An off-grid type light storage inverter system collaborative control method and system based on a multi-modal intelligent algorithm

Abstract

The application provides an off-grid type light storage inversion system cooperative control method and system based on a multi-modal intelligent algorithm, relates to the field of light storage inversion, and comprises the following steps: obtaining hardware configuration requirements, collecting multi-source data in real time, using a phase-locked algorithm for accurate phase tracking, and obtaining standardized control signals; obtaining photovoltaic voltage and current data, battery state and light mutation environment parameters, performing photovoltaic tracking, combining a fuzzy logic controller for battery management, and outputting cooperative control instructions; obtaining historical load data and real-time state flags, generating a load demand prediction curve through cloud model processing, driving a working mode program decision, and outputting seamless switching instructions; and through multi-algorithm fusion processing of real-time monitoring data, combining convolutional neural network fault diagnosis and voice feedback, and obtaining closed-loop protection instructions. The application is used to solve the problems of low energy conversion efficiency, poor dynamic response speed and poor full-working-condition adaptability of off-grid light storage.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic-storage inverter technology, and in particular to a collaborative control method and system for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms. Background Technology

[0002] As the core equipment of the photovoltaic energy storage system, the integrated photovoltaic energy storage inverter and controller (which integrates inverter, controller, energy storage management and other functions) faces technical challenges mainly in three aspects: multi-energy flow coordinated control, system efficiency optimization and safety and reliability. These challenges include, but are not limited to, dynamic power balance control, multi-mode seamless control, control algorithm optimization, hardware loss optimization, and safety and reliability.

[0003] In traditional solutions, the photovoltaic MPPT, battery management, and inverter control modules are separate, requiring multiple energy conversions for energy transfer. Under sudden changes in illumination, the MPPT algorithm's tracking delay exceeds 500ms, leading to power oscillations and poor dynamic response. It also exhibits weak adaptability to various operating conditions: fixed-threshold battery management cannot adapt to temperature / SOC changes, and low-temperature charging carries the risk of overcurrent. The standard PSO algorithm is prone to getting trapped in local optima, with a fixed particle iteration step size, resulting in a convergence time exceeding 10s under complex illumination. Furthermore, the operating mode relies on manual setting, lacks load prediction capabilities, and the off-grid to bypass switching delay exceeds 200ms.

[0004] This invention systematically solves the problems of low energy conversion efficiency, poor dynamic response speed, and poor adaptability under all operating conditions in off-grid optical energy storage by innovating hardware architecture, integrating multimodal algorithms, and using cloud-edge collaborative prediction. Summary of the Invention

[0005] This invention provides a collaborative control method and system for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms, which can solve the problems of low energy conversion efficiency, poor dynamic response speed, and poor adaptability under all operating conditions in off-grid photovoltaic-storage systems.

[0006] On one hand, this invention provides a collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms, comprising:

[0007] Obtain hardware configuration requirements and process load prediction data through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters;

[0008] Based on the optimized hardware operating parameters, multi-source data is collected in real time, and phase-locked loop algorithm is used to accurately track the phase to obtain standardized control signals. The multi-source signals include voltage, battery current, and original grid phase signals.

[0009] Based on standardized control signals, photovoltaic voltage and current data, battery status and sudden light change environmental parameters are acquired to perform photovoltaic tracking and combine with a fuzzy logic controller for battery management, and output cooperative control commands.

[0010] Based on collaborative control commands, historical load data and real-time status indicators are acquired, load demand prediction curves are generated through cloud model processing, and working mode program decisions are driven to output seamless switching commands.

[0011] Based on seamless switching instructions, real-time monitoring data is processed through multi-algorithm fusion, and closed-loop protection instructions are obtained by combining convolutional neural network fault diagnosis and voice feedback.

[0012] By integrating hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions, hierarchical processing and performance verification are performed to obtain a system performance report.

[0013] Furthermore, hardware configuration requirements are obtained, and load prediction data is processed using a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters, including:

[0014] Obtain hardware configuration requirements, which include hardware pin functions, relay status, and initial load prediction data;

[0015] Based on hardware configuration requirements, a neural network dynamic optimization algorithm is used to analyze load prediction data in real time and dynamically adjust the modulation frequency to obtain a hardware control parameter optimization strategy.

[0016] Based on the hardware control parameter optimization strategy, the power management unit dynamically adjusts the duty cycle of the modulation generator, switches between battery or grid power supply modes, and monitors temperature to trigger current limiting protection in real time, thus obtaining real-time updated operating status.

[0017] Integrate real-time updated operational status data and perform performance verification to obtain optimized hardware operating parameters.

[0018] Furthermore, based on the optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop algorithm is used for precise phase tracking to obtain a standardized control signal. The multi-source signals include volt voltage, battery current, and raw grid phase signals, including:

[0019] Based on the optimized hardware operating parameters, multi-source data is collected in real time, including raw signals of photovoltaic voltage, battery current, and grid phase, and environmental parameters, including temperature and light intensity, are also acquired simultaneously.

[0020] Based on multi-source data, an application signal processing program is used to convert the multi-source data into real physical values ​​to obtain preprocessed multi-source data.

[0021] Based on the preprocessed multi-source data, a phase-locked loop algorithm is used for precise phase tracking to obtain the tracking results;

[0022] The tracking results are compared with and corrected to the theoretical phase angle to obtain a standardized control signal.

[0023] Furthermore, based on standardized control signals, photovoltaic voltage and current data, battery status, and parameters related to sudden changes in sunlight are acquired. Photovoltaic tracking is performed, and battery management is combined with a fuzzy logic controller. Coordinated control commands are output, including:

[0024] Based on standardized control signals, photovoltaic voltage and current data, battery status and sudden light change environmental parameters are collected synchronously to form a multi-dimensional input vector.

[0025] Based on multidimensional input vectors, a neural network is used to dynamically adjust the number of particles and the iteration step size to obtain the tracking parameter set;

[0026] Based on the tracking parameter set, the particle swarm is initialized with voltage and current as dimensions, the power fitness function is calculated, the particle position is updated to the maximum power point, and the circuit duty cycle adjustment command is obtained.

[0027] Based on the circuit duty cycle adjustment command, the input variables are parsed by the fuzzy logic controller, and the circuit charging current is adjusted in real time to obtain the battery constant current control command.

[0028] The circuit duty cycle adjustment command is integrated with the battery constant current control command to output a coordinated control command.

[0029] Furthermore, based on collaborative control commands, historical load data and real-time status indicators are acquired, and load demand prediction curves are generated through cloud-based model processing. This drives the operating mode program to make decisions and outputs seamless switching commands, including:

[0030] Based on collaborative control commands, historical load data and real-time status flags are obtained, and load time dependency characteristics are analyzed through cloud models to generate load demand prediction curves.

[0031] Based on the load demand forecast curve, the working mode switching program decision logic is driven to output seamless switching instructions.

[0032] Furthermore, based on seamless switching instructions, real-time monitoring data is processed through multi-algorithm fusion, and combined with convolutional neural network fault diagnosis and voice feedback, closed-loop protection instructions are obtained, including:

[0033] Based on seamless switching instructions, parameters are monitored in real time, data verification and feature extraction are performed, and a standardized collaborative control dataset is generated.

[0034] Based on a standardized collaborative control dataset, this algorithm integrates deep reinforcement learning optimization, fuzzy control and cloud prediction algorithms, and uses a convolutional neural network to diagnose temperature anomalies in real time, generating fault risk level assessments and optimized control strategies.

[0035] Based on fault risk level assessment and optimized control strategy, the voice module is driven by pins to broadcast the status according to the warning level and generate audible status feedback signals.

[0036] By integrating and optimizing control strategies and audible status feedback signals, closed-loop protection commands are obtained.

[0037] Furthermore, by integrating hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions, hierarchical processing and performance verification are performed to obtain a system performance report, including:

[0038] Integrate hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions to generate a full-link collaborative input matrix;

[0039] Based on the end-to-end collaborative input matrix, collaborative processing is performed on the driver layer, algorithm layer, and interaction layer to output optimized control flow;

[0040] Based on optimized control flow, energy is verified step by step to obtain verification results;

[0041] Based on the verification results, dynamic response limit tests are conducted to obtain a system performance report.

[0042] On the other hand, a collaborative control system for an off-grid photovoltaic-storage-inverter system based on a multimodal intelligent algorithm includes:

[0043] The acquisition module is used to acquire hardware configuration requirements and process load prediction data through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters. Based on the optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop algorithm is used for precise phase tracking to obtain standardized control signals. The multi-source signals include volt voltage, battery current, and original grid phase signals.

[0044] The processing module is used to acquire photovoltaic voltage and current data, battery status, and environmental parameters related to sudden changes in sunlight based on standardized control signals. It performs photovoltaic tracking and battery management in conjunction with a fuzzy logic controller, outputting coordinated control commands. Based on the coordinated control commands, it acquires historical load data and real-time status flags, generates load demand prediction curves through cloud model processing, drives the operating mode program decision, and outputs seamless switching commands. Based on the seamless switching commands, it processes real-time monitoring data through multi-algorithm fusion and combines convolutional neural network fault diagnosis and voice feedback to obtain closed-loop protection commands. It integrates hardware operating parameters, standardized control signals, coordinated control commands, seamless switching commands, and closed-loop protection commands for hierarchical processing and performance verification, generating a system performance report.

[0045] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the collaborative control method for off-grid photovoltaic-storage inverter system based on multimodal intelligent algorithm as described above.

[0046] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms as described above.

[0047] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms as described above.

[0048] The present invention provides a collaborative control method and system for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms, which achieves a leapfrog improvement in the performance of off-grid photovoltaic-storage systems. In terms of energy conversion efficiency, the system efficiency and photovoltaic utilization rate are improved by dynamically optimizing hardware parameters based on neural networks and combining a photovoltaic-storage collaborative direct connection architecture. In terms of dynamic response performance, the particle swarm optimization algorithm optimized by deep reinforcement learning shortens the maximum power point tracking convergence time, reduces the mode switching delay driven by cloud prediction, and accelerates the fault protection response. In terms of battery management, the fuzzy logic controller realizes soft management of charging and discharging based on a dynamic rule base, significantly extending the battery cycle life and solving the industry problem of overcharging and over-discharging. The safety protection system has been comprehensively upgraded, and the convolutional neural network fault diagnosis combined with the voice feedback module improves the response speed compared with traditional solutions.

[0049] Compared with traditional technologies, the architecture design has been upgraded from discrete modules to direct connection between optical and energy storage, reducing energy loss; in terms of maximum power point tracking performance, the adaptive tracking algorithm is faster than the traditional perturbation and observation method; and the operation and maintenance efficiency is improved through automatic reporting via an intelligent interface, thereby reducing costs. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms provided in an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of the collaborative control system for an off-grid photovoltaic-storage inverter system based on a multimodal intelligent algorithm, provided in an embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention;

[0054] Figure 4 This is a block diagram of the off-grid 6KW photovoltaic storage inverter and control integrated system proposed in this embodiment of the invention;

[0055] Figure 5 This is a hardware scheme diagram of the off-grid 6KW photovoltaic storage inverter and controller integrated machine proposed in this embodiment of the invention;

[0056] Figure 6 This is the hardware I / O attribute table of the off-grid 6KW photovoltaic storage inverter and controller integrated machine proposed in this embodiment of the invention;

[0057] Figure 7 This is a block diagram of the off-grid inverter control strategy proposed in this embodiment of the invention;

[0058] Figure 8 This is a block diagram of the battery charging control strategy proposed in the embodiments of the present invention;

[0059] Figure 9 This is a loop program flowchart of the off-grid 6KW photovoltaic storage inverter and controller integrated machine proposed in this embodiment of the invention;

[0060] Figure 10 This is a working example diagram of the PSO-MPPT algorithm proposed in this embodiment of the invention;

[0061] Figure 11 This is a flowchart of the PSO-MPPT algorithm proposed in this embodiment of the invention;

[0062] Figure 12 This is a schematic diagram of the unipolar frequency doubling modulation principle proposed in the embodiments of the present invention;

[0063] Figure 13 This is the SOGI loop block diagram proposed in the embodiments of the present invention;

[0064] Figure 14 This is a flowchart of the human-computer interaction program proposed in the embodiments of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0066] like Figures 1 to 14 As shown in the embodiment of the present invention, the collaborative control method for off-grid photovoltaic-storage inverter system based on multimodal intelligent algorithm mainly includes the following steps:

[0067] 11. Obtain hardware configuration requirements and process load prediction data through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters;

[0068] 12. Based on the optimized hardware operating parameters, multi-source data is collected in real time, and phase-locked loop algorithm is used for precise phase tracking to obtain standardized control signals. The multi-source signals include volt voltage, battery current, and original grid phase signals.

[0069] 13. Based on standardized control signals, acquire photovoltaic voltage and current data, battery status and sudden change in light environment parameters, perform photovoltaic tracking and combine with fuzzy logic controller for battery management, and output cooperative control commands;

[0070] 14. Based on collaborative control commands, acquire historical load data and real-time status indicators, generate load demand prediction curves through cloud model processing, drive working mode program decisions, and output seamless switching commands.

[0071] 15. Based on seamless switching instructions, real-time monitoring data is processed through multi-algorithm fusion, and closed-loop protection instructions are obtained by combining convolutional neural network fault diagnosis and voice feedback.

[0072] 16. Integrate hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions, perform hierarchical processing and performance verification, and obtain a system performance report.

[0073] In this embodiment of the invention, a photovoltaic-storage collaborative control architecture is adopted to solve the problems of low energy conversion efficiency, poor dynamic response, and weak adaptability to multiple operating conditions in traditional off-grid systems. The system architecture mainly includes: a PV-side BOOST circuit supporting DC 120~480V bus; a bidirectional isolated synchronous rectifier converter circuit on the battery side supporting DC 48V~58V input; and a full-bridge inverter circuit on the AC side supporting inverter discharge and PFC charging. The overall framework is shown in the figure below. It also includes human-machine interaction functions, thermal management functions, and protection functions. The off-grid operating modes of this design mainly include: photovoltaic power supply to both the load and the battery (charging and energy storage) simultaneously, photovoltaic power supply to the load simultaneously, battery power supply to the load alone, photovoltaic power supply to the battery alone (without load), and grid power supply to the battery alone; covering all usage scenarios of off-grid inverters.

[0074] The system adopts a photovoltaic-storage collaborative control architecture to solve the problems of low energy conversion efficiency, poor dynamic response, and weak adaptability to multiple operating conditions in traditional off-grid systems. The system architecture mainly includes: a PV-side BOOST circuit, supporting DC input voltage of 120~480V to the bus; a bidirectional isolated synchronous rectifier converter circuit on the battery side, supporting DC input of 48V~58V; and a full-bridge inverter circuit on the AC side, supporting inverter discharge and PFC charging. The overall framework is as follows. It also includes human-machine interaction functions, thermal management functions, and protection functions. The software design of this system is mainly modular, and the program can be divided into six major sections: the underlying hardware driver program, including its initialization, automatic control algorithm program (affected by flag bits), signal modulation algorithm, signal processing program, working mode processing program, system protection and power-on / off program, and human-machine interaction program. The off-grid working modes of this design mainly include: photovoltaic power supply to the load and battery (charging energy storage) simultaneously, photovoltaic power supply to the load simultaneously, battery power supply to the load alone, photovoltaic power supply to the battery alone (without load), and grid power supply to the battery alone.

[0075] like Figures 1 to 14 As shown in Figure 11, the hardware configuration requirements are obtained, and the load prediction data is processed through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters, including:

[0076] 111. Obtain hardware configuration requirements, including hardware pin functions, relay status, and initial load prediction data;

[0077] 112. Based on hardware configuration requirements, a neural network dynamic optimization algorithm is used to analyze load prediction data in real time and dynamically adjust the modulation frequency to obtain a hardware control parameter optimization strategy.

[0078] 113. Based on the hardware control parameter optimization strategy, the power management unit dynamically adjusts the duty cycle of the modulation generator, switches between battery or grid power supply modes, and monitors the temperature to trigger current limiting protection in real time, thereby obtaining real-time updated operating status.

[0079] 114. Integrate real-time updated operating status data and perform performance verification to obtain optimized hardware operating parameters.

[0080] In this embodiment of the invention, based on Figure 6 Hardware I / O attribute tables (such as GPI00 controlling the full-bridge inverter arm and GPI08 driving the photovoltaic BOOST circuit) are used to analyze hardware pin functions, relay states (such as GPI045 controlling AC on / off), and initial load prediction data (estimated through ADCB3 temperature sampling). This constructs the basic hardware control framework, ensuring 100% hardware interface compatibility, eliminating configuration conflict risks, and providing accurate input for dynamic optimization. Neural network models (such as LSTM) are used to analyze load prediction data (temperature / current fluctuations) in real time, dynamically adjusting the PWM modulation frequency (increasing the frequency under high load) to generate hardware control parameter optimization strategies, reducing hardware losses and improving response speed. The power management unit executes the optimization strategy, dynamically adjusting the PWM duty cycle (GPI08 pin controls the BOOST circuit), switching between battery / grid power supply modes (such as enabling grid charging at low temperatures), and real-time temperature monitoring (ADCB3) triggering current limiting protection (power reduction >85℃), thus reducing energy consumption and lowering the over-temperature failure rate.

[0081] like Figures 1 to 14 As shown in Figure 12, based on the optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop algorithm is used for precise phase tracking to obtain a standardized control signal. The multi-source signals include volt voltage, battery current, and original grid phase signals, including:

[0082] 121. Based on the optimized hardware operating parameters, multi-source data is collected in real time, including the raw signals of photovoltaic voltage, battery current, and grid phase, and environmental parameters, including temperature and light intensity, are acquired simultaneously.

[0083] 122. Based on multi-source data, use applied signal processing programs to convert the multi-source data into real physical values ​​to obtain preprocessed multi-source data;

[0084] 123. Based on the preprocessed multi-source data, a phase-locked loop algorithm is used for precise phase tracking to obtain the tracking results;

[0085] 124. Compare and correct the tracking results with the theoretical phase angle to obtain a standardized control signal.

[0086] In this embodiment of the invention, the PV-side BOOST circuit supports input voltages of 120~450V to the DC bus of 0~500V. Its control loop provides the expected value of the MPPT output voltage to the voltage outer loop and current inner loop to stably track the maximum photovoltaic power. The battery side uses a bidirectional isolated synchronous rectifier converter circuit, supporting DC input of 48V~58V. It adopts pure hardware control and software enablement to reduce control difficulty. The battery side also uses a BUCK circuit in series for constant current control, adjusting the battery charging current in real time according to the bus voltage to achieve constant power charging. The AC side is a full-bridge inverter circuit, supporting automatic control algorithms for inverter discharge and PFC charging (affected by flag bits) and signal modulation algorithms, mainly used to ensure that the IGBT / MOS operates in a suitable switching state.

[0087] like Figures 1 to 14 As shown in Figure 13, based on standardized control signals, photovoltaic voltage and current data, battery status, and parameters of sudden changes in light intensity are acquired. Photovoltaic tracking is performed, and battery management is combined with a fuzzy logic controller. Cooperative control commands are output, including:

[0088] 131. Based on standardized control signals, photovoltaic voltage and current data, battery status and sudden change in light environment parameters are collected synchronously to form a multi-dimensional input vector;

[0089] 132. Based on multidimensional input vectors, a neural network is used to dynamically adjust the number of particles and the iteration step size to obtain the tracking parameter set;

[0090] 133. Based on the tracking parameter set, initialize the particle swarm with voltage and current as dimensions, calculate the power fitness function, update the particle position to the maximum power point, and obtain the circuit duty cycle adjustment command.

[0091] 134. Based on the circuit duty cycle adjustment instruction, the input variable is parsed by the fuzzy logic controller, and the circuit charging current is adjusted in real time to obtain the battery constant current control instruction.

[0092] 135. Integrate the circuit duty cycle adjustment command with the battery constant current control command to output a coordinated control command.

[0093] In this embodiment of the invention, the particle swarm optimization (MPPT) control algorithm includes related signal processing algorithms such as unipolar frequency doubling SPWM modulation algorithm, SOGI phase-locked loop algorithm, and root mean square (RMS) algorithm. The SOGI phase-locked loop algorithm is used to track the grid phase and generate a reference signal for the loop to perform rectification and synchronization voltage, thereby realizing PFC control.

[0094] ;

[0095] Convert to the Z-domain using MATLAB.

[0096] ;

[0097] Write the difference equation:

[0098] ;

[0099] This algorithm can extract a set frequency signal, i.e., 50Hz. The root mean square algorithm is used to calculate the effective values ​​of the grid voltage and inverter voltage and update them to the screen.

[0100] like Figures 1 to 14 As shown in Figure 14, based on collaborative control commands, historical load data and real-time status flags are acquired, load demand prediction curves are generated through cloud model processing, and the working mode program is driven to make decisions, outputting seamless switching commands, including:

[0101] 141. Based on collaborative control commands, acquire historical load data and real-time status flags, and analyze load time dependency characteristics through cloud model analysis to generate load demand prediction curves;

[0102] 142. Based on the load demand forecast curve, drive the working mode switching program decision logic and output seamless switching instructions.

[0103] In this embodiment of the invention, based on collaborative control commands (such as MPPT output power and battery charging / discharging status), historical load data and real-time status flags (such as insufficient photovoltaic power signals) are acquired. Through a cloud-based long short-term memory network model, the load time-dependent characteristics are deeply analyzed to generate a high-precision load demand prediction curve. This overcomes the static limitations of traditional offline prediction, enabling dynamic load trend capture and providing forward-looking data support for mode decision-making. The prediction curve drives the working mode program decision logic (…). Figure 9 The process automatically triggers mode switching (such as instantaneously switching to "photovoltaic + battery" hybrid power supply when photovoltaic power supply is insufficient), eliminating delays caused by manual intervention, ensuring seamless energy flow transition, and maintaining the continuity of system power supply.

[0104] like Figures 1 to 14 As shown in Figure 15, based on seamless switching instructions, real-time monitoring data is processed through multi-algorithm fusion, and combined with convolutional neural network fault diagnosis and voice feedback to obtain closed-loop protection instructions, including:

[0105] 151. Based on seamless switching instructions, monitor parameters in real time, perform data verification and feature extraction, and generate a standardized collaborative control dataset;

[0106] 152. Based on a standardized collaborative control dataset, this paper integrates deep reinforcement learning optimization algorithms, fuzzy control and cloud prediction algorithms, and uses convolutional neural networks to diagnose temperature anomalies in real time, generating fault risk level assessments and optimized control strategies.

[0107] 153. Based on fault risk level assessment and optimized control strategy, the voice module is driven by pins to broadcast the status according to the warning level and generate audible status feedback signals.

[0108] 154. Integrate and optimize the control strategy and the audible status feedback signal to obtain the closed-loop protection command.

[0109] In this embodiment of the invention, multi-dimensional parameters such as battery current and temperature are collected in real time based on seamless switching commands. A standardized collaborative control dataset is generated through data verification and feature extraction to eliminate signal noise and dimensional differences. High-precision ADC channels (such as ADCB2 for current acquisition and ADCB3 for temperature acquisition) ensure data reliability, providing clean input for multi-algorithm fusion. Deep reinforcement learning optimization algorithms, fuzzy control, and cloud prediction algorithms are integrated. A convolutional neural network is used to analyze abnormal temperature characteristics (such as localized overheating of the heat sink) in real time, generating fault risk levels and dynamic optimization strategies. Figure 4 The system architecture enables algorithmic collaboration, with the convolutional neural network diagnostic module accurately locating potential hazards. A voice module is driven via pin GPIO49 to broadcast real-time warnings based on risk level (e.g., "Abnormal temperature, power limiting upon startup"), generating audible feedback signals to improve human-computer interaction efficiency. Direct hardware connection for voice ensures feedback latency is less than 100 milliseconds. The system integrates and optimizes control strategies with voice feedback signals to generate hardware-level protection commands (e.g., overcurrent power limiting, overtemperature shutdown), forming a closed-loop "monitoring-diagnosis-response" chain. Figure 9 The protection process enables seamless execution of instructions;

[0110] Convolutional neural networks accurately locate and analyze the temperature distribution of heat sinks (such as ADCB3 data) to identify local overheating risks that traditional threshold detection cannot capture (such as heat accumulation at the edge of the aluminum substrate), achieving early screening of potential hazards. Deep reinforcement learning optimization algorithms dynamically adjust protection thresholds, and fuzzy control softens protection strategies based on load conditions to avoid false triggering. When a high temperature warning is issued, the voice module announces "Please check the heat dissipation duct" and simultaneously reduces output power, guiding manual intervention and linking with the system's self-protection mechanism. When the convolutional neural network diagnoses a temperature >90℃, it directly triggers... Figure 9 The protection program shuts down the circuit, and the response speed breaks through the bottleneck of traditional solutions.

[0111] A closed-loop protection system of "monitoring → diagnosis → feedback → protection" has been constructed: Monitoring layer: Real-time capture of current and temperature signals through high-precision sensors (ADCB2 / ADCB3) to generate standardized datasets; Diagnosis layer: Figure 4 The architecture integrates convolutional neural networks, deep reinforcement learning optimization algorithms, and fuzzy control to achieve multi-dimensional assessment of fault risk; Feedback layer: A voice module (driven by GPIO49) broadcasts warnings according to risk level, improving human-machine collaboration efficiency; Protection layer: Figure 9The protection process translates optimization strategies into hardware instructions to achieve hard protection such as over-temperature shutdown and over-current power limiting; system security is upgraded from passive threshold protection to intelligent active protection, and fault response speed and coverage achieve a leapfrog breakthrough.

[0112] like Figures 1 to 14 As shown in Figure 16, hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions are integrated, and hierarchical processing and performance verification are performed to obtain a system performance report, including:

[0113] 161. Integrate hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions to generate a full-link collaborative input matrix;

[0114] 162. Based on the end-to-end collaborative input matrix, perform collaborative processing of the driver layer, algorithm layer, and interaction layer to output optimized control flow;

[0115] 163. Based on optimized control flow, energy is verified step by step to obtain verification results;

[0116] 164. Based on the verification results, conduct dynamic response limit tests to obtain a system performance report.

[0117] In this embodiment of the invention, hardware dynamic parameters (such as adaptive pulse width modulation frequency), high-precision control signals (phase error <0.5°), optical-storage collaborative instructions, mode switching instructions, and protection instructions are integrated to generate a standardized collaborative input matrix, eliminating data barriers between modules and providing a distortion-free data foundation for system-level verification; based on Figure 4 The system architecture executes three levels of processing: Driver layer: direct execution of hardware logic (e.g., GPI08 driving a photovoltaic boost circuit); Algorithm layer: deep reinforcement learning optimization algorithms dynamically schedule resources, fuzzy controllers regulate battery charging and discharging, and long short-term memory networks verify prediction consistency; Interaction layer: through... Figure 14 The human-machine interface maps the control flow state in real time; the output optimizes the control flow to achieve cross-level resource collaboration, significantly improving system stability; along Figure 4 Energy path (PV → boost circuit → isolation converter → inverter) step-by-step testing: PV maximum power point tracking accuracy verification, DC-DC converter constant power charging efficiency analysis, inverter output waveform distortion detection, generation of a complete report on energy conversion efficiency and losses at each stage, and identification and optimization bottlenecks; injection of step load disturbances (e.g., 0-6kW instantaneous switching), based on... Figure 9 Loop process monitoring: mode switching delay, fault protection response time, algorithm convergence speed; output dynamic response performance report to verify the robustness of the system under extreme conditions.

[0118] In this embodiment of the invention, the underlying configuration program is used to set the functions of the DSP-IO; the signal processing program converts the data sampled by the ADC into real values ​​in real time, provides them to the loop and function, and confirms the system status; the operating mode processing program updates the system's operating status in real time according to the system status or user settings; the system protection program performs protection processing based on the updated real parameters, including battery-side overcurrent, overvoltage, and overtemperature protection, photovoltaic-side overcurrent, overvoltage, and overtemperature protection, AC-side overcurrent, overvoltage, and overtemperature protection, and bus overvoltage protection; the power-on / off program detects the button status, enables the auxiliary power supply, and simultaneously stops the bus power supply; the human-machine interaction program displays the operating mode and uploaded real data in real time, executes user settings in real time, and updates the saved data to the external memory. The system performs relevant operations based on the set variables and updates the data simultaneously.

[0119] The hardware control board interface and the system composition of this invention are as follows: The hardware IO attribute table of the off-grid 6KW photovoltaic storage inverter and controller is as follows. Figure 6 According to Figure 4 Off-grid 6KW photovoltaic storage inverter and controller integrated system block diagram:

[0120] The control board is used for data processing, system control, and human-machine interaction. The communication port is configured with RS232 protocol for data synchronization and cloud upload, RS485 protocol for battery BMS communication, and CAN protocol for inter-machine communication. It includes button and display updates, and user settings data is stored in an external ROM. The PV side sets drive signals to control the BOOST circuit to operate in MPP mode. The battery side controls the isolated DC / DC converter for battery charging and discharging. Both the power grid and the load are referred to as the AC side. Relays 2 and 3 determine whether the power grid is connected to the full-bridge circuit, and activating relay 1 determines whether power is supplied to the load. The auxiliary power supply has two modes: one is from the battery to the bus, where the auxiliary power is obtained from the bus; the other is grid rectification, where the auxiliary power is obtained from the grid. The auxiliary power supply converts high-voltage DC to a suitable voltage to power the entire system.

[0121] Hardware driver confirmation: The PV side single-channel drive signal is configured to operate in PWM mode, and the full-bridge side four-channel switch signals are configured to operate in PWM mode with center alignment. A unipolar frequency multiplication drive method is used, and SPWM modulation converts the bus 320V~450V to AC 220V / 50Hz. The AC side uses a lookup table to obtain the modulation reference signal for the inverter, and the loop consists of a current-side voltage outer loop and an inductor current inner loop to complete the inverter. There are two charging methods on the battery side: during PV charging, the battery is in overcurrent and overvoltage detection mode, and the charging power depends on the PV power minus the load power; during grid charging, a voltage outer loop and current inner loop are used to achieve constant power charging. The PV side uses PSO-MPPT to track the maximum power point, and the tracking result is fed to the voltage inner loop and voltage outer loop to control the PV port voltage stability (MPP). The system program determines whether the system meets the power-on conditions based on the data returned from the hardware. The system determines whether it should be in a certain state based on the port voltage status. Operating modes: Prepared mode includes photovoltaic charging of the battery and grid charging of the battery; Bypass mode includes grid-powered load only and grid-powered load charging the battery; Off-grid mode includes photovoltaic-powered load + charging battery (fully charged), photovoltaic + battery-powered load (insufficient photovoltaic), battery-powered load only (photovoltaic unavailable), and photovoltaic-powered load only (no battery charging); Human-machine interaction function: the display interface can be switched according to user button commands, and setting parameters can be changed according to user button settings; Protection range is set according to port: PV side maximum operating voltage 500V, maximum operating current 27A, MPPT voltage normal operating range 120V~450V, battery side maximum charging current 120A, exceeding 120A will trigger MPPT power limiting protection; AC side normal operating voltage is 220V / 230V / 240V, 50 / 60HZ, AC side charging power limited to 1KW, exceeding the above thresholds will trigger real-time system protection.

[0122] System integration involves several aspects: hardware drivers are unchangeable; control algorithms are also unchangeable, used to implement basic system functions, with input signals from the sampling module and output signals from PWM modulation reference values; real-time system functions are confirmed by the system's judgment / set operating mode, changing the operating mode in real time according to the system status, with response speed affected by algorithm performance; data processing is unchangeable, with hardware parameters confirming conversion coefficients used for system judgment of real-time status; human-machine interaction functions do not participate in actual control, requiring confirmation of user-provided parameters and real-time display updates; operating modes are influenced by user settings but limited by actual data, automatically making judgments based on interface conditions; protection is confirmed based on national standards and hardware limit parameters.

[0123] This embodiment confirms the software with hardware, and the system's operating mode is confirmed by the hardware interface status, making full use of system resources and fully considering real-time scenarios. The modular system design defines common parameters and interfaces, divides functional functions, streamlines processing, and prioritizes responses. This simplifies the system's collaborative control framework, improves energy conversion efficiency and dynamic response through real-time automatic algorithm control, and enhances adaptability to multiple operating conditions.

[0124] like Figure 2 As shown, a collaborative control system 20 for an off-grid photovoltaic-storage inverter system based on a multimodal intelligent algorithm includes:

[0125] The acquisition module 21 is used to acquire hardware configuration requirements and process load prediction data through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters. Based on the optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop algorithm is used for precise phase tracking to obtain standardized control signals. The multi-source signals include volt voltage, battery current, and original grid phase signals.

[0126] Processing module 22 is used to acquire photovoltaic voltage and current data, battery status, and environmental parameters related to sudden changes in light intensity based on standardized control signals, perform photovoltaic tracking, and manage the battery in conjunction with a fuzzy logic controller, outputting collaborative control commands; based on the collaborative control commands, it acquires historical load data and real-time status flags, generates load demand prediction curves through cloud model processing, drives the working mode program decision, and outputs seamless switching commands; based on the seamless switching commands, it processes real-time monitoring data through multi-algorithm fusion, and combines convolutional neural network fault diagnosis and voice feedback to obtain closed-loop protection commands; it integrates hardware operating parameters, standardized control signals, collaborative control commands, seamless switching commands, and closed-loop protection commands, performs hierarchical processing and performance verification, and obtains a system performance report.

[0127] Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention.

[0128] like Figure 3 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logic instructions from the memory 630 to execute a collaborative control method for an off-grid photovoltaic-storage inverter system based on a multimodal intelligent algorithm.

[0129] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0130] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms provided by the above methods.

[0131] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms provided by the above methods.

[0132] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A collaborative control method for an off-grid photovoltaic-storage inverter system based on a multimodal intelligent algorithm, characterized in that, include: Obtain hardware configuration requirements, which include hardware pin functions, relay status, and initial load prediction data; Based on hardware configuration requirements, a neural network dynamic optimization algorithm is used to analyze load prediction data in real time and dynamically adjust the modulation frequency to obtain a hardware control parameter optimization strategy. Based on the hardware control parameter optimization strategy, the power management unit dynamically adjusts the duty cycle of the modulation generator, switches between battery or grid power supply modes, and monitors temperature to trigger current limiting protection in real time, thus obtaining real-time updated operating status. Integrate real-time updated operational status data and perform performance verification to obtain optimized hardware operating parameters; Based on the optimized hardware operating parameters, multi-source data is collected in real time, and phase-locked loop algorithm is used to accurately track the phase to obtain a standardized control signal. The multi-source data includes volt voltage, battery current, and original grid phase signal. Based on standardized control signals, photovoltaic voltage and current data, battery status and sudden light change environmental parameters are collected synchronously to form a multi-dimensional input vector. Based on a multi-dimensional input vector, a neural network is used to dynamically adjust the number of particles and the iteration step size to obtain a tracking parameter set. Based on the tracking parameter set, the particle swarm is initialized with voltage and current as dimensions, the power fitness function is calculated, and the particle positions are updated to the maximum power point to obtain the circuit duty cycle adjustment command. Based on the circuit duty cycle adjustment command, the input variables are parsed by a fuzzy logic controller, and the circuit charging current is adjusted in real time to obtain the battery constant current control command. The circuit duty cycle adjustment command and the battery constant current control command are fused to output a cooperative control command. Based on collaborative control commands, historical load data and real-time status indicators are acquired, load demand prediction curves are generated through cloud model processing, and working mode program decisions are driven to output seamless switching commands. Based on seamless switching instructions, parameters are monitored in real time, data verification and feature extraction are performed, and a standardized collaborative control dataset is generated. Based on the standardized collaborative control dataset, deep reinforcement learning optimization algorithms, fuzzy control and cloud prediction algorithms are integrated, and temperature anomalies are diagnosed in real time through convolutional neural networks to generate fault risk level assessment and optimized control strategies. Based on the fault risk level assessment and optimized control strategies, the voice module is driven by pins to broadcast the status according to the warning level, generating audible status feedback signals. The optimized control strategies and audible status feedback signals are integrated to obtain closed-loop protection instructions. Integrate hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions to generate a full-link collaborative input matrix; Based on the end-to-end collaborative input matrix, collaborative processing is performed on the driver layer, algorithm layer, and interaction layer to output optimized control flow; Based on optimized control flow, energy is verified step by step to obtain verification results; Based on the verification results, dynamic response limit tests are conducted to obtain a system performance report.

2. The collaborative control method for off-grid photovoltaic-storage inverter system based on multimodal intelligent algorithm according to claim 1, characterized in that, Based on optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop (PLL) algorithm is used for precise phase tracking to obtain a standardized control signal. The multi-source data includes volt-voltage, battery current, and raw grid phase signals, including: Based on the optimized hardware operating parameters, multi-source data is collected in real time, including raw signals of photovoltaic voltage, battery current, and grid phase, and environmental parameters, including temperature and light intensity, are also acquired simultaneously. Based on multi-source data, an application signal processing program is used to convert the multi-source data into real physical values ​​to obtain preprocessed multi-source data. Based on the preprocessed multi-source data, a phase-locked loop algorithm is used for precise phase tracking to obtain the tracking results; The tracking results are compared with and corrected to the theoretical phase angle to obtain a standardized control signal.

3. The collaborative control method for off-grid photovoltaic-storage inverter systems based on multimodal intelligent algorithms according to claim 2, characterized in that, Based on collaborative control commands, historical load data and real-time status indicators are acquired, and load demand prediction curves are generated through cloud-based model processing. This drives the operating mode program to make decisions and outputs seamless switching commands, including: Based on collaborative control commands, historical load data and real-time status flags are obtained, and load time dependency characteristics are analyzed through cloud models to generate load demand prediction curves. Based on the load demand forecast curve, the working mode switching program decision logic is driven to output seamless switching instructions.

4. A collaborative control system for an off-grid photovoltaic-storage inverter system based on a multimodal intelligent algorithm, wherein the system implements the method as described in any one of claims 1 to 3, characterized in that, include: The acquisition module is used to acquire hardware configuration requirements and process load prediction data through a neural network dynamic optimization algorithm to obtain optimized hardware operating parameters. Based on the optimized hardware operating parameters, multi-source data is collected in real time, and a phase-locked loop algorithm is used for precise phase tracking to obtain standardized control signals. The multi-source data includes volt voltage, battery current, and original grid phase signals. The processing module is used to acquire photovoltaic voltage and current data, battery status, and parameters of sudden changes in light intensity based on standardized control signals, perform photovoltaic tracking, and manage the battery in conjunction with a fuzzy logic controller, outputting coordinated control commands; based on the coordinated control commands, it acquires historical load data and real-time status flags, generates load demand prediction curves through cloud model processing, drives the working mode program decision, and outputs seamless switching commands; based on the seamless switching commands, it processes real-time monitoring data through multi-algorithm fusion, and combines convolutional neural network fault diagnosis and voice feedback to obtain closed-loop protection commands; By integrating hardware operating parameters, standardized control signals, collaborative control instructions, seamless switching instructions, and closed-loop protection instructions, hierarchical processing and performance verification are performed to obtain a system performance report.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the collaborative control method for off-grid photovoltaic-storage inverter system based on multimodal intelligent algorithm as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the collaborative control method for off-grid photovoltaic-storage inverter system based on multimodal intelligent algorithm as described in any one of claims 1 to 3.

Images