Electric power fire alarm monitoring system based on wireless communication

Abstract

The invention discloses an electric power fire alarm monitoring system based on wireless communication, relates to the technical field of intelligent monitoring of an electric power system, and is used for solving the problems of multi-parameter detection, intelligent alarm and path steady state judgment of an electric fire. Through high-precision acquisition of multiple parameters such as zero-sequence current, temperature, voltage, grounding impedance and the like, in combination with local intelligent analysis and a remote linkage strategy, real-time identification and response to electrical fire hazards are realized. The system has a field complement sequence grouping mechanism, a path stability scoring mechanism and an extraction strategy scheduling mechanism, and the abnormal fault tolerance and communication link adaptive capacity are improved. And the power supply module adopts a main-standby dual-power-supply framework, so that stable operation of the system during a power failure period is ensured. The platform supports trend modeling, remote control and strategy management, constructs an integrated monitoring scheme of data acquisition, intelligent discrimination and closed-loop response, and is suitable for various electric power facility fire risk scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system intelligent monitoring, more particularly, the present application relates to a power fire alarm monitoring system based on wireless communication. BACKGROUND

[0002] Electrical fire is one of the most common and most serious fire types in cities and industrial sites, and its concealment, suddenness and multi-source characteristics bring challenges to traditional fire prevention and control means. Conventional electrical fire monitoring systems mostly rely on single parameter detection, lack of continuous tracking and multi-factor interaction identification capability for complex evolution paths, and are prone to monitoring failure or strategy mis-triggering under extreme conditions such as communication anomaly and power failure. In addition, the mainstream monitoring equipment lacks path consistency management of the corresponding relationship between alarm events and sensing data, and the data tag drift problem frequently occurs after restart, affecting the accuracy of platform linkage determination. The existing technology does not construct a robust data recovery and path correction mechanism for scenarios such as power failure restart and field backfill, resulting in phenomena such as historical state adsorption and label residual chain reuse that are difficult to trace and control.

[0003] In view of the above problems, the present application provides a solution. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a power fire alarm monitoring system based on wireless communication to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: In one preferred embodiment, it comprises a sensing acquisition module, an alarm processing module, a wireless communication module, a power and backup module, and a platform monitoring module, and the modules are signal connected; The sensing acquisition module is used to configure sensors and connect microcontroller ADC for synchronous sampling through an analog signal conditioning circuit; The alarm processing module is used to perform filtering processing on the sensing signal data after signal conditioning by using an embedded microcontroller, and performs alarm determination; In the wireless communication module, the acquisition terminal encapsulates the encrypted data frame as an MQTT message body through Wi-Fi or LoRa networking to report to the platform monitoring module, and buffers and retransmits during network interruption; In the power and backup module, the main power supply outputs a direct current bus through a switching power supply, the backup power supply is a battery pack, the main and backup switching is completed and the power failure is reported when power failure occurs, and in the field backfill process, a rule set is generated based on the power-on serial number, label version number and path fingerprint, a plurality of power-on sequence blocks are grouped as field, and a sequence feature vector is constructed to calculate the field attribution path stability score, the field is extracted dynamically according to the conflict mapping cluster, and the path is written after encapsulating the field meta state information. The platform monitoring module is used for receiving and analyzing the data frame uploaded by the collection terminal, performing ternary consistency judgment, generating linkage control instructions, and issuing the instructions to the collection terminal and recording behavior logs.

[0006] In a preferred embodiment, in the sensing and collecting module, a zero sequence current transformer, a Hall effect current sensor, an NTC type thermistor, a precision resistance array, an isolation amplifier, an integrated temperature and humidity sensor, and a photoelectric smoke sensor are configured in the collection terminal. The output signals of the sensors are sequentially buffered, amplified, and low-pass filtered by the analog signal conditioning circuit, and the conditioned signals are connected to the ADC module of the microcontroller for synchronous sampling. Before the collection logic is started, sensor short circuit and open circuit detection is triggered, and the least squares fitting method is used to calculate the range coefficient and zero offset coefficient based on the standard curve.

[0007] In a preferred embodiment, in the alarm processing module, an embedded microcontroller is used to sequentially read the sensing signal data of each channel in the microcontroller ADC module through timing interrupts. The mean filter algorithm is applied to the line residual current signal, and the median filter algorithm based on singular value sorting is applied to the conductor temperature signal. A circular buffer is constructed and updated in real time, and alarm judgment is performed based on the line residual current alarm threshold, the conductor temperature alarm threshold, and the pre-stored disturbance waveform feature model in the arc fault recognition mode library.

[0008] In a preferred embodiment, in the wireless communication module, a double-layer networking structure is used. In a short-distance communication scenario, the Wi-Fi wireless communication module embedded in the collection terminal is connected to the local gateway device and the local area network. In a long-distance or Wi-Fi access unavailable scenario, the LoRa transceiver embedded in the collection terminal is connected to the platform monitoring module through the LoRa gateway and the cellular network. The microcontroller constructs a unified data frame, performs AES-128 encryption, and encapsulates it as an MQTT message body to report to the platform monitoring module. The data frame containing the alarm status flag is enabled with ACK confirmation and automatic retransmission. At the same time, data is cached during network disconnection and is sent after the network is restored.

[0009] In a preferred embodiment, in the power supply and backup module, the main power supply outputs a DC bus through an industrial-grade isolated switching power supply, and the backup power supply uses a sealed lead-acid battery pack. The switching circuit completes the main-backup switching when the main power supply fails, and cuts off unnecessary loads and reports power supply failure when the battery voltage is abnormal.

[0010] In a preferred embodiment, in the power supply and backup module, before the pre-initialization stage of power supply switching completed by the backup power supply and starting the field backfill logic, the power-on sequence number, tag version number, tag loading completion flag, backfill frame number proportion, and field upload readiness time and tag table readiness time difference are extracted from the power-on log and cache index, and the minimum dependency set is generated for each field supplement source state marker; Based on the minimum dependency set, three types of abnormalities, including pre-upload of lock table not ready, backfill pre-coverage upload, and residual chain reuse mapping, are identified within the first upload window, and a rule set is dynamically generated according to the readiness time difference, source state marker, path fingerprint, and tag version difference, to apply the power-on minimum readiness window to real-time fields, add backfill dedicated markers to backfill fields and freeze strategy judgment qualifications, and forcibly cut off historical path adsorption through a tag exclusion table for residual chain reuse fields, then the fields are grouped by power-on sequence to form power-on sequence blocks of the same origin and same type.

[0011] In a preferred embodiment, in the power supply and backup module, during the field backfill process, the terminal collects the average readiness time difference, path fingerprint similarity average, frame sequence rearrangement degree, exclusion hit rate, and backfill proportion of each group of power-on sequence blocks, constructs a sequence feature vector, and calculates the field attribution path stability score based on the weighted sum of the frame sequence rearrangement degree, exclusion hit rate, readiness time difference, and backfill proportion.

[0012] In a preferred embodiment, in the power supply and backup module, when extracting fields from the main chain, the field extraction path is dynamically scheduled according to the conflict mapping cluster: fields with high field attribution path stability score and not hitting the exclusion are executed with the tag locking strategy, fields with low field attribution path stability score and not hitting the conflict mapping cluster are executed with the flexible buffer binding, fields hitting the conflict mapping cluster are executed with the path suppression strategy, and the field meta-state information is encapsulated before the field is written into the main structure object set.

[0013] In a preferred embodiment, in the platform monitoring module, a message middleware and a RESTAPI service interface are deployed to receive data frames uploaded by the collection terminal and send a reception confirmation, parse the monitoring point unique identifier ID, signal type identifier, value field, state bit, and timestamp, write the value field into a time series database, write the alarm field into a relational database, and perform field, path, and tag consistency judgment, when the alarm condition is met, issue a linkage control instruction to the collection terminal through MQTT, and simultaneously encapsulate the context data and write it into the behavior log database.

[0014] The technical effects and advantages of the wireless communication-based power fire alarm monitoring system of the present application are as follows: This invention achieves accurate identification of multi-parameter, multi-channel electrical fire hazards through the linkage of the data acquisition terminal and the platform, significantly improving the accuracy of system early warning. It introduces a field power-on sequence block and a path stability scoring mechanism to effectively solve the problems of label drift and path misjudgment during restart and recovery. A dynamic scheduling strategy for field-assigned paths is constructed using the minimum dependency set, path fingerprint, and mutual exclusion lookup table to ensure data structure continuity and linkage response stability. For communication assurance, dual-layer caching and AES-128 encryption enhance data integrity and security. The power module adopts a main / backup power switching and undervoltage protection linkage mechanism to ensure continuous system operation and proactive reporting of power failures under power outage or abnormal conditions. The platform supports time-series data modeling, strategy configuration, remote linkage, and log tracking, achieving closed-loop management from data acquisition and judgment to response. The overall solution possesses high reliability, strong adaptability, and excellent remote maintainability. Attached Figure Description

[0015] Figure 1 This is a timing diagram of the power fire alarm monitoring system based on wireless communication according to the present invention.

[0016] Figure 2 This is a schematic diagram of the power fire alarm and monitoring system module based on wireless communication according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example

[0019] This invention discloses a power fire alarm and monitoring system based on wireless communication, such as... Figure 2 As shown, it includes: a sensor acquisition module, an alarm processing module, a wireless communication module, a power supply and backup module, and a platform monitoring module, with signal connections between each module.

[0020] In the sensor acquisition module, multiple types of acquisition terminals deployed in the monitoring area collect synchronous, high-precision, and periodic data on key electrical and environmental parameters related to electrical fire risks.

[0021] Specifically, in the process of collecting electrical key parameters, the zero sequence current transformer configured in the collection terminal is installed in a sleeved manner at the main incoming line of the monitoring circuit. Through the unbalanced component of the sum of the current vectors in the induction circuit, the real-time collection of the line leakage current change signal is realized, so as to judge the potential leakage or insulation deterioration trend. At the same time, a Hall effect current sensor is deployed in series in each load branch. Based on the magnetic induction principle, a voltage signal proportional to the load current is output, which is used to synchronously collect the actual operating current and overload current of the line in the working state.

[0022] For the thermal risk monitoring of electrical connection points, NTC type thermistors are pre-stuck at typical heat points such as copper bars and wiring terminals. With the negative correlation between temperature and resistance value, a voltage signal linearly changing with the conductor temperature is output under constant current bias, which is used to obtain the real-time conductor temperature evolution trajectory, and to assist in determining hidden dangers such as poor contact and oxidation aging.

[0023] In terms of line voltage monitoring, a voltage division network composed of a precision resistor array is constructed in the collection terminal to perform non-distortion amplitude reduction on high-voltage line signals. An isolation amplifier is configured synchronously to realize electrical isolation and bandwidth compensation of high-voltage input signals, thereby completing the safe collection of line voltage parameters. For ground loop state monitoring, a dedicated ground resistance measurement module is connected in the collection terminal. Based on the small current excitation and impedance estimation model, the loop impedance value change trend of the ground end to the ground is obtained under low-frequency alternating current excitation, and hidden dangers such as ground failure and increased ground resistance are identified accordingly.

[0024] Further, in terms of the collection of environmental key parameters, integrated temperature and humidity sensors, photoelectric smoke sensors and gas sensitive sensor modules are pre-set in the monitoring space of the collection terminal. Based on the thermistor capacitor method, the scattered light intensity method and the semiconductor sensitive film conductance method, the air temperature / humidity index, the smoke particulate matter concentration and the volume fraction index of typical combustible gases such as carbon monoxide in the monitored environment are obtained, which are used to construct the dynamic parameter field of environmental incentives of electrical fire risk.

[0025] Then all sensor output signals are connected in turn to a uniformly designed analog signal conditioning circuit, which includes a signal buffer stage, a gain amplification stage and a low-pass filter stage. By configuring an input buffer amplifier to improve input impedance, using a high-linearity operational amplifier to set the gain range, and designing a low-pass filter based on RC structure to reduce high-frequency noise, a signal front-end processing channel matched with the input parameters of the back-end analog-to-digital conversion module ADC is constructed.

[0026] All conditioned signals are accessed into the 12-bit ADC module built-in the acquisition terminal microcontroller MCU, and the sampling timing is controlled by the hardware timer configured in the MCU, which has the ability of synchronous sampling with fixed period and unified phase, ensuring that all physical quantities are collected in parallel within the same sampling period, improving the consistency of data timing and the feasibility of linkage analysis.

[0027] Before the start of the acquisition logic, the microcontroller MCU automatically stimulates each sensor to perform short circuit / open circuit detection to determine the initial access state of the sensor. When powered on for the first time, the standard curve provided by the sensor manufacturer is called synchronously, and the range coefficient and zero offset coefficient of the single-point sensor are calculated by least squares fitting. The calibration results are stored in the local Flash of the MCU as a data calibration reference for the entire life cycle, ensuring the long-term traceability and consistency of comparison of various sensor data.

[0028] In the alarm processing module, the embedded microcontroller is used as the core processing unit. Through real-time analysis and comprehensive judgment of the electrical key parameters and environmental key parameters obtained by the sensor acquisition module, multi-dimensional alarm recognition and local linkage control execution of typical electrical fire inducers are realized. The confirmed alarm result is packaged as a remote reporting field as a pre-trigger basis for the upper layer platform response and intervention.

[0029] Specifically, the alarm processing module uses an embedded microcontroller MCU based on the ARM Cortex-M series kernel, which integrates timing interrupts and analog-to-digital conversion resources. The microcontroller reads the conditioned sensor signal data in the ADC module in a fixed sampling period, including line residual current, conductor temperature, load current, voltage, ground impedance, air temperature and humidity, smoke concentration, and CO volume fraction.

[0030] In the original data reading and processing stage, to eliminate the false judgment behavior caused by short-term peak interference, the alarm processing module applies a 5-cycle sliding window-based mean filter algorithm to the residual current signal to dynamically calculate the current trend in the sampling interval. For the conductor temperature signal, a median filter algorithm based on singular value sorting is used to suppress outliers caused by contact disturbance at the sensor sticking part.

[0031] Among them, the above filtering algorithms build a circular buffer in the local RAM of the embedded microcontroller and update it in real time without relying on external resources.

[0032] For the alarm triggering logic, the alarm processing module introduces a multi-channel and multi-type judgment mechanism. Specifically, the alarm threshold of the line residual current can be configured and adjusted in the range of 300mA to 500mA, and the alarm threshold of the conductor temperature can be set in the range of 45℃ to 140℃. Both are dynamically set by the terminal management program according to the risk level strategy.

[0033] The temperature threshold range is referenced to the General Technical Conditions for Low Voltage Switchgear and Controlgear Assemblies GB7251 and the copper bar temperature resistance level specification provided by the equipment manufacturer (such as the long-term working temperature is not higher than 105℃), combined with the steady-state heat accumulation curve of the conductor connection terminal under different loads in historical data, to build a ΔT (temperature rise) and environmental temperature coupling model. The platform end uses TSDB trend fitting curve to form a thermal characteristic model in the initial operation and maintenance stage of the equipment, dynamically generates the recommended alarm threshold of each monitoring point, and can be remotely adjusted through RESTAPI during the operation of the equipment.

[0034] Among them, the residual current alarm threshold configuration combines the differences of different loop load types and on-site grounding methods, and the terminal management program dynamically configures 300mA, 350mA, 400mA, 450mA or 500mA levels. The threshold setting can be dynamically adjusted by the platform end parameter issuing, and recorded in the alarm rule table in the terminal Flash for restart recovery.

[0035] For environmental parameters, the smoke concentration and CO gas concentration alarm mechanism adopts a dual condition triggering mechanism of unit time growth rate threshold and absolute concentration upper limit, that is, both the short-term sharp rising trend and the concentration exceeding the standard are required to determine the environmental pre-alarm event, thereby effectively avoiding false alarms caused by minor disturbances in life.

[0036] In the identification of electrical fire core hidden dangers, the alarm processing module has a built-in arc fault identification mode library, which pre-stores multiple groups of disturbance waveform feature models based on typical series / parallel arc samples, combines with imported current and voltage synchronous sampling values, and uses multi-dimensional features such as local extreme density, fluctuation rate, and harmonic energy proportion to build a fast matching framework to identify the initial disturbance behavior of typical arc discharge, and realize rapid response to high-risk arc hidden dangers.

[0037] It should be noted that all the above alarm rules are based on a unified continuous sampling confirmation mechanism, that is, when the corresponding alarm condition is met in N consecutive sampling periods, the alarm is confirmed, effectively avoiding false triggering caused by transient pulses and short-term disturbances.

[0038] When the alarm condition is confirmed, the alarm processing module immediately executes local multi-strategy linkage response, as follows: On the one hand, the microcontroller outputs a GPIO signal to drive the relay or circuit breaker element to act, completing the rapid shutdown of the fault branch power supply; On the other hand, the locally configured sound and light alarm module is activated synchronously, and the visual sound and light prompt is performed in the form of high-frequency flashing of the buzzer and LED, prompting the on-site personnel to pay attention to safety; at the same time, the current alarm channel number, alarm type, alarm threshold, confirmation timestamp and other fields are packaged into a standard format alarm data frame, and sent into the wireless communication module through the communication interface.

[0039] In the wireless communication module, a double-layer networking structure is adopted to construct a wireless data link between the collection terminal and the platform monitoring module, so as to adapt to the data reporting demand in different physical scenes and to ensure that the key alarm information is reliably uplinked in a low-power, high-redundancy and safe transmission mechanism.

[0040] Specifically, in a short-distance communication scene, a Wi-Fi wireless communication module is embedded in the collection terminal, which is connected to the on-site gateway device based on an 802.11 protocol stack; then the Wi-Fi wireless communication module is connected through the UART or SPI interface built in the microcontroller MCU, and accesses the local area network by means of the TCP / IP protocol stack.

[0041] For the gateway device with public network access capability, a 4G full-network module or an NB-IoT wireless communication module is integrated inside, and cellular network access is supported at the communication protocol layer, so as to realize transparent forwarding and reporting of terminal data from the local to the remote platform monitoring module in Wi-Fi and 4G modes.

[0042] When the terminal is deployed in a long-distance area or in an environment where the signal coverage is weak and Wi-Fi access is unavailable, the collection terminal uses an embedded SX1278 series LoRa low-power long-distance communication transceiver, forms a star topology based on a 433MHz frequency band and a SpreadingFactor configurable modulation scheme, and transmits data to the LoRa gateway as a slave node, and then reports the platform through the cellular network, which effectively improves the scene adaptability and network redundancy of the overall link.

[0043] At the communication protocol layer, in order to ensure the consistency of information organization across terminals and the platform scalability, a lightweight publish / subscribe message queue telemetry transport protocol stack is integrated in the wireless communication module to support stable communication between multiple nodes through Topic classification, QoS level and session maintenance mechanism.

[0044] In each collection terminal, the microcontroller constructs the data frame to be uploaded according to a unified structure, and each frame of data is composed of five core fields, as follows: collection channel number; monitoring point unique ID; Unix format timestamp; alarm status flag; current measurement value; Further, after the data frame is structured and packaged at the MCU end, it is sent to the communication module buffer through the UART / SPI bus, and the wireless communication module performs AES-128 encryption on the data frame to encapsulate it into an MQTT message body, which is then reported to the platform monitoring module specified Topic channel.

[0045] Subsequently, ACK confirmation and automatic retransmission are enabled for high-priority data frames containing alarm status flag bits. After successfully receiving the alarm data frame, the platform monitoring module actively issues a reception confirmation message. In the absence of a confirmation message, the wireless communication module automatically retransmits according to the upper limit of the retransmission number and the exponential backoff strategy until the platform confirms or times out and discards.

[0046] In the case of communication link fluctuations or temporary interruptions, to avoid important data loss, the wireless communication module is internally configured with a double-layer buffer structure: on the one hand, a RAM-level short-time cache buffer pool is constructed based on a ring array in the communication buffer area; on the other hand, all data are simultaneously backed up to the Flash non-volatile storage area, with a maximum storage period of 72 hours.

[0047] During network disconnection, a retransmission list is constructed based on local timing index, and the lost data re-reporting process is automatically completed after network recovery, ensuring data flow continuity and event trajectory integrity.

[0048] In terms of communication security mechanisms, the wireless communication module introduces a bidirectional encryption tunnel based on the TLS / SSL protocol for all uplink data channels, and enables symmetric encryption algorithm AES-128 for content protection for monitoring point ID, alarm type, and measurement value key fields in the MQTT message body.

[0049] In terms of online status monitoring mechanisms, the platform monitoring module sends a heartbeat request frame to all collection terminals every 30 seconds, and the terminals immediately return a response frame upon receiving the heartbeat request. The platform monitoring module records the number of missed responses, and if the response fails consecutively for 2 times, the collection terminal is marked as offline, and the platform-side event notification or operation and maintenance instruction is triggered, ensuring real-time visibility and disaster recovery controllability of the platform for the online status of the collection terminal.

[0050] In the power supply and backup module, a dual-power supply architecture is adopted, with the main power supply and backup power supply working together to ensure that the collection terminal, alarm processing module, and wireless communication module can operate continuously and stably under different power supply scenarios, improving the overall reliability and abnormal disaster recovery capability of the system.

[0051] Specifically, the system main power supply adopts the AC 220V alternating current input mode commonly used in standard industrial field, and generates stable 5V and 3.3V DC bus for power supply main line output through the internal integrated industrial-grade isolated switching power supply module for voltage reduction and stabilization conversion. The voltage regulation module supports the compatible access of the DC 24V input scene, and adapts to various field power supply systems.

[0052] To prevent the abnormal power failure or fault interruption scene of the main power supply, the power supply and the backup module are synchronously configured with a standby power supply system, and a 12V, 4Ah specification sealed lead-acid battery group is used as a backup power supply. During normal power supply of the main power supply, the standby power supply is maintained by the current limiting loop provided by the main power supply. The float charging current is dynamically controlled by the constant voltage current limiting device to ensure that the battery is in a trickle charging state and prevent the battery performance from deteriorating during long-term storage.

[0053] In terms of power supply switching control, seamless switching operation of the main-standby power supply is completed by the switching circuit after the main power supply is powered off. The switching structure can be based on an automatic transfer switch, a bidirectional Schottky diode array, or a voltage gate device, and has a millisecond-level switching response capability.

[0054] During the operation of the standby power supply, the battery voltage state is continuously monitored. When the voltage drops below the set threshold, the under-voltage protection mechanism is triggered: the microcontroller issues a load cutoff instruction, and the power-off MOSFET or relay module is used to realize power-off protection for non-essential loads to prevent permanent capacity degradation caused by excessive discharge of the battery. At the same time, the under-voltage state is packaged as an alarm field data frame in the form of a power failure event and reported to the platform monitoring module through the wireless communication module to remind the platform to perform manual inspection or remote command control.

[0055] When the main power supply is restored and enters a stable output state, the switching circuit automatically detects the power-on of the main power supply, performs reverse switching to the main power supply power supply path, and again puts the standby power supply into the float maintenance mode.

[0056] It should be noted that when the acquisition terminal is started by the standby power supply due to the power-off of the main power supply, and enters the LoRa broadcast channel within a short time to perform buffer data backfill reporting, since the field tag binding table is not synchronized and persisted in the power-off state, and the field buffer structure is restored by default in the FIFO mode without binding the field ownership state, the sequence drift occurs in the field tag loading and data recovery process, and then the tag remapping residual chain occurs, i.e., the channel to which the field value belongs is incorrectly mapped to the historical tag path, thereby inducing the path drift of the platform-side linkage strategy determination, and finally causing the linkage deviation phenomenon such as incorrect triggering, retriggering, or closed loop behavior conflict of the strategy triggering chain.

[0057] In the original scheme, after switching to the standby power supply, the terminal will first perform MCU-level field cache recovery, and restore the sampling channel frame stack structure based on the field pointer index recorded in the Flash. In this process, the data frames received by the platform are usually parsed according to the original channel number + tag path combination to construct the linkage rule matching chain based on the field tag. However, at this stage, the field pointer index only records the field bit sequence relationship and does not record the tag mapping table state. After restarting, the field data is recovered before the tag table, causing some fields to be uploaded before the tag table is mapped, forming a tag lag binding behavior. When parsing, the platform uses the residual tag mapping cache of the last period, rather than the tag chain corresponding to the actual sampling channel of the current field, causing the field to be misidentified as other channel events or historical alarm rewriting behavior.

[0058] Therefore, in the embodiment, as shown in Figure 1 In the pre-initialization stage of the terminal after completing power switching by the standby power supply and starting the field recovery logic, the power-on serial number, tag version number, tag loading completion flag, proportion of recovery frame number to total frame number to be transmitted, and difference between field upload readiness time and tag table readiness time are first extracted and retained from the power-on log and cache index; at the same time, a minimum dependency set is generated for each field to be uploaded, including the source state mark and the digest formed by the combination of channel number, monitoring point number, and the latest policy group number; Then, based on the minimum dependency set, the tag and field misalignment phenomenon that occurs only in the combined scenario of power failure restart, recovery reporting, and tag table lag loading is continuously captured before the field is generated, i.e., in the first upload window covered by the power-on sequence: The first type is pre-upload with table not ready: when the readiness time difference is positive and the tag loading completion flag is false, real-time fields are released in advance; its characteristic performance is that the field is marked as real-time, but the path fingerprint has high similarity with the path fingerprint of the last power-on; The second type is recovery pre-coverage upload: when the proportion of recovery frame number to total frame number to be transmitted is high, and the cache queue prioritizes historical alarm fragments into the unpacking channel in a first-in-first-out manner, the recovery frame and the real-time frame are interleaved into the same channel within three sampling periods, and the frame sequence rearrangement degree exceeds the preset recovery threshold.

[0059] Among them, the recovery threshold is set to the average value of the absolute value of the difference between the field entry sequence number in the channel cache area and the recovery scheduling sequence number, which is considered to be excessive pre-coverage if the difference exceeds 3 sampling periods. The threshold is calculated through the 95% confidence interval of the historical upload sequence sorting stability distribution, and the default value is ±3 frames, which can be dynamically adjusted by configuration.

[0060] The third type is residual chain reuse mapping: when the tag version number has not been written into the current frame header, the path fingerprint reuses the channel path corresponding to the historical version; the characteristic is that the tag version difference is at least one and the path cross matching density exceeds the preset reuse threshold, which usually occurs in the transition period when the encryption session and subscription session have not completed rebinding.

[0061] The reuse threshold is set to 0.65 by default, which is set by the platform according to the coincidence ratio of the same channel field path fingerprint in multiple historical power-on cycles, indicating that when more than 65% of the fingerprint bits in the current field path coincide with the historical path, it is considered that there is a risk of historical tag chain adsorption. The threshold can be corrected according to the path mismatch rate in actual platform operation.

[0062] When any two types of the above three types of abnormalities occur simultaneously or any one type lasts more than two sampling periods in a power-on sequence, the abnormal combination state is used as the driving force, not based on the nominal channel attribution of the field, but according to the corresponding mapping between the ready time difference, the source state marker, the path fingerprint and the tag version difference, and the path cross density, a rule set is dynamically generated, including but not limited to: For the first type of situation, a minimum ready window is applied, wherein the minimum ready window length covers at least the time for loading the tag table, and the real-time field within the minimum ready window is only written into the shadow buffer and attached with a placeholder flag for waiting for the tag version, and does not enter the main parsing chain; For the second type of situation, backfilling and real-time decoupling are performed, specifically, a backfilling special marker is added to the backfilling field, its eligibility for strategy judgment is frozen, and a historical reference pointer is added to the path fingerprint, which is only used for trend hole filling and does not merge with the real-time main chain; For the third type of situation, a tag mutual exclusion table is enabled, wherein the channel path of the current tag version and the channel path of the historical tag version are in mutual exclusion, and when the mutual exclusion hits, the historical path adsorption is forced to be cut off, and the parsing priority of the field in the unpacking channel is downgraded to the secondary queue.

[0063] Further, the behavior marker is classified and defined based on the rule set: If the ready time difference is positive and the tag is not ready, and the path fingerprint is highly similar to the historical power-on path, it is marked as the first type of situation; The high similarity of the path fingerprint to the historical path is quantified by the cosine similarity, and the calculation method is as follows: ( · ) / (|| ||·|| ||); Wherein: represents the path fingerprint vector corresponding to the current field, which is composed of a tag version number, a channel number, and a upload time sequence number; represents the path fingerprint of the nearest field in the same channel in the last power-on event in history; ∈[0, 1], the greater the value, the stronger the path inheritance; When ≥ similarity threshold , it is determined as path inheritance behavior.

[0064] If the backfill proportion is high and the frame sequence rearrangement degree exceeds the backfill threshold, it is marked as backfill first type; If the tag version difference is large and the path intersection density exceeds the multiplexing threshold, it is marked as residual chain multiplexing type.

[0065] After classification, the fields are grouped according to terminal number, channel number, tag version number, power-on sequence number, and source state marker, to ensure that each group only contains field sets with consistent source and behavior structure, and unique path analysis, forming homogenous power-on sequence blocks.

[0066] Subsequently, the average ready time difference, path fingerprint similarity, frame sequence rearrangement degree, mutual exclusion hit rate, and backfill proportion of each power-on sequence block are extracted, and are dynamically aggregated in a sliding window manner to construct a sliding updated sequence feature vector set, which is used as an index reflecting the stability of the field cluster structure in the current power-on period.

[0067] Among them, the mutual exclusion hit rate is defined as follows: = / ; is the number of times that the field path fingerprint in the current sequence block hits the mutual exclusion control table; is the total number of fields participating in mutual exclusion determination in the sequence block; The path intersection density is defined as: =( )) / (N·(N-1)); represents the path fingerprint of field i and field j; δ( ) is the path fingerprint intersection density measurement function (such as the prefix matching ratio in path encoding); N is the number of fields in the current extraction window; The higher, the more significant the path drift phenomenon.

[0068] On this basis, the attribution path stability score of each field sample is calculated. The score is based on the following four indicators: frame sequence rearrangement degree, mutual exclusion hit rate, ready time difference and backfill proportion, and after giving weighted coefficients, weighted sum is carried out, which constitutes the field power-on period attribution credibility evaluation index. The score result is used to guide whether the field has the analysis qualification of entering the main chain structure object set, but does not constitute the only determination basis.

[0069] The field attribution path stability score It is calculated by the following weighted model: ; Among them: Indicates the frame sequence rearrangement degree of the field in the power-on sequence, and the calculation method is the normalized value of the minimum exchange distance of the unpacking sequence and the original channel queue; Indicates the mutual exclusion hit rate of the path to which the field belongs in the conflict mapping cluster, and the value range is [0, 1]; Indicates the time difference (unit: ms) between the ready time on the field and the time when the tag table is loaded; Given by the RTC (Real-Time Clock) timestamp recorded after the MCU system starts, and the tag loading completion time The internal event flag triggered after the field tag management thread completes the analysis and loading is automatically registered. Both are calibrated in a unified clock domain, and periodic NTP time calibration is performed through the heartbeat timestamp issued by the platform, so as to ensure that all ready time differences Have consistency across restart periods and platform traceability.

[0070] Indicates the proportion of the backfill frame number of the sequence block where the field is located to the total frame number to be transmitted; W1, W2, W3, W4 are the weighting coefficients configured by the platform, satisfying W1+W2+W3+W4=1, and the platform sets W1, W2, W3, W4 according to the historical path reconstruction accuracy maximization principle or the minimum mis-triggering principle of linkage strategy, which can be configured and learned to automatically adjust.

[0071] Specifically, in the extraction into the main chain link, it does not directly make fixed mapping according to the score, but synchronously refers to the conflict mapping cluster composed of mutual exclusion control hit track, path fingerprint cross density and frame sequence rearrangement clustering derived in the rule generation channel, and takes the conflict mapping cluster as the dynamic scheduling constraint of the current field extraction path. On the basis of field scoring, further dynamic calibration is carried out on the current extraction path: The path fingerprint is constructed as follows: for each frame of field data, 5 flag fields are extracted from the channel number, monitoring point ID, power-on sequence number, tag version number, and policy number to which the field belongs, which are sequentially spliced into a unique fingerprint string, and then subjected to SHA-1 hash compression to generate a fixed-length fingerprint digest; the digest is used to quickly identify whether the field attribution path is consistent with the historical path or there is an intersection overlap.

[0072] Fields with high scores and mutual exclusion of the sequence block: the system performs a tag locking strategy, that is, the current tag version number is written in the field frame header, and a path closure checking rule is set, and only when the channel numbers of two consecutive frames in the same sequence block are consistent and the difference in the ready time approaches zero, the field is allowed to be written into the main chain structure object set. At the same time, the path enables forward consistency convolution channel, enhances the path continuity perception ability in a short window, and improves the recognition accuracy of short period structure stability.

[0073] Fields with medium-low scores and no hit conflict mapping cluster: the system enables a flexible buffer binding mechanism, allowing the field to be delayed into the chain within a short tolerance window; when the tag table is loaded and the difference in the field ready time does not exceed the set tolerance threshold, the field is merged into the main chain structure object set to absorb the analysis deviation caused by the path jitter and short mapping empty window period in the initial power-on period.

[0074] In the current extraction round, the field path fingerprint has a high overlap rate with the conflict mapping cluster: the system starts the extraction strategy reverse compression mechanism, that is, it suspends the binding process dominated by the regular scoring priority, and switches to a path suppression strategy centered on mutual exclusion weight ordering. In this strategy, the tag dilution process is implemented for the hit fields in the conflict sequence, and the direct mapping behavior of the field to the main chain structure object is forcibly disconnected to prevent the historical tag chain from adsorbing the field, causing pollution of the main chain path and false triggering of platform linkage.

[0075] Finally, to ensure the auditability and state traceability of the process before the field enters the main chain, the field state flag encapsulation action is uniformly performed before the field is finally written into the main structure object set, including but not limited to: The power-on sequence number to which the field belongs; The version number used for tag mapping; The current field stability score value; Whether it hits the conflict mapping cluster; Whether it enters the rule adjustment channel and other control state bits.

[0076] Then the above field meta-state information is written into the field report frame tail metadata segment, and is used as an auxiliary criterion in the field path matching, strategy path selection, and error rollback logic on the platform side, to build a bidirectional self-consistent closed loop mechanism from field anomaly recognition, rule dynamic generation, extraction scheduling decision to main chain injection writing.

[0077] The platform monitoring module serves as a remote hub node, responsible for receiving, decoding, reviewing, and responding to various field data frames uploaded by the collection terminal, and through centralized processing, linkage control, and intelligent modeling analysis of electrical fire hazard data, realizes closed-loop management and control of the overall fire risk chain and improves the multi-system collaborative linkage capability.

[0078] Specifically, the platform monitoring module backend deploys a lightweight message middleware (MQTTBroker) and a RESTAPI service interface, wherein the former receives all data frames uploaded by the collection terminal at a QoS1 service level based on the MQTT protocol stack, and sends a receiving confirmation message to the terminal; the latter is used for platform internal management instruction issuance, remote configuration synchronization, and multi-system adaptation docking, forming a south-north bidirectional communication system of control and data channels.

[0079] Further, all received data frames are parsed and unpacked according to a unified structure, and the field content includes: monitoring point unique ID, signal type identification, numerical field, state bit, and timestamp metadata. After parsing, the data enters the real-time processing channel: Numerical field: written into a time series database for trend curve display, historical comparison analysis, and modeling use; Alarm field: written into a PostgreSQL relational database simultaneously, forming a complete alarm event traceability log, supporting subsequent correlation query, causal chain analysis, and false alarm correction; Field meta state: such as power-on serial number, tag version number, field stability score, hit conflict mapping cluster flag, whether to enter the adjustment channel, etc., which are used for dynamic rule discrimination and linkage strategy path selection mechanism through structured parsing, to ensure the consistency and integrity of the platform response after the field enters the main chain.

[0080] In the alarm state judgment process, according to the field alarm threshold rule table and the field tag path mapping cache table maintained synchronously with the field terminal, the field, path, and tag three consistency judgment mechanism is executed to ensure that the alarm event is established at the same time, and the path, channel, and tag binding relationship is accurate, avoiding the deviation of linkage response due to path drift.

[0081] Specifically, when the alarm judgment condition is established, the following four types of linkage response actions are immediately executed: Front-end interface response: the alarm state interface built by the Vue.js front-end framework, in the form of icon flickering and pop-up window, directly marks the device, channel, and specific event type that triggered the alarm; Duty staff push: call the message push engine, send APP internal notification and SMS prompt to the on-duty operation and maintenance personnel according to the event level, support alarm grading, personnel scheduling mapping, and alarm feedback; Downlink linkage control: Construct linkage control instruction frame and issue it to the designated collection terminal through the MQTT downlink channel to drive the local execution of linkage equipment actions such as power-off, exhaust, and fire extinguishing. The linkage instruction contains field tag path, policy number, device execution number, and other metadata. Context recording and behavior tracing: Record the context data snapshot when the field triggers an alarm, the policy matching process, the issued instruction content, and the terminal receipt status, and write them into the behavior log database as important basis for later tracing and mechanism self-correction.

[0082] Further, the platform front-end interface is built based on the Vue.js framework, and ECharts is used as the core graphics component, supporting multiple interactive view modes, including but not limited to: Regional map status view (device arming status indicated by building / floor / region); Device status list view (real-time state sorting / filtering by monitoring points); Trend curve review view (field historical value curve and threshold crossing point analysis); Historical event search panel (supports keyword / time period / tag path multi-condition filtering); Linkage strategy configuration area (supports linkage chain structure diagram visualization configuration and rule version management).

[0083] Subsequently, to ensure data access security and operation permission boundaries, the platform monitoring module configures the interface access permission, device data browsing, and strategy configuration permission based on the background user role model, ensuring that different operation and maintenance positions only access the data and control actions within their authorized range.

[0084] Further, to improve the perception and prediction ability of electrical fire evolution trend, the platform monitoring module deploys a model training engine in the backend, uses statistical learning methods and tree model algorithms to construct device health score models and degradation trend prediction models based on long-term running field sequence data. During the training process, the model input data is derived from the normalized field historical sample set in the TSDB, and the label data is derived from the historical alarm events and manual confirmation results. The model training task is scheduled and run by the platform backend.

[0085] After the model training is completed, the platform monitoring module solidifies the model structure and parameters into a version number model package and issues it to the collection terminal through the downlink, which is loaded and executed by the terminal local microcontroller.

[0086] Finally, in terms of operation and management capabilities, the platform monitoring module supports automatic aggregation of running data and generation of reports, can export running snapshots in PDF / Excel format, sets up a daily automatic backup path and backup cycle, and supports centralized execution of operation and maintenance operation instructions such as remote OTA configuration parameter issuance, system version upgrade, remote debugging log pulling, and significantly improves the platform remote maintainability and full life cycle management capabilities.

[0087] The above formulas are dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0088] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0089] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application of the technical solution and the constraints of the application. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0090] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0091] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0092] Finally, the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A wireless communication based electric fire alarm monitoring system characterized in that: The application relates to a field data acquisition terminal. The application comprises a sensing acquisition module, an alarm processing module, a wireless communication module, a power supply and backup module and a platform monitoring module, and the modules are connected through signals. The sensing acquisition module is used for configuring sensors and performing synchronous sampling through an analog signal conditioning circuit and a microcontroller ADC. The alarm processing module is used for filtering and processing sensing signal data after signal conditioning through an embedded microcontroller and performing alarm determination. In the wireless communication module, the acquisition terminal encapsulates encrypted data frames into MQTT message bodies through Wi-Fi or LoRa networking to report the platform monitoring module, and the data is cached and retransmitted during network interruption. In the power supply and backup module, the main power supply is output as a DC bus through a switching power supply, and the backup power supply is a storage battery group; when power failure occurs, the main and backup power supplies are switched and power failure is reported; meanwhile, a rule set is generated based on a power-on serial number, a tag version number and a path fingerprint during field recovery; a plurality of power-on sequence blocks are grouped and a sequence feature vector is constructed to calculate a field attribution path stability score; the field is extracted according to a conflict mapping cluster and is dynamically scheduled; and the field is written into a front encapsulation field meta state information. The platform monitoring module is used for receiving and analyzing data frames uploaded by the acquisition terminal, performing three-element consistency determination, generating linkage control instructions, delivering the instructions to the acquisition terminal and recording behavior logs.

2. The wireless communication based electric fire alarm monitoring system as claimed in claim 1, wherein: In the sensing acquisition module, a zero sequence current transformer, a Hall effect current sensor, an NTC type thermistor, a precision resistance array, an isolation amplifier, an integrated temperature and humidity sensor and a photoelectric smoke sensor are configured in the acquisition terminal. The analog signal conditioning circuit is used for sequentially buffering, amplifying and low-pass filtering the output signals of the sensors, and the conditioned signals are connected to the ADC module of the microcontroller to perform synchronous sampling; before the acquisition logic is started, sensor short-circuit and open-circuit detection is triggered, and a standard curve is called to calculate a range coefficient and a zero offset coefficient by using a least square fitting method.

3. The wireless communication based electric fire alarm monitoring system as claimed in claim 1, wherein: In the alarm processing module, an embedded microcontroller is used to sequentially read the sensing signal data of each channel in the ADC module of the microcontroller through a timing interrupt; a mean filtering algorithm is used for line residual current signals; and a median filtering algorithm based on singular value sorting is used for conductor temperature signals. A cyclic buffer is constructed and is updated in real time; alarm determination is performed according to a line residual current alarm threshold, a conductor temperature alarm threshold and a pre-stored disturbance waveform feature model in an arc fault recognition mode library.

4. The wireless communication based electric fire alarm monitoring system as claimed in claim 1, wherein: In the wireless communication module, a double-layer networking structure is adopted; in a short-distance communication scene, a Wi-Fi wireless communication module is embedded in the acquisition terminal to access a field gateway device and a local area network; in a long-distance or Wi-Fi access unavailable scene, a LoRa transceiver is embedded in the acquisition terminal to report the platform monitoring module through a LoRa gateway and a cellular network. The microcontroller is used to construct a unified structure data frame to perform AES-128 encryption and encapsulation into an MQTT message body to report the platform monitoring module; the data frame with an alarm state flag bit is enabled for ACK confirmation and automatic retransmission; meanwhile, data is cached during network interruption and is retransmitted after network recovery.

5. The wireless communication based electric fire alarm monitoring system as claimed in claim 1, wherein: In the power supply and backup module, the main power supply outputs a DC bus through an industrial-grade isolated switching power supply, and the backup power supply adopts a sealed lead-acid battery pack. When the main power supply fails, the switching circuit completes the main-backup switching, and cuts off unnecessary loads and reports power supply failure when the battery voltage is abnormal.

6. The wireless communication based electric fire alarm monitoring system as claimed in claim 5, wherein: In the power supply and backup module, before the pre-initialization stage of the field recovery logic completed by the backup power supply for power supply switching and starting, the power-on sequence number, tag version number, tag loading completion flag, recovery frame number proportion, and field upload readiness time and tag table readiness time difference are extracted from the power-on log and cache index, and a minimum dependency set is generated for each field recovery source state marker; Based on the minimum dependency set, three types of abnormalities, including lock table not ready pre-upload, recovery pre-coverage upload, and residual chain reuse mapping, are identified within the first upload window, and a rule set is dynamically generated according to the readiness time difference, source state marker, path fingerprint, and tag version difference to apply a power-on minimum readiness window to real-time fields, add a recovery dedicated marker to recovery fields and freeze the strategy judgment qualification, and forcibly cut off historical path adsorption through a tag exclusion table for residual chain reuse fields. Then, the fields are grouped by power-on sequence to form power-on sequence blocks of the same source and type.

7. The wireless communication based electric fire alarm monitoring system as claimed in claim 6, wherein: In the power supply and backup module, during the field recovery process, the terminal extracts the average readiness time difference, path fingerprint similarity, frame sequence rearrangement degree, exclusion hit rate, and recovery proportion of each group of power-on sequence blocks, constructs a sequence feature vector, and calculates the field attribution path stability score based on the weighted sum of the frame sequence rearrangement degree, exclusion hit rate, readiness time difference, and recovery proportion.

8. The wireless communication based electric fire alarm monitoring system as claimed in claim 7, wherein: In the power supply and backup module, when extracting fields from the main chain, the field extraction path is dynamically scheduled according to the conflict mapping cluster: fields with high field attribution path stability score and no hit exclusion are executed with a tag locking strategy, fields with low field attribution path stability score and no hit conflict mapping cluster are executed with a flexible buffer binding, fields that hit the conflict mapping cluster are executed with a path suppression strategy, and field meta-state information is encapsulated before writing the field to the main structure object set.

9. The wireless communication based electric fire alarm monitoring system as claimed in claim 1, wherein: In the platform monitoring module, a message middleware and a RESTAPI service interface are deployed to receive data frames uploaded by the collection terminal and send a reception confirmation, parse the monitoring point unique identifier ID, signal type identifier, value field, status bit, and timestamp, write the value field to a time series database, write the alarm field to a relational database, and perform field, path, and tag consistency determination. When the alarm condition is met, the MQTT is used to issue a linkage control instruction to the collection terminal, and context data is encapsulated and written to the behavior log database.

Images