An early warning system for the combined operation of hydrogen production by electrolyzing water and the power grid

Abstract

The present invention belongs to the technical field of power grid early warning, and particularly relates to an early warning system for the combined operation of electrolytic water hydrogen production and the power grid, including the following modules: an electrolysis module for performing electrolysis reaction of water according to the power provided by the power grid to generate hydrogen; a monitoring module for real-time monitoring of the power grid status and the operating parameters of the electrolytic water hydrogen production unit; a data analysis module for analyzing the data collected by the monitoring module, specifically including: a power fluctuation response sub-module and an equipment monitoring sub-module; an early warning occurrence module, based on the output of the data analysis module, when the power grid status has an adverse impact on the electrolytic water hydrogen production unit or potential faults occur in the electrolytic hydrogen production unit itself, sending out an early warning signal to the operator. Through early fault warning, the present invention can help the operator to perform maintenance and repair in a timely manner, significantly reducing the downtime caused by faults and the related maintenance costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid early warning, and particularly relates to an early warning system for the combined operation of electrolytic water hydrogen production and the power grid. Background Art

[0002] In the combined operation system of electrolytic water hydrogen production and the power grid, it is crucial to make full use of electric power resources to improve energy efficiency and economic benefits. At present, as an effective way to convert electrical energy into chemical energy, electrolytic water hydrogen production can help balance the power grid load and store excess electrical energy. However, traditional electrolytic water hydrogen production systems usually operate under fixed operating conditions and cannot dynamically adapt to the power supply fluctuations of the power grid, which may lead to energy waste or an increased burden on the power grid.

[0003] Traditional early warning systems often use fixed electrolysis parameters (such as current and voltage), which limit the adjustment ability of the system when the power supply is excessive or insufficient, and cannot maximize the utilization of renewable energy or off-peak electricity.

[0004] Existing electrolysis systems often lack efficient fault prediction and maintenance strategies. Manual intervention is often required when a fault occurs, which not only increases the maintenance cost but also affects the continuous operation of the system. The inability to effectively respond to rapid changes in the power grid state, such as power surplus or shortage, may lead to a decrease in electrolysis efficiency or an additional power grid load. Summary of the Invention

[0005] In view of the above deficiencies of the prior art, the purpose of the present invention is to provide an early warning system for the combined operation of electrolytic water hydrogen production and the power grid. Through early fault warning, it can help operators perform maintenance and repair in a timely manner, significantly reducing the downtime caused by faults and related maintenance costs.

[0006] Based on the above purpose, the present invention provides an early warning system for the combined operation of electrolytic water hydrogen production and the power grid, including the following modules:

[0007] An electrolysis module for performing electrolysis reaction of water according to the power provided by the power grid to produce hydrogen;

[0008] A monitoring module for real-time monitoring of the power grid state and the operating parameters of the electrolytic water hydrogen production unit;

[0009] A data analysis module for analyzing the data collected by the monitoring module, specifically including:

[0010] A power fluctuation response sub-module, in response to the instability of the power grid power supply, adjusts the energy consumption of the electrolytic water hydrogen production unit in real time to adapt to power fluctuations. When it detects power surplus or shortage, it automatically issues an early warning and responds to the early warning by adjusting the input power of the electrolysis process;

[0011] The device monitoring sub-module uses topological data analysis to analyze the operating parameters of the electrolytic hydrogen production unit and predict potential faults;

[0012] The early warning generation module, based on the output of the data analysis module, sends an early warning signal to the operator when the grid status has an adverse impact on the electrolytic water hydrogen production unit or potential faults occur in the electrolytic hydrogen production unit itself.

[0013] Preferably, the monitoring module is equipped with a variety of sensors for collecting real-time power data of the grid, including voltage, current, and frequency;

[0014] At the same time, monitor the operating parameters of the electrolytic water hydrogen production unit, including electrolytic current, electrolytic voltage, water temperature, and the pressure and electrolyte concentration of the electrolytic cell;

[0015] The monitoring module communicates with the data analysis module through a wireless network and transmits all the collected data in real time.

[0016] Preferably, the power fluctuation response sub-module specifically includes:

[0017] Equipped with a fluctuation analysis algorithm to analyze the power supply fluctuations of the grid and judge whether the power is excessive or insufficient based on the monitored data;

[0018] When the power is excessive, automatically instruct the electrolysis module to increase the input power of the electrolytic current, accelerate the electrolysis speed of water, and increase the hydrogen production to utilize the excess power;

[0019] When the power is insufficient, also automatically instruct the electrolysis module to reduce the input power of the electrolytic current to reduce the burden on the grid and adjust the hydrogen production speed to match the available power;

[0020] When any instability in the power supply is detected, the power fluctuation response sub-module will cooperate with the early warning generation module to automatically send an early warning signal.

[0021] Preferably, the fluctuation analysis algorithm uses a graph neural network. In the graph neural network, the grid is regarded as a network graph, where each node represents a power station, a substation, or a consumption node, and the edge represents a transmission line. The graph neural network operates directly on the network graph to analyze the complex relationships and influences between nodes and is used to predict and analyze the power flow and fluctuations based on the network structure.

[0022] Preferably, the application of the graph neural network to power fluctuation analysis specifically includes:

[0023] Network modeling: Model the grid as a graph, where the nodes represent the entities in the grid, the edges represent the power connections, including transmission lines, and each node has its attributes, including power generation, consumption, and voltage level;

[0024] Feature vector setting: Set the feature vector X for each node i , which includes power supply, demand, and historical fluctuation data information. The features of the edges are transmission capacity, distance, or connection type;

[0025] Message passing mechanism: Use the message passing mechanism in the graph neural network. Each node updates its own state according to the states of its adjacent nodes, expressed as:

[0026] ;

[0027] Among them, 、 are the hidden states of node i and node j at the l-th layer respectively, e ij is the feature of the edge between node i and node j, is the set of neighbor nodes of node i, is the aggregation function over all neighbors, and f and g are learnable neural network functions;

[0028] Power fluctuation monitoring: Use the trained graph neural network model to predict the power state of each node. Compare the predicted state of the node with the actual state to determine whether the power is excessive or insufficient;

[0029] Set a threshold δ, and determine the state by calculating the power supply-demand difference d i of each node:

[0030] , if d i > δ, it is judged that the power is excessive; if d i < -δ, it is judged that the power is insufficient; among them, represents the supplied power of node i, represents the demanded power of node i;

[0031] According to the output of the graph neural network model, when it is detected that the power of a certain node or area is insufficient or excessive, an alarm is automatically triggered, and the input power parameters of the electrolytic water hydrogen production unit are adjusted.

[0032] Preferably, the topological data analysis method TDA in the device monitoring sub-module specifically includes:

[0033] Data mapping and network construction: Map the operating parameters of the electrolytic water hydrogen production unit to high-dimensional data points. The operating parameters are electrolytic current, electrolytic voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration. Use the high-dimensional data points to construct an abstract network, where the nodes represent actual measurement points, and the edges connect the nodes with similar attributes;

[0034] Topological feature extraction: Apply the persistent homology technique in TDA to analyze the topological characteristics of the data structure formed by parameters, focus on the "holes" or continuity breaks in the data, indicate abnormal or fault patterns, and transform the topological features obtained by persistent homology into feature vectors that can be used in machine learning;

[0035] Pattern recognition and fault prediction: Create a topological signature for each operating state, monitor the state changes by comparing the topological signatures at different time points, and use topological changes to detect abnormalities during operation. Abnormalities are manifested as sudden changes in the topological structure or newly emerging topological features;

[0036] Once a topological abnormality is detected, an early warning will be automatically triggered to indicate a fault or a change that requires attention. According to the results of topological analysis, dynamically adjust the thresholds of monitoring parameters to adapt to changes in the operating environment and the impact of equipment aging. The monitoring parameters include electrolytic current, electrolytic voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration.

[0037] Preferably, the persistent homology technique specifically includes:

[0038] Construct a point cloud: Consider the monitored electrolytic current, electrolytic voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration as points in a multi-dimensional space. The parameter values at each time point form a point, forming a point cloud data set;

[0039] Construct a filter: Define a "radius" parameter r to create hyperspheres that cover the point cloud. As r increases, the hyperspheres gradually expand, start to overlap with each other and form connections, and the formed connection structure is regarded as a simplified complex;

[0040] Calculate the Vietoris-Rips complex: For a given radius r, construct a Vietoris-Rips complex to represent the connections between data points. The complex includes points, lines, and triangles. As r increases, the complex becomes larger;

[0041] Generation of persistent diagrams and Betti diagrams: As the radius r changes, record the emerging and disappearing holes, that is, topological features. Holes represent circular structures or voids in the data. The emergence and disappearance of holes are related to state changes, and generate persistent diagrams and Betti diagrams to show the changes of topological features with r;

[0042] Analyze the persistent diagrams and Betti diagrams to identify key topological changes. A suddenly emerging hole indicates the emergence of a new abnormal state. According to the analysis results, predict potential faults and issue early warning signals.

[0043] Preferably, in the persistent diagram, the persistent homology results of each parameter set are represented by a series of barcodes. Each barcode corresponds to the survival time of a hole, and the set of barcodes forms the topological signature of the current state;

[0044] In the Bassett diagram, each point represents a hole. The abscissa is the scale at which the hole appears, and the ordinate is the scale at which it disappears.

[0045] Preferably, the threshold value for dynamically adjusting the monitoring parameters includes adjustment based on statistical analysis. Statistical methods are used to calculate the volatility of the historical data of each monitoring parameter. If the volatility of a certain parameter increases, the threshold range of this parameter needs to be widened.

[0046] Preferably, the statistical analysis includes calculating the mean, standard deviation, and coefficient of variation CV of each parameter, specifically including:

[0047] The mean is the average value of a set of data, used to represent the center point of the data set. For a given data set , the formula for calculating the mean μ is:

[0048] ;

[0049] where n is the number of data points, and x i is the value of a single data point;

[0050] The standard deviation is a statistic that measures the degree of dispersion of a data set, used to describe the deviation of data points from the mean. For the same data set X, the standard deviation σ is calculated as:

[0051] ;

[0052] The coefficient of variation is an index that measures relative volatility, used to compare the volatility of data of different magnitudes. The calculation of the coefficient of variation is: ; The larger the value of CV, the higher the volatility of the data.

[0053] Advantages of the present invention:

[0054] By continuously monitoring and responding to the fluctuations in the power supply of the power grid, the electrolytic water hydrogen production system of the present invention can effectively adapt to the excess or shortage of power supply. When the power supply is in excess, the system increases the input of electrolytic current to accelerate the electrolysis process of water, thereby increasing the hydrogen production rate and maximizing the utilization of surplus power. On the contrary, when the power supply is insufficient, the system can reduce the input of electrolytic current, slow down the electrolysis reaction rate, and reduce the burden on the power grid. The dynamic adjustment not only optimizes the energy use in the electrolytic water hydrogen production process, but also helps to maintain the stability of the power grid, improve the overall energy efficiency and resource utilization rate.

[0055] The device monitoring sub-module of the present invention uses advanced Topological Data Analysis (TDA) to monitor the operating parameters of the electrolytic water hydrogen production unit and predict potential failures. By continuously monitoring and analyzing the topological changes of operating parameters (electrolysis current, electrolysis voltage, water temperature, electrolyzer pressure, and electrolyte concentration), it can early identify abnormal states that may lead to equipment failures. This early fault warning can help operators perform maintenance and repairs in a timely manner, significantly reducing downtime and related maintenance costs caused by failures. At the same time, by dynamically adjusting the thresholds of monitoring parameters, the system can more flexibly adapt to changes in the operating environment and the impact of equipment aging, further enhancing the reliability of the system.

[0056] The present invention uses Graph Neural Network (GNN) to deeply analyze the power supply fluctuations of the power grid, and can simulate and analyze the power flow and fluctuations in the power grid. By modeling the power grid as a graph, where nodes represent various entities in the power grid (such as power generation stations, substations, consumption points), and edges represent power connections, GNN can directly operate on the network graph, analyze the complex relationships and influences between nodes. The graph-based analysis method not only improves the accuracy of power fluctuation analysis but also enhances the system's response ability to power fluctuations, making the operation more efficient. The system improves the automation level and efficiency of operation by automatically triggering warnings and adjusting the input power parameters of the electrolytic water hydrogen production unit, contributing to more efficient and stable operation. Brief Description of the Drawings

[0057] Figure 1 It is a schematic diagram of the data analysis module of an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of the system function module of an embodiment of the present invention. Detailed Embodiment

[0059] The present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0060] As Figure 1 - Figure 2 shown, an early warning system for the combined operation of electrolytic water hydrogen production and the power grid includes the following modules:

[0061] An electrolysis module for performing electrolysis reaction of water according to the power provided by the power grid to produce hydrogen;

[0062] The electrolysis module includes one or more electrolyzers, and each electrolyzer is filled with at least a pair of electrodes (anode and cathode) and an appropriate amount of electrolyte solution;

[0063] The electrolysis module is connected to the monitoring module, receives power supply information from the power grid, including current, voltage, and frequency; adjusts electrolysis parameters according to changes in the power grid state and power supply, such as current intensity, voltage level, and electrolyte concentration in the electrolyzer;

[0064] When there is an oversupply of electricity, the electrolysis module increases the input of electrolysis current to accelerate the decomposition of water molecules and improve the hydrogen production rate.

[0065] When there is a shortage of electricity supply, the electrolysis module reduces the input of electrolysis current through the control module to slow down the electrolysis reaction speed, so as to reduce the burden on the power grid.

[0066] The monitoring module is used to monitor the power grid status and the operating parameters of the water electrolysis hydrogen production unit in real time.

[0067] The data analysis module analyzes the data collected by the monitoring module, specifically including:

[0068] The power fluctuation response sub-module, in response to the instability of the power grid power supply, adjusts the energy consumption of the water electrolysis hydrogen production unit in real time to adapt to power fluctuations. When it detects an oversupply or shortage of electricity, it automatically issues a warning and responds to the warning by adjusting the input power of the electrolysis process.

[0069] The equipment monitoring sub-module uses topological data analysis method to analyze the operating parameters of the electrolytic hydrogen production unit, predict potential faults, and help the system reduce unexpected shutdowns and maintenance costs through predicting faults, ensuring continuous operation.

[0070] The warning occurrence module, based on the output of the data analysis module, when the power grid status has an adverse impact on the water electrolysis hydrogen production unit or potential faults occur in the electrolytic hydrogen production unit itself, sends a warning signal to the operator.

[0071] The monitoring module is equipped with a variety of sensors for collecting power data of the power grid in real time, including voltage, current and frequency.

[0072] At the same time, it monitors the operating parameters of the water electrolysis hydrogen production unit, including electrolysis current, electrolysis voltage, water temperature, and the pressure and electrolyte concentration of the electrolytic cell.

[0073] The monitoring module communicates with the data analysis module through a wireless network and transmits all the collected data in real time.

[0074] The power fluctuation response sub-module specifically includes:

[0075] Equipped with a fluctuation analysis algorithm to analyze the power supply fluctuation of the power grid, and judge whether the power is in oversupply or shortage according to the monitored data.

[0076] When the power is in oversupply, it automatically instructs the electrolysis module to increase the input power of the electrolysis current, accelerate the electrolysis speed of water, and increase the hydrogen production to utilize the excess power.

[0077] When the power is insufficient, the electrolysis module is also automatically commanded to reduce the input power of the electrolysis current to reduce the burden on the power grid and adjust the hydrogen production speed to match the available power;

[0078] When any instability in the power supply is detected, the power fluctuation response sub-module will cooperate with the warning generation module to automatically issue a warning signal.

[0079] The fluctuation analysis algorithm uses a graph neural network (GNN). In the graph neural network, the power grid is regarded as a network graph, where each node represents a power station, a substation or a consumption node, and the edges represent transmission lines. The graph neural network (GNN) operates directly on the network graph to analyze the complex relationships and influences between nodes, and is used to predict and analyze the power flow and fluctuations based on the network structure.

[0080] Applying the graph neural network to power fluctuation analysis is a tool for processing graph-structured data and is very suitable for analyzing power systems because the power network is essentially a complex network composed of nodes such as power generation stations, transmission lines, and distribution stations. By using GNN, the power flow and fluctuations in the power grid can be effectively simulated and analyzed. The following is a specific scheme for using GNN to analyze the power supply fluctuations in the power grid, which specifically includes:

[0081] Network modeling: Model the power grid as a graph, where the nodes represent the various entities in the power grid (power generation stations, substations, consumption points), and the edges represent power connections, including transmission lines. Each node has its attributes, and the attributes include power generation, consumption, and voltage level;

[0082] Feature vector setting: Set a feature vector X i for each node, which includes power supply, demand, and historical fluctuation data information. The features of the edges are transmission capacity, distance, or connection type;

[0083] Message passing mechanism: Use the message passing mechanism in the graph neural network. Each node updates its state according to the states of its neighboring nodes, which is expressed as:

[0084] ;

[0085] Among them, and are the hidden states of node i and node j at the l-th layer respectively, e ij is the feature of the edge between node i and node j, is the set of neighboring nodes of node i, is the aggregation function over all neighbors, and f and g are learnable neural network functions;

[0086] Power Fluctuation Monitoring: Use the trained graph neural network model to predict the power status of each node, and judge whether the power is excessive or insufficient by comparing the predicted status and the actual status of the node;

[0087] Set a threshold δ, and determine the status by calculating the power supply-demand difference d of each node i to determine the status:

[0088] , if d i > δ, it is judged that the power is excessive; if d i < -δ, it is judged that the power is insufficient; where, represents the supplied power of node i, represents the demanded power of node i;

[0089] According to the output of the graph neural network model, when it is monitored that the power of a certain node or area is insufficient or excessive, an early warning is automatically triggered, and the input power parameters of the electrolytic water hydrogen production unit are adjusted.

[0090] Through the above steps, the graph neural network can help the early warning system of the combined operation of electrolytic water hydrogen production and the power grid to more accurately understand and respond to the fluctuations of power supply, so as to achieve more efficient and stable operation. This method not only improves the response ability of the power system, but also enhances the intelligent level of the system.

[0091] The topological data analysis method TDA in the equipment monitoring sub-module specifically includes:

[0092] Data Mapping and Network Construction: Map the operating parameters of the electrolytic water hydrogen production unit into high-dimensional data points. The operating parameters are electrolytic current, electrolytic voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration. Use the high-dimensional data points to construct an abstract network, where the nodes represent actual measurement points, and the edges connect nodes with similar attributes;

[0093] Topological Feature Extraction: Apply the persistent homology technique in TDA to analyze the topological characteristics of the data structure formed by the parameters, focus on the "holes" or continuity breaks in the data, indicate abnormal or fault patterns, and convert the topological features obtained by persistent homology into feature vectors that can be used for machine learning;

[0094] Pattern Recognition and Fault Prediction: Create a topological signature for each operating state, monitor the change of the state by comparing the topological signatures at different time points, and use the topological change to monitor the abnormalities during operation. The abnormalities are manifested as sudden changes in the topological structure or newly emerging topological features;

[0095] Once a topological anomaly is detected, an early warning will be automatically triggered to indicate a fault or a change that requires attention. Based on the results of topological analysis, the thresholds of monitoring parameters will be dynamically adjusted to adapt to changes in the operating environment and the impact of equipment aging. The monitoring parameters include electrolysis current, electrolysis voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration.

[0096] Provide an intuitive topological map for the operator, showing the system state and the topological map that changes over time, helping the operator understand the dynamic changes of the system, and allowing the operator to input feedback to adjust the topological analysis parameters to make the algorithm more adaptable to specific operating conditions and equipment characteristics.

[0097] The persistent homology technique specifically includes:

[0098] Construct a point cloud: Consider the monitored electrolysis current, electrolysis voltage, water temperature, pressure of the electrolytic cell, and electrolyte concentration as points in a multi-dimensional space. The parameter values at each time point form a point, forming a point cloud data set;

[0099] Construct a filter: Define a "radius" parameter r to create hyperspheres that cover the point cloud. As r increases, the hyperspheres gradually expand, start to overlap with each other and form connections, and the formed connection structure is regarded as a simplified complex;

[0100] Calculate the Vietoris-Rips complex: For a given radius r, construct a Vietoris-Rips complex to represent the connections between data points. The complex includes points (0-dimensional simplices), lines (1-dimensional simplices), and triangles (2-dimensional simplices). As r increases, the complex becomes larger;

[0101] Generation of persistent diagrams and barcode diagrams: As the radius r changes, record the emerging and disappearing holes, that is, topological features. The holes represent circular structures or voids in the data, and the emergence and disappearance of the holes are related to state changes. Generate persistent diagrams and barcode diagrams to show the changes of topological features with r;

[0102] Analyze the persistent diagrams and barcode diagrams to identify key topological changes. A suddenly emerging hole indicates the emergence of a new abnormal state. Based on the analysis results, predict potential faults and issue early warning signals.

[0103] The calculations in persistent homology are simplified and represented by the following mathematical expressions: , where H m represents the m-dimensional homology group, and X r represents the Vietoris-Rips complex generated from the point cloud data under a given covering radius r.

[0104] In the persistence diagram, the persistence homology results for each parameter set are represented by a series of barcodes, where each barcode corresponds to the lifetime of a hole, and the set of barcodes forms the topological signature of the current state;

[0105] In the Betti diagram, each point represents a hole, with the abscissa being the scale at which the hole appears and the ordinate being the scale at which it disappears.

[0106] To further explain the application of the topological data analysis method (TDA) to the device monitoring sub-module of the present invention for predicting potential failures, a specific example (monitoring abnormal electrolyzer pressure) is given:

[0107] During the electrolytic water hydrogen production process, the pressure of the electrolyzer is a key parameter to ensure safety and efficiency. Too high or too low pressure may indicate potential equipment problems, such as poor sealing, failure of temperature control, or improper electrolyte concentration.

[0108] Use TDA to monitor the topological changes in the electrolyzer pressure and identify early the faults that lead to abnormal pressure.

[0109] Implementation steps:

[0110] Data collection:

[0111] Collect pressure data from the electrolyzer sensors at regular time intervals (e.g., every minute), and synchronously record the electrolysis current, electrolysis voltage, water temperature, and electrolyte concentration.

[0112] Point cloud construction:

[0113] Take the data collected each time (pressure, current, voltage, water temperature, concentration) as a point in a five-dimensional space, and the continuous data points form a point cloud.

[0114] Persistence homology calculation:

[0115] Apply the Vietoris-Rips complex construction method to construct a complex according to the set scale parameter (radius) and calculate its persistence homology.

[0116] Generate a persistence diagram and a Betti diagram, and record the appearance and disappearance of each topological feature (hole).

[0117] Topological signature analysis:

[0118] Generate the topological signature of the electrolyzer regularly (e.g., every day) and compare it with the historical signature, paying particular attention to the changes in the topological features related to pressure.

[0119] Abnormal monitoring and warning:

[0120] If a new topological hole appears or an existing hole suddenly disappears, especially the topological changes related to pressure changes, the system will automatically send out a warning signal.

[0121] The warning information will include potential fault types and recommended inspection or maintenance measures.

[0122] Maintenance and feedback: Conduct maintenance inspections according to the warnings, such as checking seals, adjusting the temperature control system, or regulating the electrolyte concentration. The maintenance results are fed back to the system for adjusting the data analysis model and optimizing the warning thresholds.

[0123] Dynamically adjusting the thresholds of monitoring parameters includes adjustments based on statistical analysis. Statistical methods are used to calculate the historical data volatility of each monitoring parameter. If the volatility of a certain parameter increases, the threshold range of this parameter needs to be widened.

[0124] Statistical analysis includes calculating the mean, standard deviation, and coefficient of variation CV for each parameter, specifically including:

[0125] The mean is the average value of a set of data, used to represent the center point of the data set. For a given data set (here the data set represents all readings of a specific parameter such as electrolysis current within a certain period of time), the formula for calculating the mean μ is:

[0126] ;

[0127] where n is the number of data points, and x i is the value of a single data point;

[0128] The standard deviation is a statistic that measures the dispersion degree of the data set, used to describe the deviation degree of data points from the mean. For the same data set X, the standard deviation σ is calculated as:

[0129] ;

[0130] The coefficient of variation is an index that measures relative volatility, used to compare the volatility of data of different magnitudes. The calculation of the coefficient of variation is: ; The larger the value of CV, the higher the volatility of the data.

[0131] In the equipment monitoring sub-module of the water electrolysis hydrogen production system, the above statistical methods are used to evaluate and adjust the monitoring thresholds of key operating parameters. The specific steps include:

[0132] Data collection: Regularly collect data of key parameters (such as electrolysis current, electrolysis voltage, water temperature, electrolyzer pressure, and electrolyte concentration).

[0133] Statistical analysis: Calculate the mean, standard deviation, and coefficient of variation for each parameter.

[0134] Evaluating volatility: Evaluate the volatility of each parameter through the coefficient of variation. If the CV of a certain parameter suddenly increases, it indicates that the operating environment of the system has changed, or the device is starting to age, affecting the stability of the parameter.

[0135] Adjusting the threshold: Based on the results of the CV, dynamically adjust the monitoring threshold of this parameter. For example, for parameters with increased volatility, appropriately expand their monitoring thresholds to reduce the occurrence of false alarms.

[0136] Implementing and monitoring: Apply the new threshold and continuously monitor its effectiveness to ensure the stability and efficiency of the system operation.

Claims

1. An early warning system for the combined operation of water electrolysis and power grid, characterized in that: Includes the following modules: An electrolysis module, used to perform an electrolysis reaction of water according to the electricity provided by the power grid to produce hydrogen; A monitoring module, used to monitor the state of the power grid and the operating parameters of the water electrolysis hydrogen production unit in real time; The data analysis module analyzes the data collected by the monitoring module, including: The power fluctuation response submodule adjusts the energy consumption of the water electrolysis hydrogen production unit in real time to adapt to power fluctuations in response to the instability of power supply from the power grid. When it detects excess or insufficient power, it automatically issues an early warning and responds to the early warning by adjusting the input power of the electrolysis process. The equipment monitoring submodule uses topological data analysis to analyze the operating parameters of the electrolytic hydrogen production unit and predict potential failures; The early warning module, based on the output of the data analysis module, sends an early warning signal to the operator when the power grid status has an adverse effect on the water electrolysis hydrogen production unit or the electrolysis hydrogen production unit itself has a potential failure; The topological data analysis method TDA in the device monitoring submodule specifically includes: Data mapping and network construction: The operating parameters of the water electrolysis hydrogen production unit are mapped to high-dimensional data points. The operating parameters are electrolysis current, electrolysis voltage, water temperature, electrolysis cell pressure and electrolyte concentration. An abstract network is constructed using high-dimensional data points, where nodes represent actual measurement points and edges connect nodes with similar attributes. Topological feature extraction: Apply the persistent homology technique in TDA to analyze the topological characteristics of the data structure formed by the parameters, pay attention to the "holes" or continuity interruptions in the data, indicate abnormal or failure modes, and convert the topological features obtained by persistent homology into feature vectors that can be used for machine learning; Pattern recognition and fault prediction: Create a topological signature for each operating state, monitor state changes by comparing topological signatures at different time points, and use topological changes to monitor anomalies in operation. Anomalies are manifested as sudden changes in the topological structure or new topological features. Once a topological anomaly is detected, an early warning will be automatically triggered to indicate a fault or a change that requires attention. Based on the topological analysis results, the thresholds of the monitoring parameters are dynamically adjusted to adapt to changes in the operating environment and the impact of equipment aging. The monitoring parameters include electrolysis current, electrolysis voltage, water temperature, electrolytic cell pressure, and electrolyte concentration. The persistent coherence technology specifically includes: Constructing point cloud: The monitored electrolysis current, electrolysis voltage, water temperature, pressure of the electrolytic cell and electrolyte concentration are regarded as points in multidimensional space. The parameter value at each time point constitutes a point, forming a point cloud data set; Build filter: define a "radius" parameter r, which is used to create a hypersphere covering the point cloud. As r increases, the hypersphere gradually expands, begins to overlap and form connections, and the resulting connection structure is considered simplified complexity; Calculate the Vietoris-Rips complex: For a given radius r, construct a Vietoris-Rips complex to represent the connection between data points. The complex includes points, lines, and triangles. As r increases, the complex becomes larger and larger. Generation of persistence graphs and Bassett graphs: As the radius r changes, the holes that appear and disappear, i.e., the topological features, are recorded. Holes represent ring structures or voids in the data. The appearance and disappearance of holes are related to state changes. Persistence graphs and Bassett graphs are generated to show the changes in topological features with r. Analyze persistence graphs and Bassett graphs to identify key topology changes. A sudden hole indicates a new abnormal state. Based on the analysis results, predict potential failures and issue early warning signals. In the persistence graph, the persistence homology result of each parameter set is represented by a series of barcodes, each barcode corresponds to the survival time of a hole, and the collection of barcodes forms a topological signature of the current state; In the Bassett diagram, each point represents a hole, the horizontal axis is the scale at which the hole appears, and the vertical axis is the scale at which it disappears; The dynamically adjusting the threshold of the monitoring parameter includes an adjustment based on statistical analysis, using a statistical method to calculate the volatility of the historical data of each monitoring parameter. If the volatility of a certain parameter increases, it is necessary to widen the threshold range of the parameter; The statistical analysis includes calculating the mean, standard deviation and coefficient of variation CV of each parameter, specifically including: The mean is the average value of a set of data, which is used to represent the center point of the data set. , the mean μ is calculated as: ； Where n is the number of data points, x i is the value of a single data point; The standard deviation is a statistic that measures the dispersion of a data set. It is used to describe the degree of deviation of a data point from the mean. For the same data set X, the standard deviation σ is calculated as: ； The coefficient of variation is an indicator of relative volatility, which is used to compare the volatility of data of different magnitudes. The coefficient of variation is calculated as: ; The larger the CV value, the higher the volatility of the data.

2. The early warning system for the combined operation of water electrolysis and power grid according to claim 1, characterized in that: The monitoring module is configured with a variety of sensors for collecting power data of the power grid in real time, including voltage, current and frequency; At the same time, the operating parameters of the water electrolysis hydrogen production unit are monitored, including electrolysis current, electrolysis voltage, water temperature, and pressure and electrolyte concentration of the electrolyzer; The monitoring module communicates with the data analysis module via a wireless network, transmitting all collected data in real time.

3. The early warning system for the combined operation of water electrolysis and power grid according to claim 2 is characterized in that: The power fluctuation response submodule specifically includes: Equipped with a fluctuation analysis algorithm to analyze the power supply fluctuations of the power grid and determine whether there is excess or insufficient power based on the monitored data; When there is excess electricity, the electrolysis module is automatically instructed to increase the input power of the electrolysis current, speed up the electrolysis of water, and increase the production of hydrogen to utilize the excess electricity; When electricity is insufficient, the electrolysis module is automatically instructed to reduce the input power of the electrolysis current to reduce the burden on the power grid and adjust the hydrogen production rate to match the available electricity; When any instability in power supply is detected, the power fluctuation response submodule will cooperate with the early warning module to automatically issue an early warning signal.

4. The early warning system for the combined operation of water electrolysis and power grid according to claim 3 is characterized in that: The fluctuation analysis algorithm adopts a graph neural network. In the graph neural network, the power grid is regarded as a network graph, in which each node represents a power station, a substation or a consumption node, and the edge represents a transmission line. The graph neural network operates directly on the network graph to analyze the complex relationships and influences between nodes, and is used to predict and analyze power flows and fluctuations based on the network structure.

5. The early warning system for the combined operation of water electrolysis and power grid according to claim 4, characterized in that: The graph neural network is applied to power fluctuation analysis, specifically including: Network modeling: The power grid is modeled as a graph, where nodes represent entities in the grid, edges represent power connections, including transmission lines, and each node has its attributes, including power generation, consumption, and voltage level; Eigenvector setting: Set the eigenvector X for each node i , which includes information on power supply, demand, and historical fluctuation data. The edge features are transmission capacity, distance, or connection type; Message passing mechanism: Using the message passing mechanism in the graph neural network, each node updates its own state according to the state of its adjacent nodes, expressed as: ； in, , are the hidden states of node i and node j in layer l, respectively. ij is the feature of the edge between node i and node j, is the set of neighbor nodes of node i, is an aggregation function over all neighbors, and f and g are learnable neural network functions; Power fluctuation monitoring: Use the trained graph neural network model to predict the power status of each node, and compare the predicted status with the actual status of the node to determine whether the power is excessive or insufficient; According to the output of the graph neural network model, when it is detected that there is insufficient or excessive power in a certain node or area, an early warning is automatically triggered and the input power parameters of the water electrolysis hydrogen production unit are adjusted.

Images