A flexible cooperative device for micro-grid AI regulation

Abstract

The present application relates to the technical field of micro-grid intelligent regulation and control equipment, in particular to a flexible collaborative device for micro-grid AI regulation and control, which is a primary and secondary deep fusion device, comprising a power conversion unit, a heterogeneous dual-core main control unit, an AI edge regulation and control unit, a multi-protocol compatible communication unit, a direct current mutual aid interface unit, a hardware level safety protection unit and a human-computer interaction unit; the power conversion unit is a primary part, which realizes bidirectional power conversion and power quality management function between direct current energy storage battery and alternating current power grid; the secondary part of the device integrates a high-precision synchronous sampling module to provide data support for control and decision, which can replace the combination of multiple traditional devices to realize multiple functions such as energy storage access, power quality management, interconnection between transformer areas, fault transfer supply and the like, greatly reducing the deployment cost and floor area, and solving the pain points of single function, high deployment cost and difficult popularization of the existing flexible interconnected device of transformer area.

Description

Technical Field

[0001] This invention relates to the field of intelligent control equipment technology for microgrids, and specifically to a flexible collaborative device for AI control of microgrids. Background Technology

[0002] my country's new power system construction has entered a period of accelerated development, with high proportions of distributed renewable energy and flexible load integration becoming typical characteristics of low-voltage distribution networks. As the core nodes at the end of distributed energy integration, distribution network substations face core problems in traditional operation modes, such as voltage exceeding limits, three-phase imbalance, reverse power flow, power quality degradation, inability to mutually support energy between substations, and insufficient fault transfer capacity, making them difficult to adapt to the operational needs of the end of the new power system.

[0003] In current technologies, energy storage devices at the distribution substation level generally suffer from asset binding issues due to deep coupling between batteries, BMS, and power electronic converters, preventing users from freely choosing or changing battery brands; long communication links and inconsistent protocols lead to delayed dispatch response, making it difficult to meet the grid's rapid adjustment needs; and the inconsistent functions of BMS result in dispersed safety risks, making direct grid supervision impossible.

[0004] Existing flexible interconnection devices for distribution substations generally suffer from high cost, large size, and limited functionality. They can only achieve power transmission between substations and lack integrated functions such as energy storage access, grid control, and AI-driven autonomous regulation, making it difficult to meet the "observable, measurable, adjustable, and controllable" requirements of new distribution networks. Furthermore, existing devices have insufficient regulation capabilities, relying heavily on rigid regulation strategies with preset fixed thresholds. This makes them unable to adapt to dynamic changes in substation loads, resulting in low energy utilization efficiency and insufficient operational stability in scenarios with high proportions of renewable energy integration. They also fail to achieve flexible energy exchange and fault emergency coordination across substations, making it difficult to support the efficient and safe operation requirements of the new power system's end points.

[0005] Therefore, there is an urgent need to develop a flexible collaborative device for AI-controlled microgrid regulation to solve the above problems. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention proposes a flexible collaborative device for AI control of microgrids. Through the collaborative design of primary and secondary fusion architecture, AI adaptive control and flexible interconnection of distribution areas, it replaces the multi-device combination scheme, adapts to the "four-fold" management requirements of the power grid, and comprehensively improves the operational resilience of the distribution network.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A flexible collaborative device for AI-driven microgrid regulation is a deeply integrated primary and secondary system, comprising a power conversion unit, a heterogeneous dual-core main control unit, an AI edge control unit, a multi-protocol compatible communication unit, a DC mutual aid interface unit, a hardware-level security protection unit, and a human-machine interaction unit. The power conversion unit, the primary component, enables bidirectional power conversion and power quality management between DC energy storage batteries and the AC grid. The secondary component integrates a high-precision synchronous sampling module to provide data support for control and decision-making. The heterogeneous dual-core main control unit employs a heterogeneous architecture with a communication management core and a real-time control core, respectively handling non-real-time and hard real-time control tasks. The AI ​​edge control unit incorporates load and photovoltaic prediction models and a dynamic scheduling optimization model to achieve adaptive regulation across multiple scenarios. The multi-protocol compatible communication unit enables secure communication with the grid master station and various distributed energy devices. The DC mutual aid interface unit enables flexible interconnection and power interaction between adjacent transformer areas of the same model. The hardware-level security protection unit provides hardware-level rapid protection. The human-machine interaction unit enables local and remote operation and maintenance functions. It can replace the multiple functions of traditional multi-unit equipment combination, such as energy storage access, power quality management, inter-unit mutual assistance, and fault transfer, significantly reducing deployment costs and floor space, and solving the pain points of existing flexible interconnection devices in transformer substations, such as single function, high deployment cost, and difficulty in promotion.

[0009] In some specific implementations, the heterogeneous dual-core main control unit includes a communication management core module and a real-time control core module. The communication management core module runs a domestically developed operating system and undertakes non-real-time tasks such as northbound communication, AI strategy calculation, and OTA upgrades. The real-time control core module runs an RTOS real-time operating system and focuses on hard real-time functions such as VSG algorithm execution, high-precision AC sampling, high-frequency current loop control, and hardware protection. The two cores interact with each other through a high-speed shared memory mechanism, and scheduling commands can be converted into precise power output within 10ms. This achieves the synergy between hard real-time control and AI edge computing, ensuring millisecond-level response of network control and protection actions while also possessing AI adaptive regulation capabilities. The scheduling command response time is ≤10ms, effectively solving the problem of the inability to balance real-time performance and computing power in traditional single-core architectures, and significantly improving the device's regulation response speed and operational stability.

[0010] In some specific implementations, the AI ​​edge control unit incorporates an LSTM load and photovoltaic output prediction model and a deep reinforcement learning dynamic scheduling optimization model. The LSTM prediction model can dynamically predict load and photovoltaic output at different time scales by integrating historical operating data, real-time sensing data, and meteorological data, with a prediction accuracy of no less than 95%. The deep reinforcement learning optimization model can dynamically optimize charging and discharging, mutual assistance, and power quality management strategies based on prediction results, real-time operating status, and scheduling requirements, achieving multi-objective adaptive control. The strategy priority can be dynamically adjusted according to user needs, enabling local autonomous operation without relying on upper-level scheduling instructions.

[0011] In some specific implementations, the power conversion unit incorporates a grid-type virtual synchronous machine algorithm, possessing functions such as autonomous voltage build-up, virtual inertia, black start, primary frequency regulation, and seamless switching between grid and off-grid operation. When the virtual synchronous machine algorithm is always active, the real-time control core module executes the VSG rotor equation every 100μs to output a virtual electromotive force as a voltage loop reference, controlling the power electronic converter to output the corresponding voltage waveform. This can provide voltage and frequency support in weak grid scenarios, and the voltage qualification rate of distribution areas in high-proportion renewable energy access scenarios can be increased to over 99.9%.

[0012] In some specific implementations, the DC power exchange interface unit supports flexible interconnection of the same type of equipment in adjacent distribution areas, enabling energy exchange, fault transfer, and SOC balancing between distribution areas. When a power exchange command is received or a power exchange requirement is detected locally, the communication management core module first exchanges status information with neighboring equipment, calculates the optimal power command based on AI strategies, and controls the power conversion unit to execute the corresponding power transmission in real time. The fault transfer response time is less than 20ms, enabling uninterrupted power supply to critical loads and improving the power supply reliability of the distribution area to 99.99%.

[0013] In some specific embodiments, the power conversion unit has power quality management functions, including low voltage management at the transformer substation end, photovoltaic overvoltage and reverse power flow management, and three-phase imbalance management. When voltage exceeding limits at the transformer substation end, reverse power flow overvoltage, or three-phase imbalance exceeding the threshold is detected, the real-time control core module independently controls the active and reactive power output of each phase to stabilize the voltage and imbalance within the qualified range. The three-phase imbalance can be reduced to less than 1%, effectively reducing transformer losses in the transformer substation, extending the service life of the transformer, and increasing the proportion of distributed photovoltaic power consumption.

[0014] In some specific implementations, the multi-protocol compatible communication unit supports power grid standard protocols such as IEC 60870-5-104, IEC61850, and MQTT over TLS, and integrates the national cryptographic SM2 / SM4 security chip to achieve secure communication with the dispatch master station, IoT platform, edge collaborative controller, smart converged terminal of distribution area, and distributed energy equipment. It meets the power grid's four requirements for distributed energy: "observable, measurable, adjustable, and controllable," with a telemetry refresh rate of not less than 1Hz and a dispatch command response time of no more than 10ms.

[0015] In some specific implementations, the hardware-level safety protection unit integrates multiple protection functions, including short-circuit protection, insulation monitoring protection, over / under voltage protection, over / under frequency protection, reverse polarity protection, phase sequence protection, phase loss protection, overload protection, anti-islanding protection, and battery emergency stop safety circuit protection. The battery emergency stop safety circuit signal directly triggers the hardware-level power switching device to lock down without going through software, with a response time of less than 1ms. Key protection logic is offloaded to hardware and does not rely on software operation, significantly improving the operational safety of the device.

[0016] In some specific implementations, the device supports multiple energy management modes, including peak shaving and valley filling, photovoltaic consumption, economic operation, and three-phase imbalance control strategies in the context of distribution area energy storage, as well as collaborative peak shaving, emergency mutual assistance, SOC balancing, and cross-distribution area voltage collaborative support strategies in the context of distribution area mutual assistance. All strategies have SOC safety boundary verification, priority management, and anti-frequent switching mechanisms to ensure the safe and stable operation of the device, and the energy utilization efficiency is improved by more than 30% compared with the traditional rigid control strategy.

[0017] In some specific embodiments, the device adopts an integrated modular design, with a total weight not exceeding 35kg. It supports both wall-mounted and pole-mounted installation methods. The device's protection rating is no less than IP54 for indoor models and no less than IP65 for outdoor models. The operating temperature range is -40℃ to 70℃, the design life is no less than 20 years, and the mean time between failures (MTBF) is no less than 50,000 hours. The device supports multiple energy management modes, including peak shaving and valley filling, photovoltaic consumption, economic operation, and three-phase imbalance management strategies in transformer substation energy storage scenarios, as well as collaborative peak shaving, emergency mutual assistance, SOC balancing, and cross-transformer voltage collaborative support strategies in transformer substation mutual assistance scenarios. All strategies have SOC safety boundary verification, priority management, and anti-frequent switching mechanisms.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. This invention adopts a modular integrated design, and a single device can replace multiple traditional devices to realize multiple functions such as energy storage access, power quality management, inter-station mutual assistance, and fault transfer. The whole machine weighs only 35kg and can be installed by a single person, which greatly reduces deployment costs and floor space, improves functional integration, and solves the problems of single function and high deployment cost of existing devices.

[0020] 2. It adopts a heterogeneous dual-core architecture with a scheduling command response time of ≤10ms, taking into account both hard real-time control and AI edge computing capabilities. It has built-in dual AI models of LSTM prediction and deep reinforcement learning, which have high source load prediction accuracy and improve energy utilization efficiency compared with traditional rigid control strategies. It integrates a grid-type virtual synchronous machine algorithm, which has excellent control performance and is adapted to the dynamic source load fluctuation characteristics and weak grid operation requirements.

[0021] 3. Equipped with an independent DC mutual assistance interface, it breaks through the limitations of traditional independent operation of distribution areas, enabling second-level power support between adjacent distribution areas, with a fault transfer response time of ≤20ms, ensuring uninterrupted power supply to important loads, improving the reliability of power supply in distribution areas, and significantly enhancing the emergency response capability of the distribution network at the end of the fault.

[0022] 4. The power conversion unit has full-scenario power quality management function, which can specifically solve typical problems such as low voltage, overvoltage, reverse power transmission, and three-phase imbalance in the transformer area. After management, the three-phase imbalance is reduced to less than 1%, effectively reducing transformer losses, increasing the proportion of distributed photovoltaic power consumption, and improving power quality.

[0023] 5. Supports multi-scenario mode adaptation and high reliability design, requires no additional civil engineering modifications, has a design life of ≥10 years, and an average mean time between failures of ≥50,000 hours; multi-protocol compatibility and national cryptographic encryption design meet the four requirements of the power grid, can be used as a virtual power plant node to participate in the power auxiliary service market, shortens the investment return cycle, and significantly reduces the full life cycle operation and maintenance costs.

[0024] 6. The hardware-level multi-layer security protection design pushes key protection logic down to the hardware, with a protection action response time of ≤1ms and the security loop does not rely on software operation; the integrated national cryptographic SM2 / SM4 security chip realizes full-link encrypted communication, which fully meets the stringent requirements of power grid security protection and greatly improves operational security. Attached Figure Description

[0025] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings:

[0026] Figure 1 This is a schematic diagram of the overall hardware architecture of the flexible collaborative device for AI regulation of microgrids according to the present invention.

[0027] Figure 2 This is a schematic diagram of the inter-station flexible collaborative mutual assistance logic of the flexible collaborative device for microgrid AI regulation according to the present invention;

[0028] Figure 3 This is an AI edge control logic block diagram of the flexible collaborative device for AI control of microgrids according to the present invention;

[0029] Figure 4 This is a schematic diagram illustrating an application scenario of Embodiment 1 of the flexible collaborative device for AI control of microgrids according to the present invention.

[0030] Figure 5 This is a hardware framework block diagram of Embodiment 1 of the flexible collaborative device for AI regulation of microgrids according to the present invention;

[0031] Figure 6 This is a block diagram of the power electronic converter portion of Embodiment 1 of the flexible collaborative device for AI regulation of microgrids according to the present invention. Detailed Implementation

[0032] 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0033] The materials, methods, and embodiments described herein are exemplary and should not be construed as limiting unless otherwise stated.

[0034] Example 1

[0035] This embodiment provides a flexible collaborative device for AI-driven microgrid regulation. It is an integrated standard model with a rated power of 100kW, supporting up to 120kWh lithium iron phosphate battery packs and adaptable to 380V low-voltage distribution substation deployment. The hardware architecture includes a power conversion unit, a heterogeneous dual-core main control unit, an AI edge control unit, a multi-protocol compatible communication unit, a DC mutual aid interface unit, a hardware-level safety protection unit, and a human-machine interaction unit.

[0036] The power conversion unit adopts a deep integration design of primary and secondary sides: the primary side includes an AC / DC converter, a DC / DC converter, a bidirectional AC / DC conversion module, and an LCL filter, enabling efficient bidirectional power conversion between DC energy storage batteries and the 380V AC grid, supporting independent active / reactive power regulation, with a power conversion efficiency of ≥96.8%; the secondary side integrates a 10kHz high-precision synchronous sampling module, which can synchronously collect full data of three-phase AC voltage and current, providing data support for real-time control and AI decision-making. It incorporates a grid-type virtual synchronous machine algorithm, featuring autonomous voltage build-up, virtual inertia, black start, primary frequency regulation, and seamless switching between grid and off-grid operations. The real-time control core performs VSG rotor equation calculations every 100μs, providing voltage and frequency support in weak grid scenarios. It also possesses full-scenario power quality management functions, including low voltage management at the distribution terminal, photovoltaic overvoltage and backfeed management, and three-phase imbalance management, controlling the three-phase imbalance to within 1%.

[0037] The heterogeneous dual-core main control unit adopts a heterogeneous architecture of a communication management core and a real-time control core. The communication management core runs a domestic embedded Linux operating system and undertakes non-real-time tasks such as northbound communication scheduling, AI strategy calculation, and OTA remote upgrade. The real-time control core runs an RTOS operating system and focuses on hard real-time functions such as VSG network algorithm execution, high-precision AC sampling and parsing, high-frequency current loop control, and hardware protection triggering. The two cores interact with each other through a high-speed shared memory mechanism between the cores. The response time from receiving the scheduling command to converting it into power output is ≤10ms.

[0038] The AI ​​edge control unit incorporates an LSTM load and photovoltaic output prediction model and a deep reinforcement learning dynamic scheduling optimization model. It can integrate historical operating data, real-time sensing data, and external meteorological data to dynamically predict the load and distributed photovoltaic output levels of the distribution area within the next 15 minutes to 24 hours, with a prediction accuracy of ≥95%. It autonomously optimizes and generates charging and discharging control strategies, inter-distribution energy mutual assistance strategies, and power quality management strategies to achieve adaptive control in multiple scenarios. It can independently complete local autonomous operation without relying on instructions issued by the upper-level dispatcher. It supports multiple energy management modes, including peak shaving and valley filling, photovoltaic consumption, economic operation, and three-phase imbalance management strategies in the distribution area energy storage scenario, as well as collaborative peak shaving, emergency mutual assistance, SOC balancing, and cross-distribution voltage collaborative support strategies in the distribution area mutual assistance scenario. All strategies have SOC safety boundary verification, priority management, and anti-frequent switching mechanisms.

[0039] The multi-protocol compatible communication unit supports power grid standard communication protocols such as IEC 60870-5-104, IEC 61850, and MQTT over TLS. It integrates the national cryptographic SM2 / SM4 security chip and can realize encrypted and secure communication with the dispatch master station, IoT management platform, adjacent transformer area collaborative controller, transformer area intelligent fusion terminal, photovoltaic inverter, charging pile, and flexible load control device. The telemetry refresh rate is ≥1Hz, which fully meets the power grid's four management requirements for distributed energy resources: "observable, measurable, adjustable, and controllable".

[0040] The DC power exchange interface unit is equipped with a dedicated DC power exchange terminal, which supports flexible interconnection with the same type of device deployed in adjacent transformer areas through the DC bus. It can complete the collaborative functions such as bidirectional power exchange between transformer areas, emergency power transfer during faults, and SOC balancing of multiple devices. The fault transfer response time is ≤20ms, which can realize uninterrupted power supply to important loads.

[0041] The hardware-level safety protection unit integrates 10 layers of hardware protection logic, including short-circuit protection, insulation monitoring protection, over / under voltage protection, over / under frequency protection, reverse polarity protection, phase sequence protection, phase loss protection, overload protection, anti-islanding protection, and battery emergency stop safety circuit protection. It is equipped with an independent battery emergency stop safety circuit interface, and the trigger signal directly triggers the hardware-level power switching device to lock down without software processing. The protection action response time is ≤1ms.

[0042] The human-machine interface unit supports two interaction methods: local operation and remote maintenance. The local terminal is equipped with status indicator lights, manual start / stop buttons, and parameter configuration buttons, enabling quick on-site operation. The remote terminal supports access to the maintenance management interface, which can perform operations such as device parameter configuration, real-time monitoring of operating status, historical event query, and firmware upgrade.

[0043] The device in this embodiment weighs 35kg and supports both wall-mounted and pole-mounted installation methods, requiring no additional civil engineering modifications. The indoor version has an IP54 protection rating, while the outdoor version has an IP65 protection rating. The operating temperature range is -40℃ to 70℃, the design life is 20 years, and the mean time between failures (MTBF) is 50,000 hours. It can be widely adapted to various complex transformer substation deployment scenarios.

[0044] Example 2

[0045] This embodiment is a simplified derivative of embodiment 1. It is designed for industrial and commercial transformer substations with peak shaving and valley filling as the core requirements. The DC mutual assistance interface and diesel generator coordination function are removed. The AI ​​control unit is optimized to load only the peak shaving and valley filling exclusive scheduling strategy. It charges at rated power during the midday peak photovoltaic / grid valley period and discharges at full power during the evening peak period. It can handle up to 30% of the peak load of the transformer substation. The energy utilization efficiency is improved compared with the traditional rigid control strategy, and the device cost is reduced compared with embodiment 1.

[0046] Example 3

[0047] This embodiment is a simplified derivative of Embodiment 1. It is designed for residential transformer substations with prominent issues such as low voltage, three-phase imbalance, and photovoltaic backfeeding. The DC mutual assistance interface and virtual power plant communication function are removed, and the power conversion unit is optimized to prioritize responding to power quality management commands. This can control the three-phase imbalance of the transformer substation to within 0.8% and improve the voltage qualification rate.

[0048] Example 4

[0049] This embodiment is a simplified derivative of embodiment 1. It is designed to meet the emergency power supply needs of important load areas such as hospitals and government units. It strengthens the grid control and diesel generator coordination functions, reduces the mutual assistance between distribution stations and the multi-scenario energy management strategy, and switches to off-grid grid mode within 10ms when the main grid loses power. It can adaptively match the economic operating range of diesel generators, and the emergency power supply time is extended compared with traditional solutions.

[0050] Example 5

[0051] This embodiment is a simplified derivative of embodiment 1. It is designed for cluster deployment scenarios with multiple adjacent transformer areas. It enhances the DC mutual assistance interface and multi-device collaborative scheduling function, and reduces the local power quality management and diesel generator collaborative functions. It can achieve power mutual assistance and SOC balance for three or more adjacent transformer areas, with a fault transfer response time of ≤18ms and the overall power supply reliability of the cluster is improved to 99.992%.

[0052] Example 6

[0053] This embodiment is a simplified derivative of embodiment 1. It is designed for virtual power plant aggregation and control scenarios, strengthens multi-protocol communication and security encryption functions, optimizes the AI ​​control unit to prioritize response to virtual power plant dispatch instructions, and achieves a dispatch instruction response time of ≤8ms. It can directly participate in the power auxiliary service market such as peak shaving, frequency regulation, and reserve, thereby increasing the owner's average annual additional income and shortening the investment return cycle.

[0054] Example 7

[0055] This embodiment is a simplified derivative of Embodiment 1. It is designed for remote weak grid areas with a high proportion of new energy access. It strengthens the network-type virtual synchronous machine algorithm and inertia support function, and removes DC mutual assistance and diesel generator coordination functions. It can respond quickly within 5ms when the grid frequency deviation exceeds ±0.05Hz, and provide instantaneous active power support within 10ms when the frequency change rate exceeds 0.2Hz / s, thereby improving the voltage stability of the distribution area.

[0056] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A flexible collaborative device for AI-controlled microgrids, which is a deep fusion device of primary and secondary systems, characterized in that: include: The power conversion unit is the primary part, which realizes the bidirectional power conversion and power quality management functions between the DC energy storage battery and the AC power grid. The secondary part of the device integrates a high-precision synchronous sampling module to provide data support for control and decision-making. The heterogeneous dual-core main control unit adopts a heterogeneous architecture of communication management core and real-time control core, which respectively undertake non-real-time tasks and hard real-time control tasks. The AI ​​edge control unit has a built-in load and photovoltaic prediction model and a dynamic scheduling optimization model to achieve adaptive control in multiple scenarios. A multi-protocol compatible communication unit, which enables secure communication with the power grid master station and various distributed energy devices; A DC interconnection interface unit, which enables flexible interconnection and power exchange of the same type of devices in adjacent transformer areas; A hardware-level security protection unit, which implements hardware-level fast protection functions; The human-computer interaction unit realizes local and remote operation and maintenance functions.

2. The flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The heterogeneous dual-core main control unit includes a communication management core module and a real-time control core module. The communication management core module runs a domestic operating system and undertakes non-real-time tasks such as northbound communication, AI strategy calculation, and OTA upgrades. The real-time control core module runs an RTOS real-time operating system and focuses on hard real-time functions such as VSG algorithm execution, high-precision AC sampling, high-frequency current loop control, and hardware protection. The communication management core module and the real-time control core module realize data interaction through a high-speed shared memory mechanism between cores, and the scheduling instructions are converted into precise power output.

3. The flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The AI ​​edge control unit incorporates an LSTM load and photovoltaic output prediction model and a deep reinforcement learning dynamic scheduling optimization model. The LSTM prediction model can dynamically predict load and photovoltaic output at different time scales by integrating historical operating data, real-time sensing data, and meteorological data. The deep reinforcement learning dynamic scheduling optimization model can dynamically optimize charging and discharging, mutual assistance, and power quality management strategies based on prediction results, real-time operating status, and scheduling requirements.

4. The flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The power conversion unit has a built-in grid-type virtual synchronous machine algorithm, which has the functions of autonomous voltage building, virtual inertia, black start, primary frequency regulation, and seamless switching between grid and off-grid. When the virtual synchronous machine algorithm is always enabled, the real-time control core module executes the VSG rotor equation every 100μs to output a virtual electromotive force as a voltage loop reference, and controls the power electronic converter to output the corresponding voltage waveform.

5. A flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The DC mutual assistance interface unit supports flexible interconnection of the same type of device in adjacent transformer areas, and can realize the functions of energy mutual assistance, fault transfer and SOC balancing between transformer areas. When a mutual assistance command is received or a mutual assistance requirement is detected locally, the communication management core module first exchanges status information with the neighboring device, calculates the optimal power command based on the AI ​​strategy, and controls the power conversion unit to execute the corresponding power transmission in real time.

6. A flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The power conversion unit has power quality management functions, including low voltage management at the end of the transformer area, photovoltaic overvoltage and reverse power transmission management, and three-phase imbalance management. When the voltage at the end of the transformer area exceeds the limit, reverse power flow overvoltage, or three-phase imbalance exceeds the threshold, the real-time control core module independently controls the active and reactive power output of each phase to stabilize the voltage and imbalance within the qualified range.

7. A flexible collaborative device for AI control of microgrids according to claim 2, characterized in that, The multi-protocol compatible communication unit supports power grid standard protocols such as IEC 60870-5-104, IEC 61850, and MQTT over TLS, and integrates the national cryptographic SM2 / SM4 security chip to achieve secure communication with the dispatch master station, IoT platform, edge collaborative controller, smart converged terminal of distribution area, and distributed energy equipment.

8. A flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The hardware-level safety protection unit integrates multiple protection functions, including short circuit protection, insulation monitoring protection, over / under voltage protection, over / under frequency protection, reverse polarity protection, phase sequence protection, phase loss protection, overload protection, anti-islanding protection, and battery emergency stop safety circuit protection. The battery emergency stop safety circuit signal directly triggers the hardware-level power switching device to block without going through software.

9. A flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The device adopts an integrated modular design, with a total weight of no more than 35kg. It supports both wall-mounted and pole-mounted installation methods. The protection level of the device is no less than IP54 for indoor models and no less than IP65 for outdoor models. The operating temperature range is -40℃ to 70℃, the design life is no less than 20 years, and the mean time between failures (MTBF) is no less than 50,000 hours. The device supports multiple energy management modes, including peak shaving and valley filling, photovoltaic consumption, economic operation, and three-phase imbalance management strategies in the context of transformer substation energy storage, as well as collaborative peak shaving, emergency mutual assistance, SOC balancing, and cross-transformer voltage collaborative support strategies in the context of transformer substation mutual assistance. All strategies have SOC safety boundary verification, priority management, and anti-frequent switching mechanisms.

10. A flexible collaborative device for AI control of microgrids according to claim 1, characterized in that, The device supports emergency power supply coordination with external diesel generators. When the main grid loses power and a diesel generator is detected to be connected, the real-time control core module automatically identifies the diesel generator mode and controls the AC / DC conversion unit to rectify the AC power output of the diesel generator into DC power and connect it to the internal bus. The communication management core module dynamically adjusts the charging power according to load demand, energy storage SOC, and the economic operating range of the diesel generator.

Images