Storage battery intelligent operation and maintenance system based on edge calculation model
The intelligent operation and maintenance system, which utilizes an edge computing model, collects and diagnoses battery data in real time. Combined with a cloud-based collaborative architecture, it solves the problems of delayed fault response and high costs in traditional operation and maintenance, achieving efficient and accurate fault identification and extended battery life.
Patent Information
- Application Number
- CN202511658543.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-02-03
AI Technical Summary
Traditional battery maintenance relies on manual inspections, which suffers from long inspection cycles, poor real-time performance, high data redundancy, and untimely fault detection. It cannot meet the demand for second-level response to emergency faults and lacks a proactive error correction mechanism, resulting in shortened battery life and increased maintenance costs.
The intelligent operation and maintenance system adopts an edge computing model, which includes a data acquisition layer, a high-speed collection layer, a field display layer, and a cloud layer. It uses sensor modules to collect data in real time, performs fault diagnosis through wavelet transform algorithms and random forest models, and realizes data aggregation and federated learning through a cloud-edge collaborative architecture, supporting local data storage and remote management.
It has reduced the fault diagnosis delay from 5 minutes to within 2 seconds, increased the fault identification accuracy to 95%, extended the battery pack life by 50%, reduced the operation and maintenance cost by 30%, and met the needs for rapid response to emergency faults.
Smart Images
Figure CN121460751A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of battery operation and maintenance technology, specifically a smart battery operation and maintenance system based on an edge computing model. Background Technology
[0002] With the widespread application of batteries in critical fields such as power, communication, and energy storage, the quality of their operation and maintenance directly determines the stability and safety of equipment systems. Traditional battery operation and maintenance largely relies on manual inspections, which suffers from problems such as long inspection cycles, poor real-time performance, and untimely fault detection. Specifically, these drawbacks include: the long cycle (typically 3-6 months / time) and poor real-time performance of traditional manual inspections, leading to delayed fault detection; the high data redundancy of centralized data processing architectures (with invalid data accounting for over 60%) and difficulties in selecting key parameters, failing to meet the second-level response requirements for emergency faults such as thermal runaway and open circuits; and the lack of proactive error correction mechanisms, failing to automatically intervene when batteries exhibit deterioration trends such as capacity decay and voltage imbalance, resulting in a 30%-50% reduction in battery life and an increase in operation and maintenance costs of over 40%. Therefore, there is an urgent need to build a smart operation and maintenance system that integrates edge computing to solve the technical problems of insufficient intelligence, delayed fault response, and excessively high maintenance costs associated with traditional operation and maintenance models. Summary of the Invention
[0003] In response to the problems pointed out in the background technology, the purpose of this invention is to provide a smart battery operation and maintenance system based on an edge computing model, which aims to solve the technical defects in existing battery operation and maintenance technologies, such as low level of intelligence (more than 70% of decisions are made manually), fault diagnosis accuracy of less than 85%, early warning response delay (average >5 minutes), and inability to achieve automatic error correction.
[0004] To achieve the above objectives, the present invention adopts the following technical solution: A smart battery operation and maintenance system based on an edge computing model includes a data acquisition layer, a high-speed collection layer, a field display layer, and a cloud layer. The data acquisition layer is deployed within 1 meter of the battery pack and includes a sensor module and an execution module. The sensor module includes a voltage resistance acquisition unit, a current sensor, and a temperature sensor, used to collect parameters of the battery, power supply, and environment. The execution module includes a charge / discharge regulation unit, a fault isolation unit, and a maintenance execution unit. The high-speed collection layer includes a convergence module; The on-site display layer includes an edge computing host built on an industrial-grade edge computing gateway. The edge computing host is equipped with a wavelet transform preprocessing algorithm and a random forest fault diagnosis model, and supports local data storage. The on-site display layer also includes a thermal runaway early warning module and an offline and open-circuit fault diagnosis module. The cloud layer includes a remote management platform; the remote management platform receives the structured data uploaded by the on-site display layer to achieve multi-site data aggregation display and iterative federated learning models; The data acquisition layer uses the sensor module to collect the voltage, internal resistance, charge and discharge current parameters of each single battery in real time, and receives the control instructions sent by the on-site display layer through the high-speed collection layer to drive the execution module to perform maintenance; the on-site display layer, as the local data processing core, receives the original data of the data acquisition layer through the high-speed collection layer. After noise reduction by the wavelet transform algorithm and feature parameter extraction, a random forest model is used for fault diagnosis, and control instructions are generated and fed back to the execution unit of the data acquisition layer. At the same time, the compressed structured data is uploaded to the cloud layer.
[0005] The number of battery packs is n groups, and each battery pack consists of m single batteries. Both m and n are positive integers greater than or equal to 1. The corresponding data acquisition layer includes n groups of sensor modules and n groups of execution modules. The n groups of battery packs are respectively connected to the user load through the power supply, and the n groups of battery packs are respectively connected to n groups of sensor modules and n groups of execution modules.
[0006] The thermal runaway warning module realizes warning by calculating the temperature rise coefficient KT = ΔT / Δt (°C / min) and the floating charge current increase coefficient KI = It / I0. When KT ≥ 5 °C / min or KI ≥ 1.5, a first-level alarm is triggered.
[0007] The off-line and open-circuit fault diagnosis module runs the following strategy: when the floating charge current of the battery pack is zero and the voltage of all single batteries is less than the set first threshold, an off-line alarm is triggered; when the floating charge current of the battery pack is zero, and there is any single battery voltage greater than the second threshold, and at the same time the voltages of the remaining single batteries are all less than the set first threshold, an open-circuit alarm is triggered.
[0008] The execution module uses an automatic equalization maintenance process to perform equalization maintenance on each battery pack. The specific automatic equalization maintenance process is as follows: collect the voltage Ui of the single battery, calculate the standard deviation σ = √[Σ(Ui - Uavg)² / m], Uavg is the average voltage of the battery pack, and m is the number of batteries in the battery pack; when it is detected that the voltage Ui of the single battery < Uavg - 3σ, start a small current compensation charge of 0.3C for the single battery until the voltage balance degree σ ≤ 20 mV.
[0009] The cloud layer adopts a MySQL + InfluxDB hybrid database architecture and receives the structured data uploaded by the edge layer through the MQTT protocol (QoS = 1).
[0010] The voltage resistance acquisition device of the data acquisition layer adopts a three-terminal measurement method, supports a voltage range of 0-16V (accuracy ±0.5%FS) and an internal resistance range of 0-200mΩ (test current 100mA ±5%), and a sampling frequency ≥1kHz.
[0011] The fault isolation unit uses a dual-coil magnetic latching relay with a contact capacity of ≥30A / DC60V, supports remote opening and closing control, and has a mechanical life of ≥100,000 cycles and an electrical life of ≥10,000 cycles.
[0012] The power supply includes a backup power supply and a power conversion module. One end of the backup power supply is connected to the user load, and the other end of the backup power supply is connected to the battery pack through the power conversion module.
[0013] Compared with existing technologies, the beneficial effects of this invention are as follows: Through localized processing at edge computing nodes, the delay in issuing fault diagnosis and maintenance commands is reduced from 5 minutes in traditional centralized architectures to within 2 seconds, meeting the rapid response requirements for emergency faults such as thermal runaway; by integrating multi-parameter feature libraries and machine learning algorithms, the fault identification accuracy is increased from 85% in existing technologies to over 95%, especially improving the identification rate of early faults such as micro-short circuits and latent capacity decay by 40%; through online automatic balancing maintenance, the cycle life of the battery pack is extended by 50%, reducing the need for complete pack replacement; the technology of mixing new and old batteries reduces the procurement cost of a single battery pack by 30%, and reduces annual maintenance manpower costs by 60 person-days per site; the cloud-edge collaborative architecture reduces transmission bandwidth requirements; and cloud-based global analysis enables dynamic optimization of maintenance strategies, reducing the average number of battery failures per year from 3 times per pack to 0.5 times per pack. Attached Figure Description
[0014] Figure 1 This is a system architecture diagram of the present invention.
[0015] Figure 2 This is a system architecture diagram for Example 1.
[0016] Figure 3 This is a system power supply diagram for Example 1.
[0017] Figure 4 This is a system communication diagram for Example 1.
[0018] Figure 5 This is a flowchart of the balanced maintenance process of the present invention. Detailed Implementation
[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. It should be noted that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0020] like Figure 1-5 As shown, Embodiment 1 provides a smart battery operation and maintenance system based on an edge computing model, including: The data acquisition layer, deployed within 1 meter of the battery pack, adopts a distributed architecture design and includes sensor modules and execution modules. The sensor modules integrate a voltage sensor (measurement accuracy ±0.5mV), an internal resistance acquisition unit (test range 0-50mΩ), a Hall current sensor (range ±100A), and an infrared temperature sensor (temperature range -40℃~125℃, accuracy ±0.5℃), enabling synchronous acquisition of battery cell voltage, internal resistance, total voltage, charging and discharging current, terminal temperature, and ambient temperature and humidity (sampling frequency 1Hz). The execution modules include: a charge and discharge regulation unit (supporting 0-5A constant current regulation), an electromagnetic fault isolation switch (response time <10ms), and a maintenance execution unit (compensation current continuously adjustable from 0.1-2A). The high-speed collection layer includes a convergence module, which is integrated inside the battery monitoring data collection device. The on-site display layer includes edge computing hosts built on industrial-grade edge computing gateways (CPU frequency ≥ 1.2GHz, memory ≥ 2GB). The edge computing hosts have the following functions: Data preprocessing: The original collected data is denoised by wavelet transform algorithm, and 12 key feature parameters such as voltage fluctuation rate, temperature gradient, and internal resistance change rate are extracted, with a data compression ratio of 10:1. Real-time analysis: A feature library containing 23 fault modes (such as thermal runaway feature value: ΔT / Δt>5℃ / min) is embedded, and a random forest algorithm is used for fault diagnosis (accuracy ≥92%), with a diagnosis delay of <2 seconds; Decision execution: Based on the diagnostic results, maintenance instructions are generated and sent to the execution module via the Modbus-RTU protocol, with an instruction response time of <500ms; (4) Local storage: Configure an 8GB industrial-grade SD card to store 3 months of key data (abnormal parameters, fault records, maintenance logs) in a loop.
[0021] The cloud layer includes a remote management platform. The edge computing host adopts a cloud-edge collaborative architecture, uploading structured data (≤50MB / site per day) processed by the field display layer to the remote management platform via the MQTT protocol (QoS=1) to achieve distributed storage. The remote management platform adopts a MySQL+InfluxDB hybrid database architecture, supporting concurrent writing of ≥100,000 nodes. The remote management platform can achieve global monitoring, realize multi-site data aggregation and display through a B / S architecture, and support GIS map positioning and device status visualization. Based on accumulated operation and maintenance data, the remote management platform updates the fault diagnosis model of the field display layer through federated learning, with a model iteration cycle of 3 months.
[0022] In the execution module of the data acquisition layer of Example 1, the charge and discharge adjustment unit adjusts the maintenance voltage of a single battery cell according to the instructions generated by the edge computing host to avoid long-term operation of undercharged batteries. The maintenance cycle and operation parameters are dynamically adjusted by the edge computing host according to the battery status. By supplementing the lagging batteries with low voltage and small current, the voltage of the lagging batteries is charged to the average level. At the same time, the voltage of overcharged batteries will automatically drop back to the average level to prevent the batteries in the group from being in an overcharged or undercharged state for a long time.
[0023] Based on the characteristics of battery temperature gradually rising, the rate of temperature rise accelerating, and the float charge current gradually increasing during battery thermal runaway, the thermal runaway early warning module of the field display layer in Example 1 incorporates a battery thermal runaway early warning model to achieve accurate early warning of battery pack thermal runaway. The specific early warning mechanism is as follows: When the battery pack is in float charge mode, if any single battery in the battery pack triggers a temperature alarm, the system will automatically retrieve the temperature and float charge current monitoring data of the single battery within the previous 2 hours (the time period can be set) and automatically calculate the temperature rise coefficient and the float charge current increase coefficient. When either the temperature rise coefficient or the float charge current increase coefficient triggers a temperature alarm, the system will automatically calculate the temperature rise coefficient and the float charge current increase coefficient. When the temperature rise coefficient and float charge current boost coefficient reach the set threshold (the alarm threshold can be set in the system alarm threshold, usually the system recommended value is used), the system will automatically issue a battery thermal runaway alarm; when the battery pack is in the charging and discharging state, if the temperature of any single battery in the battery pack alarms, the system will automatically retrieve the temperature of the single battery in the previous 10 minutes (the time period can be set) and automatically calculate the temperature rise coefficient. When the temperature rise coefficient reaches the set threshold (the temperature rise coefficient can be set in the system alarm threshold, usually the system recommended value is used), the system will automatically issue a battery thermal runaway alarm.
[0024] The offline and open - circuit fault diagnosis module of Embodiment 1 runs the following strategies: When the floating charge current of the battery pack is zero and the voltages of all individual batteries are less than the set first threshold, an offline alarm is triggered; When the floating charge current of the battery pack is zero, there is any individual battery voltage greater than the second threshold (the second threshold is greater than the first threshold), and the voltages of the remaining individual batteries are all less than the set first threshold, an open - circuit alarm is triggered.
[0025] As Figure 5 shown, the execution module adopts an automatic equalization maintenance process to perform equalization maintenance on each battery pack. The specific automatic equalization maintenance process is as follows: Collect the voltage Ui of the individual battery, calculate the standard deviation σ = √[Σ(Ui - Uavg)² / m], where Uavg is the average voltage of the battery pack and m is the number of batteries in the battery pack; When it is detected that the voltage of the individual battery Ui < Uavg - 3σ, start a small - current compensation charge of 0.3C for this individual battery until the voltage equalization degree σ ≤ 20mV.
[0026] As Figure 1 shown, the power supply of the intelligent battery operation and maintenance system of Embodiment 1 includes a backup power supply and a power conversion module. One end of the backup power supply is connected to the user load, and the other end of the backup power supply is connected to the battery pack through the power conversion module.
[0027] As Figure 2 shown, in the intelligent battery operation and maintenance system of Embodiment 1, the remote management platform includes a local terminal server and a mobile phone APP. The remote management platform supports remote configuration and status monitoring of the edge computing host, sensors, and execution modules; The remote management platform automatically generates battery operation reports, maintenance record reports, and fault diagnosis reports, providing data support for operation and maintenance decision - making.
[0028] As Figure 3 shown, in Embodiment 1, the edge computing host, sensor module, and execution module are all powered by the backup power supply.
[0029] As Figure 4 shown, in Embodiment 1, the sensor module and the execution module corresponding to each individual battery are integrated into an integrated module. Multiple integrated modules are sequentially connected for communication in accordance with the numbering order, and finally communicate with the edge computing host through the corresponding convergence module. The edge computing host communicates with the remote management platform through a switch in an interactive manner.
[0030] The intelligent battery operation and maintenance system provided by Embodiment 1 has the following core functions: 1) Online monitoring: Real - time collection of the voltage, negative - pole temperature, internal resistance of the battery, the total voltage of the battery pack, ambient temperature, charge - discharge current.
[0031] 2) Multi-parameter synchronous monitoring: Using time-division multiplexing technology, the system can simultaneously collect data on individual battery voltage (resolution 0.1mV), internal resistance (test accuracy ±2%), negative terminal temperature (sampling interval 10s), total battery voltage, charging and discharging current (range -200A~+100A), and ambient temperature and humidity (sampling interval 1min). The data upload delay is ≤200ms.
[0032] 3) Automatic maintenance: This includes a charge / discharge regulation unit, a fault isolation unit, and a maintenance execution unit. The charge / discharge regulation unit adjusts the maintenance voltage of individual batteries according to the instructions generated by the edge computing host to prevent undercharged batteries from operating for a long time. The maintenance cycle and operating parameters are dynamically adjusted by the edge computing host according to the status of the individual battery. By providing low-current supplementary charging to lagging batteries, the voltage of lagging batteries is charged to the average level. At the same time, the voltage of overcharged batteries will automatically drop back to the average level to prevent the batteries in the group from being in an overcharged or undercharged state for a long time.
[0033] 4) Adaptive Maintenance Execution: Based on the diagnostic results displayed on-site, the execution unit can automatically trigger three types of maintenance actions: A. Equalization charging: For individual batteries whose voltage is 90% lower than the average value in the group, start 0.3C small current compensation charging (cutoff voltage 13.5V). B. Overcharge protection: When the voltage of a single battery cell is detected to be >14.4V, bypass discharge is triggered (discharge current 0.5A). C. Fault Isolation: In the event of an internal short circuit (internal resistance <5mΩ) or thermal runaway (temperature >85℃), the faulty battery will be disconnected from the main circuit within 300ms.
[0034] In the intelligent battery operation and maintenance system provided in Example 1, the edge computing host also integrates a data visualization module, an equipment management module, a fault alarm module, a fault analysis and diagnosis module, a data storage module, a consistency calculation module, and a report generation module. The data visualization module supports line charts / bar charts / heatmaps. The equipment management module supports SN code binding and lifecycle management. The fault alarm module adopts a three-level alarm (early warning / alarm / emergency, with adjustable response priority). The fault analysis and diagnosis module has 18 built-in fault tree analysis models. The data storage module supports historical data traceability (supports data query within 1 year, with data granularity of 1 minute). The consistency calculation module performs consistency evaluation (calculates voltage standard deviation σ≤50mV). The report generation module supports Excel / PDF export.
[0035] The edge computing host receives battery operation data and status information uploaded by the acquisition layer, and displays the real-time changing trends of parameters such as voltage, current, and temperature of each battery pack in the form of charts; it intuitively displays the consistency of the battery packs, including deviation values and dispersion percentages; through comprehensive analysis of abnormal parameters, operating trends, and environmental parameters of battery operation, it performs fault diagnosis by comparing data with a bit database; the edge computing host judges charging and discharging characteristics, completes automatic retrieval of charging and discharging periods, and judges whether the battery capacity is qualified by comparing information such as unit time voltage difference, charging and discharging current, capacity, and temperature, and lists the lagging batteries.
[0036] The parts of this invention not described in detail are prior art.
Claims
1. A smart battery operation and maintenance system based on an edge computing model, comprising a data acquisition layer, a high-speed collection layer, a field display layer, and a cloud layer, characterized in that: The data acquisition layer is deployed within 1 meter of the battery pack, and includes a sensor module and an execution module; the sensor module includes a voltage and internal resistance collector, a current sensor, and a temperature sensor, which are used to collect parameters of the battery, power supply, and environment; the execution module includes a charge and discharge regulation unit, a fault isolation unit, and a maintenance execution unit; The high-speed collection layer includes a convergence module; The on-site display layer includes an edge computing host built based on an industrial-grade edge computing gateway. The edge computing host is configured with a wavelet transform preprocessing algorithm and a random forest fault diagnosis model, and supports local data storage; the on-site display layer also includes a thermal runaway warning module and an off-line and open-circuit fault diagnosis module; The cloud layer includes a remote management platform; the remote management platform receives the structured data uploaded by the on-site display layer, and realizes multi-site data aggregation display and federated learning model iteration; The data acquisition layer collects the voltage, internal resistance, charge and discharge current parameters of individual batteries in real time through the sensor module, and receives the control instructions sent by the on-site display layer through the high-speed collection layer to drive the execution module to perform maintenance; as the local data processing core, the on-site display layer receives the original data of the data acquisition layer through the high-speed collection layer, performs noise reduction and feature parameter extraction through the wavelet transform algorithm, uses the random forest model for fault diagnosis, generates control instructions and feeds them back to the execution unit of the data acquisition layer, and at the same time uploads the compressed structured data to the cloud layer.
2. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The number of the battery packs is n groups, and each group of battery packs consists of m individual batteries. Both m and n are positive integers greater than or equal to 1. The corresponding data acquisition layer includes n groups of sensor modules and n groups of execution modules. The n groups of battery packs are respectively connected to the user load through the power supply, and the n groups of battery packs are respectively connected to the n groups of sensor modules and the n groups of execution modules.
3. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The thermal runaway warning module realizes warning by calculating the temperature rise coefficient KT = ΔT / Δt (°C / min) and the floating charge current increase coefficient KI = It / I0. When KT ≥ 5 °C / min or KI ≥ 1.5, a first-level alarm is triggered.
4. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The off-line and open-circuit fault diagnosis module runs the following strategy: when the floating charge current of the battery pack is zero and the voltages of all individual batteries are less than the set first threshold, an off-line alarm is triggered; when the floating charge current of the battery pack is zero, there is any individual battery voltage greater than the second threshold, and the voltages of the remaining individual batteries are less than the set first threshold, an open-circuit alarm is triggered.
5. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The execution module adopts an automatic equalization maintenance process to perform equalization maintenance on each group of battery packs. The specific automatic equalization maintenance process is as follows: collect the voltage Ui of individual batteries, calculate the standard deviation σ = √[Σ(Ui - Uavg)² / m], Uavg is the average voltage of the battery pack, and m is the number of batteries in the battery pack; when it is detected that the voltage Ui of an individual battery < Uavg - 3σ, start a small current compensation charge of 0.3C for this individual battery until the voltage balance degree σ ≤ 20 mV.
6. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The cloud layer adopts a MySQL + InfluxDB hybrid database architecture and receives the structured data uploaded by the edge layer through the MQTT protocol (QoS = 1).
7. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The voltage resistance acquisition device of the data acquisition layer adopts a three-terminal measurement method, supports a voltage range of 0-16V (accuracy ±0.5%FS) and an internal resistance range of 0-200mΩ (test current 100mA ±5%), and a sampling frequency ≥1kHz.
8. The intelligent operation and maintenance system for batteries based on an edge computing model according to claim 1, characterized in that: The fault isolation unit uses a dual-coil magnetic latching relay with a contact capacity of ≥30A / DC60V, supports remote opening and closing control, and has a mechanical life of ≥100,000 cycles and an electrical life of ≥10,000 cycles.
9. A smart battery operation and maintenance system based on an edge computing model according to claim 1, characterized in that: The power supply includes a backup power supply and a power conversion module. One end of the backup power supply is connected to the user load, and the other end of the backup power supply is connected to the battery pack through the power conversion module.