Energy management control method and system for photovoltaic off-grid system based on hydrogen energy storage

Abstract

This invention discloses an energy management and control method and system for off-grid photovoltaic systems based on hydrogen energy storage. Addressing the problems of spatiotemporal imbalance between energy supply and demand, insufficient coordinated control of the electric-hydrogen hybrid energy storage, low system energy self-sufficiency, and poor operational stability in off-grid photovoltaic systems, it constructs a closed-loop energy flow model of photovoltaic-electrolyzer-hydrogen storage tank-fuel cell-load and designs a three-layer hierarchical distributed control system. The upper layer performs predictive economic scheduling, the middle layer completes real-time power allocation with rolling optimization, and the lower layer achieves device-level adaptive precision control. An intelligent control platform is built by integrating edge computing and a cloud platform, and a graded fault-tolerant mechanism for abnormal operating conditions is set up. This invention achieves intelligent management and control of the entire generation, storage, and utilization chain, significantly improving the system's energy self-sufficiency, operational economy, and safety. It supports modular expansion and is adaptable to various off-grid energy supply application scenarios.

Description

Technical Field

[0001] This invention relates to the field of energy management technology for new energy microgrids, specifically to an energy management and control method and system for off-grid photovoltaic systems based on hydrogen energy storage. Background Technology

[0002] Photovoltaic power generation, as a clean and renewable energy source, is a core energy supply solution for remote rural areas, communication base stations, border outposts, islands, and other areas without large power grid coverage. However, the inherent intermittency, volatility, and randomness of photovoltaic output lead to a common problem of energy supply and demand imbalance in off-grid photovoltaic systems. Therefore, it is necessary to match them with energy storage systems to realize the spatial and temporal transfer of energy and ensure the stability of power supply.

[0003] Existing off-grid photovoltaic systems mostly use electrochemical energy storage such as lithium-ion batteries. Although they have the advantages of fast response speed and high short-term conversion efficiency, they also have inherent shortcomings such as short cycle life, rapid degradation in deep charge and discharge scenarios, poor performance in low-temperature environments, high cost of long-term large-capacity energy storage, and environmental pollution risks after retirement. They cannot meet the requirements of long-term stable operation in off-grid scenarios.

[0004] Hydrogen energy storage boasts advantages such as large storage capacity, long storage period, strong environmental adaptability, and zero carbon emissions. It can achieve bidirectional conversion between electricity and hydrogen energy and is highly compatible with off-grid photovoltaic systems. However, existing energy management technologies for hydrogen energy storage photovoltaic off-grid systems lack sufficient research on the collaborative control mechanism of electricity-hydrogen hybrid energy storage, making it impossible to achieve cross-timescale energy collaborative scheduling. The control architecture suffers from high hierarchical coupling, a disconnect between economic scheduling and real-time control, and insufficient refined control of core hydrogen energy equipment. It also lacks a comprehensive hierarchical fault-tolerant mechanism and intelligent operation and maintenance optimization capabilities, making it difficult to guarantee the long-term efficient, safe, and stable operation of the system.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage, comprising a closed-loop energy flow module, a three-layer hierarchical distributed control unit, an IoT intelligent control platform, and an abnormal operating condition fault-tolerant unit; the energy transmission link of the closed-loop energy flow module is sequentially a photovoltaic array, an electrolyzer, a hydrogen storage tank, a fuel cell, and a load, forming a closed-loop energy flow of generation-storage-use; The three-layer distributed control unit communicates with each device of the closed-loop energy flow module to perform full-link energy scheduling and control; the IoT intelligent control platform communicates with the three-layer hierarchical distributed control unit to realize system data acquisition, remote monitoring and operation optimization; the abnormal operating condition fault-tolerant unit communicates with each device of the closed-loop energy flow module and the three-layer distributed control unit to perform hierarchical protection control of system abnormal states.

[0006] Optionally, the closed-loop energy flow module includes a photovoltaic array, a DC / DC converter, an electrolyzer, a hydrogen storage tank, a pressure sensor, a temperature sensor, a fuel cell, a DC / AC converter, a load, and a data acquisition subunit; the output of the photovoltaic array is connected to the power supply terminals of the electrolyzer and the load via the DC / DC converter; the hydrogen output terminal of the electrolyzer is connected to the inlet terminal of the hydrogen storage tank, and the outlet terminal of the hydrogen storage tank is connected to the hydrogen input terminal of the fuel cell; The output of the fuel cell is connected to the power supply of the load via a DC / AC converter; the data acquisition subunit is connected to the corresponding detection terminals of the photovoltaic array, electrolyzer, hydrogen storage tank, fuel cell, and load, respectively, for collecting the operating power, hydrogen pressure, equipment temperature, and load power data of each node; the output of the data acquisition subunit is connected to the input of the three-layer hierarchical distributed control unit.

[0007] Optionally, the three-layer hierarchical distributed control unit includes an upper-layer economic scheduling subunit, a middle-layer real-time power allocation subunit, and a lower-layer device-level control subunit; the output terminal of the upper-layer economic scheduling subunit is connected to the input terminal of the middle-layer real-time power allocation subunit, and the output terminal of the middle-layer real-time power allocation subunit is connected to the input terminal of the lower-layer device-level control subunit. The upper-level economic dispatch subunit is used to acquire weather forecast data, load demand forecast data, and equipment operating parameters. With the goal of minimizing the system's full-cycle operating cost, it performs day-ahead economic dispatch and generates energy allocation benchmark strategies for each time period within the day. The mid-level real-time power allocation subunit receives the upper-level energy allocation benchmark strategy, combines it with real-time collected photovoltaic output and load power data, and performs rolling optimization of real-time power allocation with the goal of minimizing the system supply and demand power deviation, generating operation instructions and start / stop control signals for each device; the lower-level device-level control subunit receives the operation instructions from the mid-level subunit, and uses a control algorithm combining PID and fuzzy control to perform closed-loop adjustment of the operating parameters of the electrolyzer and fuel cell.

[0008] Optionally, the objective function of the upper-level economic scheduling subunit is: In the formula, The total operating cost for the scheduling cycle, Total number of time periods The time period number. For time period Operating costs For time period Equipment maintenance costs; The parameters in the formula correspond to the operating power of the electrolyzer / fuel cell, the duration of a single time period, the unit energy consumption conversion cost, the maintenance cost coefficient for a single start-up and shutdown, and the number of start-ups and shutdowns, respectively. The constraints of the upper-level economic dispatch subunit include power balance, electrolyzer / fuel cell operating power and start-stop times, and hydrogen storage tank pressure constraints. A mixed integer linear programming algorithm is used to solve and generate an energy allocation benchmark strategy. The PID and fuzzy control composite algorithm of the lower-level device-level control subunit is used to measure power deviation. and rate of change of deviation The input is the PID parameter correction value, and the formula for the control output value is: In the formula, Sampling time, For real-time PID parameters, The sampling period; The fuzzy controller input The universe of discourse is [-6,6], and the universes of discourse for the output parameter corrections are [-3,3], [-0.06,0.06], and [-0.3,0.3], respectively. The fuzzy subsets are all [-6,6]. The membership function adopts a triangular function. After reasoning through preset fuzzy rules, the centroid method is used to defuzzify the parameters and complete the online adaptive update of PID parameters.

[0009] Optionally, the mid-level real-time power allocation subunit adopts a model predictive control algorithm to solve for the optimal power allocation solution with a preset rolling optimization window as the period. The duration of the rolling optimization window is longer than the lower-level control period and shorter than the upper-level scheduling period. The discrete state-space prediction model of the algorithm is as follows: In the formula, Let be the system state vector. To control the input vector, For the system output vector, This is the corresponding coefficient matrix; The algorithm's objective function is: In the formula, To predict the time domain, To control the time domain, These are the weighting coefficients. Due to the power deviation between supply and demand, To control the amount of input variation; The algorithm constraints are consistent with the upper-level economic scheduling constraints. A quadratic programming algorithm is used to solve the problem within each rolling window, and the control quantity in the first control time domain is issued as a real-time running instruction.

[0010] Optionally, the IoT intelligent control platform includes an edge computing gateway and a cloud platform service subunit; the edge computing gateway is communicatively connected to the data acquisition subunit, the three-layer distributed control unit, and the cloud platform service subunit, and is used to perform data preprocessing, local edge computing, and bidirectional parameter transmission. The cloud platform service sub-unit includes a remote monitoring module, a fault diagnosis module, and a self-learning optimization module. The remote monitoring module is used for remote visualization of the system's operating status and issuance of control commands. The fault diagnosis module adopts a random forest classification model, with input feature vectors including photovoltaic power output fluctuation rate, equipment power / temperature deviation, hydrogen storage tank pressure change rate, and system power imbalance. The output is the fault type and location. After training with historical fault datasets, it completes real-time fault reasoning. The self-learning optimization module adopts a deep reinforcement learning algorithm, with photovoltaic / load prediction deviation, operating cost, and equipment efficiency as the state space, upper-level scheduling weight coefficient and constraint threshold as the action space, and negative values ​​of operating cost and power imbalance as the reward function. The scheduling parameters are continuously optimized through offline training and online iteration.

[0011] Optionally, the abnormal operating condition fault-tolerant unit includes an abnormality detection module and a hierarchical fault-tolerant control module; the input of the abnormality detection module is connected to the data acquisition subunit of the closed-loop energy flow module, and is used to collect and determine in real time whether the system's operating status is in an abnormal condition, including hydrogen pressure exceeding the limit, electrolyzer temperature exceeding the limit, fuel cell temperature exceeding the limit, photovoltaic output abnormality, and load power sudden change; the input of the hierarchical fault-tolerant control module is connected to the output of the abnormality detection module, and the output of the hierarchical fault-tolerant control module is connected to the control terminals of the three-layer distributed control unit and each device of the closed-loop energy flow module; The hierarchical fault-tolerant control module is pre-set with hierarchical control strategies corresponding to the levels of abnormal operating conditions. When a minor abnormal operating condition is detected, a load reduction adjustment command is generated and sent to the three-layer hierarchical distributed control unit to adjust the operating power of the corresponding equipment. When a severe abnormal operating condition is detected, a safety mode switching command is generated to cut off the operating circuit of the corresponding faulty equipment and start the backup operating mode to ensure the power supply of the system's basic load and the safety of the equipment.

[0012] The control method for the energy management and control system of a photovoltaic off-grid system based on hydrogen energy storage includes the following steps: S1. Synchronously collect time-series operation data of the entire process of photovoltaic off-grid system power generation, electro-hydrogen conversion, hydrogen energy storage, hydrogen-electric conversion and electricity consumption, and construct a dynamic mapping model of energy conversion in the entire process; S2. Based on the dynamic mapping model, and following the time-dimensional decoupling rules of long-cycle planning, medium-cycle optimization, and short-cycle execution, generate energy allocation schemes and equipment operation control commands for the entire process. S3. Construct a collaborative control link between local real-time data processing and cloud-based model iterative optimization to complete the dynamic update of control logic; S4. Calculate the deviation between the system's operating status and the normal operating range in real time, and adjust the control strategy according to the preset hierarchical response rules.

[0013] Optionally, step S3 is as follows: S301. Perform local real-time preprocessing on the collected time-series operation data of the entire process, extract system operation characteristic quantities, and complete the verification and dynamic correction of local real-time control commands based on the characteristic quantities; the system operation characteristic quantities include photovoltaic power output fluctuation rate, electrolyzer input power deviation, electrolyzer temperature deviation, hydrogen storage tank pressure change rate, fuel cell output power deviation, fuel cell temperature deviation, and system power imbalance. S302. Synchronize the preprocessed running data and the execution results of control commands to the cloud according to the preset upload cycle to build a full-cycle running dataset of the system; S303. Based on the full-cycle running dataset, iteratively train the dynamic mapping model of energy conversion in the whole process, and update the model's feature weights and mapping relationships. Meanwhile, based on the full-cycle operation dataset, a fault classification model is trained using the random forest algorithm, and a scheduling parameter optimization model is trained using the deep reinforcement learning algorithm. S304. The optimized dynamic mapping model parameters, fault classification model, and scheduling parameter optimization model are sent to the local control terminal to complete the dynamic update of the local control logic and realize the two-way collaboration between local real-time control and cloud iterative optimization.

[0014] Optionally, step S4 is specifically as follows: S401. Compare the extracted real-time operating feature quantities with the normal operating condition range output by the dynamic mapping model, calculate the deviation of the system operating state, and classify the system operating state level based on the deviation value; the formula for calculating the deviation is: In the formula, The deviation from the operating state. For the total number of characteristic quantities, For the first Real-time values ​​of each feature quantity For the first The rated values ​​of each characteristic quantity For the first The upper and lower limits of normal operating conditions for each characteristic quantity; S402, When the operating status deviates Within the first preset range At that time, a control parameter fine-tuning instruction is generated to perform closed-loop correction on the currently executed control instruction, bringing the system back to the normal operating range; S403, When the operating status deviates... In the second preset range When necessary, suspend unnecessary control actions in the current cycle, regenerate the power allocation scheme for the current cycle, and limit the operating limits of each component; S404, When the operating status deviates... In the third preset interval When this happens, the safety operation logic is triggered, cutting off the operation path of unnecessary links and locking the operation status of the core energy conversion and power supply links; in, It is slightly off the threshold. It is a severe deviation from the threshold, and .

[0015] This invention provides an energy management and control method and system for off-grid photovoltaic systems based on hydrogen energy storage, which has the following beneficial effects: This invention uses a closed-loop energy flow model to convert surplus photovoltaic power into hydrogen energy for long-term storage via an electrolyzer. During periods of insufficient photovoltaic output or peak load, hydrogen energy is converted into electricity through fuel cells to supply the load. This can achieve real-time power balance at the second to minute level through lower-level equipment control, thus smoothing out short-term fluctuations in photovoltaic output. It can also achieve long-term energy allocation from the day to the next day through upper-level economic dispatch. This invention designs a three-layer distributed control system. The upper layer performs day-ahead economic scheduling based on weather and load forecast data, generating an energy allocation benchmark strategy with the goal of minimizing the system's full-cycle operating cost, thus ensuring the long-term economic efficiency of the system. The middle layer uses model predictive control to execute rolling optimization of real-time power allocation, correcting the scheduling strategy based on real-time operating data, achieving real-time balance between system supply and demand, and ensuring the stability of the system's power supply. The lower layer employs an adaptive control algorithm combining PID and fuzzy control, which can adjust PID control parameters online, perfectly adapting to the highly nonlinear operating characteristics of electrolyzers and fuel cells, and significantly improving the dynamic response performance and operating efficiency of the equipment.

[0016] This invention integrates IoT technology to build an intelligent control platform combining edge computing and cloud platforms. Through an edge computing gateway, it achieves real-time data acquisition, preprocessing, and local edge computing, ensuring real-time control performance. The cloud platform enables remote visual monitoring of system operation status, intelligent fault diagnosis, and self-learning optimization. Specifically, the fault diagnosis module accurately identifies the type and location of system faults based on a pre-trained fault classification model, significantly reducing system maintenance difficulty and costs. The self-learning optimization module iteratively optimizes the scheduling prediction model and control parameters based on historical operating data, allowing the system's control strategy to continuously optimize with changes in the operating environment and equipment status, ensuring long-term system performance and cost-effectiveness.

[0017] This invention also features a robust fault-tolerant mechanism for abnormal operating conditions and establishes a hierarchical fault-tolerant control system. It can monitor operating parameters such as hydrogen pressure, equipment temperature, photovoltaic output, and load power in real time, accurately identifying different levels of abnormal operating conditions. Furthermore, the system supports modular expansion, allowing for flexible adjustment of system capacity and equipment configuration according to different application scenarios and power supply requirements. It perfectly adapts to various off-grid application scenarios such as rural microgrids, communication base stations, border outposts, and islands, providing a standardized control paradigm for industrial-grade hydrogen energy storage photovoltaic off-grid systems and promoting the integration and evolution of industrial control systems towards green energy + intelligent control. Attached Figure Description

[0018] Figure 1 This diagram illustrates the system framework of the invention. Figure 2 This is a flowchart illustrating the method of the invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Please see Figure 1 This embodiment provides a specific implementation of an energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage. The core components of the system include a closed-loop energy flow module, a three-layer distributed control unit, an IoT intelligent control platform, and an abnormal operating condition fault-tolerant unit.

[0021] The closed-loop energy flow module's hardware components include a photovoltaic array, a DC / DC converter, an alkaline electrolyzer, a high-pressure hydrogen storage tank, a pressure sensor, a temperature sensor, a proton exchange membrane fuel cell, a DC / AC converter, an AC load, a DC load, and a data acquisition subunit. The output of the photovoltaic array is connected to the input of a unidirectional DC / DC converter. The output of the unidirectional DC / DC converter is divided into two paths: one connected to the power supply of the DC load, and the other connected to the power supply of the electrolyzer. The hydrogen output of the electrolyzer is connected to the inlet of the high-pressure hydrogen storage tank via a check valve, and the outlet of the high-pressure hydrogen storage tank is connected to the hydrogen input of the fuel cell via a pressure reducing valve. The output terminal connects to the DC side of the bidirectional DC / AC converter, and the AC side of the bidirectional DC / AC converter connects to the power supply terminal of the AC load. At the same time, the DC side of the bidirectional DC / AC converter can be connected to the power supply terminal of the DC load to realize bidirectional power transmission. The data acquisition subunit includes voltage sensors, current sensors, pressure sensors, and temperature sensors, which are installed at the corresponding detection positions of the photovoltaic array, electrolyzer, hydrogen storage tank, fuel cell, and load, respectively, to collect the operating voltage, current, power, hydrogen pressure, equipment temperature, and load power data of each node. The output terminal of the data acquisition subunit is connected to the input terminal of the three-layer hierarchical distributed control unit through the fieldbus.

[0022] The three-layer hierarchical distributed control unit is implemented using an industrial controller, consisting of an upper-layer economic scheduling subunit, a middle-layer real-time power allocation subunit, and a lower-layer device-level control subunit. The upper-layer economic scheduling subunit runs in the upper-layer scheduling thread of the industrial controller, with a scheduling cycle of 24 hours and a scheduling time step of 1 hour. It acquires local photovoltaic output forecast data, weather forecast data, load demand forecast data, and equipment operating efficiency parameters. With the goal of minimizing the system's total operating cost, it performs day-ahead economic scheduling and solves for energy allocation benchmark strategies such as the electrolyzer operating power benchmark, fuel cell operating power benchmark, and hydrogen storage tank pressure benchmark for 24 time periods within the day. These strategies are then distributed to the middle-layer real-time power allocation subunit.

[0023] The mid-level real-time power allocation subunit runs in the mid-level real-time control thread of the industrial controller. The control cycle is set to 15 minutes. It adopts a model predictive control algorithm and the rolling optimization window is set to 4 control cycles. It receives the energy allocation baseline strategy sent down from the upper level, and at the same time receives the real-time photovoltaic output and real-time load power data uploaded by the data acquisition subunit. With the goal of minimizing the power deviation between system supply and demand, it solves the optimal power allocation solution in the finite time domain, generates real-time operating power commands and equipment start-stop control signals for the electrolyzer and fuel cell, and sends them down to the lower-level equipment-level control subunit.

[0024] The lower-level device-level control subunit runs in the lower-level device control thread of the industrial controller, with a control cycle set to 100 milliseconds. It receives device operating power commands from the middle layer and uses an adaptive control algorithm combining PID and fuzzy control to perform closed-loop regulation of the input power of the electrolyzer and the output power of the fuel cell. The fuzzy controller uses the deviation between the device's set power and the actual acquired power... Deviation change rate As input, the fuzzy domain of the input variable is set to [-6,6]. The output variable is the correction amount of the proportional coefficient, integral coefficient, and derivative coefficient of the PID controller. Through inference using a preset fuzzy rule base, the real-time correction value of the PID parameter is obtained after defuzzification, realizing online adaptive adjustment of the PID parameter. The output control signal is sent to the matching actuator of the electrolyzer and fuel cell to complete the precise control at the equipment level.

[0025] The IoT intelligent control platform includes an edge computing gateway and a cloud platform service subunit. The edge computing gateway adopts an industrial-grade edge gateway and is connected to the field data acquisition subunit and the three-layer hierarchical distributed control unit via Ethernet communication. It performs preprocessing on the collected real-time operating data, including filtering, noise reduction, and format standardization. At the same time, it executes local edge computing tasks to monitor the device operating status in real time. The preprocessed data is uploaded to the cloud platform service subunit via 4G / 5G wireless network. It also receives optimized control parameters issued by the cloud platform service subunit and updates the scheduling and control parameters of the three-layer hierarchical distributed control unit.

[0026] The cloud platform service subunit is deployed on Alibaba Cloud servers and includes a remote monitoring module, a fault diagnosis module, and a self-learning optimization module. The remote monitoring module provides a web-based visual interface to display the system's real-time operating data, equipment status, and energy flow, and supports remote control command issuance. The fault diagnosis module uses a fault classification model trained based on the random forest algorithm, with historical fault operation data as the training set. It extracts features from the uploaded real-time operation data, identifies the system's fault type and location, generates fault handling prompts, and pushes them to maintenance personnel. The self-learning optimization module uses a reinforcement learning algorithm to iteratively optimize the photovoltaic output prediction model, load demand prediction model, and scheduling optimization parameters of the upper-level economic scheduling subunit based on the system's historical operating data and scheduling execution results. It updates the generation parameters of the energy allocation benchmark strategy, thereby achieving continuous optimization of the system control strategy.

[0027] The anomaly detection module of the fault-tolerant unit collects real-time data on hydrogen pressure in the hydrogen storage tank, operating temperature of the electrolyzer, operating temperature of the fuel cell, photovoltaic output fluctuation rate, and load power fluctuation rate. This data is compared with preset thresholds to determine the level of abnormal operating conditions. When hydrogen pressure or equipment temperature exceeds the mild threshold, or when photovoltaic output or load power experiences a small change, it is classified as a mild abnormal condition. The tiered fault-tolerant control module generates a load reduction adjustment command, which is sent to the three-layer distributed control unit to adjust the operating power of the corresponding equipment, bringing the system operating parameters back to the normal range and maintaining continuous system operation. When hydrogen pressure or equipment temperature exceeds the severe threshold, or when equipment experiences serious faults such as short circuits or leaks, it is classified as a severe abnormal condition. The tiered fault-tolerant control module generates a safety mode switching command, cutting off the operating circuit of the corresponding faulty equipment, closing the power supply circuit of the electrolyzer and the inlet valve of the hydrogen storage tank, and simultaneously activating the standby operating mode to control the fuel cell to maintain continuous power supply to the basic load, ensuring the safe operation of the equipment and the system.

[0028] Please see Figure 2 This embodiment also provides a specific implementation method for an energy management and control method for a photovoltaic off-grid system based on hydrogen energy storage, applied to the above-mentioned system. The specific steps are as follows: The first step is to establish a closed-loop energy flow from photovoltaic cells to electrolyzers, hydrogen storage tanks, fuel cells, and loads. The data acquisition subunit collects data on the operating voltage, current, power, hydrogen pressure, equipment temperature, and load power of each node in the closed-loop energy flow. The second step involves executing the system's energy scheduling and control through a three-layer distributed control architecture, which is divided into three sub-steps: (1) Upper-level economic dispatch steps: obtain weather forecast data, photovoltaic output forecast data, load demand forecast data, and equipment operating parameters; with the goal of minimizing the system's full-cycle operating cost, execute day-ahead economic dispatch and generate energy allocation benchmark strategies for each time period within the day. (2) Mid-level real-time power allocation step: Receive the energy allocation benchmark strategy generated by the upper layer, combine the real-time collected photovoltaic output and load power data, take the minimum system supply and demand power deviation as the goal, use the model predictive control algorithm to perform rolling optimization real-time power allocation, and generate the operation instructions and start / stop control signals of each device. (3) Lower-level equipment control steps: receive the running instructions generated by the middle layer, and use an adaptive control algorithm combining PID and fuzzy control to perform closed-loop adjustment of the operating parameters of the electrolyzer and fuel cell; The third step involves using an IoT control platform that integrates edge computing and cloud platforms to achieve remote monitoring, fault diagnosis, and self-learning optimization of the system. This process is divided into three sub-steps: (1) The collected real-time running data is preprocessed and local edge computing is performed through the edge computing gateway, and the preprocessed data is uploaded to the cloud platform; (2) Remote visualization monitoring of system operation status is achieved through cloud platform. A pre-trained random forest fault classification model is used to diagnose faults in the uploaded operation data, identify fault types and locations, and generate fault handling prompts. (3) Based on the historical operation data and scheduling execution results of the system, the reinforcement learning algorithm is used to iteratively learn and optimize the prediction model and scheduling parameters of economic scheduling, and update the generation parameters of energy allocation strategy; The fourth step is to monitor abnormal operating conditions of the system in real time and execute corresponding hierarchical fault-tolerant control, which is divided into three sub-steps: (1) Collect hydrogen pressure, equipment temperature, photovoltaic output and load power data of each node of the closed-loop energy flow in real time, compare them with the preset threshold, and determine whether the system is in an abnormal working condition and the level of the abnormal working condition. (2) When a slight abnormal operating condition is detected, a load reduction adjustment command is generated to adjust the operating power of the corresponding equipment so that the system operating parameters return to the normal range and the system continues to operate. (3) When a severe abnormal operating condition is detected, a safety mode switching command is generated to cut off the operating circuit of the corresponding faulty equipment and start the standby operating mode to ensure the power supply of the system's basic load and the safety of the core equipment.

[0029] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage, characterized in that, It includes a closed-loop energy flow module, a three-layer distributed control unit, an IoT intelligent control platform, and an abnormal operating condition fault-tolerant unit; the energy transmission link of the closed-loop energy flow module is, in sequence, a photovoltaic array, an electrolyzer, a hydrogen storage tank, a fuel cell, and a load, forming a closed-loop energy flow of generation-storage-use. The three-layer distributed control unit communicates with each device of the closed-loop energy flow module to perform full-link energy scheduling and control; the IoT intelligent control platform communicates with the three-layer hierarchical distributed control unit to realize system data acquisition, remote monitoring and operation optimization; the abnormal operating condition fault-tolerant unit communicates with each device of the closed-loop energy flow module and the three-layer distributed control unit to perform hierarchical protection control of system abnormal states.

2. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 1, characterized in that, The closed-loop energy flow module includes a photovoltaic array, a DC / DC converter, an electrolyzer, a hydrogen storage tank, a pressure sensor, a temperature sensor, a fuel cell, a DC / AC converter, a load, and a data acquisition subunit. The output of the photovoltaic array is connected to the power supply terminals of the electrolyzer and the load via the DC / DC converter. The hydrogen output terminal of the electrolyzer is connected to the inlet terminal of the hydrogen storage tank, and the outlet terminal of the hydrogen storage tank is connected to the hydrogen input terminal of the fuel cell. The output of the fuel cell is connected to the power supply of the load via a DC / AC converter; the data acquisition subunit is connected to the corresponding detection terminals of the photovoltaic array, electrolyzer, hydrogen storage tank, fuel cell, and load, respectively, for collecting the operating power, hydrogen pressure, equipment temperature, and load power data of each node; the output of the data acquisition subunit is connected to the input of the three-layer hierarchical distributed control unit.

3. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 1, characterized in that, The three-layer hierarchical distributed control unit includes an upper-layer economic scheduling subunit, a middle-layer real-time power allocation subunit, and a lower-layer device-level control subunit; the output terminal of the upper-layer economic scheduling subunit is connected to the input terminal of the middle-layer real-time power allocation subunit, and the output terminal of the middle-layer real-time power allocation subunit is connected to the input terminal of the lower-layer device-level control subunit. The upper-level economic dispatch subunit is used to acquire weather forecast data, load demand forecast data, and equipment operating parameters. With the goal of minimizing the system's full-cycle operating cost, it performs day-ahead economic dispatch and generates energy allocation benchmark strategies for each time period within the day. The mid-level real-time power allocation subunit receives the upper-level energy allocation benchmark strategy, combines it with real-time collected photovoltaic output and load power data, and performs rolling optimization of real-time power allocation with the goal of minimizing the system supply and demand power deviation, generating operation instructions and start / stop control signals for each device; the lower-level device-level control subunit receives the operation instructions from the mid-level subunit, and uses a control algorithm combining PID and fuzzy control to perform closed-loop adjustment of the operating parameters of the electrolyzer and fuel cell.

4. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 3, characterized in that, The objective function of the upper-level economic scheduling subunit is: In the formula, The total operating cost for the scheduling cycle, The total number of time periods. The time period number, For time period Operating costs For time period Equipment maintenance costs; The parameters in the formula correspond to the operating power of the electrolyzer / fuel cell, the duration of a single time period, the unit energy consumption conversion cost, the maintenance cost coefficient for a single start-up and shutdown, and the number of start-ups and shutdowns, respectively. The constraints of the upper-level economic dispatch subunit include power balance, electrolyzer / fuel cell operating power and start-stop times, and hydrogen storage tank pressure constraints. A mixed integer linear programming algorithm is used to solve and generate an energy allocation benchmark strategy. The PID and fuzzy control composite algorithm of the lower-level device-level control subunit is used to measure power deviation. and rate of change of deviation The input is the PID parameter correction value, and the formula for the control output value is: In the formula, Sampling time, For real-time PID parameters, The sampling period; The fuzzy controller input The universe of discourse is [-6,6], and the universes of discourse for the output parameter corrections are [-3,3], [-0.06,0.06], and [-0.3,0.3], respectively. The fuzzy subsets are all [-6,6]. The membership function adopts a triangular function. After reasoning through preset fuzzy rules, the centroid method is used to defuzzify the parameters and complete the online adaptive update of PID parameters.

5. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 3, characterized in that, The mid-level real-time power allocation subunit adopts a model predictive control algorithm to solve for the optimal power allocation solution with a preset rolling optimization window as the period. The duration of the rolling optimization window is longer than the lower-level control period and shorter than the upper-level scheduling period. The discrete state-space prediction model of the algorithm is as follows: In the formula, Let be the system state vector. To control the input vector, For the system output vector, This is the corresponding coefficient matrix; The algorithm's objective function is: In the formula, To predict the time domain, To control the time domain, These are the weighting coefficients. Due to the power deviation between supply and demand, To control the amount of input variation; The algorithm constraints are consistent with the upper-level economic scheduling constraints. A quadratic programming algorithm is used to solve the problem within each rolling window, and the control quantity in the first control time domain is issued as a real-time running instruction.

6. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 1, characterized in that, The IoT intelligent control platform includes an edge computing gateway and a cloud platform service subunit. The edge computing gateway is communicatively connected to the data acquisition subunit, the three-layer distributed control unit, and the cloud platform service subunit, and is used to perform data preprocessing, local edge computing, and bidirectional parameter transmission. The cloud platform service sub-unit includes a remote monitoring module, a fault diagnosis module, and a self-learning optimization module. The remote monitoring module is used for remote visualization of the system's operating status and issuance of control commands. The fault diagnosis module adopts a random forest classification model, with input feature vectors including photovoltaic power output fluctuation rate, equipment power / temperature deviation, hydrogen storage tank pressure change rate, and system power imbalance. The output is the fault type and location. After training with historical fault datasets, it completes real-time fault reasoning. The self-learning optimization module adopts a deep reinforcement learning algorithm, with photovoltaic / load prediction deviation, operating cost, and equipment efficiency as the state space, upper-level scheduling weight coefficient and constraint threshold as the action space, and negative values ​​of operating cost and power imbalance as the reward function. The scheduling parameters are continuously optimized through offline training and online iteration.

7. The energy management and control system for a photovoltaic off-grid system based on hydrogen energy storage according to claim 1, characterized in that, The abnormal operating condition fault-tolerant unit includes an abnormality detection module and a hierarchical fault-tolerant control module. The input of the abnormality detection module is connected to the data acquisition subunit of the closed-loop energy flow module, and is used to collect and determine in real time whether the system is in an abnormal operating condition. The abnormal operating conditions include hydrogen pressure exceeding the limit, electrolyzer temperature exceeding the limit, fuel cell temperature exceeding the limit, photovoltaic output abnormality, and load power sudden change. The input of the hierarchical fault-tolerant control module is connected to the output of the abnormality detection module, and the output of the hierarchical fault-tolerant control module is connected to the control terminals of the three-layer distributed control unit and each device of the closed-loop energy flow module. The hierarchical fault-tolerant control module is pre-set with hierarchical control strategies corresponding to the levels of abnormal operating conditions. When a minor abnormal operating condition is detected, a load reduction adjustment command is generated and sent to the three-layer hierarchical distributed control unit to adjust the operating power of the corresponding equipment. When a severe abnormal operating condition is detected, a safety mode switching command is generated to cut off the operating circuit of the corresponding faulty equipment and start the backup operating mode to ensure the power supply of the system's basic load and the safety of the equipment.

8. A control method for an off-grid photovoltaic system energy management and control system based on hydrogen energy storage as described in any one of claims 1 to 7, characterized in that, Includes the following steps: S1. Synchronously collect time-series operation data of the entire process of photovoltaic off-grid system power generation, electro-hydrogen conversion, hydrogen energy storage, hydrogen-electric conversion and electricity consumption, and construct a dynamic mapping model of energy conversion in the entire process; S2. Based on the dynamic mapping model, and following the time-dimensional decoupling rules of long-cycle planning, medium-cycle optimization, and short-cycle execution, generate energy allocation schemes and equipment operation control commands for the entire process. S3. Construct a collaborative control link between local real-time data processing and cloud-based model iterative optimization to complete the dynamic update of control logic; S4. Calculate the deviation between the system's operating status and the normal operating range in real time, and adjust the control strategy according to the preset hierarchical response rules.

9. The control method for the energy management and control system of a photovoltaic off-grid system based on hydrogen energy storage according to claim 8, characterized in that, Step S3 is as follows: S301. Perform local real-time preprocessing on the collected time-series operation data of the entire process, extract system operation characteristic quantities, and complete the verification and dynamic correction of local real-time control commands based on the characteristic quantities; the system operation characteristic quantities include photovoltaic power output fluctuation rate, electrolyzer input power deviation, electrolyzer temperature deviation, hydrogen storage tank pressure change rate, fuel cell output power deviation, fuel cell temperature deviation, and system power imbalance. S302. Synchronize the preprocessed running data and the execution results of control commands to the cloud according to the preset upload cycle to build a full-cycle running dataset of the system; S303. Based on the full-cycle running dataset, iteratively train the dynamic mapping model of energy conversion in the whole process, and update the model's feature weights and mapping relationships. Meanwhile, based on the full-cycle operation dataset, a fault classification model is trained using the random forest algorithm, and a scheduling parameter optimization model is trained using the deep reinforcement learning algorithm. S304. The optimized dynamic mapping model parameters, fault classification model, and scheduling parameter optimization model are sent to the local control terminal to complete the dynamic update of the local control logic and realize the two-way collaboration between local real-time control and cloud iterative optimization.

10. The control method for the energy management and control system of a photovoltaic off-grid system based on hydrogen energy storage according to claim 8, characterized in that, Step S4 is as follows: S401. Compare the extracted real-time operating feature quantities with the normal operating condition range output by the dynamic mapping model, calculate the deviation of the system operating state, and classify the system operating state level based on the deviation value; the formula for calculating the deviation is: In the formula, The deviation from the operating state. For the total number of characteristic quantities, For the first Real-time values ​​of each feature quantity For the first The rated values ​​of each characteristic quantity For the first The normal operating limits of each characteristic quantity; S402, When the operating status deviates Within the first preset range At that time, a control parameter fine-tuning instruction is generated to perform closed-loop correction on the currently executed control instruction, bringing the system back to the normal operating range; S403, When the operating status deviates... In the second preset range When necessary, suspend unnecessary control actions in the current cycle, regenerate the power allocation scheme for the current cycle, and limit the operating limits of each component; S404, When the operating status deviates... In the third preset interval When this happens, the safety operation logic is triggered, cutting off the operation path of unnecessary links and locking the operation status of the core energy conversion and power supply links; in, It is slightly off the threshold. It is a severe deviation from the threshold, and .

Images