Optimized scheduling method and system for power allocation of electrolytic cell group under wind power fluctuation conditions

Abstract

This invention provides an optimized scheduling method and system for power allocation of an electrolyzer group under wind power fluctuation conditions, relating to the field of wind power grid-connected power generation control technology. The method includes collecting wind farm power data and electrolyzer operating parameters, adaptively segmenting the power data, constructing a fluctuation characteristic space to predict fluctuation risks, establishing an electrolyzer regulation capability assessment system and dynamically grouping them, constructing a power regulation matrix to calculate the optimal power allocation ratio, and realizing coordinated regulation of wind power fluctuations by the electrolyzer group, thereby improving system operational stability and energy utilization efficiency.

Description

Technical Field

[0001] This invention relates to wind power grid-connected power generation control technology, and more particularly to an optimized scheduling method and system for power allocation of electrolytic cell groups under wind power fluctuation conditions. Background Technology

[0002] As the scale of new energy grid connection continues to expand, the volatility of renewable energy sources such as wind power poses a severe challenge to the safe and stable operation of the power grid. Large-scale hydrogen production, as an important means of absorbing new energy sources, requires addressing the key issue of how electrolyzer clusters can adapt to fluctuations in wind power output.

[0003] Existing electrolyzer power allocation methods mainly employ fixed grouping strategies, which cannot be optimized and adjusted in real time according to wind power fluctuation characteristics and the dynamic response capability of the electrolyzers. Furthermore, traditional allocation algorithms do not fully consider the coupling relationships within the electrolyzer group, making it difficult to accurately track wind power fluctuations and reducing system operating efficiency.

[0004] As a crucial load for absorbing renewable energy, the power allocation method of electrolyzer clusters directly impacts wind power absorption. Existing methods cannot accurately predict wind power fluctuation trends, nor have they established a mapping relationship between the dynamic response characteristics of electrolyzers and control requirements, making it difficult to achieve coordinated control of wind power fluctuations by the electrolyzer cluster. Therefore, a power optimization allocation method for electrolyzer clusters based on dynamic grouping is urgently needed. Summary of the Invention

[0005] This invention provides an optimized scheduling method and system for power allocation of an electrolytic cell group under wind power fluctuation conditions, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides an optimized scheduling method for power allocation of an electrolyzer group under wind power fluctuation conditions, comprising:

[0007] Collect wind farm output power data and electrolytic cell group operating parameters;

[0008] The output power data is adaptively segmented, the power fluctuation feature vector in each time period is calculated, a three-dimensional fluctuation feature space is constructed based on the power fluctuation feature vector, the fluctuation trend trajectory is extracted in the three-dimensional fluctuation feature space, the future fluctuation risk is predicted and time-segmented control instructions are generated.

[0009] A multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of the electrolyzer. A mapping relationship between dynamic response characteristics and time-segmented regulation requirements is established. The comprehensive regulation capability index of each electrolyzer is calculated using the mapping relationship. Based on the comprehensive regulation capability index, the electrolyzers are classified, screened, and prioritized to generate a dynamic grouping scheme for the electrolyzers.

[0010] A power regulation matrix is ​​constructed based on the dynamic grouping scheme of electrolyzers. The optimal power allocation ratio of each electrolyzer in the power regulation matrix is ​​calculated by a recursive iterative method, and a control command containing the real-time power setpoint of each electrolyzer is generated.

[0011] The real-time power setpoint is transmitted to the control unit of each electrolyzer to perform power adjustment, and the real-time response data of the electrolyzer is collected. Based on the real-time response data, the control instructions are optimized to achieve coordinated control of wind power fluctuations by the electrolyzer group.

[0012] In one alternative embodiment,

[0013] The output power data is adaptively segmented, and the power fluctuation feature vector for each time period is calculated. Based on the power fluctuation feature vector, a three-dimensional fluctuation feature space is constructed, including:

[0014] The power change rate sequence is obtained by calculating the power difference between adjacent sampling points in the wind power data. The power change rate sequence is then processed by a sliding window, and the mean and standard deviation of the power change rate within each window are calculated.

[0015] A first-order difference sequence is constructed based on the mean of the power change rate, and a fluctuation amplitude sequence is constructed based on the standard deviation of the power change rate. The first-order difference sequence and the fluctuation amplitude sequence are weighted and combined to obtain a segmentation criterion. An adaptive threshold is calculated on the segmentation criterion. A benchmark threshold is determined based on the statistical distribution characteristics of the power change rate sequence. The threshold coefficient is dynamically adjusted in combination with the cumulative distribution function of the fluctuation amplitude sequence to obtain the segmented threshold.

[0016] Sampling points exceeding the segmentation threshold are marked as segmentation points. The wind power data sequence is divided into multiple time periods based on the segmentation points. Power fluctuation features are extracted from the wind power data sequence in each time period, including average power value, power change rate, and power fluctuation frequency. The power fluctuation features are combined to construct a power fluctuation feature vector. The feature contribution of the power fluctuation feature vector is calculated using the principal component analysis method. The three feature directions with the largest cumulative contribution are selected to construct a three-dimensional fluctuation feature space.

[0017] In one alternative embodiment,

[0018] Extracting volatility trend trajectories from a three-dimensional volatility feature space, predicting future volatility risks, and generating time-segmented control instructions include:

[0019] The power fluctuation feature vectors are connected sequentially in the three-dimensional fluctuation feature space to form a fluctuation trend trajectory. The distance and direction angle between adjacent trajectory points in the fluctuation trend trajectory are calculated to determine the nearest neighbor set of each trajectory point. Based on the nearest neighbor set, the local evolution direction of each trajectory point is calculated. The ratio of the displacement change between adjacent trajectory points to the time interval is used as the evolution rate.

[0020] The fluctuation trend trajectory is divided into partitions according to the evolution direction. The product of the proportion of trajectory points in each partition and the average evolution rate is calculated to determine the development probability of each evolution direction. The direction with the highest development probability is selected as the predicted trend. The power fluctuation risk value is obtained by combining the average evolution rate of that direction.

[0021] Based on the power fluctuation risk value, the electrolytic cell control response requirements are set. The power regulation rate and steady-state error of each electrolytic cell are calculated using the electrolytic cell operating parameters. The control response requirements are compared with the power regulation rate and steady-state error of the electrolytic cells to calculate the response matching degree. Based on the response matching degree, the electrolytic cells are grouped and sorted. The electrolytic cell combination with the highest matching degree is selected as the target electrolytic cell combination. For the target electrolytic cell combination, the maximum regulation power and response time constraint are calculated based on its operating parameters. Combined with the power fluctuation risk value, the target regulation power and execution sequence for each time period are determined, and time-segmented control instructions are generated.

[0022] In one alternative embodiment,

[0023] A multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of electrolyzers. A mapping relationship between dynamic response characteristics and time-segmented regulation requirements is established. This mapping relationship is used to calculate the comprehensive regulation capability index of each electrolyzer. Based on the comprehensive regulation capability index, electrolyzers are graded, screened, and prioritized to generate a dynamic grouping scheme for electrolyzers, including:

[0024] Based on the obtained electrolyzer operation data, a multi-dimensional evaluation index system including response time, adjustment accuracy and efficiency loss is established, and the response characteristic curve, accuracy index value and loss index value of the electrolyzer are calculated based on the electrolyzer operation data.

[0025] Acquire system regulation and control demand data, perform matching analysis between the regulation and control demand data and the response characteristic curve, accuracy index value and loss index value of the electrolyzer, obtain the adaptability evaluation value of the electrolyzer to the regulation and control demand, perform dynamic evaluation of the electrolyzer based on the adaptability evaluation value, and calculate the comprehensive regulation capability index value of the electrolyzer according to the importance weight of regulation in each regulation period.

[0026] Acquire the cumulative number of adjustments and adjustment intensity data of the electrolytic cell, calculate the fatigue index value; and classify the electrolytic cell into a rapid response level, a conventional adjustment level, and a standby level based on the comprehensive adjustment capability index value.

[0027] Within each level of electrolytic cell, the comprehensive adjustment capability index value and the fatigue index value are weighted to obtain a priority coefficient, and the electrolytic cells are sorted according to the priority coefficient; when the fatigue index value exceeds a preset fatigue threshold, the electrolytic cell is downgraded and the priority ranking is updated, generating a dynamic grouping scheme for electrolytic cells containing grading information and priority information.

[0028] In one alternative embodiment,

[0029] The electrolyzer is dynamically evaluated based on its adaptability assessment value. The comprehensive regulation capability index of the electrolyzer is calculated according to the weighted importance of regulation in each regulation period, including:

[0030] Collect wind power fluctuation data, obtain power fluctuation inflection points through extreme value detection, divide the evaluation period into multiple adjustment periods based on the power fluctuation inflection points, and calculate the control difficulty coefficient based on the power change trend in each adjustment period.

[0031] The power fluctuation amplitude, fluctuation frequency and system stability margin of each adjustment period are obtained, and the adjustment period weight evaluation matrix is ​​constructed. The control importance weight of each adjustment period is calculated based on the weight evaluation matrix.

[0032] The operating data of the electrolyzer during each adjustment period is collected. Based on the operating data and the adjustment difficulty coefficient, the adaptability evaluation value of the electrolyzer during each adjustment period is calculated. The adaptability evaluation value is normalized and combined with the weight of the importance of adjustment to obtain the comprehensive adjustment capability index value of the electrolyzer.

[0033] In one alternative embodiment,

[0034] A power regulation matrix is ​​constructed based on a dynamic grouping scheme for electrolyzers. A recursive iterative method is used to calculate the optimal power allocation ratio for each electrolyzer in the power regulation matrix, generating control instructions containing real-time power setpoints for each electrolyzer.

[0035] The operating power, power fluctuation, and response time of the electrolytic cell group are collected, and the dynamic response characteristics of the electrolytic cells are calculated based on the operating power, power fluctuation, and response time.

[0036] A power regulation matrix is ​​constructed based on the dynamic grouping scheme of the electrolyzer. The inter-group power coupling coefficient is calculated based on the dynamic response characteristics of the electrolyzer. The inter-group power coupling coefficient is written into the corresponding position of the power regulation matrix to obtain the initial power regulation matrix.

[0037] The power adjustment deviation matrix is ​​obtained by performing a difference operation between the element values ​​of the initial power adjustment matrix and the rated power value of the electrolyzer. The inter-group adaptive weights are calculated based on the power adjustment deviation matrix, and the inter-group adaptive weights are multiplied by the initial power adjustment matrix to obtain the optimized power adjustment matrix.

[0038] A multidimensional optimization objective function is constructed based on the optimized power adjustment matrix and the power adjustment deviation matrix. The power adjustment deviation, inter-group power coupling coefficient and dynamic response characteristics are set as optimization variables. The weight coefficients of the optimization variables are calculated according to the system operation requirements.

[0039] A recursive iterative method is used to calculate the power allocation ratio of each electrolytic cell in the power adjustment matrix. The adaptive step size is calculated by multiplying the weight coefficients of the optimization variables with the gradient of the multidimensional optimization objective function. The dynamic compensation factor is calculated based on historical iteration data to obtain the optimal power allocation ratio. The optimal power allocation ratio is multiplied by the rated power of the electrolytic cell to obtain the real-time power setpoint of each electrolytic cell, and a control command is generated.

[0040] A second aspect of this invention provides an optimized scheduling system for power allocation of an electrolyzer group under wind power fluctuation conditions, comprising:

[0041] The first unit is used to collect wind farm output power data and electrolytic cell group operating parameters;

[0042] The second unit is used to perform adaptive segmentation processing on the output power data, calculate the power fluctuation feature vector in each time period, construct a three-dimensional fluctuation feature space based on the power fluctuation feature vector, extract the fluctuation trend trajectory in the three-dimensional fluctuation feature space, predict future fluctuation risks, and generate time-segmented control instructions.

[0043] The third unit is used to construct a multi-dimensional regulation capability evaluation system based on the operating parameters of the electrolyzer, establish a mapping relationship between dynamic response characteristics and time-segmented regulation requirements, use the mapping relationship to calculate the comprehensive regulation capability index of each electrolyzer, classify and prioritize the electrolyzers according to the comprehensive regulation capability index, and generate a dynamic grouping scheme for the electrolyzers.

[0044] The fourth unit is used to construct a power regulation matrix based on the dynamic grouping scheme of electrolyzers, and to calculate the optimal power allocation ratio of each electrolyzer in the power regulation matrix using a recursive iterative method, thereby generating a control command containing the real-time power setpoint of each electrolyzer.

[0045] The fifth unit is used to transmit the real-time power setpoint to the control unit of each electrolyzer to perform power regulation, and to collect the real-time response data of the electrolyzers. Based on the real-time response data, the control instructions are optimized to realize the coordinated control of wind power fluctuations by the electrolyzer group.

[0046] A third aspect of the present invention provides an electronic device, comprising:

[0047] processor;

[0048] Memory used to store processor-executable instructions;

[0049] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0050] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0051] In this embodiment, by adaptively segmenting wind power data and constructing a three-dimensional fluctuation feature space, accurate identification and prediction of wind power fluctuation trends are achieved, effectively reducing the blindness of electrolyzer group regulation and improving the system's adaptability to different types of power fluctuations. A multi-dimensional regulation capability evaluation system and dynamic grouping mechanism for electrolyzers are established, enabling reasonable classification and regulation priority ranking of electrolyzers with different response characteristics. This fully leverages the differentiated regulation potential of each electrolyzer, significantly improving the overall coordination and flexibility of the electrolyzer group. The optimal power allocation ratio is calculated using a recursive iterative method, and regulation commands are optimized through real-time response data. This achieves accurate tracking and smooth control of wind power fluctuations by the electrolyzer group, improving hydrogen production efficiency, reducing electrolyzer equipment wear, extending system lifespan, and providing technical support for the stable operation of the wind power-hydrogen energy system. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the optimized scheduling method for power allocation of an electrolytic cell group under wind power fluctuation conditions, according to an embodiment of the present invention.

[0053] Figure 2 This is a thermogram showing the relationship between the adjustment accuracy and efficiency loss of the electrolytic cell in an embodiment of the present invention.

[0054] Figure 3 This is a bar chart comparing the adjustment capabilities of different electrolytic cell types in embodiments of the present invention. Detailed Implementation

[0055] 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 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 are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the optimized scheduling method for power allocation of an electrolytic cell group under wind power fluctuation conditions, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0058] Collect wind farm output power data and electrolytic cell group operating parameters;

[0059] The output power data is adaptively segmented, the power fluctuation feature vector in each time period is calculated, a three-dimensional fluctuation feature space is constructed based on the power fluctuation feature vector, the fluctuation trend trajectory is extracted in the three-dimensional fluctuation feature space, the future fluctuation risk is predicted and time-segmented control instructions are generated.

[0060] A multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of the electrolyzer. A mapping relationship between dynamic response characteristics and time-segmented regulation requirements is established. The comprehensive regulation capability index of each electrolyzer is calculated using the mapping relationship. Based on the comprehensive regulation capability index, the electrolyzers are classified, screened, and prioritized to generate a dynamic grouping scheme for the electrolyzers.

[0061] A power regulation matrix is ​​constructed based on the dynamic grouping scheme of electrolyzers. The optimal power allocation ratio of each electrolyzer in the power regulation matrix is ​​calculated by a recursive iterative method, and a control command containing the real-time power setpoint of each electrolyzer is generated.

[0062] The real-time power setpoint is transmitted to the control unit of each electrolyzer to perform power adjustment, and the real-time response data of the electrolyzer is collected. Based on the real-time response data, the control instructions are optimized to achieve coordinated control of wind power fluctuations by the electrolyzer group.

[0063] In one optional implementation, the output power data is adaptively segmented, and a power fluctuation feature vector is calculated for each time period. A three-dimensional fluctuation feature space is then constructed based on the power fluctuation feature vector, including:

[0064] The power change rate sequence is obtained by calculating the power difference between adjacent sampling points in the wind power data. The power change rate sequence is then processed by a sliding window, and the mean and standard deviation of the power change rate within each window are calculated.

[0065] A first-order difference sequence is constructed based on the mean of the power change rate, and a fluctuation amplitude sequence is constructed based on the standard deviation of the power change rate. The first-order difference sequence and the fluctuation amplitude sequence are weighted and combined to obtain a segmentation criterion. An adaptive threshold is calculated on the segmentation criterion. A benchmark threshold is determined based on the statistical distribution characteristics of the power change rate sequence. The threshold coefficient is dynamically adjusted in combination with the cumulative distribution function of the fluctuation amplitude sequence to obtain the segmented threshold.

[0066] Sampling points exceeding the segmentation threshold are marked as segmentation points. The wind power data sequence is divided into multiple time periods based on the segmentation points. Power fluctuation features are extracted from the wind power data sequence in each time period, including average power value, power change rate, and power fluctuation frequency. The power fluctuation features are combined to construct a power fluctuation feature vector. The feature contribution of the power fluctuation feature vector is calculated using the principal component analysis method. The three feature directions with the largest cumulative contribution are selected to construct a three-dimensional fluctuation feature space.

[0067] For example, the output power data of the wind farm is collected through a SCADA system at a sampling frequency of 1 minute per sampling, acquiring continuous 24-hour wind power data to form a time series of 1440 sampling points. The collected wind power data is adaptively segmented to identify time periods with different fluctuation characteristics, providing a basis for subsequent power allocation of the electrolyzer group.

[0068] When processing wind power data, the power difference between adjacent sampling points is calculated to obtain a power change rate sequence. Taking a wind farm with a rated capacity of 100MW as an example, in the power data of a certain day, the power values ​​of two adjacent sampling points are 45.6MW and 47.2MW respectively, and the corresponding power change rate is 1.6MW / min. Performing this operation on the entire power data sequence yields a power change rate sequence of length 1439.

[0069] A sliding window processing method was applied to the power change rate sequence, with a window size of 30 sampling points and a sliding step size of 1 sampling point. Within each window, the mean and standard deviation of the power change rate were calculated. Taking the aforementioned wind farm as an example, within a certain time window, the calculated mean power change rate was 0.85 MW / min, and the standard deviation was 1.24 MW / min. Through sliding window processing, a sequence of the mean and standard deviation of the power change rate was obtained, both with a length of 1410.

[0070] A first-order difference sequence is constructed based on the mean power change rate sequence. The difference in the mean power change rate between adjacent windows is calculated to obtain the first-order difference sequence. For example, if the mean power change rates of two adjacent windows are 0.85 MW / min and 1.03 MW / min, the corresponding first-order difference value is 0.18 MW / min. This process is performed on the entire mean power change rate sequence to obtain a first-order difference sequence of length 1409.

[0071] A fluctuation amplitude sequence is constructed based on the standard deviation sequence of the power change rate. The fluctuation amplitude sequence directly uses the standard deviation of the power change rate. A weighted combination of the first-order difference sequence and the fluctuation amplitude sequence is used to obtain the segmentation criterion. The weighting coefficients are determined through historical data analysis; the weight of the first-order difference sequence is 0.6, and the weight of the fluctuation amplitude sequence is 0.4. Taking a certain moment as an example, if the first-order difference value is 0.18 MW / min and the fluctuation amplitude value is 1.24 MW / min, then the segmentation criterion value is 0.6 × 0.18 + 0.4 × 1.24 = 0.604 MW / min.

[0072] An adaptive threshold calculation is performed on the segmentation criterion. A baseline threshold is determined based on the statistical distribution characteristics of the power change rate sequence, using twice the standard deviation of the power change rate sequence as the baseline threshold. Taking the aforementioned wind farm as an example, the standard deviation of the power change rate sequence is 1.5 MW / min, so the baseline threshold is 3.0 MW / min. The threshold coefficient is dynamically adjusted based on the cumulative distribution function of the fluctuation amplitude sequence. The fluctuation amplitude of the cumulative distribution function value of the fluctuation amplitude sequence at the 0.9 quantile is 2.1 MW / min, corresponding to a threshold coefficient of 0.7. The final segmented threshold is the baseline threshold multiplied by the threshold coefficient, i.e., 3.0 × 0.7 = 2.1 MW / min.

[0073] Sampling points with values ​​greater than the segmentation threshold were marked as segmentation points. Sampling points with segmentation criteria values ​​greater than 2.1 MW / min were also marked as segmentation points. In this example, a total of 8 segmentation points were identified, dividing the wind power data sequence into 9 time periods with lengths of 126 minutes, 204 minutes, 153 minutes, 118 minutes, 217 minutes, 189 minutes, 152 minutes, 164 minutes, and 117 minutes, respectively.

[0074] Power fluctuation characteristics are extracted from the wind power data sequence for each time period. The average power value is calculated as the arithmetic mean of all power data within the time period. Taking the first time period as an example, the average power value is 36.4 MW. The power change rate is calculated as the mean of the power change rate sequence within the time period. Taking the first time period as an example, the power change rate is 0.42 MW / min. The power fluctuation frequency is calculated as the ratio of the number of times the power change direction changes within the time period to the length of the time period. Taking the first time period as an example, the number of times the power change direction changes is 47, and the time period length is 126 minutes, so the power fluctuation frequency is 47 / 126 = 0.37 times / minute.

[0075] The extracted power fluctuation features are combined to construct a power fluctuation feature vector. For each time period, the power fluctuation feature vector consists of three components: average power value, power change rate, and power fluctuation frequency. Taking the first time period as an example, the power fluctuation feature vector is [36.4, 0.42, 0.37].

[0076] Principal component analysis (PCA) was used to calculate the feature contribution of power fluctuation eigenvectors. PCA was performed on the power fluctuation eigenvectors for nine time periods, yielding eigenvalues ​​of 2.45, 0.38, and 0.17, with corresponding feature contributions of 81.7%, 12.7%, and 5.6%. A three-dimensional fluctuation feature space was constructed by selecting the three feature directions with the largest cumulative contribution, enabling a visual representation of wind power fluctuation characteristics and providing an intuitive basis for power allocation in electrolytic cell clusters.

[0077] In this embodiment, by calculating the power change rate and performing sliding window processing, combined with a weighted combination of the first-order difference sequence and the fluctuation amplitude sequence, the segmentation threshold can be adaptively determined, thereby accurately capturing the key turning points of wind power fluctuations. Compared with traditional segmentation methods using fixed thresholds or fixed time windows, this method is more adaptable to the randomness and volatility characteristics of wind power. The constructed three-dimensional fluctuation feature space intuitively displays the power fluctuation characteristics of different time periods, providing a scientific basis for subsequent power allocation of the electrolyzer group. By extracting features such as average power value, power change rate, and power fluctuation frequency for each time period, the characteristics of wind power fluctuations can be comprehensively characterized, thereby realizing differentiated scheduling strategies for different fluctuation scenarios. This improves the adaptability of the electrolyzer group to wind power fluctuations, reduces the adverse impact of power fluctuations on the operational stability of the electrolyzers, and simultaneously improves wind power absorption efficiency and the overall economic efficiency of the system.

[0078] In one optional implementation, extracting the fluctuation trend trajectory in the three-dimensional fluctuation feature space, predicting future fluctuation risks, and generating time-segmented control instructions includes:

[0079] The power fluctuation feature vectors are connected sequentially in the three-dimensional fluctuation feature space to form a fluctuation trend trajectory. The distance and direction angle between adjacent trajectory points in the fluctuation trend trajectory are calculated to determine the nearest neighbor set of each trajectory point. Based on the nearest neighbor set, the local evolution direction of each trajectory point is calculated. The ratio of the displacement change between adjacent trajectory points to the time interval is used as the evolution rate.

[0080] The fluctuation trend trajectory is divided into partitions according to the evolution direction. The product of the proportion of trajectory points in each partition and the average evolution rate is calculated to determine the development probability of each evolution direction. The direction with the highest development probability is selected as the predicted trend. The power fluctuation risk value is obtained by combining the average evolution rate of that direction.

[0081] Based on the power fluctuation risk value, the electrolytic cell control response requirements are set. The power regulation rate and steady-state error of each electrolytic cell are calculated using the electrolytic cell operating parameters. The control response requirements are compared with the power regulation rate and steady-state error of the electrolytic cells to calculate the response matching degree. Based on the response matching degree, the electrolytic cells are grouped and sorted. The electrolytic cell combination with the highest matching degree is selected as the target electrolytic cell combination. For the target electrolytic cell combination, the maximum regulation power and response time constraint are calculated based on its operating parameters. Combined with the power fluctuation risk value, the target regulation power and execution sequence for each time period are determined, and time-segmented control instructions are generated.

[0082] For example, firstly, based on the power fluctuation feature vectors obtained from the aforementioned adaptive segmentation processing, these vectors are sequentially connected in a three-dimensional fluctuation feature space to form a fluctuation trend trajectory. Taking the power fluctuation feature vectors of nine time periods as an example, these nine vectors are [36.4, 0.42, 0.37], [42.1, 0.56, 0.42], [38.7, 0.48, 0.39], [51.3, 0.71, 0.45], [47.2, 0.63, 0.41], [53.8, 0.78, 0.49], [49.5, 0.68, 0.43], [55.7, 0.82, 0.52], and [52.3, 0.75, 0.47]. Connecting these vectors in three-dimensional space yields the fluctuation trend trajectory.

[0083] Calculate the distance and direction angle between adjacent trajectory points in the fluctuation trend trajectory. The distance between adjacent trajectory points is calculated using Euclidean distance. Taking the first two trajectory points as an example, the distance is 8.44. The direction angle is determined by calculating the angle between the line connecting the adjacent trajectory points and the coordinate axes. For the first two trajectory points, the angle with the x-axis is 41.6 degrees, the angle with the y-axis is 121.2 degrees, and the angle with the z-axis is 83.7 degrees.

[0084] Determine the set of nearest neighbors for each trajectory point. Nearest neighbors are defined as trajectory points whose distance is less than a threshold. In this example, the distance threshold is set to 12.0. For the first trajectory point [36.4, 0.42, 0.37], its set of nearest neighbors includes the second trajectory point [42.1, 0.56, 0.42] and the third trajectory point [38.7, 0.48, 0.39].

[0085] The local evolution direction of each trajectory point is calculated based on the set of nearest neighbors. The local evolution direction is determined by the average vector between the current point and its nearest neighbors. For the first trajectory point, its local evolution direction is [4.0, 0.1, 0.035]. The ratio of the displacement change between adjacent trajectory points to the time interval is used as the evolution rate. The time interval between adjacent trajectory points is the length of each time interval. Taking the first two trajectory points as an example, the displacement change is 8.44, the time interval is 126 + 204 = 330 minutes, and the evolution rate is 8.44 / 330 = 0.0256 / minute.

[0086] The fluctuation trend trajectory is statistically analyzed by partitioning it according to the evolution direction. The three-dimensional space is divided into 8 directional regions, corresponding to combinations of increasing or decreasing power value, increasing or decreasing rate of change, and increasing or decreasing fluctuation frequency. The proportion of trajectory points in each directional region is counted. In this example, there are 3 trajectory points in the directional region where power value increases, rate of change increases, and fluctuation frequency increases, with a proportion of 3 / 8 = 0.375. The average evolution rate in this directional region is calculated to be 0.0283 / minute. The development probability of each evolution direction is obtained by multiplying the proportion of trajectory points in each directional region by the average evolution rate. For the above directional region, the development probability is 0.375 × 0.0283 = 0.0106.

[0087] The direction with the highest probability of development is selected as the predicted trend. In this example, the direction with the highest probability of development (0.0106) is characterized by increasing power value, increasing rate of change, and increasing fluctuation frequency. Combined with the average evolution rate of 0.0283 / minute in this direction, the power fluctuation risk value is predicted for the next 180 minutes. The power fluctuation risk value is defined as the displacement change within the prediction time, calculated as 0.0283 × 180 = 5.094.

[0088] The control response requirements for the electrolytic cells are set based on the power fluctuation risk value. The control response requirements corresponding to a power fluctuation risk value of 5.094 are: a power adjustment range of not less than 6MW and a response time of not more than 5 minutes. The system has 10 electrolytic cells, each with a rated power of 8MW. The power adjustment rate and steady-state error of each electrolytic cell are calculated using the cell operating parameters. The power adjustment rate of electrolytic cell 1 is 1.2MW / min, and the steady-state error is 0.5%; the power adjustment rate of electrolytic cell 2 is 1.1MW / min, and the steady-state error is 0.6%; and so on.

[0089] The required control response is compared with the power regulation rate and steady-state error of the electrolyzer to calculate the response matching degree. The response matching degree is determined by combining the ratio of the electrolyzer's power regulation rate to the required response rate, and the ratio of the steady-state error to the allowable error. The response matching degree of electrolyzer 1 is 0.92, that of electrolyzer 2 is 0.87, and so on. Based on the response matching degree, the electrolyzers are grouped and ranked, and the combination of electrolyzers with the highest matching degree is selected as the target combination. In this example, the combination of electrolyzer 1, electrolyzer 3, and electrolyzer 5 is selected, and its overall matching degree is 0.89.

[0090] For the target electrolyzer combination, the maximum regulating power and response time constraint are calculated based on its operating parameters. The maximum regulating power of the target electrolyzer combination is 24MW, and the response time constraint is 4.2 minutes. Combining the power fluctuation risk value of 5.094, the target regulating power and execution sequence for each time period are determined. The 180-minute prediction period is divided into three 60-minute periods. The target regulating power for the first period is 2MW, for the second period it is 3MW, and for the third period it is 1MW. The execution sequence is as follows: the first period starts from the prediction start time, the second period starts at the 60th minute, and the third period starts at the 120th minute. The generated time-segmented control instructions include information on the target electrolyzer combination, the target regulating power for each period, and the execution time point.

[0091] Based on the above technical solutions, accurate prediction and optimized scheduling of power allocation for electrolyzer clusters under wind power fluctuation conditions can be achieved. By extracting fluctuation trend trajectories in a three-dimensional fluctuation feature space, an intuitive expression and quantitative analysis of wind power fluctuation characteristics are realized; predicting future fluctuation risks based on the evolution direction and rate of the trend trajectory improves the predictability of wind power fluctuations; through electrolyzer response matching degree evaluation and grouping and sorting, differentiated power allocation of the electrolyzer cluster is realized, improving the system's adaptability to fluctuations; the generation of time-segmented control commands allows the electrolyzer cluster to prepare for power adjustments in advance, avoiding the impact of large power fluctuations on the electrolyzers and extending equipment life; the overall solution improves wind power absorption efficiency, enhances the system's economy and reliability, and provides effective technical support for the coordinated operation of wind power and hydrogen energy systems.

[0092] In one optional implementation, a multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of the electrolyzers. A mapping relationship is established between dynamic response characteristics and time-segmented regulation requirements. The comprehensive regulation capability index of each electrolyzer is calculated using this mapping relationship. Based on the comprehensive regulation capability index, the electrolyzers are graded, screened, and prioritized to generate a dynamic grouping scheme for the electrolyzers, including:

[0093] Based on the obtained electrolyzer operation data, a multi-dimensional evaluation index system including response time, adjustment accuracy and efficiency loss is established, and the response characteristic curve, accuracy index value and loss index value of the electrolyzer are calculated based on the electrolyzer operation data.

[0094] Acquire system regulation and control demand data, perform matching analysis between the regulation and control demand data and the response characteristic curve, accuracy index value and loss index value of the electrolyzer, obtain the adaptability evaluation value of the electrolyzer to the regulation and control demand, perform dynamic evaluation of the electrolyzer based on the adaptability evaluation value, and calculate the comprehensive regulation capability index value of the electrolyzer according to the importance weight of regulation in each regulation period.

[0095] Acquire the cumulative number of adjustments and adjustment intensity data of the electrolytic cell, calculate the fatigue index value; and classify the electrolytic cell into a rapid response level, a conventional adjustment level, and a standby level based on the comprehensive adjustment capability index value.

[0096] Within each level of electrolytic cell, the comprehensive adjustment capability index value and the fatigue index value are weighted to obtain a priority coefficient, and the electrolytic cells are sorted according to the priority coefficient; when the fatigue index value exceeds a preset fatigue threshold, the electrolytic cell is downgraded and the priority ranking is updated, generating a dynamic grouping scheme for electrolytic cells containing grading information and priority information.

[0097] In this embodiment, a multi-dimensional evaluation index system, including response time, adjustment accuracy, and efficiency loss, is first established based on the acquired electrolyzer operating data. The response time index describes the time required for the electrolyzer to reach the target power value from receiving a power adjustment command, expressed in seconds. The adjustment accuracy index describes the deviation between the actual output power and the target power of the electrolyzer, expressed as a percentage. The efficiency loss index describes the degree of reduction in the energy conversion efficiency of the electrolyzer during power adjustment, expressed as a percentage. Taking 10 electrolytic cells in a certain system as an example, electrolytic cell 1 has a response time of 120 seconds, an adjustment accuracy of 0.5%, and an efficiency loss of 1.2%; electrolytic cell 2 has a response time of 135 seconds, an adjustment accuracy of 0.6%, and an efficiency loss of 1.0%; electrolytic cell 3 has a response time of 115 seconds, an adjustment accuracy of 0.4%, and an efficiency loss of 1.3%; electrolytic cell 4 has a response time of 150 seconds, an adjustment accuracy of 0.7%, and an efficiency loss of 0.9%; electrolytic cell 5 has a response time of 125 seconds, an adjustment accuracy of 0.5%, and an efficiency loss of... The efficiency loss is 1.1%; the response time of electrolytic cell 6 is 140 seconds, the adjustment accuracy is 0.6%, and the efficiency loss is 1.0%; the response time of electrolytic cell 7 is 130 seconds, the adjustment accuracy is 0.5%, and the efficiency loss is 1.2%; the response time of electrolytic cell 8 is 145 seconds, the adjustment accuracy is 0.7%, and the efficiency loss is 0.9%; the response time of electrolytic cell 9 is 120 seconds, the adjustment accuracy is 0.5%, and the efficiency loss is 1.1%; the response time of electrolytic cell 10 is 155 seconds, the adjustment accuracy is 0.8%, and the efficiency loss is 0.8%.

[0098] The response characteristic curve, accuracy index, and loss index of the electrolyzer were calculated based on the electrolyzer's operating data. The response characteristic curve describes the change in electrolyzer power over time and is obtained by fitting power change data collected from the electrolyzer under different operating conditions. For electrolyzer 1, during the power adjustment from 6MW to 7MW, the collected power change data points were: 6.00MW at 0 seconds, 6.32MW at 20 seconds, 6.58MW at 40 seconds, 6.75MW at 60 seconds, 6.88MW at 80 seconds, 6.95MW at 100 seconds, and 6.99MW at 120 seconds. The fitted response characteristic curve is an exponential curve with a time constant of 85 seconds as the characteristic parameter. The accuracy index was calculated by statistically analyzing the power deviation of the electrolyzer under stable operating conditions. For electrolyzer 1, with a target power of 7MW, the average actual power was 7.02MW, the standard deviation was 0.035MW, and the calculated accuracy index was 0.5%. The loss index value was calculated by comparing the energy conversion efficiency of the electrolyzer before and after adjustment. For electrolyzer 1, during the process of adjusting the power from 6MW to 7MW, the energy conversion efficiency decreased from 68.5% to 67.3%, and the calculated loss index value was 1.2%.

[0099] Acquire system control requirement data, and perform a matching analysis between this data and the response characteristic curves, accuracy index values, and loss index values ​​of the electrolyzers to obtain an evaluation value of the electrolyzers' adaptability to the control requirements. The system control requirement data includes control time requirements, accuracy requirements, and efficiency requirements. Taking a certain time period as an example, the control time requirement is 130 seconds, the accuracy requirement is 0.6%, and the efficiency requirement is 1.1%. Calculate the matching degree between the electrolyzer's response characteristics and the control time requirement. For electrolyzer 1, the response time is 120 seconds, which is less than the control time requirement of 130 seconds, and the matching degree is 130 / 120 = 1.08, which is greater than 1, indicating that the requirement is met. Calculate the matching degree between the electrolyzer's accuracy index and the control accuracy requirement. For electrolyzer 1, the accuracy index value is 0.5%, which is less than the control accuracy requirement of 0.6%, and the matching degree is 0.6 / 0.5 = 1.2, which is greater than 1, indicating that the requirement is met. Calculate the matching degree between the electrolyzer's loss index and the control efficiency requirement. For electrolyzer 1, the loss index is 1.2%, which is greater than the control efficiency requirement of 1.1%, and the matching degree is 1.1 / 1.2 = 0.92. A value less than 1 indicates that the requirement is not fully met. Considering the matching degree of the three aspects, the adaptability evaluation value of electrolyzer 1 to this control requirement is calculated as 1.08×0.4 + 1.2×0.4 + 0.92×0.2 = 1.076, where 0.4, 0.4, and 0.2 are the weighting coefficients of each index.

[0100] The electrolyzer is dynamically evaluated based on its adaptability evaluation value. The comprehensive regulation capacity index of the electrolyzer is calculated according to the weighted importance of regulation in each regulation period. A day is divided into six regulation periods, each lasting four hours. The weighted importance of regulation in each period is 0.15, 0.2, 0.25, 0.2, 0.15, and 0.05, respectively. The adaptability evaluation value of the electrolyzer in each period is calculated, multiplied by the weighted importance of regulation, and summed to obtain the comprehensive regulation capacity index of the electrolyzer. For electrolyzer 1, the adaptability evaluation values ​​for the six time periods are 1.076, 1.102, 1.054, 1.089, 1.065, and 1.033, respectively. The calculated comprehensive regulation capacity index is 1.076×0.15+1.102×0.2+1.054×0.25+1.089×0.2+1.065×0.15+1.033×0.05=1.074. Similarly, the comprehensive regulation capacity index values ​​for other electrolyzers are calculated as follows: electrolyzer 2 is 0.989, electrolyzer 3 is 1.095, electrolyzer 4 is 0.921, electrolyzer 5 is 1.035, electrolyzer 6 is 0.978, electrolyzer 7 is 1.002, electrolyzer 8 is 0.934, electrolyzer 9 is 1.056, and electrolyzer 10 is 0.912.

[0101] Obtain the cumulative number of adjustments and adjustment intensity data for the electrolytic cells, and calculate the fatigue index value. The cumulative number of adjustments refers to the cumulative number of times the electrolytic cell participates in power adjustment. The adjustment intensity refers to the ratio of the power change amplitude of each adjustment to the rated power of the electrolytic cell. For electrolytic cell 1, the cumulative number of adjustments is 156, the average adjustment intensity is 0.15, and the calculated fatigue index value is 156 × 0.15 = 23.4. Similarly, calculate the fatigue index values ​​for other electrolytic cells: electrolytic cell 2 is 19.8, electrolytic cell 3 is 25.2, electrolytic cell 4 is 17.1, electrolytic cell 5 is 21.5, electrolytic cell 6 is 18.7, electrolytic cell 7 is 20.3, electrolytic cell 8 is 16.9, electrolytic cell 9 is 22.6, and electrolytic cell 10 is 15.8.

[0102] Electrolyzers are classified into three levels based on their comprehensive regulation capability index: fast response, conventional regulation, and standby. The standard for fast response is a comprehensive regulation capability index value greater than 1.05; for conventional regulation, a value between 0.95 and 1.05; and for standby, a value less than 0.95. Based on this standard, electrolyzers 1, 3, and 9 belong to the fast response level; electrolyzers 2, 5, 6, and 7 belong to the conventional regulation level; and electrolyzers 4, 8, and 10 belong to the standby level.

[0103] Within each electrolytic cell, the priority coefficient is calculated by weighting the comprehensive adjustment capability index value and the fatigue index value. The formula for calculating the priority coefficient is: comprehensive adjustment capability index value multiplied by 0.7 plus normalized fatigue margin multiplied by 0.3. The fatigue margin is defined as a preset fatigue threshold minus the fatigue index value, with the preset fatigue threshold set to 30. For electrolytic cell 1, the fatigue margin is 30 - 23.4 = 6.6, the normalized fatigue margin is 6.6 / 30 = 0.22, and the calculated priority coefficient is 1.074 × 0.7 + 0.22 × 0.3 = 0.818. Similarly, the priority coefficients for other electrolytic cells are calculated and ranked within each level. The internal order of the fast response stage is: electrolyzer 3 (0.834), electrolyzer 9 (0.815), electrolyzer 1 (0.818); the internal order of the conventional regulation stage is: electrolyzer 5 (0.797), electrolyzer 7 (0.760), electrolyzer 2 (0.734), electrolyzer 6 (0.721); the internal order of the standby stage is: electrolyzer 10 (0.684), electrolyzer 8 (0.679), electrolyzer 4 (0.670).

[0104] When the fatigue index value exceeds the preset fatigue threshold, the electrolytic cell is downgraded and its priority ranking is updated. The preset fatigue threshold is 30. If the fatigue index value of electrolytic cell 3 increases to 32.1, exceeding the preset fatigue threshold of 30, then electrolytic cell 3 is downgraded from the fast response level to the normal regulation level, and the priority coefficient and ranking are recalculated within the normal regulation level. A dynamic grouping scheme for electrolytic cells containing classification and priority information is generated: the fast response level includes electrolytic cells 9 and 1; the normal regulation level includes electrolytic cells 3, 5, 7, 2, and 6; and the standby level includes electrolytic cells 10, 8, and 4.

[0105] Based on the above technical solution, efficient and optimized scheduling of power allocation for electrolyzer groups under wind power fluctuation conditions can be achieved. Existing technologies often only consider single indicators such as response speed or power capacity when evaluating the regulation capability of electrolyzers, failing to comprehensively reflect the overall performance of electrolyzers during dynamic regulation. Furthermore, scheduling strategies typically employ fixed grouping methods, making it difficult to adapt to changes in electrolyzer operating states and wind power fluctuation characteristics. This application constructs a multi-dimensional evaluation system encompassing response time, regulation accuracy, and efficiency loss, achieving a comprehensive evaluation of electrolyzer regulation capability. By establishing a mapping relationship between electrolyzer response characteristics and regulation requirements, precise matching between electrolyzers and regulation needs is achieved. Introducing fatigue indicators and setting a dynamic degradation mechanism effectively balances the service life of electrolyzers with system regulation performance. Through dynamic grading and priority ranking, the adaptability of the electrolyzer group to wind power fluctuations is enhanced, wind power absorption efficiency is improved, and the stability and economy of system operation are ensured.

[0106] Figure 2 This is a thermogram showing the relationship between the adjustment accuracy and efficiency loss of the electrolytic cell in an embodiment of the present invention. Figure 2 As shown, this heatmap illustrates the relationship between electrolytic cell regulation accuracy and load regulation rate, and its impact on efficiency loss. The heatmap visually represents the degree of loss through different color shades, from a light-colored low-loss zone (0-2%) to a dark-colored high-loss zone (>6%). The heatmap clearly shows that efficiency loss exhibits a non-linear growth trend as the load regulation rate increases and the regulation accuracy requirement rises. The optimal operating region is located within the dashed box in the lower left corner, representing the ideal operating state. This technical solution, with its multi-dimensional evaluation index system, can achieve low efficiency loss while maintaining high accuracy; traditional methods, such as voltage-controlled PID feedback algorithms, while meeting basic regulation requirements, suffer from high efficiency loss. The zoning analysis of the heatmap provides an intuitive basis for formulating electrolytic cell operating strategies, helping to select the most suitable operating point based on accuracy requirements and load conditions in practical applications, thereby balancing regulation performance and energy efficiency.

[0107] In one optional implementation, the electrolyzer is dynamically evaluated based on the adaptability evaluation value, and the comprehensive regulation capability index value of the electrolyzer is calculated according to the weight of the importance of regulation in each regulation period, including:

[0108] Collect wind power fluctuation data, obtain power fluctuation inflection points through extreme value detection, divide the evaluation period into multiple adjustment periods based on the power fluctuation inflection points, and calculate the control difficulty coefficient based on the power change trend in each adjustment period.

[0109] The power fluctuation amplitude, fluctuation frequency and system stability margin of each adjustment period are obtained, and the adjustment period weight evaluation matrix is ​​constructed. The control importance weight of each adjustment period is calculated based on the weight evaluation matrix.

[0110] The operating data of the electrolyzer during each adjustment period is collected. Based on the operating data and the adjustment difficulty coefficient, the adaptability evaluation value of the electrolyzer during each adjustment period is calculated. The adaptability evaluation value is normalized and combined with the weight of the importance of adjustment to obtain the comprehensive adjustment capability index value of the electrolyzer.

[0111] In this implementation, wind power fluctuation data is first collected. Power fluctuation inflection points are then obtained through extreme value detection. Based on these inflection points, the evaluation period is divided into multiple adjustment periods. Wind power fluctuation data is collected through the wind farm's SCADA system at a sampling frequency of 1 minute per instance, with a collection period of 24 hours, resulting in a power data sequence of 1440 sampling points. Taking a wind farm as an example, some of the collected power data are: 42.3MW, 43.5MW, 45.1MW, 46.8MW, 47.2MW, 46.9MW, 45.8MW, 44.2MW, 42.9MW, and 41.5MW. Power fluctuation inflection points are obtained through extreme value detection using a sliding window method with a window size of 5 sampling points. If the power value of a sampling point is greater than the power values ​​of other sampling points within the window, it is marked as a maximum point; if the power value of a sampling point is less than the power values ​​of other sampling points within the window, it is marked as a minimum point. Maximum and minimum points are collectively referred to as power fluctuation inflection points. Extreme value detection was performed on the above data, identifying a maximum of 47.2 MW at the 5th sampling point and a minimum of 41.5 MW at the 10th sampling point. Extreme value detection was performed on the entire day's power data, identifying a total of 18 maximum points and 17 minimum points, for a total of 35 power fluctuation inflection points. Based on these inflection points, the 24-hour assessment period was divided into 36 adjustment periods, each lasting the time interval between two consecutive inflection points. Taking the first two inflection points as an example, the first inflection point occurred at the 82nd minute, and the second at the 127th minute; therefore, the first adjustment period lasted 45 minutes.

[0112] The control difficulty coefficient is calculated based on the power change trend within each regulation period. The power change trend is characterized by the slope and fluctuation amplitude of the power data within the period. The slope is calculated as the ratio of the power change to the duration of the period. The fluctuation amplitude is calculated as the difference between the maximum and minimum power values ​​within the period. The control difficulty coefficient is determined by combining the absolute value of the slope and the fluctuation amplitude, and the calculation formula is: absolute value of slope multiplied by 0.6 plus the ratio of fluctuation amplitude to rated power multiplied by 0.4. Taking the first regulation period as an example, the starting power is 35.6MW, the ending power is 47.2MW, the period duration is 45 minutes, the slope is (47.2-35.6) / 45 = 0.26MW / minute, and the absolute value of the slope is 0.26MW / minute. The maximum power during the period is 47.2MW, the minimum is 35.6MW, and the fluctuation range is 47.2-35.6 = 11.6MW. The wind farm's rated power is 100MW, and the ratio of the fluctuation range to the rated power is 11.6 / 100 = 0.116. The calculated regulation difficulty coefficient is 0.26×0.6 + 0.116×0.4 = 0.2024. Similarly, the regulation difficulty coefficients for other regulation periods are calculated. The regulation difficulty coefficient for the second regulation period is 0.1853, the regulation difficulty coefficient for the third regulation period is 0.2145, and so on.

[0113] The power fluctuation amplitude, fluctuation frequency, and system stability margin for each regulation period are obtained, and a weighted evaluation matrix for each regulation period is constructed. The power fluctuation amplitude is defined as the ratio of the power change to the rated power within the period. The fluctuation frequency is defined as the number of times the power change direction changes per unit time. The system stability margin is defined as the ratio of the grid frequency deviation to the allowable deviation, characterizing system stability. Taking the first regulation period as an example, the power fluctuation amplitude is (47.2-35.6) / 100 = 0.116, the fluctuation frequency is 3 times / 45 minutes = 0.067 times / minute, and the system stability margin is 0.8. These three indicators are assigned weights according to their importance: power fluctuation amplitude has a weight of 0.5, fluctuation frequency has a weight of 0.3, and system stability margin has a weight of 0.2. The three indicators are then normalized so that their values ​​range from 0 to 1. Taking power fluctuation amplitude as an example, the largest power fluctuation amplitude among the 36 time periods is 0.215, and the smallest is 0.042. The normalized power fluctuation amplitude of the first time period is (0.116-0.042) / (0.215-0.042)=0.427. Similarly, the normalized fluctuation frequency is calculated to be 0.568, and the normalized system stability margin is 0.625. A weighted evaluation matrix for the adjustment period is constructed, with rows representing each adjustment period and columns representing three evaluation indicators. The row vector of the weighted evaluation matrix for the first time period is [0.427, 0.568, 0.625].

[0114] The importance weight of regulation for each adjustment period is calculated based on the weight evaluation matrix. The importance score for each adjustment period is obtained by multiplying each indicator value in the weight evaluation matrix by its corresponding weight and summing the results. The importance score for the first period is 0.427×0.5 + 0.568×0.3 + 0.625×0.2 = 0.5126. Similarly, the importance scores for other periods are calculated. The importance scores for each period are normalized so that the sum of the importance weights for all periods is 1. The importance weight for the first period is 0.5126 / 18.4536 = 0.0278, where 18.4536 is the sum of the importance scores for all periods. Similarly, the importance weights for other periods are calculated. The importance weight for the second period is 0.0246, the importance weight for the third period is 0.0312, and so on.

[0115] Operating data of the electrolyzers during each adjustment period is collected. Based on the operating data and the adjustment difficulty coefficient, the adaptability evaluation value of the electrolyzers during each adjustment period is calculated. The electrolyzer operating data includes response time, adjustment accuracy, and efficiency loss. Taking electrolyzer 1 in a certain system as an example, in the first adjustment period, the response time is 125 seconds, the adjustment accuracy is 0.5%, and the efficiency loss is 1.2%. The adaptability evaluation value is calculated by combining the electrolyzer operating data with the adjustment difficulty coefficient for each period. The formula for calculating the adaptability evaluation value is: response time adaptability multiplied by 0.4, adjustment accuracy adaptability multiplied by 0.4, and efficiency loss adaptability multiplied by 0.2. The response time adaptability is calculated by dividing the baseline response time by the actual response time, with the baseline response time set to 120 seconds. The adjustment accuracy adaptability is calculated by dividing the baseline accuracy by the actual accuracy, with the baseline accuracy set to 0.6%. The efficiency loss adaptability is calculated by dividing the baseline loss by the actual loss, with the baseline loss set to 1.0%. Taking electrolytic cell 1 in the first adjustment period as an example, the response time fitness is 120 / 125 = 0.96, the adjustment accuracy fitness is 0.6 / 0.5 = 1.2, and the efficiency loss fitness is 1.0 / 1.2 = 0.833; the calculated fitness evaluation value is 0.96 × 0.4 + 1.2 × 0.4 + 0.833 × 0.2 = 1.007. Multiplying the fitness evaluation value by the adjustment factor of the control difficulty coefficient (calculated as 1 plus the control difficulty coefficient), the control difficulty coefficient for the first period is 0.2024, the adjustment factor is 1 + 0.2024 = 1.2024, and the adjusted fitness evaluation value is 1.007 × 1.2024 = 1.211. Similarly, the fitness evaluation values ​​for electrolytic cell 1 in other adjustment periods are calculated. In the second adjustment period, the fitness evaluation value is 1.15; in the third adjustment period, the fitness evaluation value is 1.23; and so on.

[0116] The adaptability evaluation values ​​were normalized, and the comprehensive regulation capacity index of the electrolyzer was calculated by combining the weights of the regulation importance. The adaptability evaluation values ​​of the electrolyzer in each regulation period were normalized to a range of 0 to 1. Taking electrolyzer 1 as an example, among the 36 regulation periods, the highest adaptability evaluation value was 1.45, and the lowest was 0.92. The normalized adaptability evaluation value for the first period was (1.211-0.92) / (1.45-0.92) = 0.547. Similarly, the normalized adaptability evaluation value for electrolyzer 1 in the second period was 0.434, and in the third period it was 0.585, and so on. The normalized adaptability evaluation value was multiplied by the regulation importance weight of the corresponding regulation period, and the sum was calculated over all periods to obtain the comprehensive regulation capacity index of the electrolyzer. The comprehensive adjustment capability index of electrolytic cell 1 is 0.547×0.0278+0.434×0.0246+0.585×0.0312+...=0.526. Similarly, the comprehensive adjustment capability index values ​​of the other electrolytic cells in the system are calculated. The comprehensive adjustment capability index value of electrolytic cell 2 is 0.489, the comprehensive adjustment capability index value of electrolytic cell 3 is 0.542, and so on.

[0117] Based on the above technical solution, accurate assessment and optimized scheduling of power allocation for electrolyzer clusters under wind power fluctuation conditions can be achieved. Existing technologies typically use fixed time-period division methods to assess the regulation capacity of electrolyzers, which cannot reflect the actual fluctuation characteristics of wind power; moreover, the assessment index system is singular and cannot comprehensively characterize the adaptability of electrolyzers under different fluctuation conditions; furthermore, the weight allocation of each time period lacks a scientific basis, leading to a disconnect between assessment results and actual needs. This application identifies power fluctuation inflection points through extreme value detection technology, achieving adaptive division of the assessment period, closely integrating time-period division with the actual fluctuation characteristics of wind power; it constructs a multi-dimensional weighted evaluation matrix including power fluctuation amplitude, fluctuation frequency, and system stability margin, scientifically quantifying the importance of each regulation period; and it introduces a regulation difficulty coefficient to dynamically adjust the electrolyzer adaptability evaluation value, making the assessment results more objective and accurate. Through these improvements, this solution achieves accurate assessment of the regulation capacity of electrolyzers, enhances the adaptability of the electrolyzer cluster to wind power fluctuations, improves the stability and economy of system operation, and provides a more scientific scheduling basis for wind power consumption and hydrogen energy production.

[0118] Figure 3 This is a bar chart comparing the adjustment capabilities of different electrolytic cell types in embodiments of the present invention, such as... Figure 3As shown in the bar chart, this comparison details the comprehensive regulatory capacity indices of three types of electrolyzers (alkaline electrolyzers, PEM electrolyzers, and solid oxide electrolyzers) under three different evaluation methods. The solid gray bars represent the evaluation results of this technical solution, the light gray bordered bars represent the results of the uniform time period method, and the white dashed bordered bars represent the results of the empirical threshold method. The data shows that under all evaluation methods, the PEM electrolyzer exhibits the best regulatory capacity, followed by the alkaline electrolyzer, while the solid oxide electrolyzer performs the weakest. Specifically, under the evaluation of this technical solution, the comprehensive regulatory capacity index of the PEM electrolyzer reaches 0.79, significantly higher than the 0.59 of the alkaline electrolyzer and the 0.43 of the solid oxide electrolyzer; while under the traditional uniform time period method, the indices for the three electrolyzers are 0.65, 0.49, and 0.31, respectively; and under the empirical threshold method, the indices for the three electrolyzers are even lower, at 0.61, 0.45, and 0.24, respectively. This scheme can more accurately reflect the differences in the actual regulation capabilities of various electrolyzers, especially the differences in adaptability under extreme operating conditions. Compared with traditional methods that simply divide the time period into uniform periods or use fixed empirical thresholds for judgment, this technical scheme accurately divides the regulation period based on the inflection point of power fluctuation and constructs a multi-dimensional weighted evaluation matrix, which can more comprehensively evaluate the adaptability of electrolyzers to wind power fluctuations.

[0119] In one optional implementation, a power regulation matrix is ​​constructed based on a dynamic grouping scheme for electrolyzers. A recursive iterative method is used to calculate the optimal power allocation ratio for each electrolyzer in the power regulation matrix, generating a control command containing the real-time power setpoint for each electrolyzer.

[0120] The operating power, power fluctuation, and response time of the electrolytic cell group are collected, and the dynamic response characteristics of the electrolytic cells are calculated based on the operating power, power fluctuation, and response time.

[0121] A power regulation matrix is ​​constructed based on the dynamic grouping scheme of the electrolyzer. The inter-group power coupling coefficient is calculated based on the dynamic response characteristics of the electrolyzer. The inter-group power coupling coefficient is written into the corresponding position of the power regulation matrix to obtain the initial power regulation matrix.

[0122] The power adjustment deviation matrix is ​​obtained by performing a difference operation between the element values ​​of the initial power adjustment matrix and the rated power value of the electrolyzer. The inter-group adaptive weights are calculated based on the power adjustment deviation matrix, and the inter-group adaptive weights are multiplied by the initial power adjustment matrix to obtain the optimized power adjustment matrix.

[0123] A multidimensional optimization objective function is constructed based on the optimized power adjustment matrix and the power adjustment deviation matrix. The power adjustment deviation, inter-group power coupling coefficient and dynamic response characteristics are set as optimization variables. The weight coefficients of the optimization variables are calculated according to the system operation requirements.

[0124] A recursive iterative method is used to calculate the power allocation ratio of each electrolytic cell in the power adjustment matrix. The adaptive step size is calculated by multiplying the weight coefficients of the optimization variables with the gradient of the multidimensional optimization objective function. The dynamic compensation factor is calculated based on historical iteration data to obtain the optimal power allocation ratio. The optimal power allocation ratio is multiplied by the rated power of the electrolytic cell to obtain the real-time power setpoint of each electrolytic cell, and a control command is generated.

[0125] The system collects the operating power, power fluctuation, and response time of the electrolytic cell group. Based on these data, the dynamic response characteristics of the electrolytic cells are calculated. Taking 10 electrolytic cells in a system as an example, each with a rated power of 8MW, the current operating power of the electrolytic cells is collected through the electrolytic cell control system. The current operating power of electrolytic cell 1 is 6.5MW, 2 is 7.2MW, 3 is 5.8MW, 4 is 6.8MW, 5 is 7.5MW, 6 is 6.2MW, 7 is 7.0MW, 8 is 5.6MW, 9 is 6.4MW, and 10 is 6.0MW. Power fluctuation refers to the magnitude of power change during the adjustment process, obtained through statistical analysis of historical operating data. The power fluctuation of electrolytic cell 1 is 0.8MW, electrolytic cell 2 is 0.6MW, electrolytic cell 3 is 0.9MW, electrolytic cell 4 is 0.5MW, electrolytic cell 5 is 0.7MW, electrolytic cell 6 is 0.6MW, electrolytic cell 7 is 0.8MW, electrolytic cell 8 is 0.5MW, electrolytic cell 9 is 0.7MW, and electrolytic cell 10 is 0.4MW. Response time refers to the time required for an electrolytic cell to reach the target power value after receiving a power adjustment command. The response time of electrolytic cell 1 is 125 seconds, electrolytic cell 2 is 135 seconds, electrolytic cell 3 is 115 seconds, electrolytic cell 4 is 150 seconds, electrolytic cell 5 is 125 seconds, electrolytic cell 6 is 140 seconds, electrolytic cell 7 is 130 seconds, electrolytic cell 8 is 145 seconds, electrolytic cell 9 is 120 seconds, and electrolytic cell 10 is 155 seconds.

[0126] The dynamic response characteristics of the electrolyzer were calculated based on the collected data. These characteristics include three indicators: power regulation rate, steady-state error, and response sensitivity. The power regulation rate is calculated as the ratio of power fluctuation to response time, expressed in MW / second. The power regulation rate of electrolytic cell 1 is 0.8 / 125 = 0.0064 MW / s, that of electrolytic cell 2 is 0.6 / 135 = 0.0044 MW / s, that of electrolytic cell 3 is 0.9 / 115 = 0.0078 MW / s, that of electrolytic cell 4 is 0.5 / 150 = 0.0033 MW / s, that of electrolytic cell 5 is 0.7 / 125 = 0.0056 MW / s, that of electrolytic cell 6 is 0.6 / 140 = 0.0043 MW / s, that of electrolytic cell 7 is 0.8 / 130 = 0.0062 MW / s, that of electrolytic cell 8 is 0.5 / 145 = 0.0034 MW / s, that of electrolytic cell 9 is 0.7 / 120 = 0.0058 MW / s, and that of electrolytic cell 10 is 0.4 / 155 = 0.0026 MW / s. Steady-state error refers to the deviation ratio between the actual power and the target power of the electrolytic cell after power adjustment, obtained through statistical analysis of historical operating data. The steady-state error is 0.5% for electrolytic cell 1, 0.6% for electrolytic cell 2, 0.4% for electrolytic cell 3, 0.7% for electrolytic cell 4, 0.5% for electrolytic cell 5, 0.6% for electrolytic cell 6, 0.5% for electrolytic cell 7, 0.7% for electrolytic cell 8, 0.5% for electrolytic cell 9, and 0.8% for electrolytic cell 10. Response sensitivity refers to the degree to which the electrolytic cell responds to power adjustment commands, calculated as the ratio of power fluctuation to rated power. The response sensitivity of electrolytic cell 1 is 0.8 / 8 = 0.1, that of electrolytic cell 2 is 0.6 / 8 = 0.075, that of electrolytic cell 3 is 0.9 / 8 = 0.1125, that of electrolytic cell 4 is 0.5 / 8 = 0.0625, that of electrolytic cell 5 is 0.7 / 8 = 0.0875, that of electrolytic cell 6 is 0.6 / 8 = 0.075, that of electrolytic cell 7 is 0.8 / 8 = 0.1, that of electrolytic cell 8 is 0.5 / 8 = 0.0625, that of electrolytic cell 9 is 0.7 / 8 = 0.0875, and that of electrolytic cell 10 is 0.4 / 8 = 0.05.

[0127] A power regulation matrix is ​​constructed based on a dynamic grouping scheme for electrolyzers. The inter-group power coupling coefficient is calculated based on the dynamic response characteristics of the electrolyzers, and this coefficient is written into the corresponding positions in the power regulation matrix to obtain the initial power regulation matrix. According to the aforementioned dynamic grouping scheme, the electrolyzers are divided into three groups: a fast response stage including electrolyzers 1, 3, and 9; a conventional regulation stage including electrolyzers 2, 5, 6, and 7; and a standby stage including electrolyzers 4, 8, and 10. The power regulation matrix is ​​a 10×10 matrix, with rows and columns corresponding to 10 electrolyzers. The elements in the matrix represent the power coupling relationship between the electrolyzers. The inter-group power coupling coefficient is calculated as a weighted average of the ratio of the power regulation rate to the response sensitivity of two electrolyzers. The weighting coefficients are 0.6 for power regulation rate and 0.4 for response sensitivity. Taking the inter-group power coupling coefficient of electrolytic cell 1 and electrolytic cell 2 as an example, the power regulation rate ratio is 0.0064 / 0.0044 = 1.455, the response sensitivity ratio is 0.1 / 0.075 = 1.333, and the inter-group power coupling coefficient is 1.455 × 0.6 + 1.333 × 0.4 = 1.407. Similarly, the inter-group power coupling coefficients between other electrolytic cell pairs are calculated, and these coefficients are filled into the corresponding positions in the power regulation matrix to obtain the initial power regulation matrix.

[0128] The power regulation deviation matrix is ​​obtained by subtracting the rated power value of the electrolyzer from the element values ​​of the initial power regulation matrix. Based on this matrix, the inter-group adaptive weights are calculated, and the optimized power regulation matrix is ​​obtained by multiplying these weights by the initial power regulation matrix. The power regulation deviation matrix is ​​calculated by subtracting the rated power of the electrolyzer in the corresponding row from each element of the initial power regulation matrix. For example, the element in the first row and second column of the power regulation matrix has a value of 1.407. Since the rated power of electrolyzer 1 is 8MW, the power regulation deviation is 1.407 - 8 = -6.593MW. Similarly, the other elements of the power regulation deviation matrix are calculated. The inter-group adaptive weights are calculated by normalizing the mean deviation values ​​between groups in the power regulation deviation matrix. The mean deviation between the fast response stage and the conventional regulation stage is -6.285MW, the mean deviation between the fast response stage and the standby stage is -6.748MW, and the mean deviation between the conventional regulation stage and the standby stage is -6.532MW. After normalization, the adaptive weights between the fast response stage and the conventional regulation stage are 0.322, between the fast response stage and the standby stage are 0.346, and between the conventional regulation stage and the standby stage are 0.332. Multiplying these inter-group adaptive weights by the initial power regulation matrix yields the optimized power regulation matrix. Taking the element in the first row and second column of the optimized power regulation matrix as an example, its value is 1.407 × 0.322 = 0.453.

[0129] A multidimensional optimization objective function is constructed based on the optimized power regulation matrix and power regulation deviation matrix. Power regulation deviation, inter-group power coupling coefficient, and dynamic response characteristics are set as optimization variables, and the weight coefficients of these variables are calculated according to system operating requirements. The multidimensional optimization objective function comprises three sub-objective functions: a power balance objective function, a response characteristic objective function, and a coupling optimization objective function. The power balance objective function, defined as the sum of squares of the elements of the power regulation deviation matrix, is used to ensure system power balance. The response characteristic objective function, defined as the sum of squares of the differences between the power regulation rates of each electrolyzer and the target rate, is used to optimize the power coupling relationship between electrolyzers and is defined as the sum of squares of the elements of the optimized power regulation matrix. The weight coefficients of the optimization variables are determined according to system operating requirements: the weight coefficient for the power balance objective function is 0.5, the weight coefficient for the response characteristic objective function is 0.3, and the weight coefficient for the coupling optimization objective function is 0.2.

[0130] A recursive iterative method is used to calculate the power allocation ratio of each electrolyzer in the power regulation matrix. The adaptive step size is calculated by multiplying the optimization variable weight coefficients with the gradient of the multidimensional optimization objective function. A dynamic compensation factor is calculated based on historical iteration data to obtain the optimal power allocation ratio. This optimal power allocation ratio is then multiplied by the rated power of each electrolyzer to obtain the real-time power setpoint for each electrolyzer, generating control commands. During the recursive iteration process, the initial power allocation ratio is set as the ratio of the current operating power to the rated power of each electrolyzer. The initial power allocation ratio for electrolyzer 1 is 6.5 / 8 = 0.8125, for electrolyzer 2 it is 7.2 / 8 = 0.9, for electrolyzer 3 it is 5.8 / 8 = 0.725, and so on. The gradient of the multidimensional optimization objective function is calculated by taking the partial derivative of the objective function with respect to each optimization variable. The weight coefficients of the optimization variables are multiplied by the gradient to obtain the weighted gradient. The adaptive step size is calculated by multiplying the magnitude of the weighted gradient by a preset step size coefficient. The preset step size coefficient is 0.05. The dynamic compensation factor is calculated based on the convergence trend of historical iterative data and is adaptively adjusted. The initial dynamic compensation factor is set to 1.0 and adjusted according to the changing trend of the objective function value over three consecutive iterations. If the objective function value continuously decreases, the compensation factor is increased; if the objective function value fluctuates, the compensation factor is decreased. After 20 iterations, the optimal power allocation ratio is obtained. The optimal power allocation ratio is 0.85 for electrolytic cell 1, 0.82 for electrolytic cell 2, 0.88 for electrolytic cell 3, 0.75 for electrolytic cell 4, 0.84 for electrolytic cell 5, 0.78 for electrolytic cell 6, 0.81 for electrolytic cell 7, 0.72 for electrolytic cell 8, 0.86 for electrolytic cell 9, and 0.70 for electrolytic cell 10. Multiplying the optimal power allocation ratio by the rated power of the electrolytic cell yields the real-time power setpoint for each electrolytic cell. The real-time power setpoints for electrolytic cell 1 are 0.85 × 8 = 6.8 MW, for electrolytic cell 2 it is 0.82 × 8 = 6.56 MW, for electrolytic cell 3 it is 0.88 × 8 = 7.04 MW, for electrolytic cell 4 it is 0.75 × 8 = 6 MW, for electrolytic cell 5 it is 0.84 × 8 = 6.72 MW, for electrolytic cell 6 it is 0.78 × 8 = 6.24 MW, for electrolytic cell 7 it is 0.81 × 8 = 6.48 MW, for electrolytic cell 8 it is 0.72 × 8 = 5.76 MW, for electrolytic cell 9 it is 0.86 × 8 = 6.88 MW, and for electrolytic cell 10 it is 0.70 × 8 = 5.6 MW. A control command containing the above real-time power setpoints is generated. The command includes information such as the electrolytic cell identifier, power setpoint, and execution timestamp.

[0131] Based on the above technical solutions, precise optimization and efficient scheduling of power allocation in electrolyzer groups under wind power fluctuation conditions can be achieved. Existing electrolyzer power allocation methods typically employ equal distribution strategies or simple proportional allocation, failing to consider the dynamic response differences and power coupling relationships between electrolyzers, resulting in low overall system response efficiency. Furthermore, optimization algorithms often use fixed-step-size iterative methods, leading to slow convergence and susceptibility to local optima. This application constructs a power adjustment matrix based on the dynamic response characteristics of electrolyzers, achieving precise characterization of the power coupling relationships between them; introduces an inter-group adaptive weighting mechanism to effectively balance the adjustment load of electrolyzers with different response levels; designs a multi-dimensional objective function encompassing power balance, response characteristics, and coupling optimization, comprehensively considering various system operation constraints; and employs a recursive iterative method with adaptive step size and dynamic compensation factors, significantly improving the algorithm's convergence speed and optimization accuracy. Through the synergistic effect of the above technical means, this solution achieves precise power allocation and dynamic adjustment of the electrolyzer group, improving the system's adaptability to wind power fluctuations, reducing the impact of power fluctuations on the electrolyzers, enhancing wind power absorption capacity, and ensuring the stability and economy of hydrogen energy production.

[0132] A second aspect of the present invention provides an optimized scheduling system for power allocation of an electrolyzer group under wind power fluctuation conditions, the system comprising:

[0133] The first unit is used to collect wind farm output power data and electrolytic cell group operating parameters;

[0134] The second unit is used to perform adaptive segmentation processing on the output power data, calculate the power fluctuation feature vector in each time period, construct a three-dimensional fluctuation feature space based on the power fluctuation feature vector, extract the fluctuation trend trajectory in the three-dimensional fluctuation feature space, predict future fluctuation risks, and generate time-segmented control instructions.

[0135] The third unit is used to construct a multi-dimensional regulation capability evaluation system based on the operating parameters of the electrolyzer, establish a mapping relationship between dynamic response characteristics and time-segmented regulation requirements, use the mapping relationship to calculate the comprehensive regulation capability index of each electrolyzer, classify and prioritize the electrolyzers according to the comprehensive regulation capability index, and generate a dynamic grouping scheme for the electrolyzers.

[0136] The fourth unit is used to construct a power regulation matrix based on the dynamic grouping scheme of electrolyzers, and to calculate the optimal power allocation ratio of each electrolyzer in the power regulation matrix using a recursive iterative method, thereby generating a control command containing the real-time power setpoint of each electrolyzer.

[0137] The fifth unit is used to transmit the real-time power setpoint to the control unit of each electrolyzer to perform power regulation, and to collect the real-time response data of the electrolyzers. Based on the real-time response data, the control instructions are optimized to realize the coordinated control of wind power fluctuations by the electrolyzer group.

[0138] A third aspect of the present invention provides an electronic device, comprising:

[0139] processor;

[0140] Memory used to store processor-executable instructions;

[0141] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0142] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0143] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimized scheduling method for power allocation of an electrolytic cell group under wind power fluctuation conditions, characterized in that, include: Collect wind farm output power data and electrolytic cell group operating parameters; The output power data is adaptively segmented, the power fluctuation feature vector in each time period is calculated, a three-dimensional fluctuation feature space is constructed based on the power fluctuation feature vector, the fluctuation trend trajectory is extracted in the three-dimensional fluctuation feature space, the future fluctuation risk is predicted and time-segmented control instructions are generated. A multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of the electrolyzer. A mapping relationship between dynamic response characteristics and time-segmented regulation requirements is established. The comprehensive regulation capability index of each electrolyzer is calculated using the mapping relationship. Based on the comprehensive regulation capability index, the electrolyzers are classified, screened, and prioritized to generate a dynamic grouping scheme for the electrolyzers. A power regulation matrix is ​​constructed based on the dynamic grouping scheme of electrolyzers. The optimal power allocation ratio of each electrolyzer in the power regulation matrix is ​​calculated by a recursive iterative method, and a control command containing the real-time power setpoint of each electrolyzer is generated. The real-time power setpoint is transmitted to the control unit of each electrolyzer to perform power adjustment, and the real-time response data of the electrolyzer is collected. The control instructions are optimized based on the real-time response data to achieve coordinated control of wind power fluctuations by the electrolyzer group. The output power data is adaptively segmented, and the power fluctuation feature vector for each time period is calculated. Based on the power fluctuation feature vector, a three-dimensional fluctuation feature space is constructed, including: The power change rate sequence is obtained by calculating the power difference between adjacent sampling points in the wind power data. The power change rate sequence is then processed by a sliding window, and the mean and standard deviation of the power change rate within each window are calculated. A first-order difference sequence is constructed based on the mean of the power change rate, and a fluctuation amplitude sequence is constructed based on the standard deviation of the power change rate. The first-order difference sequence and the fluctuation amplitude sequence are weighted and combined to obtain a segmentation criterion. An adaptive threshold is calculated on the segmentation criterion. A benchmark threshold is determined based on the statistical distribution characteristics of the power change rate sequence. The threshold coefficient is dynamically adjusted in combination with the cumulative distribution function of the fluctuation amplitude sequence to obtain the segmented threshold. Sampling points exceeding the segmentation threshold are marked as segmentation points. The wind power data sequence is divided into multiple time periods based on the segmentation points. Power fluctuation features are extracted from the wind power data sequence in each time period, including average power value, power change rate, and power fluctuation frequency. The power fluctuation features are combined to construct a power fluctuation feature vector. The feature contribution of the power fluctuation feature vector is calculated using the principal component analysis method. The three feature directions with the largest cumulative contribution are selected to construct a three-dimensional fluctuation feature space.

2. The method according to claim 1, characterized in that, Extracting volatility trend trajectories from a three-dimensional volatility feature space, predicting future volatility risks, and generating time-segmented control instructions include: The power fluctuation feature vectors are connected sequentially in the three-dimensional fluctuation feature space to form a fluctuation trend trajectory. The distance and direction angle between adjacent trajectory points in the fluctuation trend trajectory are calculated to determine the nearest neighbor set of each trajectory point. Based on the nearest neighbor set, the local evolution direction of each trajectory point is calculated. The ratio of the displacement change between adjacent trajectory points to the time interval is used as the evolution rate. The fluctuation trend trajectory is divided into partitions according to the evolution direction. The product of the proportion of trajectory points in each partition and the average evolution rate is calculated to determine the development probability of each evolution direction. The direction with the highest development probability is selected as the predicted trend. The power fluctuation risk value is obtained by combining the average evolution rate of that direction. Based on the power fluctuation risk value, the electrolytic cell control response requirements are set. The power regulation rate and steady-state error of each electrolytic cell are calculated using the electrolytic cell operating parameters. The control response requirements are compared with the power regulation rate and steady-state error of the electrolytic cells to calculate the response matching degree. Based on the response matching degree, the electrolytic cells are grouped and sorted. The electrolytic cell combination with the highest matching degree is selected as the target electrolytic cell combination. For the target electrolytic cell combination, the maximum regulation power and response time constraint are calculated based on its operating parameters. Combined with the power fluctuation risk value, the target regulation power and execution sequence for each time period are determined, and time-segmented control instructions are generated.

3. The method according to claim 1, characterized in that, A multi-dimensional regulation capability evaluation system is constructed based on the operating parameters of electrolyzers. A mapping relationship between dynamic response characteristics and time-segmented regulation requirements is established. This mapping relationship is used to calculate the comprehensive regulation capability index of each electrolyzer. Based on the comprehensive regulation capability index, electrolyzers are graded, screened, and prioritized to generate a dynamic grouping scheme for electrolyzers, including: Based on the obtained electrolyzer operation data, a multi-dimensional evaluation index system including response time, adjustment accuracy and efficiency loss is established, and the response characteristic curve, accuracy index value and loss index value of the electrolyzer are calculated based on the electrolyzer operation data. Acquire system regulation and control demand data, perform matching analysis between the regulation and control demand data and the response characteristic curve, accuracy index value and loss index value of the electrolyzer, obtain the adaptability evaluation value of the electrolyzer to the regulation and control demand, perform dynamic evaluation of the electrolyzer based on the adaptability evaluation value, and calculate the comprehensive regulation capability index value of the electrolyzer according to the importance weight of regulation in each regulation period. Acquire the cumulative number of adjustments and adjustment intensity data of the electrolytic cell, calculate the fatigue index value; and classify the electrolytic cell into a rapid response level, a conventional adjustment level, and a standby level based on the comprehensive adjustment capability index value. Within each level of electrolytic cell, the comprehensive adjustment capability index value and the fatigue index value are weighted to obtain a priority coefficient, and the electrolytic cells are sorted according to the priority coefficient; when the fatigue index value exceeds a preset fatigue threshold, the electrolytic cell is downgraded and the priority ranking is updated, generating a dynamic grouping scheme for electrolytic cells containing grading information and priority information.

4. The method according to claim 3, characterized in that, The electrolyzer is dynamically evaluated based on its adaptability assessment value. The comprehensive regulation capability index of the electrolyzer is calculated according to the weighted importance of regulation in each regulation period, including: Collect wind power fluctuation data, obtain power fluctuation inflection points through extreme value detection, divide the evaluation period into multiple adjustment periods based on the power fluctuation inflection points, and calculate the control difficulty coefficient based on the power change trend in each adjustment period. The power fluctuation amplitude, fluctuation frequency and system stability margin of each adjustment period are obtained, and the adjustment period weight evaluation matrix is ​​constructed. The control importance weight of each adjustment period is calculated based on the weight evaluation matrix. The operating data of the electrolyzer during each adjustment period is collected. Based on the operating data and the adjustment difficulty coefficient, the adaptability evaluation value of the electrolyzer during each adjustment period is calculated. The adaptability evaluation value is normalized and combined with the weight of the importance of adjustment to obtain the comprehensive adjustment capability index value of the electrolyzer.

5. The method according to claim 1, characterized in that, A power regulation matrix is ​​constructed based on a dynamic grouping scheme for electrolyzers. A recursive iterative method is used to calculate the optimal power allocation ratio for each electrolyzer in the power regulation matrix, generating control instructions containing real-time power setpoints for each electrolyzer. The operating power, power fluctuation, and response time of the electrolytic cell group are collected, and the dynamic response characteristics of the electrolytic cells are calculated based on the operating power, power fluctuation, and response time. A power regulation matrix is ​​constructed based on the dynamic grouping scheme of the electrolyzer. The inter-group power coupling coefficient is calculated based on the dynamic response characteristics of the electrolyzer. The inter-group power coupling coefficient is written into the corresponding position of the power regulation matrix to obtain the initial power regulation matrix. The power adjustment deviation matrix is ​​obtained by performing a difference operation between the element values ​​of the initial power adjustment matrix and the rated power value of the electrolyzer. The inter-group adaptive weights are calculated based on the power adjustment deviation matrix, and the inter-group adaptive weights are multiplied by the initial power adjustment matrix to obtain the optimized power adjustment matrix. A multidimensional optimization objective function is constructed based on the optimized power adjustment matrix and the power adjustment deviation matrix. The power adjustment deviation, inter-group power coupling coefficient and dynamic response characteristics are set as optimization variables. The weight coefficients of the optimization variables are calculated according to the system operation requirements. A recursive iterative method is used to calculate the power allocation ratio of each electrolytic cell in the power adjustment matrix. The adaptive step size is calculated by multiplying the weight coefficients of the optimization variables with the gradient of the multidimensional optimization objective function. The dynamic compensation factor is calculated based on historical iteration data to obtain the optimal power allocation ratio. The optimal power allocation ratio is multiplied by the rated power of the electrolytic cell to obtain the real-time power setpoint of each electrolytic cell, and a control command is generated.

6. An optimized scheduling system for power allocation of an electrolytic cell group under wind power fluctuation conditions, used to implement the method described in any one of claims 1-5, characterized in that, include: The first unit is used to collect wind farm output power data and electrolytic cell group operating parameters; The second unit is used to perform adaptive segmentation processing on the output power data, calculate the power fluctuation feature vector in each time period, construct a three-dimensional fluctuation feature space based on the power fluctuation feature vector, extract the fluctuation trend trajectory in the three-dimensional fluctuation feature space, predict future fluctuation risks, and generate time-segmented control instructions. The third unit is used to construct a multi-dimensional regulation capability evaluation system based on the operating parameters of the electrolyzer, establish a mapping relationship between dynamic response characteristics and time-segmented regulation requirements, use the mapping relationship to calculate the comprehensive regulation capability index of each electrolyzer, classify and prioritize the electrolyzers according to the comprehensive regulation capability index, and generate a dynamic grouping scheme for the electrolyzers. The fourth unit is used to construct a power regulation matrix based on the dynamic grouping scheme of electrolyzers, and to calculate the optimal power allocation ratio of each electrolyzer in the power regulation matrix using a recursive iterative method, thereby generating a control command containing the real-time power setpoint of each electrolyzer. The fifth unit is used to transmit the real-time power setpoint to the control unit of each electrolyzer to perform power regulation, and to collect the real-time response data of the electrolyzers. Based on the real-time response data, the control instructions are optimized to realize the coordinated control of wind power fluctuations by the electrolyzer group.

7. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 5.

Images