Optimization method for hydrogen production through multi-cell series-parallel connection of alkaline electrolytic cells

Abstract

The invention relates to an alkaline electrolytic cell multi-cell series-parallel hydrogen production optimization method, and the method comprises the steps: predicting photovoltaic power generation data in a certain time through an LSTM-CNN recurrent neural network, and enabling the interval of prediction time points to be 15 minutes; the predicted photovoltaic power generation power data are combined with a particle swarm optimization (PSO) algorithm to optimize the state of the electrolytic cell in the multi-cell series-parallel connection of the alkaline electrolytic cell, and the interval of optimization time points is 1 hour; each alkaline electrolytic cell is set to be in the optimal state at each time point for hydrogen production, the number of times of startup and shutdown of the alkaline electrolytic cell is reduced as much as possible, the hydrogen production efficiency of the alkaline electrolytic cell is improved to the maximum extent, meanwhile, the service life of the alkaline electrolytic cell is prolonged, the operation cost is reduced, and the safety problem of gas cross permeation is effectively avoided; and the stability and the safety of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic hydrogen production, and in particular to an optimization method for hydrogen production by multiple hybrid alkaline electrolyzers. Background Art

[0002] Alkaline electrolyzer hydrogen production (AWE) is a mature water electrolysis technology that is widely used in the field of industrial hydrogen production. By combining water with an alkaline electrolyte such as potassium hydroxide (KOH), direct current is applied in the electrolyzer to cause water molecules to undergo an electrochemical decomposition reaction to produce hydrogen and oxygen. The advantages of this alkaline electrolyzer are low cost, long life and relatively simple structural design, which makes it highly economical and feasible in hydrogen energy applications.

[0003] In recent years, with the increase in global demand for clean energy, hydrogen energy has received widespread attention as an important renewable energy source. Alkaline electrolyzer hydrogen production has become an important means to promote the large-scale application of hydrogen energy because it can effectively utilize the surplus electricity of renewable energy sources such as wind power and photovoltaics. However, alkaline electrolyzers have the disadvantages of slow aerodynamic response and poor dynamic adaptability, especially when combined with unstable renewable energy sources, facing challenges such as energy efficiency optimization and safe operation.

[0004] In response to these problems, research in recent years has mainly focused on improving the startup efficiency of alkaline electrolyzers, reducing energy consumption, and improving safety and stability. Currently, there are methods for predicting alkaline electrolyzer failures by using algorithms, such as CN113724804A, which uses the Surrogate model to predict the failure rate of the electrolyzer, and uses Latin hypercube sampling and covariance matrix to achieve efficient sampling and error control of sample data. However, a large amount of computing resources are required to establish and optimize the Surrogate model, especially when modeling high-dimensional sample spaces between different units, and the computational complexity is high. Therefore, there are methods for optimizing hydrogen production methods in unstable power supply scenarios such as wind power and photovoltaics, such as CN117421878A, which constructs a current-voltage empirical relationship model based on electrode kinetics, simplifies the solution of logarithmic terms through piecewise linear fitting, reduces the complexity of the solution and improves the calculation speed, but this method Although the segmented fitting method can reduce the computational complexity, it cannot fully capture complex nonlinear behaviors in some fine optimization scenarios, affecting the accuracy of the final model; there are also studies that perform multi-objective optimization for multiple optimization targets such as start-up time, energy consumption and hydrogen production. For example, CN117724338A transforms the multi-objective optimization model into a free terminal problem. The optimization method is more adaptable in dealing with complex scenarios, especially for the different requirements of high-purity and low-purity hydrogen production. However, this method does not optimize safety; there are also mathematical models for fluctuating power inputs such as wind power and photovoltaics to simulate cold start behaviors under different conditions. For example, CN118094866A simulates the cold start duration and operating conditions, selects the optimal parameters for optimizing the cold start strategy, reduces the start-up time and energy consumption, and improves the overall start-up efficiency, but its real-time regulation has a lag and does not have a predictive function. Summary of the invention

[0005] In view of this, the present invention provides an optimization method for hydrogen production by multiple mixed alkaline electrolyzers, which effectively solves the safety and production problems caused by the frequent start and stop times, excessive power or too low power of the alkaline electrolyzers in the prior art, while improving the utilization rate of renewable energy and the efficiency of hydrogen production by alkaline electrolyzers.

[0006] To achieve the above purpose, an optimization method for hydrogen production by multiple alkaline electrolyzers is provided, which comprises the following two parts: A: Use LSTM-CNN recurrent neural network to predict photovoltaic power generation data, with an interval of 15 minutes; B: The predicted photovoltaic power generation data results are combined with the particle swarm algorithm PSO to optimize the alkaline electrolyzer control strategy. The optimization time interval is 1 hour, including the following steps: Define the state of the alkaline electrolyzer, using three state constants 0, 1, and 2 to represent shutdown, cold start, and startup states; Set the objective function score, the expression is: in, represents the total revenue from hydrogen production, It represents the penalty cost of the number of power on and off times on revenue. It indicates the operation and maintenance costs of the alkaline electrolyzer during normal operation and standby. represents the penalty factor; Set constraints; The discrete particle swarm algorithm (PSO) is used to optimize the state of the alkaline electrolyzer stack. The optimization is performed every hour to obtain the state of the alkaline electrolyzer and evenly distribute the photovoltaic power generation to each operating alkaline electrolyzer. The alkaline electrolyzer is set to the state corresponding to each point in time for hydrogen production.

[0007] Preferably, a CNN neural network is used to extract local features of the input data.

[0008] Preferably, the constraint conditions include energy conservation constraint, operating power constraint, power overload constraint, power-on time constraint, state constraint, cold start process constraint and operating time difference constraint.

[0009] Preferably, the energy conservation constraint follows the following formula: Photovoltaic power output = power absorbed by the alkaline electrolyzer during operation + power consumed to maintain the cold start state + power consumed by turning the alkaline electrolyzer on and off + wasted power; The operating power limit is: when operating, the operating power should be between the upper and lower limits of the operating power; when in standby mode, the operating power is 5% of the rated power; when stopped, the operating power is 0; The power overload limit is: the alkaline electrolytic cell that is in an overload state and reaches the maximum overload time is forced to return to a cold start state or a shutdown state, and the overload state is that the operating power of the alkaline electrolytic cell is between the rated power and the maximum overload power; The startup time limit is: the startup time of the alkaline electrolytic cell is 1 hour; The state restriction is that at any time point, the alkaline electrolyzer can only be in one of the three states of shutdown, cold start and startup; The cold start process restriction is that the alkaline electrolytic cell must go through a cold start state from shutdown to startup, and the cold start state lasts for 1 hour; The operating time difference limit is that the operating time difference of each alkaline electrolytic cell in the predicted total time cannot exceed 20% of the predicted total time.

[0010] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention predicts the photovoltaic power generation at each time point through the LSTM-CNN neural network, fully considering the relationship between each alkaline electrolyzer, thereby achieving a better optimization effect, significantly increasing the hydrogen production efficiency and the utilization rate of renewable energy; 2. The present invention takes into account the cold start state of the alkaline electrolytic cell, reduces the number of starts and stops, prolongs the service life of the alkaline electrolytic cell, and reduces the operating cost; 3. The present invention improves system stability and effectively avoids safety issues such as gas cross-permeation by controlling different alkaline electrolytic cells to operate at optimal power; 4. The method provided by the present invention has strong robustness and low computational complexity and can be used in different scenarios such as wind power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Fig. 1 is a flow chart of the present invention; Figure 2 It is the photovoltaic power generation power curve diagram of the present invention; Figure 3 A comparison diagram of the photovoltaic power generation curve predicted by the present invention and the actual photovoltaic power generation curve; Figure 4 It is the particle swarm algorithm PSO iterative objective function of the present invention Change graph; Figure 5 A state diagram of the alkaline electrolytic cell stack at each time point of the present invention; Figure 6 The state diagram of the alkaline electrolytic cell stack at each time point obtained by the simple start-stop algorithm of the present invention. DETAILED DESCRIPTION

[0012] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0013] Alkaline electrolyzers can operate safely and stably only when they are running between 25% and 120% of the rated power. If the rated power is lower than 25%, problems such as gas cross-permeation and reduced gas separation efficiency will occur. In addition, renewable energy sources such as wind power and PV have certain volatility, which will lead to a low utilization rate of renewable energy for a single electrolyzer, resulting in energy waste and potential safety hazards. Therefore, multiple alkaline electrolyzers with lower rated power are mixed together. By controlling the states of different alkaline electrolyzers, the lower limit of energy absorption is reduced while the upper limit of energy absorption remains unchanged. In order to reduce the number of times the alkaline electrolyzer is turned on and off, increase the service life of the alkaline electrolyzer and reduce the operating cost, avoid safety problems and production problems caused by excessive power or too low power, and improve the utilization rate of renewable energy and the efficiency of hydrogen production by the alkaline electrolyzer, the present invention provides the following technical solutions: An optimization method for producing hydrogen by connecting multiple alkaline electrolyzers, comprising the following two parts: A: Use LSTM-CNN recurrent neural network to predict photovoltaic power generation data within a certain period of time, with the prediction time interval being 15 minutes; Since photovoltaic power generation is not as unpredictable as wind power generation, its output power curve is as follows: Figure 2 As shown in the figure, by training the long short-term memory network LSTM to process the time dependency of time series data, the photovoltaic power output value for a long time can be predicted by only inputting the input value of the starting point; Obtain local photovoltaic power generation historical data and divide the data into training set and test set; The training set data is used to train the LSTM-CNN recurrent neural network. The input data includes time point, total solar irradiance, direct solar irradiance, global horizontal irradiance, temperature, pressure and relative humidity. The time interval between data is 15 minutes. Then use CNN neural network to further extract local features of input data to improve the accuracy of prediction; The output data is the predicted photovoltaic power generation; The test set data is used to predict the photovoltaic power generation data within a certain period of time. It only needs to input data with the same dimension as the training set to the starting point of the neural network, and then the photovoltaic power generation power within the next certain period of time can be predicted at intervals of 15 minutes. The photovoltaic power generation power is different from the solar power. After calculation, the photovoltaic power generation is directly supplied to the alkaline electrolyzer for use; In this embodiment, 2000 data are used as the training set of the neural network. After the neural network training is completed, the total solar irradiance, direct solar irradiance, global horizontal irradiance, temperature, pressure and relative humidity at 0 o'clock are input to predict the photovoltaic power generation power in the next 24 hours. The prediction is made every 15 minutes. The predicted results are as follows: Figure 2 As shown in the figure, the predicted photovoltaic power generation curve has a high degree of fit with the actual photovoltaic power generation curve, indicating that the LSTM-CNN neural network prediction has a good prediction ability for photovoltaic power generation; B: The predicted photovoltaic power generation data results are combined with the particle swarm algorithm PSO to optimize the control strategy of the alkaline electrolyzer, and the interval of the optimization time point is 1 hour; The reasons for choosing 1 hour as the optimization time point interval are: (1) The time interval of one optimization point includes the time interval of four prediction points. Since the photovoltaic power generation has obvious regularity and shows a continuous trend, the values ​​of the four prediction points within the optimization point time interval will not undergo a large mutation, resulting in the state of the optimized alkaline electrolyzer stack being able to handle the photovoltaic power generation of these four prediction points with higher benefits; (2) Since the calculation mode of the heuristic algorithm is multiple populations and multiple iterations, the increase in the number of calculations in a single calculation will lead to an exponential increase in the overall calculation amount and the increase in the calculation amount will also result in smaller benefits; The particle swarm optimization PSO optimization of alkaline electrolyzer includes the following steps: Define the state of the alkaline electrolyzer, and use three state constants of 0, 1, and 2 to represent the shutdown, cold start, and startup states. When the alkaline electrolyzer is shut down, it consumes no power and produces no hydrogen. When it is cold started, it consumes 5% of the rated power and produces no hydrogen (actually a small amount of hydrogen is produced, but it is difficult to quantify so it is simplified). When it is started, according to the photovoltaic power generation power at the current time point and the number of normal operation of the current alkaline electrolyzer, the photovoltaic power generation power is evenly distributed to each alkaline electrolyzer to produce corresponding hydrogen; The objective function score is set to evaluate the quality of the alkaline electrolyzer stack from multiple aspects. The expression is: in, represents the total revenue from hydrogen production, It represents the penalty cost of the number of power on and off times on revenue. It indicates the operation and maintenance costs of the alkaline electrolyzer during normal operation and standby. represents the penalty factor; From the perspective of engineering application, the direct way to evaluate the working condition of alkaline electrolyzer is to measure the hydrogen production. Multiplying it by the price coefficient is the total revenue of hydrogen production. Therefore, the cost of producing hydrogen is The expression is: in, represents the market price of hydrogen, Indicates the power consumed to produce each unit of hydrogen, represents the power consumed by the electrolyzer to produce hydrogen, Indicates a time interval of 15 minutes. The value of is a positive number; Frequent start and stop can easily reduce the life of the alkaline electrolyzer, affect its performance and consume extra energy. Therefore, the number of starts and stops of the alkaline electrolyzer should be reduced as much as possible. The expression is: in, Indicates the power-on penalty coefficient, Indicates the number of times the machine has been turned on. Represents the shutdown penalty coefficient, Indicates the number of shutdowns. The reason for multiplying it by 4 is that although the prediction point is once every 15 minutes, the optimization point is once every 1 hour. The value of is a positive number; If the alkaline electrolyzer is always turned on, when the photovoltaic power generation is insufficient, the operation of the alkaline electrolyzer will not only fail to produce hydrogen, but also reduce the service life of the alkaline electrolyzer and consume extra energy. Therefore, it is necessary to avoid turning on the alkaline electrolyzer to absorb power when the photovoltaic power generation is insufficient. The cost of operating and maintaining the alkaline electrolyzer is high. The expression is: in, represents the operation penalty coefficient (without this penalty coefficient, all alkaline electrolyzers will be kept in the open state and cannot be properly regulated). Indicates the number of alkaline electrolytic cells that have been operated, represents the standby penalty coefficient, Indicates the number of alkaline electrolytic cells that have been in standby mode (if this parameter is not present, the alkaline electrolytic cells will remain in standby mode for easy startup). The value of is a positive number; Costs of operating and maintaining alkaline electrolyzers during cold start The expression is: Among them, when the alkaline electrolyzer is operating normally When the value is 0, the operating power of the alkaline electrolyzer is less than 20% of the rated power or greater than the overload power, severe penalties will be imposed. The value of is a positive number; The cold start time of the alkaline electrolyzer is long, and no hydrogen is produced, resulting in low returns. Try to reduce the cold start time and set constraints to ensure the normal operation of the alkaline electrolyzer; The constraints include energy conservation constraints, operating power constraints, power overload constraints, startup time constraints, state constraints, cold start process constraints, and operating time difference constraints; The energy conservation constraint follows the formula: Photovoltaic power output = power absorbed by the alkaline electrolyzer during operation + power consumed to maintain the cold start state + power consumed by turning the alkaline electrolyzer on and off + wasted power; The operating power limit is: when operating, the operating power should be between the upper and lower limits of the operating power; when in standby mode, the operating power is 5% of the rated power; when stopped, the operating power is 0; The power overload limit is: the alkaline electrolyzer that is in the overload state and reaches the maximum overload time is forced to return to the cold start state or shutdown state. The overload state is that the operating power of the alkaline electrolyzer is between the rated power and the maximum overload power; The startup time limit is: the startup time of the alkaline electrolyzer is 1 hour; The state restriction is: at any point in time, the alkaline electrolyzer can only be in one of the three states: shutdown, cold start, and startup; The cold start process restriction is: the alkaline electrolyzer must go through the cold start state from shutdown to startup, and the cold start state lasts for 1 hour; The operating time difference limit is: the operating time difference of each alkaline electrolytic cell in the predicted total time cannot exceed 20% of the predicted total time, so as to avoid the production chain being broken due to the fact that some alkaline electrolytic cells have a longer working time than other alkaline electrolytic cells; in this embodiment, the predicted total time is 24 hours; The discrete particle swarm algorithm PSO is used to optimize the state of the alkaline electrolyzer stack, and the optimization is performed every 1 hour. The objective function As the number of iterations changes Figure 4 As shown, the iteration stops until the preset number of iterations is reached, and the state of the alkaline electrolytic cell at each time point is obtained as shown in Figure 5 As shown, Figure 6 The state of the alkaline electrolytic cell at each time point obtained by the simple start-stop algorithm is compared. It can be seen that the optimization method adopted in this embodiment has a smaller difference in the operation time of the alkaline electrolytic cell, less wasted energy, fewer start-stop times, and higher energy utilization; By setting the alkaline electrolyzer to produce hydrogen in the state corresponding to each time point, the efficiency of hydrogen production in the alkaline electrolyzer can be maximized.

[0014] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An optimization method for producing hydrogen by connecting multiple alkaline electrolyzers, characterized in that: The method includes: A: Use LSTM-CNN recurrent neural network to predict photovoltaic power generation data, with an interval of 15 minutes; B: The predicted photovoltaic power generation data results are combined with the particle swarm algorithm PSO to optimize the alkaline electrolyzer state. The optimization time interval is 1 hour, including the following steps: Define the state of the alkaline electrolyzer, using three state constants 0, 1, and 2 to represent shutdown, cold start, and startup states; Set the objective function score, the expression is: in, represents the total revenue from hydrogen production, It represents the penalty cost of the number of power on and off times on revenue. It indicates the operation and maintenance costs of the alkaline electrolyzer during normal operation and standby. represents the penalty factor; Set constraints; The discrete particle swarm algorithm (PSO) is used to optimize the state of the alkaline electrolyzer stack. The optimization is performed every hour until the preset number of iterations is reached to obtain the optimal state of the alkaline electrolyzer at each time point, and the photovoltaic power generation power is evenly distributed to each operating alkaline electrolyzer. The alkaline electrolyzer is set to the state corresponding to each point in time for hydrogen production.

2. The method for optimizing hydrogen production by multiple alkaline electrolyzers according to claim 1, characterized in that: Step A also includes obtaining local photovoltaic power generation historical data and training the LSTM-CNN recurrent neural network. The input data includes time point, total solar irradiance, direct solar irradiance, global horizontal irradiance, temperature, pressure and relative humidity. The output data is the predicted photovoltaic power generation power.

3. The method for optimizing hydrogen production by multiple alkaline electrolyzers according to claim 2, characterized in that: Use CNN neural network to extract local features of input data.

4. The method for optimizing hydrogen production by multiple alkaline electrolyzers according to claim 1, characterized in that: Total revenue from hydrogen production The expression is: in, represents the market price of hydrogen, Indicates the power consumed to produce each unit of hydrogen, Indicates the power consumed by the alkaline electrolyzer to produce hydrogen, Indicates a time interval of 15 minutes. The value of is a positive number; Penalty costs for power on / off times The expression is: in, Indicates the power-on penalty coefficient, Indicates the number of times the machine has been turned on. Represents the shutdown penalty coefficient, Indicates the number of shutdowns. The value of is a positive number; Operation and maintenance costs of alkaline electrolyzers during normal operation and standby The expression is: in, represents the running penalty coefficient, Indicates the number of alkaline electrolytic cells that have been operated, represents the standby penalty coefficient, Indicates the number of alkaline electrolytic cells that have been in standby mode. The value of is a positive number; Penalty Factor The expression is: Among them, when the alkaline electrolyzer is operating normally The value is 0, when the operating power of the alkaline electrolyzer is less than 20% of the rated power or greater than the overload power The value of is a positive number.

5. The method for optimizing hydrogen production by multiple alkaline electrolyzers according to claim 1, characterized in that: The constraint conditions include energy conservation constraints, operating power constraints, power overload constraints, power-on time constraints, state constraints, cold start process constraints, and operating time difference constraints.

6. The method for optimizing hydrogen production by multiple alkaline electrolyzers according to claim 5, characterized in that: The energy conservation constraint follows the formula: Photovoltaic power output = power absorbed by the alkaline electrolyzer during operation + power consumed to maintain the cold start state + power consumed by turning the alkaline electrolyzer on and off + wasted power; The operating power limit is: when operating, the operating power should be between the upper and lower limits of the operating power; when in standby mode, the operating power is 5% of the rated power; when stopped, the operating power is 0; The power overload limit is: the alkaline electrolytic cell that is in an overload state and reaches the maximum overload time is forced to return to a cold start state or a shutdown state, and the overload state is that the operating power of the alkaline electrolytic cell is between the rated power and the maximum overload power; The startup time limit is: the startup time of the alkaline electrolytic cell is 1 hour, at which time the alkaline electrolytic cell is in a cold start state; The state restriction is that at any time point, the alkaline electrolyzer can only be in one of the three states of shutdown, cold start and startup; The cold start process restriction is that the alkaline electrolytic cell must go through a cold start state from shutdown to startup, and the cold start state lasts for 1 hour; The operating time difference limit is that the operating time difference of each alkaline electrolytic cell in the predicted total time cannot exceed 20% of the predicted total time.

Images