CN-121986194-A - Distributed learning of optimal operation for electrolyser modules

Abstract

The invention relates to a system (1) for improving the operation of an electrolyser module (2), the system (1) comprising a plurality of electrolysers (3) and a central part (4) connected to the electrolysers (3) for exchanging data, the electrolysers (3) each having a plurality of electrolyser modules (2), the central part being configured to collect model parameters of the electrolyser modules (2) of the electrolysers (3), compress the model parameters and update the model for the electrolyser modules (2), wherein the electrolyser modules (3) each comprise a controller (5) connected to the electrolyser modules (2) and a local training module (6) connected to the controller (5), the local training module being configured to update a statistical model of the electrolyser modules (2) based on the controller data, to transmit the updated model parameters of the statistical model to the central part (4), to receive the model parameters processed in the central part (4) and to collate the model parameters to the respective controllers (5) for controlling the electrolyser modules (2). The invention also relates to a method for improving the operation of an electrolyser module (2) in an electrolyser (3).

Inventors

• C pillen

Assignees

• 西门子能源国际公司

Dates

Publication Date

20260505

Application Date

20240828

Priority Date

20231010

Claims (14)

1. 1. System (1) for improving the operation of an electrolyser module (2), the system (1) comprising a plurality of electrolysers (3) and a central part (4) connected to the electrolysers (3) for exchanging data, the electrolysers each having a plurality of electrolyser modules (2), the central part being configured to collect model parameters of the electrolyser modules (2) of the electrolysers (3), compress the model parameters and update the model for the electrolyser modules (2), wherein the electrolyser (3) each comprises a controller (5) connected to the electrolyser modules (2) and a local training module (6) connected to the controller (5), the local training module being configured to update the statistical model of the electrolyser modules (2) based on the controller data, to transmit the model parameters of the updated statistical model to the central part (4), to receive the model parameters processed in the central part (4) and to collate the model parameters to the corresponding controller (5) for controlling the electrolyser modules (2).
2. 2. The system (1) for improving the operation of an electrolyser module (2) according to claim 1, wherein said local training module (6) is configured to learn the relation between the operation of the electrolyser module (2) and performance indicators.
3. 3. System (1) for improving the operation of an electrolyser module (2) according to claim 1 or 2, wherein said local training module (6) is configured to generate and/or update a statistical model for the electrolyser module (2) based on data characterizing the operation characteristics of the electrolyser module (2) and to optimize the operation of the electrolyser module (2) based on said statistical model.
4. 4. A system (1) for improving the operation of electrolyser modules (2) according to claim 3, wherein said optimization comprises one or more of the total efficiency of said electrolyser (3), degradation or service life of each electrolyser module (2), operational safety and maintenance mode.
5. 5. System (1) for improving the operation of an electrolyser module (2) according to any of the preceding claims, wherein said local training module (6) can be synchronized by said central section (4).
6. 6. System (1) for improving the operation of an electrolyser module (2) according to any of the preceding claims, wherein said system is configured to periodically transmit said data to said central section (4).
7. 7. System (1) for improving the operation of an electrolyser module (2) according to any of the preceding claims, wherein the data are summarized by arithmetic mean.
8. 8. Method for improving the operation of an electrolyser module (2) in an electrolyser (3), wherein in the electrolyser (3) a controller (5) controls and monitors the electrolyser module (2), wherein a local training module (6) of the electrolyser (3) updates a statistical model of the electrolyser module (2) based on controller data and transmits the model parameters thus obtained to a central part (4), wherein the central part (4) collects, compresses and updates the model parameters of the electrolyser module (2) of the electrolyser (3) for the electrolyser module (2), wherein the updated model parameters are distributed to the local training module (6) of the electrolyser (3) and are transmitted to the controller (5) after finishing.
9. 9. The method according to claim 8, wherein the relation between the operating mode of the electrolyser module (2) and the performance index is learned in the electrolyser (3).
10. 10. Method according to claim 8 or 9, wherein a statistical model for an electrolyser module (2) is generated and/or updated in the electrolyser (3) based on data characterizing the operating characteristics of the electrolyser module (2), and the operating mode of the electrolyser module (2) is optimized based on the statistical model.
11. 11. The method according to claim 10, wherein the optimization comprises one or more of the total efficiency of the electrolysis device (3), degradation or service life of each electrolyser module (2), operational safety and maintenance mode.
12. 12. The method according to any one of claims 8 to 11, wherein the local training module (6) is synchronized by the central portion (4).
13. 13. The method according to any one of claims 8 to 12, wherein the data is periodically transmitted to the central portion (4) by the local training module (6).
14. 14. The method according to any one of claims 8 to 13, wherein the data is arithmetically averaged in the central portion (4).

Description

Distributed learning of optimal operation for electrolyser modules Technical Field The invention relates to a system for improving the operation of an electrolyser module and a corresponding method. Background The manner in which the electrolyser module operates has a very large impact on the degradation performance of components such as membranes. Excessive switching has been shown to accelerate degradation of the electrolyser module. A gentle mode of operation can provide remediation. Various schemes for data analysis and control optimization are known in the art. Model Predictive Control (MPC) is, for example, a control optimization method for complex, typically multi-variable, processes. The method is based on a discrete time dynamic model of the system and utilizes real-time data to calculate an optimal control signal. MPC can be used to optimize the operating parameters of the electrolyzer and ensure high efficiency. However, this method requires a huge computational power. Alternatively, by developing detailed physical models and simulations of the electrolyzer, virtual tests can be performed on different operating conditions and optimal settings determined. The model can be used to predict performance under different conditions and thus optimize control. Many electrochemical processes and their association with the manner of operation are currently unknown and are extremely difficult to model physically. Data driven schemes with respect to static models or machine learning methods may provide a solution. Here, a correlation between degradation characteristics and the operating mode is determined, and a data-driven model is developed. Models may be developed herein to identify anomalies and predict when maintenance is required, or to monitor the operating conditions of the electrolyzer generally. Thus, by modeling, a competitive advantage can be achieved, such as longer run times, stable operation of the device, or a generally deeper understanding of extremely complex electrochemical behavior patterns within the cell. Machine learning requires a large amount of data, which is typically collected centrally to build up a dataset for training. But it is precisely the exclusive acquisition of the operational data required for machine learning that constitutes an obstacle. For example, the operator of the electrolysis device may view information about its manner of operation as sensitive information, without providing raw operational data for further analysis. Alternative models are currently determined by experimental research, physical modeling, or data driven modeling, which allow for predictions of important performance metrics such as degradation, and thus allow for more intelligent and milder operating strategies. However, many different environmental impacts make it impossible to implement customized modeling for a particular device having an associated environmental impact. Disclosure of Invention The object of the present invention is to provide a system for improving the operation of an electrolyser module. The invention also aims to provide a method for improving the operation mode of an electrolytic cell module in an electrolytic device. The invention achieves the object proposed for a system for improving the operation of an electrolyser module by providing that such a system comprises a plurality of electrolysers and a central part connected to the electrolysers for data exchange, the electrolysers each having a plurality of electrolyser modules, the central part being configured to collect model parameters of the electrolyser modules of the electrolysers, compress the model parameters and update the model for the electrolyser modules, wherein the electrolysers each comprise a controller connected to the electrolyser modules and a local training module connected to the controller, respectively, the local training module being configured to update the statistical model of the electrolyser modules based on controller data, to transmit the model parameters of the updated statistical model to the central part, to receive the model parameters processed in the central part and to collate them to the corresponding controllers for controlling the electrolyser modules. Authors H. Brendan McMahan, eider Moore, DANIEL RAMAGE, seth Hampson and Blaise Ag u ERA Y AREAS describe in an article whose topic is "Communication-EFFICIENT LEARNING of Deep Networks from Decentralized Data (deep network Communication efficient learning based on decentralized data)" (website: https:// arxiv. Org/abs/1602.05629) a method that enables building a learning model for a specific application with massive data of modern mobile devices, which enables better user experience, e.g. based on an improved speech model for speech recognition or a simplified text input implementation. Although there are differences in mobile device hardware, this does not prevent the same application from being improved. Because this massive data is