplished by hardware, or by instructing related hardware through a program.

[0293] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. Unattended site intelligent power supply, characterized by: include: A multi-source power supply scheduling unit (100) is used to realize multi-energy hybrid power supply of photovoltaic, mains, and battery, unify power supply formats through adaptive rectification technology, and dynamically call low-cost energy in combination with an electricity price data prediction model to realize intelligent switching of power supply paths; An intelligent energy storage management unit (200) is used for logical management of battery charging and discharging, estimating the state of charge based on data collected by multi-source sensors in combination with a Kalman filter algorithm, and adopting a state of charge dynamic balancing strategy in combination with an EIS electrochemical impedance spectroscopy frequency domain analysis technique to extract internal resistance trends, thereby achieving battery life prediction and adaptive correction of aging parameters; A dual-link communication unit (300) is used to construct multiple redundant communication links of wired, wireless and power line carriers. The dual-link communication unit (300) uses an automatic switching strategy based on link quality data to ensure data transmission continuity, uses an edge-side data clustering algorithm to pre-process monitoring data, realizes disconnected data caching and transmission flow optimization, and realizes synchronous transmission of power and data through a POE network port; The protection and control integrated unit (400) is used to provide sealing protection and intelligent control functions, adopt a gradient heat dissipation structure to optimize equipment thermal management, based on network status monitoring, utilize a fault self-healing strategy to restore communication connections, realize local control through a waterproof operation panel, and integrate multi-sensor fusion technology to realize comprehensive environmental status assessment.

2. The unmanned site intelligent power supply according to claim 1, characterized in that: The multi-source power supply scheduling unit (100) comprises a main control module (110), wherein: The main control module (110) generates a power supply path switching instruction based on real-time energy data, supports a two-group battery charge and discharge alternating mode, integrates an electricity price prediction model and an adaptive rectifier control algorithm, and implements mains access status monitoring and battery charge and discharge parameter management. The main control module (110) includes an energy scheduling submodule (111), a battery management submodule (112), and a mains monitoring submodule (113), wherein: The energy scheduling submodule (111) generates priority scheduling strategies for photovoltaic, mains, and battery power based on historical electricity price data, weather forecasts, and light intensity data of the LSTM neural network model; and adapts peak and valley electricity price rules of power grids in different regions through a transfer learning algorithm to minimize energy costs. The battery management submodule (112) uses a bidirectional DC-DC converter to achieve independent charge and discharge control of two battery groups, automatically switches the charging path based on the state of charge threshold; integrates a battery balancing circuit, and achieves capacity consistency management of the two battery groups through dynamic impedance matching technology; The mains monitoring submodule (113) uses a voltage transformer and a frequency detection chip to collect mains voltage and frequency data in real time; when it is detected that the voltage exceeds a preset threshold range, an audible and visual alarm is triggered and an alarm log is generated.

3. The unmanned site intelligent power supply according to claim 2, characterized in that: The multi-source power supply scheduling unit (100) further includes a communication module (120), wherein: The communication module (120) is connected to the photovoltaic inverter, the battery pack and the mains monitoring submodule (113) via multiple electrical isolation serial ports, and adopts a dual-link communication architecture to achieve data interaction; the communication module (120) includes a data cache submodule (121), wherein: The data cache submodule (121) is used to store energy data when the network is disconnected and automatically resume transmitting the energy data after the network is restored.

4. The unmanned site intelligent power supply according to claim 3, characterized in that: The intelligent energy storage management unit (200) comprises a multi-source data perception module (210) and an EIS detection module (220), wherein: The multi-source data perception module (210) includes a data acquisition submodule (211) and a signal conditioning submodule (212), wherein: The data acquisition submodule (211) integrates a voltage sensor, a current sensor, and a temperature sensor, which are respectively used to collect the battery terminal voltage, charge and discharge current, and surface temperature in real time; and realizes three-channel synchronous sampling through a high-precision analog-to-digital conversion chip, and the sampling frequency meets the requirements of real-time monitoring of the battery status; The signal conditioning submodule (212) is used to filter, reduce noise and adapt the level of the sensor output signal, eliminate signal interference and convert it into a standardized digital signal; and supports sensor fault self-detection, triggering a hardware interrupt when any sensor data is abnormal and reporting it to the main control module (110); The EIS detection module (220) comprises a sweep frequency excitation submodule (221), an impedance decoupling submodule (222) and an aging parameter generation submodule (223), wherein: The sweep frequency excitation submodule (221) is used to inject a sinusoidal excitation signal of a preset frequency band into the battery pack to stimulate the electrochemical impedance response of the battery; the signal amplitude is dynamically adjusted according to the current state of the battery to ensure that the excitation intensity is adapted to different charge states; The impedance decoupling submodule (222) is used to synchronously collect voltage and current responses corresponding to the excitation signal, separate and calculate the battery impedance spectrum through a frequency domain analysis algorithm; and extract the internal resistance value of the characteristic frequency point to form an impedance characteristic vector reflecting the internal state of the battery; The aging parameter generation submodule (223) establishes an internal resistance-aging model based on the impedance characteristic vector and temperature data, and outputs battery health status correction parameters for dynamically adjusting the energy storage management strategy.

5. The unmanned site intelligent power supply according to claim 4, characterized in that: The intelligent energy storage management unit (200) further comprises a dynamic balancing control module (230), wherein the dynamic balancing control module (230) is used to achieve accurate estimation of battery pack status, environmental impact compensation and energy balancing control, and the dynamic balancing control module (230) comprises a state of charge estimation submodule (231), a temperature compensation submodule (232) and an active balancing execution submodule (233), wherein: The state of charge estimation submodule (231) fuses the voltage, current, and temperature data of the multi-source sensor modules and iteratively estimates the state of charge of the battery pack through the Kalman filter algorithm; introduces the internal resistance trend output by the EIS detection module (220) to compensate the state of charge estimation result in real time, thereby reducing the long-term estimation error; and adopts a multi-source data fusion architecture to achieve time synchronization and error calibration of different sensor data, and outputs a high-precision state of charge value to the active balancing execution submodule (233) as a decision basis for the balancing strategy; The temperature compensation submodule (232) is used to construct a nonlinear correction model of temperature and state of charge, and adjust the state of charge estimation result according to real-time temperature data to eliminate the influence of temperature drift; the nonlinear correction model parameters are generated by self-learning of battery historical data to adapt to state estimation under different ambient temperatures; and the internal resistance value detected by EIS is temperature compensated to establish a temperature-internal resistance mapping relationship; The active balancing execution submodule (233) uses a bidirectional DC-DC converter to realize bidirectional energy transfer between battery packs, and dynamically adjusts the balancing current according to the difference in charge state and internal resistance of the battery packs, so as to achieve the adaptation of batteries with different capacities and aging degrees; and constructs a multi-objective optimization function, comprehensively formulates the balancing priority according to the charge state deviation, internal resistance difference and temperature distribution, and gives priority to processing battery packs with significant state differences; and introduces a time-sharing balancing scheduling mechanism, combined with load prediction to avoid peak power consumption, thereby reducing the interference of the balancing process on the system power supply.

6. The unmanned site intelligent power supply according to claim 1, characterized in that: The dual-link communication unit (300) comprises a link management module (310), and the link management module (310) comprises a link monitoring submodule (311) and a switching control submodule (312), wherein: The link monitoring submodule (311) is used to collect signal strength, bit error rate, transmission delay and packet loss rate data of wired links, wireless links and power line carrier links; The switching control submodule (312) has a built-in multi-dimensional link quality evaluation algorithm and calculates the link comprehensive quality score Q using the following formula: Q=w1×(1-SNR / SNR max )+w2×BER+w3×(RTT / RTT max )+w4× PLR; Among them, SNR is the signal-to-noise ratio, SNR max is the maximum acceptable signal-to-noise ratio, BER is the bit error rate, RTT is the round-trip delay, and RTT is the max is the maximum acceptable delay, PLR is the packet loss rate, w1, w2, w3, w4 are weight coefficients and satisfy w1+w2+w3+w4=1; The switching control submodule (312) performs trend analysis on the link comprehensive quality scores of M consecutive sampling periods based on a sliding time window algorithm, and calculates the quality change trend factor ΔQ using the following formula: ΔQ=(Q m -Q1) / (M×T); Among them, Q1 is the link quality score at the start of the window, Q m is the link quality score at the end of the window, T is the sampling period; when the comprehensive quality score Q of a link is lower than the preset threshold Q th When the trend factor ΔQ is less than 0, the link pre-switching mechanism is triggered and the connection establishment process of the backup link is started.

7. The unmanned site intelligent power supply according to claim 6, characterized in that: The dual-link communication unit (300) further comprises a data processing module (320) and a POE transmission module (330), wherein: The data processing module (320) includes an edge computing submodule (321) and a cache scheduling submodule (322), wherein: The edge computing submodule (321) integrates a spatiotemporal correlation feature extraction algorithm, which calculates the monitoring data point x by the following formula: i The spatiotemporal correlation ST(x i ,t): ST(x i ,t)=α×Sim s (x i )+β×Sim t (x i ,t)+γ×Corr(x i ,t); Among them, Sim s (x i ) is the spatial similarity function, which calculates the spatial similarity between the current data point and the historical cluster center through the Euclidean distance; Sim t (x i ,t) is the time similarity function, which calculates the similarity of data change trends within the time window through the exponential decay model; Corr(x i ,t) is the spatiotemporal correlation function, which calculates the joint probability distribution of data points in the spatiotemporal dimension; α, β, γ are weight coefficients and α+β+γ=1; x i represents the i-th data sample; t represents the current evaluation time; The cache scheduling submodule (322) constructs the following three-level cache queue based on the time-space correlation ST value: When ST≥ST high When , it is stored in the high priority queue; When ST low ≤ST <ST high When , store it in the medium priority queue; When ST <ST low When , it is stored in the low priority queue; Among them, ST high Indicates high correlation threshold, ST low indicates a low association threshold; The POE transmission module (330) comprises a power line carrier communication submodule (331) and an Ethernet power supply submodule (332), wherein: The power line carrier communication submodule (331) has a built-in OFDM modulator and uses OFDM modulation technology to divide the power line channel into N subcarriers and dynamically allocate the number of bits b on the subcarriers through an adaptive bit loading algorithm. i : Among them, SNR i is the signal-to-noise ratio of the i-th subcarrier, SNR0 is the reference signal-to-noise ratio; The OFDM modulator is equipped with an impedance matching interface for connecting to an external impedance matching network to achieve dynamic impedance matching of the power line channel; The Ethernet power supply submodule (332) integrates a power negotiation protocol stack, and dynamically adjusts the output voltage V after parsing the classification request of the powered device. out and current I out , so that the total output power P out satisfy: Among them, P base is the basic power supply power of the equipment, n is the number of powered devices, P i is the power requirement of the i-th powered device, w i is the dynamic weight factor.

8. The unmanned site intelligent power supply according to claim 7, characterized in that: The power line carrier communication submodule (331) integrates an adaptive equalization circuit and uses an LMS algorithm to calculate the equalizer tap coefficient h(n). The calculation formula is as follows: h(n)=h(n-1)+μ·e(n)·x(n); Where μ is the step size factor, e(n) is the error signal, and x(n) is the input signal. The error signal e(n) satisfies the following equation: e(n) = d(n) - y(n), where d(n) is the desired response signal and y(n) is the equalizer output signal. The characteristic impedance Z of the impedance matching network c satisfy: Among them, Z0 is the inherent characteristic impedance of the power line; Z L (n) is the real-time load impedance, measured by injecting a 20kHz-500kHz sweep signal. The dynamic correction factor γ(n) is calculated based on the channel estimation result: Wherein, H(n) is the channel frequency response estimation value at the nth moment.

9. The unmanned site intelligent power supply according to claim 1, characterized in that: The protection and control integrated unit (400) comprises a gradient heat dissipation module (410), a fault self-healing module (420), a waterproof interaction module (430) and a multi-sensor fusion module (440), wherein: The gradient heat dissipation module (410) uses a layered equidistant fin array to optimize the heat flow path, the spacing between adjacent fin groups is based on a reference value and is expanded layer by layer with a fixed tolerance, and a serpentine fluid channel is built in; The fault self-recovery module (420) identifies communication anomalies by real-time monitoring of link status and executes a three-level recovery process of "diagnosis-switching-calibration" to achieve autonomous detection and recovery of communication faults; The waterproof interactive module (430) adopts a multi-layer sealing structure to block liquid intrusion, the touch area is divided into multiple levels of operation permissions to achieve access control, and the sensing mode is automatically switched during wet operation to ensure the reliability of human-computer interaction; The multi-sensor fusion module (440) is used to construct a multi-dimensional environmental monitoring network to collect heterogeneous data, identify environmental anomalies through joint criteria, trigger hierarchical early warning responses and realize comprehensive evaluation of environmental status.

10. The unattended site intelligent power supply according to claim 9, characterized in that: The fault self-healing module (420) includes a link monitoring subunit (421), a diagnosis and decision subunit (422), a redundancy switching subunit (423) and a strategy optimization subunit (424), wherein: The link monitoring subunit (421) collects the operating parameters of the physical layer, link layer and application layer in real time by deploying a multi-dimensional probe matrix, and constructs a full link status portrait; The diagnosis decision subunit (422) calculates the fault confidence level using a fuzzy logic algorithm based on a fault feature library and an inference engine, and triggers a corresponding recovery strategy when the confidence level exceeds a preset threshold; The redundant switching subunit (423) is used to maintain dynamic link priorities and comprehensively evaluate the multi-dimensional indicators of the candidate links through the hierarchical analysis method to achieve rapid switching of the optimal link; The strategy optimization subunit (424) is used to build a fault case knowledge base and adopt a reinforcement learning algorithm to dynamically adjust the decision weight according to the historical recovery effect to form a strategy iteration optimization closed loop.

Images