Cross-time-scale multi-electrolytic cell collaborative scheduling method

Abstract

The invention discloses a cross-time-scale multi-electrolytic-cell collaborative scheduling method which is applied to an alternating-current coupling photovoltaic hydrogen production system. The method comprises the following steps: acquiring day-ahead photovoltaic predicted power data and real-time photovoltaic output power data of the alternating-current coupling photovoltaic hydrogen production system, and calculating an upper limit of the number of electrolytic cells planned to run based on the day-ahead photovoltaic predicted power data and pre-configured rated power of a single electrolytic cell, based on the real-time photovoltaic output power data, the upper limit of the number of the planned operation electrolytic cells and a preset power distribution rule, the target electrolytic cell operation number is determined, and based on the target electrolytic cell operation number and the real-time photovoltaic output power data, hydrogen production distribution power is calculated; and a converter control instruction is generated based on the target electrolytic cell operation number, the hydrogen production distribution power and a preset equipment balance rotation mechanism. According to the invention, frequent start-stop oscillation at the operation boundary is eliminated, and multi-dimensional space-time equalization of a bottom-layer voltage stabilization control role and physical equipment can be realized.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic off-grid hydrogen production scheduling and control technology, and in particular to a method for coordinated scheduling of multiple electrolyzers across time scales. Background Technology

[0002] The off-grid photovoltaic hydrogen production system directly supplies fluctuating photovoltaic power to an alkaline water electrolyzer cluster through a multi-stage power electronic converter.

[0003] Currently, real-time power allocation for multi-electrolyte clusters typically employs a stateless algebraic mapping strategy based on real-time power ranges or a segmented startup strategy. In this approach, the controller directly divides the currently collected total photovoltaic power by the minimum operating power or rated power of a single device, determining the number of devices to be put into operation based on a preset fixed power threshold or a simple fixed proportional bandwidth. At the underlying control topology configuration, the system architecture typically configures multiple sets of parallel buck converters at the hydrogen production station. A specific converter with a fixed number is selected and operates in input voltage stabilization control mode to maintain the microgrid AC bus voltage, while the remaining parallel converters are uniformly configured in constant power point tracking mode to passively absorb the allocated power.

[0004] However, existing technologies struggle to address the high-frequency physical oscillations of equipment at operational boundaries, as well as the localized thermal stress concentration and accelerated aging of components caused by fixed control topologies. Summary of the Invention

[0005] Purpose of the invention: In view of the above-mentioned problems in the prior art, this application provides a cross-timescale multi-electrolyte collaborative scheduling method to solve the above-mentioned technical problems.

[0006] Technical Solution: A cross-timescale multi-electrolyzer collaborative scheduling method is applied to an AC-coupled photovoltaic hydrogen production system. The system is configured as a source-load direct connection architecture where the photovoltaic power station directly supplies power to the hydrogen production station via AC transmission lines, including:

[0007] Acquire the system's day-ahead photovoltaic power forecast data and real-time photovoltaic output power data;

[0008] Based on the current photovoltaic power forecast data and the rated power of a single electrolytic cell, calculate the upper limit of the planned number of electrolytic cells to be in operation;

[0009] Based on real-time photovoltaic output power data, the upper limit of the planned number of electrolyzers to be operated, and the preset power allocation rules, the target number of electrolyzers to be operated is determined.

[0010] Based on the target number of operating electrolyzers and real-time photovoltaic output power data, calculate the hydrogen production allocation power.

[0011] Based on the target number of operating electrolyzers, hydrogen production power allocation, and equipment balancing rotation mechanism, converter control commands are generated.

[0012] Beneficial effects: This invention eliminates frequent start-stop oscillations at the operating boundary and achieves multi-dimensional spatiotemporal balance between the underlying voltage regulation control role and physical devices. Attached Figure Description

[0013] Figure 1 This is a flowchart of the cross-timescale multi-electrolyte collaborative scheduling method of this application.

[0014] Figure 2 This is a flowchart for determining the target number of operating electrolyzers in this application.

[0015] Figure 3 This is a flowchart illustrating the generation of photovoltaic control commands in this application.

[0016] Figure 4 This is a flowchart illustrating the generation of hydrogen production control instructions in this application. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] To address the aforementioned issues, the applicant conducted in-depth searches and analyses, and discovered:

[0020] Because the power allocation mechanism uses stateless real-time numerical mapping, it does not record the system's historical state or assess the physical costs of starting and stopping. When the photovoltaic output power fluctuates slightly around the fixed threshold for equipment start-up and shutdown, the control system frequently issues commands to add or remove generators. This boundary oscillation generates reactive power consumption for warm-up and also causes the electrolytic cell plates to experience frequent alternating hot and cold stress, accelerating plate degradation.

[0021] Going forward, the task of stabilizing the AC bus of the microgrid will be undertaken by a single step-down converter for a long time. The converter alone absorbs all the transient power surges in the system. This asymmetrical division of control will cause the converter to age faster than other follower devices in the system, affecting the system's lifespan.

[0022] To solve these problems, combined with Figures 1 to 4 The present invention will be specifically described through the following embodiments.

[0023] In some embodiments, an infrastructure and system environment for a multi-electrolyte cooperative scheduling method across time scales are proposed, including the underlying hardware topology, communication control layer, and basic physical and mathematical models.

[0024] In this embodiment, the photovoltaic power station, the medium-voltage AC transmission line, and the hydrogen production station together constitute the physical topology of the AC-coupled photovoltaic hydrogen production system. Specifically, the photovoltaic power station includes a photovoltaic array composed of multiple photovoltaic modules connected in series and parallel to collect DC power. A two-stage inverter unit is used to convert DC power to AC power, consisting of a front-end boost converter and a rear-end three-phase inverter. The AC power output from multiple inverter units is combined, stepped up to the medium-voltage level by a boost transformer, and fed into the medium-voltage AC transmission line for long-distance transmission. At the hydrogen production station, the medium-voltage AC power passes sequentially through a step-down transformer, a multi-pulse uncontrolled rectifier circuit, and a step-down converter to be converted into low-voltage DC power required by multiple alkaline water electrolyzers.

[0025] Building upon this foundation, and to accommodate the coordinated operation of the physical architecture, the system is also configured with a hierarchical intelligent communication and control network. This network uses high-speed fiber optic communication as its data transmission backbone, constructing bidirectional data channels. Logically, the system is divided into an energy management layer and a local control layer. The energy management layer operates on a slow timescale of minutes, responsible for collecting meteorological data, output forecast data, and generating global scheduling instructions. The local control layer operates on a fast timescale of milliseconds, responsible for collecting real-time voltage and current states of the underlying power electronic equipment and directly executing closed-loop control logic for each converter. In other words, the system is configured as a direct-load architecture where the photovoltaic power plant supplies power directly to the hydrogen production plant via AC transmission lines.

[0026] Accordingly, to reduce the initial investment cost and lifecycle maintenance complexity of off-grid hydrogen production systems, the physical topology of this embodiment is a direct source-load connection architecture lacking large-capacity energy storage buffers. Under this architecture, conventional electrochemical battery packs no longer serve as physical buffers for energy throughput. The transient electrical energy generated by the photovoltaic power station, after deducting line losses, must be fully absorbed by the hydrogen production station in real time on a millisecond or minute scale. The inherent second- or minute-level drastic fluctuations of photovoltaic power will impact the electrolyzer cluster without attenuation, thus requiring coordinated control of the collaborative scheduling of multiple electrolyzers, resistance to boundary oscillations, and voltage management of the isolated microgrid.

[0027] Furthermore, the operation of an AC-coupled photovoltaic hydrogen production system relies on several underlying physical and mathematical models. For the photovoltaic power station side, a standard single-diode engineering mathematical model is used to characterize the output characteristics of the photovoltaic array. Its output voltage and output current are calculated together from preset basic parameters after compensation for ambient temperature and spatial irradiance. The basic parameters include short-circuit current, open-circuit voltage, rated power point current, and rated power point voltage.

[0028] For the hydrogen production plant side, the electrochemical dynamic characteristics of the alkaline water electrolyzer are characterized by an overvoltage model, and the specific logic for calculating its terminal voltage is as follows:

[0029] U _E =U _rev +U _act +U _ohm ;

[0030] Among them, U _E U is the terminal voltage of the alkaline water electrolyzer. _rev U is the reversible voltage required for the water electrolysis reaction. _act U is the activation overvoltage during the electrode reaction process. _ohm This refers to the ohmic overvoltage generated by the electrolyte and the diaphragm.

[0031] Accordingly, based on the high calorific value of hydrogen and the hydrogen production rate, an energy conversion efficiency model for the electrolyzer is established, and the corresponding efficiency calculation logic is expressed as follows:

[0032] η _E =(HHV _H2 ×n _H2 ) / P _E ;

[0033] Where, η _E HHV represents the hydrogen production efficiency of a single electrolyzer. _H2 For the pre-configured high calorific value parameters of hydrogen, n _H2 The hydrogen production rate is expressed as P, which can be expressed in terms of mass or volume. _EThis represents the actual input active power of the electrolyzer. The efficiency model shows that, at the same operating temperature, the hydrogen production efficiency of an alkaline water electrolyzer monotonically decreases with increasing input power. This nonlinear physical characteristic can be used to support power balancing and rotation scheduling of multiple electrolyzers.

[0034] Step 101: Obtain the system's day-ahead photovoltaic power forecast data and real-time photovoltaic output power data.

[0035] In this embodiment, the energy management layer obtains forecast data for the complete operating day in advance, including solar irradiance and temperature forecast curves, through an external meteorological data interface or a local historical database. This forecast data is then input into the photovoltaic cell engineering mathematical model to calculate the day-ahead photovoltaic power forecast data. Correspondingly, during the system's intraday real-time operation phase, the local control layer uses voltage and current sensors to continuously collect actual power generation data at millisecond-level sampling periods. The product result is uploaded to the energy management layer as real-time photovoltaic output power data. The voltage and current sensors are deployed on the photovoltaic array's combiner side. These two types of cross-time dimension data together constitute the global initial input boundary for system scheduling.

[0036] Step 102: Based on the day-ahead photovoltaic power forecast data and the rated power of a single electrolyzer, calculate the upper limit of the planned number of electrolyzers to be in operation;

[0037] Specifically, repeatedly starting and stopping multiple electrolyzers results in physical losses and additional warm-up energy consumption. Therefore, on operating days with lower peak photovoltaic output, it is not necessary to operate all equipment. Before the start of each day's operation, the system extracts the maximum predicted peak power for the day from the previous day's photovoltaic power forecast data. Next, this peak power is divided by the pre-configured rated power of a single electrolyzer. Since the actual number of devices required to be activated is an integer, and the system's total rated absorption capacity must cover the photovoltaic power peak, the system performs an up-rounding operation on the division result to obtain the upper limit of the planned number of electrolyzers to operate. This upper limit serves as a hard constraint for intraday regulation, which can shield against ineffective capacity increases caused by short-term power spikes.

[0038] Step 103: Determine the target number of operating electrolytic cells based on real-time photovoltaic output power data, the upper limit of the planned number of operating electrolytic cells, and power allocation rules.

[0039] In this embodiment, the system combines macroscopic day-ahead planning constraints with microscopic real-time input status to determine the specific number of devices that should be in operation at the current moment. A preset power allocation rule serves as a configurable logical judgment criterion, with its input parameter being continuously fluctuating real-time photovoltaic output power data. The system uses this rule to assess whether the current photovoltaic power level can support the operation of newly added electrolyzers, or whether it has fallen below the minimum threshold required to maintain the current number of operations. After rule evaluation, a preliminary number is calculated and compared with the determined upper limit of the planned number of electrolyzers to prevent the actual number of operational cells from exceeding the day-ahead planning expectation. Finally, the system converges and outputs the target number of operating electrolyzers.

[0040] Optionally, the preset power allocation rules can be implemented using a basic algebraic interval mapping method, or using an adaptive hysteresis state strategy that includes historical state memory.

[0041] Step 104: Calculate the hydrogen production allocation power based on the target number of operating electrolyzers and real-time photovoltaic output power data; so that each operating electrolyzer can fully absorb the power and balance the real-time photovoltaic output power.

[0042] After determining the number of devices required to operate, the system needs to distribute the total energy to each execution unit. Specifically, the system divides the current real-time photovoltaic output power data by the target number of operating electrolyzers to obtain the power quota that each device should handle, i.e., the hydrogen production allocation power. Through an average allocation mechanism, all operating electrolyzers maintain the same power load condition. According to the electrolyzer efficiency model, the same input power indicates that the hydrogen production efficiency of each device is consistent, avoiding efficiency degradation due to differences in operating conditions within the system.

[0043] Step 105: Generate converter control commands based on the target number of operating electrolyzers, hydrogen production allocation power, and equipment balancing rotation mechanism.

[0044] As the control endpoint of the entire collaborative scheduling method, the energy management layer, after completing the computing power allocation, invokes a pre-defined equipment balancing rotation mechanism to specify the specific physical execution targets. This mechanism considers the historical cumulative operating time of each electrolyzer, the daily startup priority sequence, and the single-peak variation characteristics of photovoltaic power. It selects target equipment to be started from the shutdown queue or target equipment to be shut down from the operating queue. Next, the system encapsulates and issues converter control commands to the selected target equipment and its corresponding hydrogen production-side step-down converter. These commands not only include the equipment's start / stop enable signals but also directly carry the calculated hydrogen production allocation power. This allocation power serves as a reference benchmark for the underlying closed-loop control, driving the AC-coupled photovoltaic hydrogen production system to achieve closed-loop, balanced operation.

[0045] For example, this application also proposes an implementation mechanism for the basic power allocation strategy and the power-average constraint, specifically including:

[0046] Step 201: Extract the daily maximum predicted power from the day-ahead photovoltaic predicted power data.

[0047] Specifically, before the start of each daily operating cycle, the energy management system receives a time series of photovoltaic (PV) output forecasts from the weather forecasting system. The system iterates through all discrete power forecast values ​​in this time series, uses an optimization comparison algorithm to select the power value with the highest amplitude, and takes this as the daily maximum forecast power. This data reflects the maximum peak electrical energy that the PV power plant is expected to output that day, and can serve as the starting point for macro-level equipment capacity planning.

[0048] Step 202: Divide the maximum predicted daily power by the rated power of a single electrolytic cell, and round up the result to obtain the upper limit of the planned number of electrolytic cells in operation. This avoids redundant electrolytic cell operating costs within the day.

[0049] Specifically, some operating areas have poor sunlight conditions, and starting all hydrogen production equipment will result in standby and warm-up losses. To determine the scale of equipment that should be put into operation on a given day, the system uses the following arithmetic logic:

[0050] N_ plan =ceil(P _pre_max / P _E_max );

[0051] Where, N _plan The maximum number of electrolytic cells to be operated is set to ceil, where ceil is the floor function and P is the maximum number of cells to be operated. _pre_max P is the maximum predicted power for the day. _E_max This refers to the pre-configured rated power of a single electrolytic cell.

[0052] For example, assuming the pre-configured rated power of a single electrolytic cell is 1 MW, if the predicted maximum daily power is 3.4 MW, dividing 3.4 by 1 gives 3.4. After rounding up, the maximum planned number of electrolytic cells is 4. This means that the system allows a maximum of 4 electrolytic cells to be in standby mode on any given day. The 5th and subsequent cells will be forcibly locked. On a macro level, this can reduce the overall start-up and shutdown frequency and operation and maintenance costs of the equipment cluster.

[0053] Step 203: Divide the real-time photovoltaic output power data by the minimum operating power of a single electrolytic cell, and round down to obtain the initial number of cells in operation.

[0054] During the intraday real-time control phase, the system continuously acquires real power generation data at millisecond intervals and performs mapping calculations:

[0055] N _init=floor(P _PV_current / P _E_min );

[0056] Where, N _init P represents the initial number of runs, floor is the floor function, and P is the number of runs. _PV_current For real-time photovoltaic output power data, P _E_min This is the minimum operating power of a pre-configured single electrolytic cell.

[0057] Because alkaline water electrolyzers have minimum operating load requirements, the system must allocate no less than the safety threshold of electrical energy to each electrolyzer in operation. Based on the physical characteristics of the electrolyzer, its hydrogen production efficiency in the low-power range is significantly higher than in the full-load range. Therefore, allocating the current photovoltaic power to the maximum number of devices ensures that each device operates at a low load close to its minimum operating power, maximizing the overall hydrogen conversion rate.

[0058] Step 204: Take the smaller value between the initial number of operating cells and the upper limit of the planned number of electrolytic cells as the basic number of operating cells;

[0059] After obtaining the initial number of runs, the system compares and intercepts it with the macro-constraints planned in the previous day. The calculation logic is as follows:

[0060] N _op =min(N _init N _plan );

[0061] Where, N _op The base number of runs, min is the function to find the minimum value, N _init N is the initial number of runs. _plan This represents the upper limit of the planned number of electrolytic cells in operation.

[0062] In this embodiment, a minimum value function is introduced to limit real-time aggressive capacity expansion actions within the safe framework of the day-ahead plan. For example, if real-time photovoltaic power surges and the calculated initial operating number reaches 5 units, but the maximum number of electrolyzers planned to operate in the day-ahead plan is 4, then the basic operating number will be forcibly truncated to 4 units.

[0063] Step 205: When the basic number of operating cells equals the upper limit of the planned number of operating electrolytic cells, if the real-time photovoltaic output power data is greater than the product of the basic number of operating cells and the rated power of a single electrolytic cell, it is determined that a photovoltaic output prediction error has occurred.

[0064] Furthermore, due to the uncertainty of actual meteorological conditions, recent forecasts may underestimate actual power generation capacity. In this case, without overcoming existing macroeconomic constraints, a large amount of high-quality green electricity will be wasted. Therefore, the system monitors the relationship between the actual generated power and the rated absorption capacity of the equipment in real time. When the number of basic operating cells approaches the upper limit of the planned number of operating electrolyzers, and each device is operating at full load, it still cannot fully absorb the real-time photovoltaic output power, triggering the correction criteria for forecast errors.

[0065] Step 206: Based on the prediction error, increase the basic operating number of electrolyzers sequentially until the real-time photovoltaic output power data is fully absorbed or the total number of electrolyzers is reached, obtaining the corrected operating number, which is then used as the target operating number of electrolyzers. In other words, the corrected operating number is used as the target operating number of electrolyzers.

[0066] Specifically, when a photovoltaic (PV) output forecast is underestimated, the system automatically removes day-ahead planning restrictions. The system employs a cyclical, cumulative correction logic, sequentially activating forcibly locked standby electrolyzers. Each time a unit is added, the system reassesses the current total energy absorption capacity. This cyclical addition process continues until the newly added units can absorb excess PV power, or all electrolyzers in the system are operational. The number of units at this point is taken as the target number of operating electrolyzers. Through this prediction error correction mechanism, the system can balance equipment wear protection with energy absorption.

[0067] Furthermore, in the absence of any prediction error in photovoltaic power output, the basic operating quantity is directly determined as the target electrolyzer operating quantity.

[0068] Step 207: When the target number of operating electrolyzers is greater than zero, the real-time photovoltaic output power data is evenly distributed to the corresponding number of operating electrolyzers to obtain the target hydrogen production allocation power per unit, which is used as the hydrogen production allocation power. In other words, when the target number of operating electrolyzers is greater than zero, the real-time photovoltaic output power data is evenly distributed to the corresponding number of operating electrolyzers to obtain the target hydrogen production allocation power per unit, which is used as the hydrogen production allocation power, so that multiple operating electrolyzers can maintain the same power load conditions and have consistent hydrogen production efficiency.

[0069] After determining the number of devices, the system allocates power quotas:

[0070] P _E =P _PV_current / N _op ;

[0071] Among them, P _E To allocate power to a single target hydrogen production unit, P _PV_current For real-time photovoltaic output power data, N _op The target number of electrolyzers in operation.

[0072] In this embodiment, an average power distribution mechanism is adopted based on the nonlinear power efficiency characteristic curve of the electrolyzer. If the same total power is unevenly distributed among multiple devices, high-load devices will enter the inefficient operating zone, while low-load devices will be in the efficient operating zone. According to the nonlinear efficiency characteristics of the electrolyzer, when multiple devices have the same input power, the total hydrogen production rate of the system reaches its maximum, causing a nonlinear reduction in the total hydrogen production rate. Therefore, by averaging power distribution, the system ensures that the devices are at the same operating coordinate point on the efficiency curve. The system can also use a basic simple start-stop strategy for power distribution. Specifically, when the input power of the hydrogen production system is greater than the minimum operating limit of a single device, the first electrolyzer is started; when the input power exceeds the rated power of the first electrolyzer, and the excess power exceeds the minimum operating limit of a single device, the system adds a second electrolyzer, and so on, until all electrolyzers are started. However, this strategy will cause the lower-numbered devices to operate at high loads for extended periods and experience frequent start-stop cycles.

[0073] In other alternative implementations, the system can also employ a segmented start-up strategy. This strategy leverages the higher hydrogen production efficiency of the electrolyzer in its low-power range by pre-setting a power ceiling for the high-efficiency operating range of the electrolyzer. This ceiling can be set to 0.5 times the rated power. When allocating power, the system ensures that all started equipment reaches this high-efficiency operating range power ceiling. Once all equipment has reached this ceiling, any subsequent increases in redundant power are allocated in a second round in the same order until each piece of equipment is operating at full rated power.

[0074] Correspondingly, the segmented startup strategy improves the utilization efficiency of the low-power range, but there are still problems such as excessive total number of system start-ups and shutdowns due to fixed priorities and large differences in single-unit losses.

[0075] As an alternative implementation, this application proposes an adaptive hysteresis switching strategy based on an economic model to address the problem of frequent start-ups and shutdowns and boundary oscillations in electrolyzers caused by using stateless calculation methods.

[0076] In microgrid systems equipped with large-capacity energy storage, rapid battery charging and discharging can suppress photovoltaic power fluctuations. However, this invention employs a direct-connect architecture without buffering, lacking battery hardware buffering. Power boundary fluctuations directly affect the electrolyzer, easily triggering high-frequency ineffective start-ups and shutdowns, causing the electrolyzer to passively act as a mechanical buffer unit. Therefore, this invention introduces an adaptive hysteresis switching strategy based on a trade-off between net hydrogen production benefits and start-up / shutdown costs, effectively replacing hardware battery buffering with algorithmic soft buffering to eliminate boundary oscillations in the energy storage-free architecture.

[0077] Step 301: Using the electrolyzer power-efficiency characteristic curve, calculate the difference in hydrogen production rate before and after adding a single operating electrolyzer, as represented by the real-time photovoltaic output power data, and construct a marginal hydrogen production gain model. In other words, based on the pre-constructed electrolyzer power-efficiency characteristic curve, calculate the total system input power represented by the real-time photovoltaic output power data, calculate the difference in hydrogen production rate before and after adding a single operating electrolyzer, and construct a marginal hydrogen production gain model.

[0078] In this embodiment, the scheduling technology primarily employs stateless algebraic operations to determine the number of electrolyzers, ensuring that the same photovoltaic power corresponds to a fixed number of electrolyzers, ignoring the current equipment operating state and switching costs. This stateless decision-making is highly susceptible to boundary oscillations. For example, when the total power fluctuates slightly around an integer multiple of the minimum power of a single electrolyzer, the system frequently issues commands to add or remove machines. Therefore, this application introduces an economic cost-benefit assessment mechanism. Specifically, the system quantifies the positive benefits of adding electrolyzers, utilizes pre-stored equipment physical characteristics, evaluates the overall efficiency improvement after the current total photovoltaic power allocation, converts this efficiency improvement into an increase in hydrogen production rate, and establishes a marginal hydrogen production gain model.

[0079] Step 302: Obtain the pre-configured high calorific value of hydrogen.

[0080] To calculate the incremental hydrogen production rate, the system pre-configures the fundamental physicochemical constants of hydrogen in the controller's storage unit. Specifically, the system reads the pre-configured hypercalorific value of hydrogen, which represents the heat released by the complete combustion of a unit mass of hydrogen. For example, in engineering implementation, the preset hypercalorific value of hydrogen is 39.4 kWh per kilogram.

[0081] Step 303: Divide the total system input power by the current electrolytic cell operating status variable to obtain the power allocated per unit before the increase, and divide by the value of the current electrolytic cell operating status variable plus one to obtain the power allocated per unit after the increase.

[0082] The system employs arithmetic division to assess the load level of electrolytic cells under different operational numbers. Using the current real-time total photovoltaic power as the dividend, and the number of currently operating electrolytic cells and the number of added units as divisors, the theoretical power capacity per unit before and after adding equipment is calculated. In other words, the theoretical power capacity per unit can also be calculated using the number of currently operating electrolytic cells and the number of units after adding one unit as divisors.

[0083] Step 304: Based on the pre-constructed power-efficiency characteristic curve of the electrolyzer, extract the first hydrogen production efficiency corresponding to the single unit power allocation before the increase, and the second hydrogen production efficiency corresponding to the single unit power allocation after the increase.

[0084] Based on the operating characteristics of the electrolyzer, the hydrogen production efficiency of the equipment monotonically increases as the input power decreases. The system calculates two single-unit allocated power values, which serve as a retrieval index. By extracting data from a pre-constructed efficiency curve mapping table, the actual hydrogen production efficiency of the system before and after the unit upgrade operation can be obtained.

[0085] Step 305: Divide the total system input power by the pre-configured hydrogen high calorific value, and multiply it by the difference between the second hydrogen production efficiency and the first hydrogen production efficiency to output the corresponding marginal hydrogen production gain value, thus completing the construction of the marginal hydrogen production gain model.

[0086] Furthermore, after extracting all the basic physical parameters, the system performs the final calculation of the gain model, the expression of which is designed as follows:

[0087] ΔQ(n,P)=(P / HHV _H2 )×(η _E (P / (n+1))-η _E (P / n));

[0088] Where ΔQ(n,P) is the marginal hydrogen production gain rate brought about by adding one electrolyzer to the system under the total input power, P is the total input power of the system, and HHV _H2 For the pre-configured high calorific value of hydrogen, η _E Let n be the hydrogen production efficiency function based on a specific power variable, where n is the current operating state variable of the electrolyzer. Since the hydrogen production efficiency function is monotonically decreasing, the power allocated to a single unit after the addition of the electrolyzer is less than before the addition. Therefore, the marginal hydrogen production gain value is always greater than zero, meaning that adding more electrolyzers is always beneficial in terms of efficiency.

[0089] Step 306: Obtain the hydrogen equivalent of the warm-up cost and the hydrogen equivalent of the equipment loss cost.

[0090] After clarifying the benefits of adding capacity, the system needs to assess the negative costs of start-up and shutdown operations. Each start-up and shutdown results in quantifiable losses in the electrolyzer. The system extracts two cost parameters from the configuration library: warm-up cost and equipment depreciation cost. The warm-up cost refers to the amount of hydrogen equivalent to the reactive power consumed during the warm-up process from cold start to stable hydrogen production. The equipment depreciation cost refers to the additional physical aging of components such as electrodes and diaphragms caused by frequent start-ups and shutdowns. This physical aging is amortized per operation based on the equipment's life-cycle replacement cost and is equivalent to the weight of hydrogen.

[0091] Step 307: Summing the hydrogen equivalent of warm-up cost and the hydrogen equivalent of equipment loss cost to construct an equivalent hydrogen loss model. In other words, the pre-configured hydrogen equivalent of warm-up cost and the pre-configured hydrogen equivalent of equipment loss cost are summed to construct an equivalent hydrogen loss model, which is used to quantify the cost of a single electrolyzer performing a single start-up and shutdown operation.

[0092] Accordingly, the system algebraically adds the two types of cost parameters to form a unified loss metric, the formula of which is:

[0093] C _sw =C _warm +C _deg ;

[0094] Among them, C _sw To calculate the equivalent hydrogen loss for the total cost of performing a single start-stop operation, C _warm The cost of warming up is in hydrogen equivalent, C _deg The system converts complex electrical and mechanical losses into a single, intuitive amount of hydrogen loss weight, allowing for direct comparison with hydrogen production gains.

[0095] Step 308: Obtain the shortest payback period. This means obtaining the pre-configured shortest payback period.

[0096] In this embodiment, the system introduces an adjustable design parameter with a time dimension, namely the shortest payback period. This parameter indicates the minimum stable operating time that a newly started electrolyzer must meet. The larger the parameter value, the more conservative the system startup strategy, which can suppress the start-up and shutdown frequency; the smaller the parameter value, the more aggressive the scheduling strategy.

[0097] Step 309: The product of the marginal hydrogen production gain model and the shortest payback period is equal to the equivalent hydrogen loss model to construct the net hydrogen production benefit balance equation. The balance equation is then solved to obtain the dynamic capacity increase threshold. In other words, the product of the output value of the marginal hydrogen production gain model and the shortest payback period is equal to the output value of the equivalent hydrogen loss model to construct the net hydrogen production benefit balance equation. The balance equation is then solved for the current electrolyzer operating state variables to obtain the dynamic capacity increase threshold.

[0098] In this embodiment, the requirement for the system to perform the booster operation is that the cumulative hydrogen production gain during the expected operating time must fully cover or exceed the preset start-up and shutdown costs. Therefore, the system constructs an implicit balance equation, the expression of which is:

[0099] ΔQ(n,P _up (n))×T _min =C _sw ;

[0100] Where ΔQ is the marginal hydrogen production gain function, n is the current electrolyzer operating state variable, and P _up (n) represents the dynamic boost power threshold to be solved, T _min To minimize the payback period, C _swThis represents the equivalent hydrogen loss. By solving this equation, the system can obtain the minimum trigger power threshold required to activate the next device in the current state; in other words, it obtains the minimum trigger power limit worth activating the next device in the current state.

[0101] Step 310: Obtain the minimum operating power and safety margin power of a single electrolytic cell. This means obtaining the pre-configured minimum operating power and pre-configured safety margin power of a single electrolytic cell.

[0102] Unlike the process of increasing capacity to seek economic optimization, reducing the number of electrolyzers increases the power load on the remaining equipment, leading to a passive decrease in hydrogen production efficiency. Therefore, the system should not actively perform capacity reduction operations; it should only trigger capacity reduction when the input power cannot maintain physically safe operation. Based on this, the system extracts the minimum physical operating power of the equipment and introduces a safety margin power value. This safety margin power is a pre-configured positive power parameter used to prevent abnormal operating conditions of the equipment near the minimum physical power boundary.

[0103] Step 311: Multiply the current electrolytic cell operating status variable by the minimum operating power of a single electrolytic cell, and add the safety margin power to obtain the dynamic reduction power threshold.

[0104] Based on physical safety baselines, the system calculates the trigger limit for reducing the number of operations, i.e., the trigger threshold, which is calculated as follows:

[0105] P _down (n) = n × P _E_min +P _buf ;

[0106] Among them, P _down (n) represents the dynamic power reduction threshold for decreasing the number of electrolytic cells from the current operating quantity by one, where n is the current operating state variable of the electrolytic cell, and P _E_min P is the minimum operating power of a single electrolytic cell. _buf For safety margin power.

[0107] Step 312, wherein the dynamic increase power threshold is greater than the dynamic decrease power threshold corresponding to the increased number of operating units, and the difference constitutes the switching hysteresis bandwidth. That is, the difference between the two adaptively constitutes the switching hysteresis bandwidth.

[0108] Because the increase threshold requires overcoming the additional physical costs of starting and stopping, its value is inevitably greater than the decrease threshold calculated solely based on the physical lower limit. This naturally creates an asymmetric difference between the two values, expressed as follows:

[0109] ΔP _hyst (n)=P _up (n)-P _down (n+1);

[0110] Where, ΔP _hyst (n) represents the switching hysteresis bandwidth adaptively constructed by the system in the corresponding state, P _up (n) represents the dynamic engine power threshold, P _down (n+1) represents the dynamic power reduction threshold after increasing the target number of machines. This bandwidth is not a fixed constant set arbitrarily, but is determined by a deeply adaptive joint approach based on the system's efficiency characteristics, loss costs, expected runtime, and other physical parameters.

[0111] Step 313: Obtain the current electrolyzer operating state variable at the current moment from the controller state storage unit of the AC-coupled photovoltaic hydrogen production system.

[0112] In the real-time control loop of the system, the controller no longer directly maps power to the number of devices. Instead, it reads the operating status counter left over from the previous control cycle from the local cache and uses it as the current electrolytic cell operating status variable. The initial value of this variable is zero.

[0113] Step 314: Based on the current electrolytic cell operating state variables and the switching determination parameters, obtain the corresponding dynamic increase power threshold and dynamic decrease power threshold respectively. The dynamic increase power threshold is greater than the dynamic decrease power threshold, forming an asymmetric hysteresis loop.

[0114] The system uses the read state variables and pre-configured switching decision parameters, which are a set of parameters pre-configured in the controller, including the electrolyzer power-efficiency characteristic curve, the hydrogen equivalent of warm-up cost, the hydrogen equivalent of equipment loss cost, the shortest benefit recovery time, the minimum operating power of a single electrolyzer, and the safety margin power, to call the generated asymmetric threshold pairs as the comparison benchmark for determining the state transition within the current control cycle.

[0115] Step 315: When the real-time photovoltaic output power data is greater than the dynamic capacity increase power threshold and the current electrolytic cell operating status variable is less than the upper limit of the planned number of electrolytic cells, increase the current electrolytic cell operating status variable.

[0116] When the system detects a real-time increase in photovoltaic power exceeding the capacity increase threshold, and the current operating scale has not reached the upper limit set by macro-control, the system performs a state transition operation, incrementing the value of the current electrolytic cell operating state variable by one.

[0117] Step 316: When the real-time photovoltaic output power data is less than the dynamic power reduction threshold and the current electrolytic cell operating status variable is greater than zero, decrease the current electrolytic cell operating status variable.

[0118] When the real-time photovoltaic power drops sharply below the reduction threshold including a safety margin, in order to avoid damage to the equipment due to underpower operation, and since there is currently equipment in operation, the system performs a reverse state transition operation, decrementing the value of the current electrolytic cell operating state variable by one.

[0119] Furthermore, if the real-time photovoltaic output power data is less than or equal to the dynamic power increase threshold and greater than or equal to the dynamic power decrease threshold, the current electrolytic cell operating status variable is kept unchanged.

[0120] Step 317: After performing the operation of increasing or decreasing the current electrolytic cell operating status variable, repeatedly execute the process of obtaining new dynamic power increase threshold and dynamic power decrease threshold and over-limit judgment until the real-time photovoltaic output power data no longer triggers the over-limit condition, in order to adapt to the cross-level change of photovoltaic power.

[0121] In some alternative implementations, when photovoltaic power is heavily obstructed by clouds or suddenly clears, its power output may jump by seconds. After adjusting the state variable once, the system does not immediately exit the current control cycle. Instead, it uses the updated state variable to re-acquire the corresponding next-level threshold and makes another determination. This recursive update mechanism allows the system state to follow the sudden changes in power output, avoiding control lag caused by the slow start-up and shutdown of multiple devices in batches.

[0122] Step 318: When the real-time photovoltaic output power data is less than or equal to the dynamic power increase threshold and greater than or equal to the dynamic power decrease threshold, the current electrolyzer operating state variable is kept unchanged, forming a switching state machine based on historical path decisions to suppress oscillations at the operating boundary.

[0123] When the photovoltaic power output remains stable within the range defined by the threshold for increasing and decreasing capacity, the system refuses to execute any state change commands. At this point, the system exhibits typical memory properties; the number of operating devices at the same power level is entirely determined by the historical evolution path of the power output—either falling from a high level or rising from a low level. The switching state machine mechanism, based on historical paths, can eliminate equipment oscillation and flutter phenomena near the critical point.

[0124] Step 319: Update the current electrolytic cell operating status variable to represent the target number of electrolytic cells in operation. That is, the updated current electrolytic cell operating status variable is determined as the target number of electrolytic cells in operation.

[0125] After undergoing a complete state machine evaluation and recursive evolution, the system will finally lock the internal state variables that have converged and stabilized. The target number of electrolytic cells in operation will be used as the real-time control output and sent to the underlying physical device execution queue.

[0126] In some scenarios, assuming a hydrogen production station is equipped with five electrolyzers with a rated power of 1 MW, and the minimum operating power of each electrolyzer is set to 0.2 MW. Under a traditional stateless strategy, when the total system power fluctuates slightly around 1.0 MW, the system will frequently switch between operating four and five electrolyzers because 0.99 / 0.2 rounded down equals 4, and 1.01 divided by 0.2 rounded down equals 5.

[0127] In one aspect of the technical solution of this invention, assuming the comprehensive start-up and shutdown cost is 1.5 kg of equivalent hydrogen and the shortest payback period is 0.5 hours, by solving the payback balance equation, the system obtains a dynamic power threshold of 1.1 MW for increasing capacity from 4 to 5 cells, while the dynamic power threshold for decreasing capacity from 5 to 4 cells is slightly higher than the sum of the minimum operating power of 5 electrolyzers. Therefore, when the photovoltaic power fluctuates around 1.0 MW, since it does not exceed the threshold range on either side, the system will maintain its original operating state, which can extend the service life of the physical electrolyzers.

[0128] As one implementation method, this application provides an offline construction and lookup mechanism for adaptive engine boosting thresholds to maintain the real-time performance of the control system.

[0129] The system relies on the hydrogen production net benefit balance equation to derive the dynamic power threshold for increasing capacity. This balance equation contains a nonlinear relationship between electrolyzer power and efficiency. If the microcontroller were to directly execute online real-time numerical solutions within a millisecond-level control cycle, it would consume computational resources and cause control delays, disrupting the rapid response to photovoltaic power output fluctuations. To eliminate the contradiction between algorithm complexity and engineering real-time requirements, this embodiment provides an architectural scheme that shifts the complex nonlinear optimization process to an offline stage, specifically including:

[0130] Step 401: In the offline construction phase of the system, the set of electrolyzer operating state variables is traversed, and the balance equation of hydrogen production net benefit is solved offline using numerical calculation methods to generate a multidimensional boosting threshold lookup table with electrolyzer operating state variables as index items.

[0131] Specifically, system engineers utilize high-performance offline servers to perform threshold calculations during equipment manufacturing or software initialization configuration phases. The system obtains a traversal set of electrolyzer operating state variables based on the physical topology of the hydrogen production station. For example, when the system contains 5 electrolyzers, this set is set as an integer sequence from 0 to 4. For each state variable value in the set, when the electrolyzer operating state variable is 0, the system does not have a pre-upgrade operating state, and the marginal hydrogen production gain model is not applicable. For this initial startup scenario, the offline server directly sets the dynamic upgrade power threshold to the sum of the minimum operating power of a single electrolyzer and the safety margin power, i.e., the lower limit for safe startup of the first electrolyzer. The offline server extracts the corresponding pre-configured shortest revenue recovery time and equivalent hydrogen loss model, and uses numerical calculation methods such as Newton's iteration method to solve the hydrogen production net revenue balance equation, obtaining the optimal upgrade trigger power corresponding to the current number of operating cells.

[0132] After completing the traversal calculations, the system encapsulates all discrete solution results and compiles them into a pre-built multidimensional capacity expansion threshold lookup table. This lookup table is structured as a read-only memory matrix. Furthermore, to adapt to different weather patterns or different electricity price periods and scheduling strategy preferences, the lookup table is expanded into a two-dimensional matrix data structure. The row indices of this two-dimensional matrix are set to the current electrolyzer operating state variables, and the column indices are set to the discrete gradients of the pre-configured shortest payback period. Its memory mapping logic is as follows:

[0133] P _up_table (n,j)=Matrix _ROM (n,T _min_idx (j));

[0134] Among them, P _up_table (n,j) represents the dynamic boost power threshold extracted by matrix mapping. _ROM The multidimensional booster threshold lookup table is pre-built and programmed into the control chip, where n is the row index parameter corresponding to the current electrolytic cell operating state variable, and T... _min_idx (j) is the column index parameter corresponding to the shortest payback period level currently activated in the system.

[0135] Step 402, in the online real-time control stage of the system, the dynamic boost power threshold is obtained by solving the balance equation. Alternatively, the dynamic boost power threshold can be obtained by directly extracting the value that matches the current electrolytic cell operating state variable by querying a pre-built multidimensional boost power threshold lookup table, in order to avoid real-time nonlinear calculation.

[0136] After the equipment is officially put into operation, during the online real-time control phase, the local control layer, which operates at the millisecond level, abandons complex equation-solving code modules. Within the control cycle executing the state transition logic, the system only performs low-computational-complexity direct memory read operations. Specifically, the system uses the real-time acquired current electrolytic cell operating state variable as a pointer address to directly read the content at the corresponding address in the read-only memory matrix, extracts the value, directly assigns it to the dynamic boost power threshold, and performs an arithmetic comparison with the real-time photovoltaic output power.

[0137] In this embodiment, by combining offline construction with online table lookup, the system compresses the complex nonlinear physical optimization process into a single spatial mapping action with constant time complexity. This avoids the risk of system paralysis due to computing power overflow when the microcontroller experiences severe photovoltaic fluctuations, and also allows theoretical economic models to be losslessly reduced in dimensionality and deployed in conventional industrial-grade digital signal processors.

[0138] As an optional implementation scheme, this application also provides a device spatiotemporal multidimensional equilibrium and control role binding mechanism.

[0139] Specifically, it includes:

[0140] Step 501: Obtain the total number of pre-configured electrolytic cells.

[0141] In this embodiment, during the initialization phase, the energy management layer reads the total physical scale of the hydrogen production equipment actually connected to the site from the local system configuration file or hardware bus. This value, as a global static constant, can be used as the modulus benchmark for executing the equipment rotation algorithm.

[0142] Step 502: Between consecutive operating days, the reference unit of the day is cyclically shifted in ascending order of the physical number of the electrolytic cells, so that each electrolytic cell takes turns to assume the highest start-up priority once every N operating days, where N equals the total number of electrolytic cells.

[0143] Specifically, to avoid fixed-number devices being in a priority startup state for extended periods, resulting in uneven runtime, the system introduces a macro-level rotation mechanism spanning entire days. The system maintains a global runtime counter, triggering cross-day switching logic at midnight each day. Modular arithmetic is used to determine the device with the highest startup priority each day. The conversion logic is expressed as follows:

[0144] ID _ref =(D _run mod N _total )+1;

[0145] Among them, ID _ref D is the calculated physical number of the reference unit for that day. _run The number of consecutive days the system has been running, where mod is the modulo operator, and N is the number of consecutive days the system has been running._total This represents the total number of pre-configured electrolytic cells. Using a modulo-shift algorithm based on consecutive days, the system ensures that each physical device can be selected exactly once within any consecutive N operating days, serving as the leader for that day. This algorithm achieves absolute macro-level amortization over time.

[0146] Step 503: Using the baseline unit of the day as the first step, construct the start-up priority sequence of the day to balance the cumulative running time of each electrolyzer.

[0147] After determining the reference unit for the day, the system dynamically constructs a one-dimensional data structure in memory to store the startup order of all devices. This data structure contains an array of priority sequences of all device numbers. The system places the reference unit for the day at the first index of the array, and the remaining devices are filled into the subsequent positions of the array in ascending order of their physical numbers. When the number exceeds the pre-configured total number of electrolytic cells, the process returns to the lowest number to continue filling. For example, in a system with 5 devices, if the reference unit for the day is unit number 3, the generated priority sequence array would be: 3, 4, 5, 1, 2. This sequence array remains locked for the day and is used as the addressing basis for subsequent real-time millisecond-level control.

[0148] Step 504: Obtain the startup priority sequence for the day.

[0149] Once the system enters the intraday real-time photovoltaic power consumption phase, in each control cycle, the local control layer requests the energy management layer to read the locked array, which is used to provide equipment identification guidance for the physical switching actions that will occur.

[0150] Step 505: When the number of target electrolytic cells in operation is greater than the current number of cells in operation of the system, start-up control commands are generated by sequentially selecting the electrolytic cells to be shut down according to the positive order of the start-up priority sequence of the day.

[0151] When the real-time photovoltaic power increases, triggering the capacity expansion condition, the system scans the priority sequence array sequentially according to the forward traversal pointer. The system checks the current physical state of the device corresponding to each element in the array, locating the device that is at the front of the sequence and in a shutdown / dormant state. The system encapsulates a startup message and sends the startup control command to the hydrogen production-side converter corresponding to the selected device.

[0152] Step 506: When the target number of operating electrolyzers is less than the current number, based on the last-in-first-out (LIFO) principle, the last electrolyzer started is selected to generate a shutdown control command. This is used to adapt to the single-peak curve characteristics of photovoltaic power and to minimize the total number of start-ups and shutdowns per day.

[0153] When real-time photovoltaic power declines and triggers the shutdown condition, this embodiment abandons the first-in-first-out (FIFO) exit logic and enforces the last-in-first-out (LIFO) principle. Specifically, the system can reversely traverse the activation timestamp logs of the already started devices, or follow the reverse order of the priority sequence array to lock the last awakened edge device and send it a shutdown control command. Since the photovoltaic output power in nature exhibits a bell-shaped single-peak curve characteristic of rising in the morning, peaking at noon, and declining in the evening, using the FIFO principle to shut down the earliest started reference device when power declines would lead to severe fragmentation of daily operation, causing frequent device start-ups and shutdowns. However, using the LIFO principle allows the priority-started reference devices to maintain continuous and uninterrupted operation for a long time throughout the entire cycle of the single-peak curve. Based on this, the total number of system start-ups and shutdowns on a given day is limited to its theoretical minimum. The limit formula for the number of start-ups and shutdowns is:

[0154] N _switch_LIFO =2×N _op_max ;

[0155] Where, N _switch_LIFO N represents the total number of start-stop actions of the entire system within a day, calculated using the Last-In-First-Out (LIFO) principle. _op_max This represents the maximum number of devices operating simultaneously that day.

[0156] Step 507: The AC bus voltage stabilization control role is dynamically assigned to the hydrogen production side converter corresponding to the reference unit of the day. The hydrogen production side converter executes the input voltage stabilization control command to maintain the voltage stability of the AC bus.

[0157] In this embodiment, the underlying power electronic control division of labor is deeply coupled with the daytime rotation mechanism to address the more subtle micro-level uneven operating conditions within the system. Specifically, in an islanded AC-coupled hydrogen production system, there must be one and only one buck converter in constant voltage mode to smooth bus fluctuations. This mode requires the converter to have extremely fast response and withstand severe transient shocks. Instead of fixing this function to a specific physical node, the system dynamically attaches it to the hydrogen production-side converter corresponding to the reference unit for the day via communication messages. Since the reference unit is the earliest to start and the latest to shut down equipment on the day, the voltage regulation control core is online throughout the operating day, ensuring that system stability and equipment operating conditions are evenly distributed.

[0158] Step 508: The power control role is assigned to the remaining hydrogen production-side converters corresponding to the non-reference unit. The remaining hydrogen production-side converters execute voltage tracking control commands to follow the target hydrogen production power allocation, so that each electrolyzer bears the cumulative duration of the voltage stabilization condition in an even manner.

[0159] Correspondingly, for devices following the reference unit in the priority sequence array, when the system issues control commands, its underlying modulation mode is uniformly configured to constant power control mode. This type of hydrogen production-side converter only needs to follow the system and allocate the target value for hydrogen production power, without needing to withstand transient impacts on the AC bus. Because the reference unit cycles daily, the voltage regulation control role and corresponding harsh operating conditions can be evenly rotated among all physical devices. Through this multi-dimensional deep binding mechanism, the system can achieve apparent balance not only in operating time and start-stop frequency, but also in the deep fatigue dimension of the core power devices, including thermal stress and insulation loss.

[0160] For example, this application also proposes a system-level power electronic conversion collaborative control, specifically including:

[0161] Step 601: When the real-time photovoltaic output power data is less than the system's minimum safe operating power, a shutdown disconnection command is generated.

[0162] In other words, when the real-time photovoltaic output power data is less than the pre-configured minimum safe operating power of the system, a shutdown disconnection command is generated. In this embodiment, the underlying controller on the photovoltaic side monitors the bus voltage and output current in real time. Based on the minimum modulation threshold of the inverter hardware and the reverse current protection safety margin, the minimum safe operating power of the system is pre-configured. When the photovoltaic output is extremely weak and falls below this safety threshold, the controller immediately blocks the drive pulses of the switching devices and generates a physical disconnection command to prevent the inverter from entering an uncontrollable oscillation region or drawing active power back from the AC bus.

[0163] Step 602: Based on the real-time power acquisition, calculate the power-voltage change rate, and calculate the dynamic adjustment step size using the product of the first step length coefficient and the power-voltage change rate. Based on this, generate the rated power point tracking control command. In other words, when the real-time photovoltaic output power data is within the pre-configured system safe operating power range, based on the real-time acquired output voltage and output power of the photovoltaic converter, calculate the power-voltage change rate, calculate the dynamic adjustment step size using the product of the pre-configured first step length coefficient and the power-voltage change rate, and generate the rated power point tracking control command based on the dynamic adjustment step size.

[0164] Under sunlight conditions, the system needs to extract the output energy of the photovoltaic array. Therefore, the system employs a variable-step-size incremental conductance method for optimization. Unlike the fixed-step-size method, this embodiment dynamically changes the disturbance amplitude by capturing the instantaneous derivative of power with respect to voltage. The logic for dynamically adjusting the step size is as follows:

[0165] ΔU _mppt =a×(dP / dU);

[0166] Wherein, ΔU _mpptTo calculate the generated dynamic adjustment step size, 'a' is the pre-configured first step size coefficient, 'dP' is the differential change in the photovoltaic array output power, and 'dU' is the differential change in the photovoltaic array output voltage. Therefore, when the operating point is far from the rated power point, the absolute value of the power-voltage change rate is large, and the resulting step size increases accordingly, making tracking faster; when the operating point approaches the rated power point, the change rate tends to zero, and the step size automatically converges to a minimum value, which can eliminate power optimization oscillations under steady state.

[0167] In one alternative implementation, the first step length coefficient can be determined using conventional control parameter tuning methods based on the slope range of the photovoltaic array's output characteristic curve and the controller's sampling period, to balance tracking speed and steady-state oscillation amplitude. The second step length coefficient can be adaptively adjusted based on the system's allowable bus voltage regulation rate. The hysteresis control threshold can be determined using conventional engineering tuning based on the system's rated power absorption and required disturbance rejection margin.

[0168] Step 603: When the real-time photovoltaic output power data cannot be absorbed, the constant direction adjustment step size is calculated using the second step size coefficient, and a power limiting control command is generated accordingly.

[0169] In other words, when all electrolyzers in all operating states reach the pre-configured rated power of a single electrolyzer and the source-load direct connection architecture lacks available energy buffer margin, resulting in the inability to further absorb real-time photovoltaic output power data, the constant direction adjustment step size is calculated using the pre-configured second step size coefficient, and a power limiting control command is generated based on the constant direction adjustment step size.

[0170] In some embodiments, when all operating electrolyzers reach the pre-configured single-unit rated power and the source-load direct connection architecture has no available energy buffer margin, making it impossible to further absorb the real-time photovoltaic output power data, the system uses the pre-configured second step size coefficient to calculate the constant direction adjustment step size and generate a power limiting control command.

[0171] Under abundant sunlight, when all equipment at the hydrogen production station is started and operating at full load, the overall system absorption capacity reaches its limit. This system is a direct-load source architecture without energy storage, meaning excess photovoltaic power cannot be transferred to the battery bank for time-shifting. Continuing to execute rated power point tracking would cause the bus voltage to rise uncontrollably, leading to overvoltage on the equipment. Therefore, the controller forcibly exits the optimization mode and enters a power-limited operation mode. The corresponding adjustment logic is as follows:

[0172] ΔU _cpg =k;

[0173] Wherein, ΔU _cpgThe constant-direction adjustment step size is defined by k, which is a pre-configured second step size coefficient. By superimposing this constant-direction adjustment step size, the system actively shifts the operating voltage of the photovoltaic array towards the open-circuit voltage direction, forcibly clamping the photovoltaic output power to the system's current rated absorption level, thus protecting the entire equipment.

[0174] Step 604 involves comparing the real-time photovoltaic output power data with a hysteresis control threshold. Based on the comparison result, it is determined whether to switch between the rated power point tracking control command and the power limiting control command. In other words, when switching between the rated power point tracking control command and the power limiting control command, the real-time photovoltaic output power data is compared with a preset hysteresis control threshold. Based on the comparison result, it is determined whether to switch the operating mode to achieve anti-oscillation control. That is, the preset hysteresis control threshold is used for anti-oscillation determination.

[0175] To prevent the system from frequently switching operating modes near the rated power absorption limit, which could lead to fatigue damage to power electronic switching devices, this embodiment introduces a hysteresis comparator at the switching boundary between the two instruction logics. Specifically, the system only switches from rated power point tracking to power limiting mode when the real-time photovoltaic power exceeds the sum of the maximum absorption level and a preset hysteresis control threshold; and only when the power falls back to the rated absorption level minus the threshold is the system allowed to resume rated power point tracking mode.

[0176] Step 605: Divide the hydrogen production-side converter into a first-class converter and a second-class converter, with the number of first-class converters set to one. That is, based on the physical characteristics of the source-load direct-connection architecture lacking an independent grid-supported power supply, the hydrogen production-side converter in the AC-coupled photovoltaic hydrogen production system is divided into a first-class converter and a second-class converter, with the number of first-class converters set to one.

[0177] The system uses a medium-voltage AC bus for energy collection and operates in an islanded state when disconnected from the external power grid. To establish and maintain a stable AC bus voltage and frequency, the system requires asymmetric topology partitioning of the underlying buck converters. Conventional islanded microgrids rely on power storage converters (PCS) for grid-type voltage and frequency support. However, this system lacks energy storage devices, and due to these physical constraints, the microgrid exhibits passive network characteristics without physical inertia. The system forcibly classifies one specific buck converter as a Class I converter, serving as the system's absolute voltage reference source; all remaining operating buck converters are uniformly defined as Class II converters, serving as controlled current sources for follow-up scheduling.

[0178] Step 606: For the first type of converter, generate an input voltage stabilization control command to maintain the stability of the AC bus voltage.

[0179] Specifically, the first type of converter is used to smooth active power fluctuations across the entire islanded microgrid. Its underlying controller employs a dual-closed-loop proportional-integral (PI) control structure with an outer loop for DC bus voltage and an inner loop for inductor current. The outer loop acquires the input-side DC bus voltage, compares it with a preset reference value, and outputs a reference current command. The inner loop adjusts the duty cycle of the switching transistors according to this command. By forcibly anchoring the voltage level on the input side of the first type of converter, the system utilizes the clamping effect of the multi-pulse rectifier to effectively stabilize the line voltage of the upstream medium-voltage AC transmission line in reverse.

[0180] Step 607: Based on the hydrogen production power distribution and the electrolyzer power-voltage characteristic curve, perform data mapping to obtain the corresponding electrolyzer voltage reference value.

[0181] The second type of converter, acting as a follower, executes power quotas issued by the upper energy management layer. Based on the power allocated for hydrogen production, the system uses locally stored electrolyzer power-voltage characteristic curves for data mapping, losslessly converting power commands into underlying electrical reference signals, namely electrolyzer voltage reference values, to determine the given objectives of the control loop.

[0182] Step 608: Based on the electrolytic cell voltage reference value, generate a voltage tracking control command for the second type of converter, which includes interleaved parallel modulation and average current sharing control logic. Alternatively, based on the electrolytic cell voltage reference value, generate a voltage tracking control command for the second type of converter, which includes interleaved parallel modulation and average current sharing control logic, to achieve steady-state power distribution control for each target operating electrolytic cell.

[0183] To meet the low-ripple DC power requirements of large-capacity alkaline water electrolyzers, this embodiment refines the physical topology and control logic of the second type of converter. The second type of converter employs a multi-phase interleaved parallel DC-DC buck converter topology. For example, four identical DC-DC buck chopper modules can be connected in parallel for power supply. The system is configured with drive pulse phase angles such that the switching transistor drive signals of the four modules are staggered by 90 electrical degrees. This interleaved parallel modulation technique can reduce the current ripple amplitude injected into the electrolyzer on the output side without increasing the switching frequency, utilizing the multi-phase ripple cancellation effect. This can extend the life of the electrolytic electrodes and improve the purity of hydrogen production.

[0184] Furthermore, due to manufacturing deviations in the parasitic parameters within the parallel modules, using only a uniform voltage tracking closed loop could lead to excessive current in one module, causing overheating and damage. Therefore, the system embeds average current sharing control logic when generating voltage tracking control commands.

[0185] In the implementation, the system collects the actual inductor current of each parallel module and calculates the global average value. The calculation formula is as follows:

[0186] i _Lavg=(i _L1 +i _L2 +i _L3 +i _L4 ) / 4;

[0187] Among them, i _Lavg i is the calculated global average value of the inductor current of the four modules. _L1 i _L2 i _L3 and i _L4 These are the real-time inductor currents of the first, second, third, and fourth parallel buck modules, respectively.

[0188] Accordingly, based on the unified duty cycle reference value generated by the outer voltage loop, the system introduces a current deviation correction mechanism for each independent buck module. The duty cycle correction formula is expressed as:

[0189] d _j =d _ref +G _c ×(i _Lavg -i_Lj);

[0190] Where, d _j d represents the actual duty cycle command ultimately output by a specific buck module. _ref G is the reference duty cycle command for the unified output of the outer loop voltage controller. _c For the pre-configured gain parameters of the current sharing loop proportional regulator, i _Lavg For the global average, i _Lj The value of variable j represents the real-time inductor current of the specific module itself, and can be any integer from 1 to 4.

[0191] Based on this, when the current of a module is higher than the global average, the system automatically reduces its drive duty cycle; conversely, it increases its duty cycle. This average current sharing control logic is based on feedback fine-tuning, which ensures that the total hydrogen production current is evenly distributed to each hardware branch, suppresses internal circulating current between modules, and enables multiple electrolyzers to distribute power in a steady state.

[0192] As an example, this embodiment presents a typical operating condition calculation case and system benefit verification based on meteorological data from Inner Mongolia.

[0193] In this embodiment, an application case of an AC-coupled photovoltaic hydrogen production system for a wind and solar resource base in Inner Mongolia is constructed. The pre-configured hardware physical parameters are as follows: the total installed capacity of the photovoltaic power station is 6.3 MW; the medium-voltage AC transmission line uses 95 mm² steel-cored aluminum stranded wire, with a total line length of 50 km and a rated transmission line voltage of 35 kV; the hydrogen production station is equipped with 5 identical alkaline water electrolyzers, each with a rated power of 1 MW, and the pre-configured minimum operating power of a single electrolyzer is set at 0.2 MW. Regarding the bus voltage reference parameters, the DC bus voltage on the photovoltaic side is stabilized at 1500 V, the AC output voltage of the three-phase inverter is stabilized at 690 V, and the DC bus voltage after multi-pulse rectification on the hydrogen production side is stabilized at 675 V.

[0194] Step 701: Obtain the day-ahead photovoltaic power forecast data and real-time photovoltaic output power data of the AC-coupled photovoltaic hydrogen production system.

[0195] Specifically, the system accesses a year-round historical meteorological dataset for Inner Mongolia. This dataset contains hourly solar irradiance and ambient temperature sequences. The energy management layer inputs the meteorological sequences into the photovoltaic cell engineering mathematical model to calculate the day-ahead photovoltaic power forecast data for the entire year. In the intraday closed-loop control verification, the system generates real-time photovoltaic output power data with high-frequency perturbations based on small time-step dynamics. This data is used to evaluate the system's ability to issue commands in the face of complex weather changes.

[0196] Step 702: Based on the day-ahead photovoltaic power forecast data and the pre-configured rated power of a single electrolyzer, calculate the upper limit of the planned number of electrolyzers to be in operation.

[0197] In real-world scenarios, each morning, the system extracts the theoretical maximum peak photovoltaic output for the day. For example, on a sunny summer day in Inner Mongolia, after deducting losses, the hydrogen production power is 5.2 MW. The system divides 5.2 MW by the rated 1 MW for a single electrolyzer, resulting in a quotient of 5.2. Next, the system performs a rounding up operation, yielding a planned maximum of 6 electrolyzers. Since the total number of pre-configured electrolyzers in this embodiment is only 5, the system automatically truncates this maximum to 5, ensuring all equipment is in a ready-to-wake state. On cloudy winter days, if the predicted peak is only 2.8 MW, the maximum number is limited to 3, with the remaining 2 forced into deep sleep mode to reduce unnecessary warm-up energy consumption.

[0198] Step 703: Determine the target number of operating electrolytic cells based on real-time photovoltaic output power data, the upper limit of the planned number of operating electrolytic cells, and preset power allocation rules.

[0199] Specifically, the system invokes three different control rules to conduct parallel data comparisons across all grades. The first set of comparative verifications uses a traditional simple start-stop strategy. When the total photovoltaic power exceeds the rated power limit of the currently operating equipment and is greater than the minimum operating limit of the next equipment by 0.2 MWh, the system issues an increase-capacity command. The second set of comparative verifications uses a segmented start-up strategy, starting a new equipment only after the already started equipment reaches the 0.5 MWh operating boundary. In the third core verification, the system fully activates the adaptive hysteresis switching strategy based on an economic model proposed in this invention. The controller loads a pre-built multi-dimensional increase-capacity threshold lookup table. When determining whether to increase capacity, it introduces the minimum revenue recovery time of 0.5 hours and the equivalent hydrogen loss parameter of 1.5 kg, and generates an asymmetric hysteresis loop by calculating the net hydrogen production revenue balance equation. Through this comparison, when the photovoltaic power frequently crosses thresholds such as 1.0 MWh and 2.0 MWh, the system, based on the state machine memory characteristics, shields invalid increase / decrease actions and can output a stable and disturbance-resistant target number of operating electrolyzers.

[0200] Step 704: Calculate the hydrogen production allocation power based on the target number of operating electrolyzers and real-time photovoltaic output power data.

[0201] After obtaining a stable target number of electrolyzers, the system executes power allocation logic. For example, if the real-time photovoltaic output power data suddenly changes to 4.2 MW, and the determined target number of operating electrolyzers is 5, the system divides 4.2 MW by 5, resulting in a single-cell hydrogen production allocation power of 0.84 MW. The underlying reference target for all 5 electrolyzers is clamped to this value. This mechanism ensures that no equipment in the simulation system falls into a state of full load or underload, and all equipment is aligned to the same operating point on the electrolyzer power and efficiency characteristic curves.

[0202] Step 705: Generate converter control commands based on the target number of operating electrolyzers, hydrogen production allocation power, and preset equipment balancing rotation mechanism.

[0203] The system combines advanced-last-out (IF) logic with a daytime shift mechanism for the reference unit to convert power ratings into control levels for specific hardware. After a year of continuous, all-weather simulation, the system compiles overall performance data under the three strategies.

[0204] According to the data exported from the data platform, under the simple start-stop strategy, the total annual hydrogen production was 159,361.05 kg, and the total number of start-stop operations for the five electrolyzers was 1,425. Under the segmented start-up strategy, the efficiency of the low-load range was utilized, and the total hydrogen production increased to 165,742.15 kg. However, this triggered drastic changes in equipment operating conditions, resulting in a total of 1,780 start-stop operations for the year, leading to mechanical fatigue in the equipment.

[0205] After adopting the collaborative scheduling method of this invention, the system achieved a peak annual cumulative hydrogen production of 165,923.72 kg, while controlling the total number of system start-ups and shutdowns to 1,464. Compared with the segmented start-up strategy, this invention reduces the number of ineffective start-ups and shutdowns by 17.7% while improving the green electricity conversion rate. Furthermore, based on the synergistic effect of in-first-out and daytime circulation shifting, the variance of the cumulative operating time of the five electrolyzers approaches zero.

[0206] It should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

Claims

1. A cross-timescale multi-electrolyzer collaborative scheduling method, applied to an AC-coupled photovoltaic hydrogen production system, wherein the system is configured as a source-load direct connection architecture where the photovoltaic power station directly supplies power to the hydrogen production station via AC transmission lines, characterized in that, include: Acquire the system's day-ahead photovoltaic power forecast data and real-time photovoltaic output power data; Based on the current photovoltaic power forecast data and the rated power of a single electrolytic cell, calculate the upper limit of the planned number of electrolytic cells to be in operation; Based on real-time photovoltaic output power data, the upper limit of the planned number of electrolyzers to be operated, and the preset power allocation rules, the target number of electrolyzers to be operated is determined. Based on the target number of operating electrolyzers and real-time photovoltaic output power data, calculate the hydrogen production allocation power. Based on the target number of operating electrolyzers, hydrogen production power allocation, and equipment balancing rotation mechanism, converter control commands are generated.

2. The method according to claim 1, characterized in that, Determine the target number of operating electrolyzers, specifically including: Divide the real-time photovoltaic output power data by the minimum operating power of a single electrolytic cell and round down to obtain the initial number of operating cells. The smaller of the initial number of operating cells and the upper limit of the planned number of electrolytic cells to be operated is taken as the basic number of operating cells; If the number of basic operating cells equals the upper limit of the planned number of operating electrolyzers, and the real-time photovoltaic output power data is greater than the product of the number of basic operating cells and the rated power of a single electrolyzer, it is determined that a photovoltaic output prediction error has occurred. Based on the prediction error, the basic operating number is increased sequentially until the real-time photovoltaic output power data is fully absorbed or the total number of electrolyzers is limited, and the corrected operating number is obtained, which is then used as the target operating number of electrolyzers.

3. The method according to claim 1, characterized in that, Generate converter control commands, including generating photovoltaic terminal control commands for the photovoltaic converter in the control system, specifically including: Generate a shutdown disconnect command; Based on the real-time power acquisition, the power voltage change rate is calculated, and the step size is dynamically adjusted by using the product of the first step length coefficient and the power voltage change rate. Based on this, the rated power point tracking control command is generated. When the real-time photovoltaic output power data cannot be absorbed, the constant direction adjustment step size is calculated using the second step size coefficient, and a power limiting control command is generated accordingly. The process involves comparing real-time photovoltaic output power data with hysteresis control thresholds, and determining whether to switch between rated power point tracking control and power limiting control based on the comparison results.

4. The method according to claim 1, characterized in that, Generate converter control commands, including generating hydrogen production end control commands for controlling the hydrogen production side converter in an AC-coupled photovoltaic hydrogen production system, specifically including: The hydrogen production-side converter is divided into a first-class converter and a second-class converter, with the number of first-class converters set to one. For the first type of converter, generate input voltage stabilization control commands; Based on the hydrogen production power distribution and the power-voltage characteristic curve of the electrolyzer, the corresponding reference value of the electrolyzer voltage is obtained by data mapping. Based on the reference value of the electrolytic cell voltage, a voltage tracking control command containing interleaved parallel modulation and average current sharing control logic is generated for the second type of converter.

5. The method according to claim 1, characterized in that, Based on real-time photovoltaic output power data, the upper limit of the planned number of electrolyzers in operation, and preset power allocation rules, the target number of electrolyzers in operation is determined, and an adaptive hysteresis switching strategy is adopted for execution, specifically including: Obtain the current operating state variables of the electrolytic cell at the current moment; Based on the current electrolytic cell operating status variables and switching determination parameters, the corresponding dynamic increase power threshold and dynamic decrease power threshold are obtained respectively. The dynamic increase power threshold is greater than the dynamic decrease power threshold, forming an asymmetric hysteresis loop. When the real-time photovoltaic output power data is greater than the dynamic capacity increase threshold and the current electrolyzer operating status variable is less than the upper limit of the planned number of electrolyzers, the current electrolyzer operating status variable is increased. When the real-time photovoltaic output power data is less than the dynamic power reduction threshold and the current electrolytic cell operating status variable is greater than zero, the current electrolytic cell operating status variable is reduced. Update the current electrolytic cell operating status variable to determine the target number of electrolytic cells in operation.

6. The method according to claim 5, characterized in that, Before obtaining the corresponding dynamic capacity increase and dynamic capacity decrease thresholds, the process also includes model construction for quantifying the net hydrogen production revenue and physical start-up and shutdown costs, specifically including: By using the power-efficiency characteristic curve of the electrolyzer, the difference in hydrogen production rate before and after adding a single operating electrolyzer, as represented by the real-time photovoltaic output power data, is calculated, and a marginal hydrogen production gain model is constructed. Obtain the hydrogen equivalent of the warm-up cost and the hydrogen equivalent of the equipment wear and tear cost; An equivalent hydrogen loss model is constructed by summing the hydrogen equivalent of the warm-up cost and the hydrogen equivalent of the equipment loss cost.

7. The method according to claim 6, characterized in that, Based on the current electrolyzer operating status variables, the corresponding dynamic power increase threshold and dynamic power decrease threshold are obtained respectively, including: Achieve the shortest payback period; The product of the marginal hydrogen production gain model and the shortest revenue recovery time is equal to the equivalent hydrogen loss model. A balance equation for the net hydrogen production revenue is constructed, and the balance equation is solved to obtain the dynamic booster power threshold. Obtain the minimum operating power and safety margin power of a single electrolytic cell; Multiply the current electrolytic cell operating status variable by the minimum operating power of a single electrolytic cell, and add the safety margin power to obtain the dynamic reduction power threshold; Among them, the dynamic increase power threshold is greater than the dynamic decrease power threshold corresponding to the increased number of operating units after the target is achieved, and the difference constitutes the switching hysteresis bandwidth.

8. The method according to claim 1, characterized in that, Generate converter control commands, including executing intraday start / stop logic based on a preset equipment balancing and rotation mechanism, specifically including: Get the startup priority sequence for the day; When the number of target electrolytic cells in operation exceeds the current number of cells in operation of the system, start-up control commands are generated by sequentially selecting the electrolytic cells to be shut down according to the positive order of the start-up priority sequence of the day. When the number of target electrolytic cells in operation is less than the current number in operation, based on the last-in-first-out principle, the last electrolytic cell to be started is selected to generate a shutdown control command.

9. The method according to claim 1, characterized in that, Based on recent photovoltaic power forecasts and the pre-configured rated power of a single electrolyzer, the upper limit of the planned number of electrolyzers in operation is calculated, specifically including: Extract the daily maximum predicted power from the current day's photovoltaic power forecast data; Divide the maximum predicted daily power by the rated power of a single electrolytic cell, and round up the result to obtain the upper limit of the planned number of electrolytic cells in operation.

10. The method according to claim 1, characterized in that, Based on the target number of operating electrolyzers and real-time photovoltaic output power data, the hydrogen production power allocation is calculated, specifically including: When the number of target electrolyzers in operation is greater than zero, the real-time photovoltaic output power data is evenly distributed to the corresponding number of operating electrolyzers to obtain the single-unit target hydrogen production allocation power, which is used as the hydrogen production allocation power.

Images