A method for suppressing DC bus voltage fluctuations in a renewable energy hydrogen production system

Abstract

This invention discloses a method for suppressing DC bus voltage fluctuations in a renewable energy hydrogen production system, relating to the field of renewable energy electrolysis hydrogen production technology. The method includes: real-time acquisition of system operating parameters; estimation of bus voltage disturbance power based on a disturbance observer and frequency decomposition; prediction of load power using ARIMA and LSTM models and calculation of prediction deviation; dynamic setting of upper and lower limits of the bus voltage control range based on the prediction deviation and operating status; construction of a multi-objective optimization function, comprehensively considering bus voltage disturbance power, prediction deviation, and control limits, to solve for the power regulation commands of battery storage, fuel cells, and electrolyzers; and execution of power regulation control to achieve bus voltage stability. This invention can improve the graded response capability to voltage disturbances, the accuracy of load power change prediction, the adaptability of control boundaries, and the overall operational stability of the system, and is suitable for power quality control of hydrogen production systems under multi-source fluctuation scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of renewable energy electrolytic hydrogen production, and particularly relates to a direct-current bus voltage fluctuation suppression method of a renewable energy hydrogen production system. BACKGROUND

[0002] With the acceleration of energy structure transformation, the penetration rate of renewable energy (such as photovoltaic and wind power) in the power system is increasing. However, due to its inherent volatility and unpredictability, the power quality problems (such as voltage fluctuation and frequency disturbance) are increasingly prominent, especially in the renewable energy driven hydrogen production system, which has a significant impact on the stability of the system direct-current bus voltage.

[0003] In such a system, the direct-current bus is connected to multiple core power components, including photovoltaic arrays, wind turbines, fuel cells, electrolytic cells, energy storage batteries and control inverters. Due to the great influence of environmental factors (such as sunlight intensity, temperature, wind speed, etc.) on renewable energy output, the load side power of the system fluctuates frequently, and the bus voltage is prone to severe fluctuations, which in turn affects the efficiency of electrolytic hydrogen production, the safety of the electrolytic cell and the service life of the energy storage device.

[0004] In the prior art, voltage limiting control with fixed threshold or charge-discharge scheduling strategy based on simple rules is usually used, which lacks the ability to perceive the system operating state; the traditional disturbance response method does not distinguish the disturbance frequency characteristics, resulting in the same response mechanism for high-frequency and low-frequency disturbances, and unreasonable allocation of control resources; when facing sensor failure, communication anomaly and other situations, the control system often lacks autonomous fault tolerance and voltage stability maintenance capability.

[0005] Therefore, there is an urgent need for a comprehensive method that can perform multi-objective adaptive optimization control based on system operating state, disturbance prediction and load dynamic change, to improve the operating stability and energy scheduling efficiency of the renewable energy hydrogen production system under complex working conditions. SUMMARY

[0006] To solve the above problems, the present application provides a direct-current bus voltage fluctuation suppression method of a renewable energy hydrogen production system, which reasonably allocates energy storage resources according to disturbance frequency, load prediction and other factors to realize the rapid stability of the direct-current bus voltage.

[0007] The present application achieves the above-mentioned purposes through the following technical solutions:

[0008] A direct-current bus voltage fluctuation suppression method of a renewable energy hydrogen production system, the method comprising:

[0009] Real-time acquisition of key operating parameters of the system, including direct-current bus voltage V dc , current, renewable energy power, current load power P load(t), battery state of charge SOC, hydrogen storage pressure P H2 , ambient temperature S I and light intensity T.

[0010] The bus voltage disturbance power d(t) is estimated by a disturbance observer, and is decomposed into a high-frequency disturbance component d high (t) and a low-frequency disturbance component d low (t).

[0011] Based on the load power data in the historical time, the load power P load (t+1) at the next moment is predicted by a hybrid load prediction algorithm, and a prediction deviation ΔP diff is calculated.

[0012] According to the system key operating parameters and the prediction deviation ΔP diff , the upper and lower limits V high (t), V low (t) of the bus voltage control interval are dynamically set.

[0013] A multi-objective optimization function is constructed, taking the bus voltage disturbance power d(t), the prediction deviation ΔP diff , and the upper and lower limits V high (t), V low (t) of the bus voltage control interval as inputs, to solve the power regulation instruction values of the battery energy storage, the fuel cell, and the electrolytic tank.

[0014] According to the power regulation instruction values , control instructions are generated and power output regulation is performed.

[0015] As a preferred scheme of the present application, the disturbance observer estimates the bus voltage disturbance power d(t) by using a dynamic gain model as follows:

[0016]

[0017] In the formula, C is the equivalent capacitance of the bus; is the current voltage rate of change, t represents time; K d (t) is the observer gain coefficient; sat() is the saturation function; V ref is the target reference voltage of the DC bus;

[0018] The expression of the observer gain coefficient K d (t) is as follows:

[0019]

[0020] wherein K0 is a basic gain constant; K1 is a voltage change rate gain coefficient; K2 is a voltage deviation gain coefficient; V dc (t) is a current bus voltage.

[0021] As a preferred scheme of the present application, the bus voltage disturbance power d(t) is subjected to frequency decomposition by wavelet transform or Fourier transform, and the disturbance with frequency components higher than a cutoff frequency fc is defined as a high-frequency disturbance component d high (t) to be responded by the battery energy storage system; and the disturbance with frequency components lower than the cutoff frequency fc is defined as a low-frequency disturbance component d low (t) to be responded by the electrolytic cell and fuel cell combination; and the cutoff frequency fc is adaptively set according to a power fluctuation period.

[0022] As a preferred scheme of the present application, the hybrid load prediction algorithm is specifically an ARIMA-LSTM dynamic fusion model, and the implementation method is as follows:

[0023] The historical load power data {P load (t-N+1),..., P load (t)} collected at N time instants are inputted and subjected to data preprocessing operations including normalization, filtering of spike abnormal values and smoothing processing, and the cleaned sequence data are outputted for modeling;

[0024] A disturbance amplitude factor D s and a dominant period length C p are constructed, and the formulae are as follows:

[0025]

[0026] wherein i is a sliding index, representing the i-th time difference value; P load (t-i+1) and P load (t-i) are the load powers at the t-i+1 time instant and the t-i time instant respectively; f peak is the maximum non-DC frequency component obtained by applying fast Fourier transform to the historical load power data;

[0027] A period identification range set Z={6h, 12h, 24h} is set, and the tolerance is ±15%;

[0028] If the disturbance amplitude factor D s is less than a preset load disturbance amplitude threshold and the dominant period length C p satisfies: and ∣C p -z∣≤0.15·z, then the ARIMA model is preferentially selected as the main model, and the LSTM model is selected as the secondary prediction path, otherwise the LSTM model is selected as the main model, and the ARIMA model is selected as the secondary prediction path.

[0029] The historical residual standard deviation and confidence weights of the ARIMA model and LSTM model are calculated respectively, and the predicted load power value P for the next time step is finally generated. load (t+1), the formula is:

[0030]

[0031] In the formula, σ ARIMA ω ARIMA , These represent the historical residual standard deviation, confidence weights, and prediction results of the ARIMA model, respectively; σ LSTM ω LSTM , These are the historical residual standard deviation, confidence weights, and prediction results of the LSTM model, respectively.

[0032] Prediction bias △P diff The calculation formula is: △P diff =P load (t+1)-P load (t).

[0033] As a preferred embodiment of the present invention, the upper and lower limits V of the dynamically set bus voltage control range are... high (t),V low (t), the methods include:

[0034] Initialization settings: Set the target reference voltage of the DC bus to V. ref And set the base voltage tolerance as ΔV base Therefore, the initial control interval is determined as: V low (0)=V ref -△V base V high (0)=V ref +△V base V high (0), V low (0) represents the initial upper and lower limits of the bus voltage, respectively;

[0035] Real-time parameter acquisition: During each control cycle, the battery state of charge (SOC) and hydrogen storage pressure (P) are acquired in real time. H2 and prediction deviation △P diff ;

[0036] Range adjustment rule setting: Based on the judgment results of the following three types of operating conditions, the upper and lower limits of the bus voltage control range are adjusted:

[0037] 1) If the battery state of charge (SOC) is less than the lower limit threshold SOC of the battery state of charge. low Then set Vlow (t) = V ref +△V SOC , where△V SOC is the voltage lower limit moving-up amount in the SOC protection state;

[0038] 2) If the hydrogen storage pressure P H2 is greater than the maximum allowable pressure P H2,max of the hydrogen tank, set V high (t) = V ref -△V H2 , where△V H2 is the voltage upper limit down-regulation amount when the hydrogen pressure is too high;

[0039] 3) If the predicted deviation△P diff is greater than the load power disturbance threshold P thresh and lasts for more than three sampling periods, expand the bus voltage control interval as follows:

[0040] V low (t) = V ref -(△V base +△V ext );

[0041] V high (t) = V ref +(△V base +△V ext );

[0042] where△V ext is the expanded voltage deviation interval;

[0043] Hysteresis interval judgment: if the interval change in the current period satisfies:

[0044] |V low (t) - V low (t-1)| < ε v ;

[0045] |V high (t) - V high (t-1)| < ε v ;

[0046] then the upper and lower limits of the bus voltage in the previous period are maintained unchanged;

[0047] In the formula, V high (t-1) and V low (t-1) are the upper and lower limits of the bus voltage control interval in the previous period, and ε v is the minimum adjustment threshold of the voltage lower limit.

[0048] As a preferred scheme of the present application, the expression of the multi-objective optimization function is:

[0049]

[0050] In the formula, J(t) is an optimized target function value, indicating the total control cost of the control system at the current time; is the target charge-discharge power of the battery energy storage at the current time, is the target output power of the fuel cell at the current time, is the power that the electrolyzer should absorb at the current time; V dc (t) is the DC bus voltage at the current time, V ref is the target reference voltage of the DC bus; and α, β, γ and δ are respectively a battery power response weight coefficient, a hydrogen energy system response weight coefficient, a voltage stability weight coefficient and a battery power change rate penalty weight.

[0051] As a preferred scheme of the application, the generating control instructions and performing power output adjustment specifically comprises:

[0052] Smooth sliding mode control of the battery energy storage system:

[0053] The hyperbolic tangent function is used to replace the traditional sign function to generate charge-discharge instructions, and to suppress chattering of the control instructions;

[0054] The charge-discharge power is dynamically limited according to the battery state of charge SOC, and only discharging is allowed when the SOC is higher than 90%, and only charging is allowed when the SOC is lower than 20%;

[0055] Hydrogen flow closed-loop control of the fuel cell:

[0056] By monitoring the fuel cell output current and hydrogen flow in real time, the theoretical hydrogen demand is calculated, and a low-frequency disturbance component d low (t) is added to generate a feedforward compensation flow instruction;

[0057] The PID controller is used for closed-loop adjustment of the hydrogen flow rate, so as to ensure that the hydrogen utilization rate is not lower than 95%, and to avoid fuel starvation or waste;

[0058] The output power instruction is dynamically adjusted according to the stack voltage and conversion efficiency, and the bus voltage demand is matched through the boost DC-DC converter;

[0059] MPPT- feedforward composite control of the electrolyzer:

[0060] The input voltage of the electrolyzer is periodically disturbed based on the maximum power point tracking (MPPT) algorithm, and a low-frequency disturbance component d low (t) is used for dynamic correction of hydrogen production power;

[0061] The hydrogen production power is limited in real time according to the hydrogen storage tank pressure, and the system is forced to stop when the pressure is out of limit, and the power is increased to the rated power when the pressure is insufficient.

[0062] As a preferred embodiment of the present invention, the method further includes an embedded PI control circuit to adjust the bus voltage deviation ΔV = V in real time. ref -V dc As input, the regulating power ΔP is generated. PI This is superimposed on the control commands;

[0063] When any of the following abnormal states are detected:

[0064] Sensor data interruption or exceeding limits;

[0065] The controller communication interruption exceeded the set time.

[0066] The energy storage unit has no power response for more than two sampling periods;

[0067] Automatically activate fault-tolerant logic to control the battery or fuel cell to maintain the bus voltage at a preset power level until the main control system recovers.

[0068] As a preferred embodiment of the present invention, the regulating power ΔP PI satisfy:

[0069]

[0070] In the formula, △P PI (t) represents the current regulation power of the PI feedback; K p V is the proportional gain coefficient; ref K is the target reference voltage for the DC bus. j τ is the integral gain coefficient; τ is the integral variable.

[0071] A computer-readable storage medium storing a computer program that, when executed by a processor, implements a DC bus voltage fluctuation suppression method for a renewable energy hydrogen production system as described above.

[0072] The beneficial effects of this invention are as follows: By comprehensively introducing disturbance observation, frequency decomposition response, load prediction, dynamic control interval setting, and multi-objective optimization control strategies, it can achieve accurate identification and graded response to bus voltage disturbances, improve the foresight and accuracy of load prediction, dynamically adapt to voltage control limits under different operating conditions, and effectively improve the system's operational stability, control intelligence, and response robustness through multi-objective function collaborative optimization of power regulation commands for battery storage, fuel cells, and electrolyzers. Furthermore, this method possesses anomaly detection and fault-tolerance mechanisms, ensuring bus voltage stability in the event of critical equipment failure or signal interruption, demonstrating good engineering adaptability and application value. Attached Figure Description

[0073] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] in:

[0075] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0077] like Figure 1 As shown, this is an embodiment of the present invention, which provides a method for suppressing DC bus voltage fluctuations in a renewable energy hydrogen production system, including the following steps:

[0078] S1: Real-time acquisition of key system operating parameters, including DC bus voltage V dc Current, renewable energy power, current load power P load (t), battery state of charge (SOC), hydrogen storage pressure P H2 Ambient temperature S I Light intensity T, etc.;

[0079] S2: Estimate the bus voltage disturbance power d(t) using a disturbance observer and decompose it into high-frequency disturbance components d high (t) and low-frequency disturbance component d low (t);

[0080] Specifically, the disturbance observer uses the following dynamic gain model to estimate the bus voltage disturbance power d(t):

[0081]

[0082] In the formula, C is the equivalent capacitance of the busbar; K represents the current rate of change of voltage, where t represents time; d (t) is the observer gain coefficient; sat() is the saturation function; V ref The target reference voltage for the DC bus;

[0083] Wherein, the observer gain coefficient K d The expression for (t) is:

[0084]

[0085] In the formula, K0 is the basic gain constant; K1 is the voltage change rate gain coefficient; K2 is the voltage deviation gain coefficient; V dc (t) represents the current bus voltage.

[0086] Frequency decomposition of the bus voltage disturbance power d(t) is performed using wavelet transform or Fourier transform, defining disturbances with frequency components higher than the cutoff frequency fc as high-frequency disturbance components d. high (t), the response of the battery energy storage system; disturbances with frequency components below the cutoff frequency fc are defined as low-frequency disturbance components d. low (t), is a joint response of the electrolyzer and the fuel cell; the cutoff frequency fc is adaptively set according to the power fluctuation period.

[0087] S3: Based on historical load power data, predict the load power P at the next moment using a hybrid load forecasting algorithm. load (t+1), and calculate the prediction deviation △P diff .

[0088] This embodiment's hybrid load forecasting algorithm introduces a time-adaptive discrimination mechanism to analyze the disturbance amplitude and periodicity of the input data, dynamically selecting either ARIMA or LSTM as the main predictor. ARIMA is used to reduce delay in the stable prediction region, while LSTM is invoked to enhance nonlinear modeling capabilities in the abrupt change region. A confidence-weighted mechanism is also introduced to improve prediction accuracy at abrupt change edges. The final prediction results are used for feedforward regulation in bus voltage control, enhancing the system's anticipatory response capability.

[0089] In one embodiment, the hybrid load forecasting algorithm is specifically an ARIMA-LSTM dynamic fusion model, implemented as follows:

[0090] S31: Input the historical load power data collected at N time points {P load (t-N+1), ..., P load (t)}, and perform data preprocessing operations, including normalization, filtering out spikes and outliers, and smoothing (such as double exponential weighted moving average), and output cleaned sequence data for modeling;

[0091] S32: Construct the perturbation amplitude factor D s and dominant cycle length C p The formula is:

[0092]

[0093] In the formula, i is the sliding index, representing the i-th time difference value; P load (t-i+1), P load (ti) represent the load power at time t-i+1 and time ti, respectively; f peak The maximum non-DC frequency component obtained after applying a fast Fourier transform to historical load power data;

[0094] Set the periodic recognition range set Z = {6h, 12h, 24h}, with a tolerance of ±15%;

[0095] S33: Judgment Rule (Adjustable): If the disturbance amplitude factor D s The load disturbance amplitude is less than the preset threshold and the dominant period length C p satisfy: Make |C p -z∣≤0.15·z (i.e., the dominant period length C) p If the prediction path falls within ±15% of any typical cycle in the cycle identification range set, then the ARIMA model is preferred as the main model and the LSTM model is preferred as the secondary prediction path; otherwise, the LSTM model is preferred as the main model and the ARIMA model is preferred as the secondary prediction path.

[0096] If the sampling period is 1 minute, the preset load disturbance amplitude threshold can be set to 5kW (i.e., load changes of no more than 5kW within one minute are considered stable). The load disturbance amplitude threshold can be set between 3kW and 10kW based on the actual load fluctuation amplitude statistics. ARIMA excels at periodic modeling but is not good at handling sudden fluctuations, while LSTM is good at nonlinear modeling but may overfit or be insensitive to short-term fluctuations. Combining the two can achieve better prediction performance under multiple operating conditions, especially in: renewable energy scenarios; mixed load (constant power + random load) scenarios; and systems where complex weather changes affect load fluctuations.

[0097] S34: Calculate the historical residual standard deviation and confidence weights for the ARIMA and LSTM models respectively, and finally generate the predicted load power value P for the next time step. load (t+1), the formula is:

[0098]

[0099] In the formula, σ ARIMA ω ARIMA , These represent the historical residual standard deviation, confidence weights, and prediction results of the ARIMA model, respectively; σ LSTM ω LSTM , These are the historical residual standard deviation, confidence weights, and prediction results of the LSTM model, respectively.

[0100] S35: Prediction bias △P diff The calculation formula is: △P diff =P load (t+1)-P load (t).

[0101] S4: Based on key system operating parameters and prediction deviation △P diff Dynamically set the upper and lower limits V of the bus voltage control range. high (t),V low (t).

[0102] Further, step S4 includes:

[0103] S41: Initialization settings

[0104] Set the target reference voltage of the DC bus to V. ref And set the base voltage tolerance as ΔV base Therefore, the initial control interval is determined as: V low (0)=V ref -△V base V high (0)=V ref +△V base V high (0), V low (0) are the initial upper and lower limits of the bus voltage, which are the upper and lower limits of the bus voltage control set when the system starts up or enters the voltage control logic for the first time;

[0105] S42: Real-time parameter acquisition

[0106] During each control cycle, the battery state of charge (SOC) and hydrogen storage pressure (P) are collected in real time. H2 and prediction deviation △P diff ;

[0107] S43: Setting Interval Adjustment Rules

[0108] Based on the following three types of operating condition judgments, the upper and lower limits of the bus voltage control range are adjusted:

[0109] 1) If the battery state of charge (SOC) is less than the lower limit threshold SOC of the battery state of charge. low Then set V low (t)=V ref +△V SOC Used to suppress further battery discharge; where ΔV SOC This is the upward shift of the lower voltage limit under SOC protection conditions. When the actual SOC is lower than the SOC... lowWhen necessary, disable or limit battery discharge to prevent damage from excessive discharge; reduce battery load by manually raising the minimum control value of the bus.

[0110] 2) If the hydrogen storage pressure P H2 Greater than the maximum allowable pressure P of the hydrogen tank H2,max Then set V high (t)=V ref -△V H2 Used to limit the continued hydrogen production of the electrolyzer; where ΔV H2 This is the amount by which the upper limit of the voltage is lowered when the hydrogen pressure is too high; P H2,max Used to prevent the electrolyzer from continuing to operate when the hydrogen storage system is saturated, ensuring system safety; when the hydrogen pressure is detected to exceed the upper limit, the bus voltage drop ΔV is applied to prevent continued hydrogen production. H2 This is used to suppress the electrolyzer from continuing to receive power and produce hydrogen, indirectly reducing the bus voltage;

[0111] 3) If the prediction deviation ΔP diff Greater than the load power disturbance threshold P thresh If the threshold value at which the system judges a drastic load change is reached and remains so for more than three sampling cycles, then the bus voltage control range is expanded as follows:

[0112] V low (t)=V ref -(△V base +△V ext );

[0113] V high (t)=V ref +(△V base +△V ext );

[0114] Wherein, △V ext To extend the voltage deviation range and enhance the system's dynamic response to high-power disturbances;

[0115] S44: Hysteresis Interval Judgment

[0116] If the interval change in the current period satisfies:

[0117] |V low (t)-V low (t-1)|<ε v ;

[0118] |V high (t)-V high (t-1)|<ε v ;

[0119] The upper and lower limits of the bus voltage from the previous cycle will remain unchanged to avoid frequent switching due to minor fluctuations.

[0120] In the formula, V high (t-1), V low (t-1) represent the upper and lower limits of the bus voltage control range in the previous cycle, respectively, and ε v This is the minimum adjustment threshold for the lower voltage limit.

[0121] Updated voltage upper and lower limits V high (t), V low (t) is used as the upper and lower limits of the bus disturbance mode identification and judgment, the adjustment trigger conditions of each energy storage device, and the reference signal of the feedback controller.

[0122] S5: Construct a multi-objective optimization function, taking the bus voltage disturbance power d(t) and prediction error ΔP as the parameters. diff And the upper and lower limits V of the bus voltage control range high (t),V low (t) is the input, and the power regulation command values ​​of the battery storage, fuel cell, and electrolyzer are solved.

[0123] The expression for the multi-objective optimization function is:

[0124]

[0125] In the formula, J(t) is the value of the objective function, representing the total control cost of the control system at the current moment; The target charge / discharge power of the battery energy storage at the current moment. The target output power of the fuel cell at the current moment. V represents the power that the electrolytic cell should absorb at the current moment. dc (t) represents the DC bus voltage at the current moment, V ref α represents the target reference voltage of the DC bus; α, β, γ, and δ represent the weighting coefficients for battery power response, hydrogen energy system response, voltage stability, and battery power change rate penalty, respectively.

[0126] This multi-objective optimization function can be solved using methods such as weighted linear method, particle swarm optimization algorithm, and NSGA-II (non-dominated sorting genetic algorithm).

[0127] S6: Based on the power adjustment command value Generate control commands and execute power output regulation.

[0128] Step S6 specifically includes:

[0129] Smooth sliding mode control for battery energy storage systems:

[0130] The hyperbolic tangent function is used to replace the traditional sign function in generating charge and discharge commands, thus suppressing control command chattering.

[0131] The charging and discharging power is dynamically limited based on the battery's state of charge (SOC). When the SOC is above 90%, only discharging is allowed, and when it is below 20%, only charging is allowed.

[0132] Closed-loop control of hydrogen flow in fuel cells:

[0133] The theoretical hydrogen demand is calculated by real-time monitoring of the fuel cell output current and hydrogen flow rate, and a low-frequency disturbance component d is added. low (t) Generate feedforward compensation flow command;

[0134] A PID controller is used to regulate the hydrogen flow rate in a closed loop to ensure that the hydrogen utilization rate is not less than 95% and to avoid fuel starvation or waste.

[0135] The output power command is dynamically adjusted according to the stack voltage and conversion efficiency, and the bus voltage requirement is matched by a boost DC-DC converter.

[0136] MPPT-feedforward composite control of electrolyzers:

[0137] The electrolytic cell input voltage is periodically perturbed based on the maximum power point tracking (MPPT) algorithm, combined with the low-frequency perturbation component d. low (t) Dynamically correct hydrogen production power;

[0138] The hydrogen production capacity is limited in real time based on the pressure of the hydrogen storage tank. When the pressure exceeds the limit, the machine is forced to shut down. When the pressure is insufficient, the production capacity is increased to the rated power.

[0139] In another specific embodiment, the method further includes an embedded PI control circuit to adjust the bus voltage deviation ΔV = V in real time. ref -V dc As input, the regulating power ΔP is generated. PI This is superimposed on the control commands;

[0140] Adjusting power ΔP PI satisfy:

[0141]

[0142] In the formula, △P PI (t) represents the current regulation power of the PI feedback; K p V is the proportional gain coefficient; ref K is the target reference voltage for the DC bus. j τ is the integral gain coefficient; τ is the integral variable;

[0143] When any of the following abnormal states are detected:

[0144] Sensor data interruption or exceeding limits;

[0145] The controller communication interruption exceeded the set time.

[0146] The energy storage unit has no power response for more than two sampling periods;

[0147] Automatically activate fault-tolerant logic to control the battery or fuel cell to maintain the bus voltage at a preset power level until the main control system recovers.

[0148] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any other combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product, which includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0149] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc. Therefore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the DC bus voltage fluctuation suppression method for a renewable energy hydrogen production system as described above.

[0150] In summary, this invention achieves accurate estimation and classification response to DC bus voltage disturbance power. By introducing a disturbance observer and performing frequency decomposition on the disturbance power, high-frequency disturbances and low-frequency disturbances can be separated and processed, with the battery energy storage system and fuel cell / electrolyte system undertaking the response tasks respectively. This achieves reasonable allocation of disturbance energy and effectively suppresses high-frequency pulsation and low-frequency drift of the bus voltage.

[0151] This invention improves the dynamic load prediction capability and enhances the forward-looking nature of control. It employs a load prediction algorithm that fuses ARIMA and LSTM deep learning models, combined with a dominant cycle determination and disturbance amplitude factor extraction mechanism. While ensuring prediction accuracy, it significantly improves the system's ability to perceive future load change trends, thereby enhancing the predictability and robustness of the control strategy.

[0152] This invention enables adaptive dynamic adjustment of the bus voltage control range. By comprehensively considering various operating parameters such as battery state of charge (SOC), hydrogen storage pressure, and load power prediction deviation, the upper and lower limits of bus voltage control can be dynamically set under different operating conditions. This avoids frequent false triggering or control failure caused by improper setting in traditional static limit control, thus improving the flexibility and reliability of bus voltage control.

[0153] This invention constructs a multi-objective optimization function to achieve coordinated power regulation control. The proposed multi-objective optimization function simultaneously considers multiple factors such as battery power response cost, hydrogen energy system response cost, voltage deviation stability, and power change rate penalty. Based on a comprehensive balance between system operating cost and control effect, it can dynamically solve for the optimal power regulation command, realizing coordinated and efficient operation among the electrolyzer, fuel cell, and battery energy storage unit.

[0154] This invention enhances the anomaly handling and fault tolerance capabilities of the control system. By setting up an embedded PI feedback regulation loop and multiple anomaly identification mechanisms (such as sensor data interruption, control communication failure, energy storage response failure, etc.), fault-tolerant logic can be automatically triggered when critical equipment in the system fails, and backup control strategies can be invoked to maintain the stability of the bus voltage, significantly improving the system's safety and continuous operation capability.

[0155] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for suppressing direct current bus voltage fluctuation of a renewable energy hydrogen production system, characterized by, The method comprises: Real-time acquisition of key operating parameters of the system, including DC bus voltage , current, renewable energy power, current load power , battery state of charge SOC, hydrogen storage pressure , ambient temperature and light intensity T; Estimating bus voltage disturbance power by disturbance observer and decomposed into a high frequency disturbance component and a low frequency disturbance component ; Based on the load power data in the historical time, the load power at the next time is predicted by a hybrid load prediction algorithm , and the prediction deviation is calculated ; According to the system key operating parameters and the predicted deviation , dynamically setting the upper and lower limits of the bus voltage control interval ; Construct a multi-objective optimization function to take bus voltage disturbance power , prediction deviation , and upper and lower limits of bus voltage control interval as input to solve the power regulation instruction value of the battery energy storage system, fuel cell, and electrolytic tank , , According to the power regulation instruction value , , , generate control instructions and perform power output regulation; The expression of the multi-objective optimization function is: ； In the formula, is a multi-objective optimization function value, indicating the total control cost of the control system at the current time; is the target charging and discharging power of the battery energy storage system at the current time, is the target output power of the fuel cell at the current time, is the power that the electrolyzer should absorb at the current time; is the DC bus voltage at the current time, is the target reference voltage of the DC bus; , , , are the battery energy storage system power response weight coefficient, the hydrogen energy system response weight coefficient, the voltage stability weight coefficient, and the battery energy storage system power change rate penalty weight, respectively. The upper and lower limits of the dynamic setting bus voltage control range The method comprises: Initialization setting: set the target reference voltage of the DC bus as , and set the basic voltage tolerance as , so as to determine the initial control interval as: , wherein , are the initial upper limit value and the lower limit value of the bus voltage, respectively; Real-time parameter acquisition: in each control cycle, real-time acquisition of battery state of charge SOC, hydrogen storage pressure and prediction deviation ; Interval adjustment rule setting: based on the following three types of working condition judgment results, adjust the upper and lower limits of the bus voltage control interval: 1) if the battery state of charge SOC is less than the lower threshold of the battery state of charge then set where is the lower voltage shift of the SOC protection state; 2) if the hydrogen storage pressure is greater than the maximum allowable pressure of the hydrogen tank then set where is the voltage upper limit reduction amount when the hydrogen pressure is too high; 3) If the prediction deviation is greater than the load power disturbance threshold and lasts for more than three sampling periods, then the bus voltage control band is enlarged to: ； ； wherein is an extended voltage deviation interval; Hysteresis interval judgment: if the interval change of the current period meets: ； ； Then maintain the upper and lower limits of the bus voltage control interval of the last period unchanged; In the formula, , are the upper and lower limits of the bus voltage control range of the previous cycle, respectively, is the minimum adjustment threshold for the lower voltage limit.

2. The method of claim 1, wherein the DC bus voltage fluctuation is suppressed by, The disturbance observer adopts a dynamic gain model as follows for the bus voltage disturbance power Estimation is made: ； In the formula, is the bus equivalent capacitance; is the current time DC bus voltage rate of change, t represents time; is the observer gain coefficient; is the saturation function; is the target reference voltage of the DC bus; wherein the observer gain coefficient is expressed as: ； In the formula, is a base gain constant; is a voltage rate of change gain coefficient; is a voltage deviation gain coefficient; is a current DC bus voltage at the current time.

3. The method of claim 2, wherein the DC bus voltage fluctuation is suppressed by, The bus voltage disturbance power is analyzed using wavelet transform or Fourier transform. Perform frequency decomposition, placing frequency components above the cutoff frequency. The disturbance is defined as the high-frequency disturbance component. The response is provided by the battery energy storage system; perturbations with frequency components below a cutoff frequency are defined as low frequency perturbation components in response to the electrolytic cell in combination with the fuel cell; the cutoff frequency is set adaptively according to the power fluctuation period.

4. The method of claim 1, wherein, The hybrid load prediction algorithm is specifically an ARIMA-LSTM dynamic fusion model, and the implementation method is as follows: Inputting the historical load power data of N time points that have been collected And performing data preprocessing operations, including normalization, filtering out spike outliers and smoothing, outputting the cleaned sequence data for modeling; Constructing a perturbation amplitude factor and a dominant period length , which is given by ； ； In the formula, i is a sliding index, representing the i th moment; , are the load power at the i th moment and the j th moment, respectively; is the maximum non-direct current frequency component obtained by applying fast Fourier transform to the historical load power data.​​ Set periodic identification range set with a tolerance of ± 15%; If the disturbance amplitude factor is less than a preset load disturbance amplitude threshold and the dominant cycle length satisfies: the ARIMA model is selected as the main model, and the LSTM model is selected as the secondary prediction path, otherwise the LSTM model is selected as the main model, and the ARIMA model is selected as the secondary prediction path. Calculate the historical residual standard deviation and confidence weight of the ARIMA model and the LSTM model respectively, and finally generate the next time load power prediction value The formula is: ， ； ； In the formula, , , are the historical residual standard deviation, confidence weight, and prediction result of the ARIMA model, respectively; , , are the historical residual standard deviation, confidence weight, and prediction result of the LSTM model, respectively; Prediction bias The formula for the calculation is: .

5. The method of claim 1, wherein, The generation control instruction and the implementation of power output adjustment specifically include: Smooth sliding mode control of the battery energy storage system: The hyperbolic tangent function is used to replace the traditional sign function to generate charge and discharge instructions, and to suppress control instruction chattering; According to the battery state of charge SOC, the charging and discharging power is dynamically limited, when the SOC is higher than 90%, only discharging is allowed, and when the SOC is lower than 20%, only charging is allowed; Hydrogen flow closed-loop control of the fuel cell: By monitoring the fuel cell output current and hydrogen flow in real time, the theoretical hydrogen demand is calculated, and a low-frequency disturbance component is superimposed A feedforward compensation flow instruction is generated; A PID controller is used for closed-loop regulation of hydrogen flow rate to ensure that the hydrogen utilization rate is not less than 95%, avoiding fuel starvation or waste; According to the stack voltage and conversion efficiency, the output power instruction is dynamically adjusted, and the bus voltage demand is matched through the boost DC-DC converter; MPPT-Feedforward composite control of the electrolytic cell: Based on the maximum power point tracking (MPPT) algorithm, the input voltage of the electrolytic cell is periodically disturbed, combined with a low-frequency disturbance component Dynamic correction of hydrogen production power According to the hydrogen storage tank pressure, the hydrogen production power is limited in real time, when the pressure is out of limit, the system is forced to stop, and when the pressure is insufficient, the power is increased to the rated power.

6. The method of claim 1, wherein, The method further comprises an embedded PI regulation link, which regulates the bus voltage deviation in real time As input, the regulation power is generated and superimposed to the control command; When any of the following abnormal states is detected: Sensor data interruption or out of limit; Controller communication interruption exceeds the set time length; Energy storage unit without power response for more than two sampling periods; Automatic start fault tolerance logic, control the battery energy storage system or fuel cell to maintain the bus voltage within the limit interval at a preset power until the main control system recovers.

7. The method of claim 6, wherein the DC bus voltage fluctuation suppression method of a renewable energy hydrogen production system is characterized by, The regulating power satisfies: ； wherein is the current regulated power for PI feedback; is the proportional gain coefficient; is the target reference voltage for the DC bus; is the integral gain coefficient; is the integral variable.

8. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 7. The computer program is executed by the processor to realize the direct current bus voltage fluctuation suppression method of the renewable energy hydrogen production system according to any one of claims 1-7.

Images