Energy storage cabinet remote control method and device based on internet of things

Abstract

The application discloses a kind of based on Internet of Things energy storage cabinet remote control method and device, it is related to remote control technical field.The method comprises: in cloud server, using federal learning mechanism cooperatively trains energy storage cabinet anomaly identification model, Internet of Things anomaly communication detection model and energy storage cabinet control strategy generation model, and is deployed to edge computing gateway;In edge computing gateway, energy storage cabinet anomaly identification, Internet of Things anomaly communication detection and energy storage cabinet control strategy generation are generated, and are uploaded to cloud server by state-instruction Internet of Things double channel;In edge computing gateway, energy storage cabinet control strategy is transmitted to the industrial computer of corresponding energy storage cabinet, and based on industrial computer, executes energy storage cabinet control strategy, starts the control equipment of corresponding energy storage cabinet and realizes remote control.Solve the problems of insufficient model generalization ability, unbalanced communication resource utilization, insufficient adaptability of control strategy and low model training efficiency in prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote control, and in particular to a remote control method and device for energy storage cabinets based on the Internet of Things. BACKGROUND

[0002] With the rapid development of the new energy industry, energy storage cabinets, as key equipment for grid peak regulation, frequency regulation, backup power supply, and renewable energy consumption, are increasingly widely used. In order to achieve efficient and centralized monitoring and management of large-scale and distributed energy storage cabinets, remote control technology based on the Internet of Things has emerged.

[0003] However, the existing remote control method for energy storage cabinets still has the following technical bottlenecks:

[0004] 1) Insufficient model generalization ability: Traditional centralized machine learning methods need to collect data from all energy storage sites to the cloud for model training, which not only exposes sensitive operating data of each site, causing data privacy and security problems, but also due to differences in operating environment, load characteristics, and battery aging degree of different energy storage cabinets, the model trained on a single data set has insufficient generalization ability and is difficult to adapt to all scenarios.

[0005] 2) Unbalanced utilization of communication resources: The communication data volume between energy storage cabinets and the cloud is huge, and existing technologies usually use a single communication channel and protocol, which cannot be differentiated according to the importance and urgency of the data. Under normal operating conditions, it may cause waste of communication resources; under abnormal operating conditions, it may delay critical instructions and alarm information due to network congestion, affecting real-time performance and reliability.

[0006] 3) Insufficient adaptability of control strategies: The generation of control strategies for energy storage cabinets often relies on pre-set rules or simple models, making it difficult to cope with complex and changing grid environments and internal states of energy storage cabinets. For example, in complex scenarios such as grid frequency fluctuations, battery inconsistencies, and temperature abnormalities, traditional control strategies cannot achieve real-time and accurate adaptive adjustment, affecting the operating efficiency and safety and stability of the energy storage system.

[0007] 4) Low efficiency of model training: In the model training process, the optimization of hyperparameters usually relies on manual experience or grid search methods, which have high computational cost and low efficiency, making it difficult to find the globally optimal combination of hyperparameters, limiting the upper limit of model performance. SUMMARY

[0008] The present application provides a remote control method and device for energy storage cabinets based on the Internet of Things, which solves the problems of insufficient model generalization ability, unbalanced utilization of communication resources, insufficient adaptability of control strategies, and low efficiency of model training in the prior art.

[0009] In a first aspect, the embodiments of the present application provide a remote control method for an energy storage cabinet based on an Internet of Things, which comprises the following steps:

[0010] In the cloud server, a federated learning mechanism is used to collaboratively train an energy storage cabinet anomaly identification model, an Internet of Things anomaly communication detection model, and an energy storage cabinet control strategy generation model, and the models are deployed to an edge computing gateway.

[0011] In the edge computing gateway, the energy storage cabinet anomaly identification model, the Internet of Things anomaly communication detection model, and the energy storage cabinet control strategy generation model are used to perform energy storage cabinet anomaly identification, Internet of Things anomaly communication detection, and energy storage cabinet control strategy generation, and the obtained energy storage cabinet anomaly identification result, Internet of Things anomaly communication detection result, and energy storage cabinet control strategy are uploaded to the cloud server through a state-instruction Internet of Things double-channel.

[0012] In the edge computing gateway, the energy storage cabinet control strategy is transmitted to the industrial computer of the corresponding energy storage cabinet, and the energy storage cabinet control strategy is executed based on the industrial computer to start the remote control of the control equipment corresponding to the energy storage cabinet.

[0013] The technical solutions provided by the embodiments of the present application have at least the following beneficial effects:

[0014] By introducing the federated learning mechanism, the original data of each training server does not need to leave the local, and only the model parameter update is uploaded to the cloud for aggregation, which effectively protects the data privacy and security of each energy storage site. At the same time, the federated learning aggregates model knowledge from different scenarios, so that the trained global model has stronger generalization ability and robustness. A state-instruction Internet of Things double-channel mechanism is designed, which intelligently switches between a "key channel" with high reliability and low latency and a "regular channel" with high efficiency and low overhead according to the running state and communication health status of the energy storage cabinet, ensuring that critical data can be transmitted preferentially and reliably in abnormal situations, while in normal situations, a large amount of regular data can be efficiently uploaded, realizing fine management and utilization of communication resources. The edge computing gateway deploys three core models based on advanced artificial intelligence algorithms: the energy storage cabinet anomaly identification model can accurately identify complex energy storage cabinet anomalies, the Internet of Things anomaly communication detection model can effectively detect anomalies in the Internet of Things communication layer, and the energy storage cabinet control strategy generation model can generate the optimal and adaptive control strategy according to the real-time state and anomaly information. This "perception-decision-execution" closed loop is completed on the edge side, greatly reducing the control delay and improving the intelligent level and operation efficiency of the energy storage control in complex working conditions. In the local training process of federated learning, an improved optimization algorithm is innovatively used to optimize the model hyperparameters synchronously, which can quickly and accurately find the globally optimal hyperparameter combination, avoiding the blindness and inefficiency of traditional methods and significantly improving the convergence speed and final performance of model training.

[0015] In an optional implementation, the cloud server is used to cooperatively train the energy storage cabinet anomaly identification model, the IoT abnormal communication detection model, and the energy storage cabinet control strategy generation model using a federated learning mechanism, and is deployed to the edge computing gateway, including:

[0016] The cloud server is used to build an initial model using an artificial intelligence algorithm; the initial model includes an initial energy storage cabinet anomaly identification model, an initial IoT abnormal communication detection model, and an initial energy storage cabinet control strategy generation model.

[0017] The first model metadata of the initial model is distributed to several training servers in the federated learning range, and the initial model is reconstructed on each training server.

[0018] The local iteration training is performed on the initial model according to the first local data set of the local energy storage cabinet in different scenarios on the training server, and a local model is obtained; the local model includes a local energy storage cabinet anomaly identification model, a local IoT abnormal communication detection model, and a local energy storage cabinet control strategy generation model.

[0019] In each local iteration training process, an improved optimization algorithm is used to synchronously optimize the hyperparameters of the local model, and the obtained optimized hyperparameters are applied to the next local iteration training.

[0020] If the local iteration training is ended or the first comprehensive loss value of the local model meets the requirement, the local model is output, the corresponding model parameter update amount and data amount are extracted, and are uploaded to the cloud server.

[0021] The cloud server is used to perform safe aggregation adjustment on the model parameters of the initial model using a federated aggregation weight dynamic adjustment mechanism according to the model parameter update amount and the data amount of all training servers, and a global model is obtained; the global model includes a global energy storage cabinet anomaly identification model, a global IoT abnormal communication detection model, and a global energy storage cabinet control strategy generation model.

[0022] The second model metadata of the global model is distributed to all edge computing gateways in the remote control range, and the global model is reconstructed on each edge computing gateway.

[0023] The edge computing gateway is used to perform migration fine-tuning on the global model according to the second local data set of the local energy storage cabinet in the current scenario, and a final model is obtained; the final model includes a final energy storage cabinet anomaly identification model, a final IoT abnormal communication detection model, and a final energy storage cabinet control strategy generation model.

[0024] If the migration fine-tuning is ended or the second comprehensive loss value of the final model meets the requirement, the deployment of the final model in the edge computing gateway is completed.

[0025] In an alternative implementation, the improved optimization algorithm is an IFPA algorithm.

[0026] In an alternative implementation, in each local iterative training process, the hyperparameters of the local model are synchronously optimized using the IFPA algorithm, and the obtained optimized hyperparameters are applied to the next local iterative training, including:

[0027] Based on the hyperparameter synchronous optimization problem, a multi-objective optimization function is set, and the multi-objective optimization function is used as the fitness function of the IFPA algorithm;

[0028] The hyperparameters of the local model are encoded into the individual vector of the IFPA algorithm, and the IFPA population parameters and the maximum number of iterations are set;

[0029] According to the IFPA population parameters, a Tent chaotic mapping sequence is used for initialization to obtain an initial IFPA population; each IFPA individual in the IFPA population corresponds to an alternative adapter parameter;

[0030] According to the first local data set, the fitness function is used to obtain the fitness value of each initial IFPA individual, and the initial IFPA individual with the optimal fitness value is used as the optimal solution;

[0031] With a conversion probability q , cross-pollination and self-pollination are selected to iteratively update the initial IFPA population to obtain an updated IFPA population;

[0032] A dynamic reverse mechanism is introduced to reversely update the updated IFPA individuals with a reverse probability p , to obtain a reversed IFPA population;

[0033] According to the first local data set, the fitness function is used to obtain the fitness value of each updated IFPA individual and reversed IFPA individual, and the updated IFPA individual or reversed IFPA individual with the optimal fitness value is updated as the optimal solution;

[0034] When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, the iterative updating of the IFPA population is terminated, and the optimal solution of the current iteration is output;

[0035] The individual vector of the IFPA individual corresponding to the optimal solution is decoded to obtain the optimal hyperparameters, and the optimized hyperparameters are used as the initial hyperparameters for the next local iterative training.

[0036] In an optional implementation, at the cloud server, the model parameters of the initial model are safely aggregated and adjusted according to the model parameter update amount and the data amount of all training servers using a federated aggregation weight dynamic adjustment mechanism to obtain a global model, including:

[0037] At the cloud server, the corresponding federated aggregation weight is generated using the federated aggregation weight dynamic adjustment mechanism according to the data amount of each training server;

[0038] The model parameters of the initial model are safely aggregated according to the federated aggregation weight and the corresponding model parameter update amount of all training servers to obtain global model parameters;

[0039] The initial model is adjusted according to the global model parameters to obtain a global model.

[0040] In an optional implementation, the state-instruction Internet of Things dual channel includes a key state and instruction Internet of Things channel and a regular state and data Internet of Things channel, the key state and instruction Internet of Things channel is provided with a first Internet of Things communication protocol with high reliability and low latency characteristics, and the regular state and data Internet of Things channel is provided with a second Internet of Things communication protocol with low overhead and high efficiency characteristics.

[0041] In an optional implementation, at the edge computing gateway, the energy storage cabinet abnormality identification model, the Internet of Things abnormal communication detection model, and the energy storage cabinet control strategy generation model are used to perform energy storage cabinet abnormality identification, Internet of Things abnormal communication detection, and energy storage cabinet control strategy generation, and the obtained energy storage cabinet abnormality identification result, Internet of Things abnormal communication detection result, and energy storage cabinet control strategy are uploaded to the cloud server through the state-instruction Internet of Things dual channel, including:

[0042] The multi-dimensional monitoring data of the energy storage cabinet are collected using the energy storage cabinet monitoring device, and the multi-dimensional monitoring data are uploaded to the edge computing gateway within the communication range through the bottom local area network;

[0043] At the edge computing gateway, the multi-dimensional monitoring data are input to the energy storage cabinet abnormality identification model to perform energy storage cabinet abnormality identification to obtain an energy storage cabinet abnormality identification result;

[0044] The energy storage cabinet abnormality identification result is input to the energy storage cabinet control strategy generation model to perform energy storage cabinet control strategy generation to obtain an energy storage cabinet control strategy;

[0045] At the edge computing gateway, the communication data of the Internet of Things communication channel between the edge computing gateway and the cloud server are captured, and the communication data are input to the Internet of Things abnormal communication detection to perform Internet of Things abnormal communication detection to obtain an Internet of Things abnormal communication detection result;

[0046] If the abnormality identification result of the energy storage cabinet is that there is an abnormality of the energy storage cabinet, or the abnormal communication detection result of the Internet of Things is that there is abnormal communication of the Internet of Things, the key state and instruction Internet of Things channel in the state-instruction Internet of Things dual channel is switched to, otherwise, the default is maintained to the regular state and data Internet of Things channel in the state-instruction Internet of Things dual channel.

[0047] The multi-dimensional monitoring data, the abnormality identification result of the energy storage cabinet, the abnormal communication detection result of the Internet of Things, and the energy storage cabinet control strategy are uploaded to the cloud server through the key state and instruction Internet of Things channel or the regular state and data Internet of Things channel in the state-instruction Internet of Things dual channel.

[0048] In an optional implementation, the energy storage cabinet abnormality identification model is constructed based on a CNN-LSTM-Attention-MLP algorithm.

[0049] The Internet of Things abnormal communication detection model is constructed based on an AE algorithm.

[0050] The energy storage cabinet control strategy generation model is constructed based on a MADDPG algorithm.

[0051] In an optional implementation, at the edge computing gateway, the energy storage cabinet control strategy is transmitted to the industrial computer of the corresponding energy storage cabinet, and based on the industrial computer, the energy storage cabinet control strategy is executed to start remote control of the control device corresponding to the energy storage cabinet, including:

[0052] At the edge computing gateway, the energy storage cabinet control strategy is transmitted to the industrial computer of the corresponding energy storage cabinet.

[0053] At the industrial computer, the energy storage cabinet control strategy is converted into a remote control instruction of the corresponding control device, and the remote control instruction is sent to the corresponding control device.

[0054] At the control device, the remote control instruction is executed to realize remote control of the energy storage cabinet.

[0055] In a second aspect, the embodiments of the present application provide an energy storage cabinet remote control device based on the Internet of Things, which is used to realize the energy storage cabinet remote control method. The device comprises:

[0056] A model deployment unit is configured to use a federated learning mechanism to collaboratively train an energy storage cabinet abnormality identification model, an Internet of Things abnormal communication detection model, and an energy storage cabinet control strategy generation model on a cloud server, and deploy them to an edge computing gateway.

[0057] The automatic function unit is used for identifying the abnormality of the energy storage cabinet, detecting the abnormal communication of the Internet of Things and generating the control strategy of the energy storage cabinet by using the energy storage cabinet abnormality identification model, the Internet of Things abnormal communication detection model and the energy storage cabinet control strategy generation model, and uploading the obtained energy storage cabinet abnormality identification result, the Internet of Things abnormal communication detection result and the energy storage cabinet control strategy to the cloud server through the state-command Internet of Things double channel.

[0058] The remote control unit is used for transmitting the energy storage cabinet control strategy to the industrial computer of the corresponding energy storage cabinet in the edge computing gateway, and executing the energy storage cabinet control strategy based on the industrial computer to start the control equipment corresponding to the energy storage cabinet to realize remote control.

[0059] The electronic device includes:

[0060] At least one processor, and a memory connected with the at least one processor in communication; wherein,

[0061] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method proposed in the first aspect of the embodiment.

[0062] The fourth aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to realize the method proposed in the first aspect of the embodiment of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 The electronic device structure schematic diagram of the hardware running environment related to the embodiment of the present application;

[0064] Figure 2 It is a step flow chart of the energy storage cabinet remote control method based on the Internet of Things provided by the embodiment of the present application;

[0065] Figure 3 It is a functional unit schematic diagram of the energy storage cabinet remote control device based on the Internet of Things provided by the embodiment of the present application. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below in combination with the drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0067] The scheme of the present application will be further described below in combination with the drawings.

[0068] Referring to Figure 1 , Figure 1 The electronic device structure schematic diagram of the hardware running environment involved in the embodiment of the present application.

[0069] As Figure 1 shown, the electronic device can include: a processor 1001, for example, a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen, an input unit such as a keyboard, and an optional user interface 1003 can also include a standard wired interface, a wireless interface. The network interface 1004 can optionally include a standard wired interface, a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 can be a high-speed random access memory (RAM) memory, or a stable non-volatile memory (NVM) such as a disk memory. The memory 1005 can also be an independent storage device from the aforementioned processor 1001.

[0070] Those skilled in the art can understand that Figure 1 the structure shown in the above description does not constitute a limitation on the electronic device, and can include more or fewer components than the diagram, or combine certain components, or different component arrangements.

[0071] As Figure 1 shown, the memory 1005 as a storage medium can include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program of the remote control device of the energy storage cabinet based on the Internet of Things.

[0072] In Figure 1 the electronic device shown, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the electronic device of the present application can be arranged in the electronic device. The electronic device calls the electronic program of the remote control device of the energy storage cabinet based on the Internet of Things stored in the memory 1005 through the processor 1001, and executes the remote control method of the energy storage cabinet based on the Internet of Things provided by the embodiment of the present application.

[0073] Referring to Figure 2The embodiment of the application provides a remote control method for an energy storage cabinet based on an Internet of Things, and the method comprises the following steps:

[0074] In the cloud server, a federal learning mechanism is used to cooperatively train an energy storage cabinet anomaly identification model, an Internet of Things anomaly communication detection model and an energy storage cabinet control strategy generation model, and the models are deployed to an edge computing gateway.

[0075] In the edge computing gateway, the energy storage cabinet anomaly identification model, the Internet of Things anomaly communication detection model and the energy storage cabinet control strategy generation model are used to perform energy storage cabinet anomaly identification, Internet of Things anomaly communication detection and energy storage cabinet control strategy generation, and the obtained energy storage cabinet anomaly identification result, Internet of Things anomaly communication detection result and energy storage cabinet control strategy are uploaded to the cloud server through a state-instruction Internet of Things double-channel.

[0076] In the edge computing gateway, the energy storage cabinet control strategy is transmitted to an industrial computer of a corresponding energy storage cabinet, and the energy storage cabinet control strategy is executed based on the industrial computer to start a control device corresponding to the energy storage cabinet to realize remote control.

[0077] The technical scheme provided by the embodiment of the application at least brings the following beneficial effects:

[0078] By introducing the federal learning mechanism, the original data of each training server does not need to leave the local, and only the model parameter update amount is uploaded to the cloud end for aggregation, thereby effectively protecting the data privacy and security of each energy storage site. Meanwhile, the federal learning aggregates model knowledge from different scenes, so that the trained global model has stronger generalization ability and robustness. The state-instruction Internet of Things double-channel mechanism is designed, and according to the running state and communication health status of the energy storage cabinet, the mechanism is intelligently switched between a high-reliability, low-latency 'key channel' and a high-efficiency, low-overhead'regular channel', so that in the abnormal situation, the key data can be preferentially and reliably transmitted, and in the normal situation, the massive regular data can be efficiently uploaded, thereby realizing fine management and utilization of communication resources. The edge computing gateway deploys three core models based on advanced artificial intelligence algorithms, the energy storage cabinet anomaly identification model can accurately identify complex energy storage cabinet anomalies, the Internet of Things anomaly communication detection model can effectively detect anomalies in the Internet of Things communication layer, and the energy storage cabinet control strategy generation model can generate the optimal and adaptive control strategy according to the real-time state and anomaly information. The closed loop of 'perception-decision-execution' is completed on the edge side, thereby greatly reducing the control delay and improving the intelligent level and operation efficiency of the energy storage control in dealing with complex working conditions. In the local training process of the federal learning, an improved optimization algorithm is innovatively used to synchronously optimize the model hyperparameters, so that the globally optimal hyperparameter combination can be quickly and accurately found, the blindness and inefficiency of the traditional method are avoided, and the convergence speed and final performance of the model training are significantly improved.

[0079] In an optional implementation, in the cloud server, the energy storage cabinet abnormality identification model, the Internet of Things abnormal communication detection model, and the energy storage cabinet control strategy generation model are collaboratively trained using a federated learning mechanism and deployed to the edge computing gateway, including:

[0080] S2011: In the cloud server, an initial model is constructed using an artificial intelligence algorithm; the initial model includes an initial energy storage cabinet abnormality identification model, an initial Internet of Things abnormal communication detection model, and an initial energy storage cabinet control strategy generation model;

[0081] S2012: First model metadata of the initial model is distributed to several training servers in a federated learning range, and the initial model is reconstructed on each training server;

[0082] S2013: In the training server, the initial model is locally iteratively trained according to a first local data set of a local energy storage cabinet in different scenarios to obtain a local model; the local model includes a local energy storage cabinet abnormality identification model, a local Internet of Things abnormal communication detection model, and a local energy storage cabinet control strategy generation model;

[0083] S2014: In each local iterative training process, an improved optimization algorithm is used to synchronously optimize hyperparameters of the local model, and the obtained optimized hyperparameters are applied to the next local iterative training;

[0084] S2015: If the local iterative training ends or a first comprehensive loss value of the local model meets the requirement, the local model is output, corresponding model parameter update and data volume are extracted, and uploaded to the cloud server;

[0085] S2016: In the cloud server, model parameters of the initial model are safely aggregated and adjusted using a federated aggregation weight dynamic adjustment mechanism according to model parameter updates and data volumes of all training servers to obtain a global model; the global model includes a global energy storage cabinet abnormality identification model, a global Internet of Things abnormal communication detection model, and a global energy storage cabinet control strategy generation model;

[0086] S2017: Second model metadata of the global model is distributed to all edge computing gateways in a remote control range, and the global model is reconstructed on each edge computing gateway;

[0087] S2018: In the edge computing gateway, the global model is transferred and fine-tuned according to a second local data set of a local energy storage cabinet in a current scenario to obtain a final model; the final model includes a final energy storage cabinet abnormality identification model, a final Internet of Things abnormal communication detection model, and a final energy storage cabinet control strategy generation model;

[0088] S2019: If the fine-tuning of migration is completed or the second comprehensive loss value of the final model meets the requirement, the deployment of the final model in the edge computing gateway is completed.

[0089] In an optional implementation, the improved optimization algorithm is an improved flower pollination algorithm (IFPA).

[0090] In an optional implementation, in each local iterative training process, the hyperparameters of the local model are optimized synchronously using the IFPA algorithm, and the obtained optimized hyperparameters are applied to the next local iterative training, including:

[0091] S20141: Based on the hyperparameter synchronous optimization problem, a multi-objective optimization function is set, and the multi-objective optimization function is used as the fitness function of the IFPA algorithm;

[0092] The formula is:

[0093]

[0094] In the formula, is the fitness value of the IFPA individual; X is the precision objective function; is the generalization ability objective function; is the communication overhead objective function; is the fitness weight coefficient, which can be adjusted according to actual requirements; is the IFPA individual reference parameter; X

[0095]

[0096] In the formula, is the classification accuracy obtained by evaluating the global model constructed by the hyperparameters of the IFPA individual on the validation set of the edge computing gateway; X is the edge computing gateway indicator; is the total number of edge computing gateways; k is the validation set of the edge computing gateway; K is used to evaluate the average accuracy of the global model on the local validation sets of all edge computing gateways participating in training, and the goal is to maximize the model precision; k

[0097]

[0098] In the formula, is the classification accuracy obtained by evaluating the global model constructed by the hyperparameters of the IFPA individual on the validation set of all edge computing gateways; X is the total number of edge computing gateways;​​​K The variance of the accuracy of each edge computing gateway using the validation set is used to evaluate the generalization ability of the global model, that is, the stability and consistency of the model's performance on different edge computing gateways (representing different data distributions). Its goal is to maximize the generalization ability and avoid the model from overfitting to certain specific data sources. The variance of the accuracy on the validation set of each edge gateway is used to measure the generalization ability. The smaller the variance, the stronger the generalization ability.

[0099]

[0100] In the formula, Number of iterations t The k Local model metadata corresponding to the hyperparameters of an edge computing gateway; For IFPA individuals X The hyperparameters construct the global model metadata; used to evaluate the communication cost caused by model updates during the aggregation process. In federated learning, the ideal aggregation result should make the local model update amplitude of each edge gateway smaller in the next round of training, thereby reducing the amount or frequency of data that needs to be uploaded. The communication overhead is quantified by the average distance between all local model metadata received in the current round and the global model metadata to be evaluated. The goal is to minimize this distance. The smaller the value of this function, the smaller the average "discordance" between the global model to be evaluated and all local models in this round. This means that the local training direction of each edge gateway tends to be consistent, the model converges faster in the next round, and the potential communication overhead is lower. t This represents the current iteration number;

[0101] S20142: Encode the hyperparameters of the local model into individual vectors for the IFPA algorithm, and set the IFPA population parameters and maximum number of iterations;

[0102] In the embodiment, the hyperparameters of the local model include the first learning rate, batch size, number of CNN convolutional kernels, number of LSTM hidden units and dropout rate of the energy storage cabinet anomaly identification model, the number of encoder / decoder layers, number of neurons and second learning rate of the IoT anomaly communication detection model, and the third learning rate, experience replay pool size, exploration noise parameters and message passing dimension of the energy storage cabinet control strategy generation model.

[0103] S20143: Based on the IFPA population parameters, the initial IFPA population is obtained by initializing using the Tent chaotic mapping sequence; each IFPA individual in the IFPA population corresponds to a candidate adapter parameter.

[0104] The formula is:

[0105]

[0106] wherein, is the initial IALA population, and i is the initial IALA individual; is the initial IALA population, and i is the initial IALA individual; is the upper and lower bounds of the search space; i is the individual indicator;

[0107]

[0108] wherein, is the initial IALA population, and i- 1 chaotic variable; compared with random initialization, chaotic initialization can ensure the uniform distribution of the population in the solution space and enhance diversity;

[0109] S20144: According to the first local data set, the fitness function is used to obtain the fitness value of each initial IFPA individual, and the initial IFPA individual with the optimal fitness value is taken as the optimal solution;

[0110] S20145: With the conversion probability q , cross-pollination and self-pollination are selected to iteratively update the initial IFPA population to obtain the updated IFPA population;

[0111] According to the conversion probability q and the threshold L , the update mode is determined, if q ≤ L , it is the global search stage, and cross-pollination is performed, otherwise, it is the local search stage, and self-pollination is performed;

[0112] The formula is:

[0113]

[0114] wherein, is the iteration number t+ 1 cross-pollination to obtain the i th updated IFPA individual; is the convergence factor; is the IFPA individual of the iteration number t , and is the initial IFPA individual when the first iteration update is performed; is the optimal solution of the iteration number t ; is the Levy distributed random number; b is the Levy step length, and b ∈[1,2];

[0115]

[0116] In the formula, These are the maximum and minimum values ​​of the convergence factor; This represents the maximum number of iterations. t This represents the current iteration number; , To adjust the parameters; It is the hyperbolic tangent function;

[0117]

[0118] In the formula, Number of iterations t+ The first self-pollination of 1 yielded the first i One updated IFPA individual; A random number between [-1, 1]; For randomly selected different IFPA individuals;

[0119] S20146: Introduces a dynamic reversal mechanism, using reversal probability. p The updated IFPA individuals are dynamically reversed to obtain the reversed IFPA population.

[0120] The formula is:

[0121]

[0122] In the formula, Number of iterations t+ 1 of i A reverse IFPA individual; For the updated IFPA population, the first i One updated IFPA individual; For the updated IFPA population, the updated first i The reverse solution for each IFPA individual; The fitness function;

[0123] S20147: Based on the first local dataset, use the fitness function to obtain the fitness value of each updated IFPA individual and the reversed IFPA individual, and update the updated IFPA individual or the reversed IFPA individual with the best fitness value as the optimal solution.

[0124] S20148: When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, terminate the iterative update of the IFPA population and output the optimal solution of the current iteration.

[0125] S20149: Decode the individual vector of the IFPA individual corresponding to the optimal solution to obtain the optimal hyperparameters, and use the optimized hyperparameters as the initial hyperparameters for the next local iterative training.

[0126] In an optional implementation, on the cloud server, the model parameters of the initial model are safely aggregated and adjusted using a federated aggregation weight dynamic adjustment mechanism according to the model parameter update amount and the data amount of all training servers to obtain a global model, including:

[0127] S20161: On the cloud server, the corresponding federated aggregation weight is generated using a federated aggregation weight dynamic adjustment mechanism according to the data amount of each training server.

[0128] The formula is:

[0129]

[0130] In the formula, is the federated aggregation weight of the i-th training server at the j-th iteration; t is the data amount of the i-th training server; is the training server indication amount; is the total number of training servers; S20162: The model parameters of the initial model are safely aggregated using a federated aggregation weight dynamic adjustment mechanism according to the federated aggregation weight and the corresponding model parameter update amount of all training servers to obtain global model parameters. The formula is:

[0131]

[0132]

[0133]

[0134] In the formula, is the global model parameter at the j-th iteration; t+ is the model parameter update amount of the i-th training server; t

[0135] S20163: The initial model is adjusted according to the global model parameters to obtain a global model.

[0136] In an optional implementation, the state-instruction Internet of Things dual channel includes a key state and instruction Internet of Things channel and a regular state and data Internet of Things channel, the key state and instruction Internet of Things channel is provided with a first Internet of Things communication protocol with high reliability and low delay characteristics, and the regular state and data Internet of Things channel is provided with a second Internet of Things communication protocol with low overhead and high efficiency characteristics.​​​​​​

[0137] In this embodiment, the first Internet of Things communication protocol is the Message Queuing Telemetry Transport (MQTT) protocol based on the Transmission Control Protocol (TCP), and the second Internet of Things communication protocol is the Constrained Application Protocol (CoAP) based on the User Datagram Protocol (UDP).

[0138] In an optional implementation, at the edge computing gateway, the energy storage cabinet abnormality identification model, the Internet of Things abnormal communication detection model, and the energy storage cabinet control strategy generation model are used to perform energy storage cabinet abnormality identification, Internet of Things abnormal communication detection, and energy storage cabinet control strategy generation, and the obtained energy storage cabinet abnormality identification result, Internet of Things abnormal communication detection result, and energy storage cabinet control strategy are uploaded to the cloud server through the state-instruction Internet of Things double-channel, including:

[0139] S2021: Using the energy storage cabinet monitoring device, multi-dimensional monitoring data of the energy storage cabinet are collected, and the multi-dimensional monitoring data are uploaded to the edge computing gateway within the communication range through the bottom local area network;

[0140] In this embodiment, the energy storage cabinet monitoring device collects multi-dimensional monitoring data from the battery management system, power conversion system, temperature and humidity sensor, and other devices of the energy storage cabinet, and the multi-dimensional monitoring data include voltage, current, temperature, internal resistance, state of charge (SOC), and state of health (SOH) of the energy storage cabinet;

[0141] S2022: At the edge computing gateway, the multi-dimensional monitoring data are input to the energy storage cabinet abnormality identification model to perform energy storage cabinet abnormality identification, and an energy storage cabinet abnormality identification result is obtained;

[0142] In this embodiment, the energy storage cabinet abnormality identification result includes over-temperature, over-voltage, micro-short circuit, communication failure, and the like;

[0143] S2023: The energy storage cabinet abnormality identification result is input to the energy storage cabinet control strategy generation model to perform energy storage cabinet control strategy generation, and an energy storage cabinet control strategy is obtained;

[0144] In this embodiment, the energy storage cabinet control strategy includes energy storage cabinet operation control decision, energy storage cabinet start-stop control decision, energy storage cabinet charge-discharge control decision, and the like, for example, the energy storage cabinet operation control decision includes actions such as “reduced power operation”, “start fan to dissipate heat”, “execute alarm and report”, and the like;

[0145] S2024: At the edge computing gateway, the communication data of the Internet of Things communication channel between the edge computing gateway and the cloud server is captured, and the communication data is input into the Internet of Things abnormal communication detection to perform the Internet of Things abnormal communication detection, and an Internet of Things abnormal communication detection result is obtained;

[0146] In this embodiment, the communication data includes the data packet size, the sending / receiving frequency, the data packet interval time, the signal strength, the protocol type, and the source / destination IP address of the data transmission of the Internet of Things communication channel between the edge computing gateway and the cloud server; and the Internet of Things abnormal communication detection result includes a distributed denial of service (DDoS) attack, a man-in-the-middle attack, and data tampering, etc.

[0147] S2025: If the energy storage cabinet abnormality identification result is that there is an energy storage cabinet abnormality, or the Internet of Things abnormal communication detection result is that there is an Internet of Things abnormal communication, the key state and instruction Internet of Things channel in the state-instruction Internet of Things dual channel is switched to, otherwise, the default to the regular state and data Internet of Things channel in the state-instruction Internet of Things dual channel is maintained.

[0148] In this embodiment, if the abnormal communication detection result is that there is an abnormality, a preset Internet of Things security communication defense mechanism is triggered immediately, which is not simply cutting off the connection, but dynamically adjusting the defense strategy according to the detected abnormal type and severity:

[0149] Alarm and log: immediately send a security alarm to the cloud and operation and maintenance personnel, and record the abnormal characteristics in detail;

[0150] Dynamic protocol switching: if a specific attack against the current communication protocol is detected, automatically switch to a backup or more secure communication protocol;

[0151] Traffic shaping and speed limiting: if a DDoS attack is detected, automatically start traffic cleaning and speed limiting strategy to guarantee the bandwidth of key control instructions;

[0152] Data encryption enhancement: temporarily increase the encryption level and key update frequency of subsequent communication data;

[0153] S2026: The multi-dimensional monitoring data, the energy storage cabinet abnormality identification result, the Internet of Things abnormal communication detection result, and the energy storage cabinet control strategy are uploaded to the cloud server through the key state and instruction Internet of Things channel or the regular state and data Internet of Things channel in the state-instruction Internet of Things dual channel.

[0154] In an optional implementation, the energy storage cabinet anomaly identification model is constructed based on a Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM)-Attention-Multilayer Perceptron (MLP) algorithm;

[0155] In this embodiment, the energy storage cabinet anomaly identification model first extracts local spatial features from multi-dimensional monitoring data at a single time step using a CNN (Convolutional Neural Network); then captures long-term time sequence dependency of the data sequence using a BiLSTM (Bidirectional Long Short-Term Memory Network); then assigns dynamic weights to features at different time steps and different dimensions through Attention (Attention Mechanism) to highlight key information; finally, the MLP (Multilayer Perceptron) is used for classification, and the energy storage cabinet anomaly identification result is output; this structure can effectively fuse the local and global, spatial and temporal features of multi-dimensional monitoring data, and improve the accuracy and robustness of energy storage cabinet anomaly identification;

[0156] The Internet of Things anomaly communication detection model is constructed based on an Autoencoder (AE) algorithm; the autoencoder is selected to train on normal and abnormal communication data, and the input is features such as packet size and interval time; the autoencoder includes an encoder and a decoder, the encoder is composed of multiple fully connected layers, the number of neurons between layers gradually decreases, and the high-dimensional input data is compressed into a low-dimensional latent space representation, the decoder: composed of multiple fully connected layers, the number of neurons between layers gradually increases, trying to reconstruct the original input data from the low-dimensional encoding vector, the output is the same dimension as the input, that is, the reconstructed communication data, the reconstruction error is obtained as the anomaly score, and if the error exceeds the threshold, it is determined as an anomaly, and according to different thresholds corresponding to the anomaly score, the Internet of Things anomaly communication detection result is divided into alarm and log levels, dynamic protocol switching levels, traffic shaping and speed limiting levels, and data encryption enhancement levels;

[0157] The energy storage cabinet control strategy generation model is constructed based on a Multi-Agent Deep Deterministic Policy Gradient (MADDPG);

[0158] In this embodiment, the energy storage cabinet control strategy generation model regards fault handling as a multi-agent collaborative decision problem, MADDPG as the core, contains multiple agents, each agent represents a control device of an energy storage cabinet, learns the optimal joint energy storage cabinet control strategy through interaction with the environment in a shared global state space, includes a plurality of agents and an experience replay pool, and the plurality of agents are provided with a global state space and a global action space; the global state space includes the SOC, SOH, energy storage cabinet abnormality identification result and power grid operation state and load demand of the energy storage cabinet; and the global action space includes a combination of actions that can be performed by the energy storage cabinet control device.

[0159] In an alternative implementation, at the edge computing gateway, the energy storage cabinet control strategy is transmitted to the industrial computer of the corresponding energy storage cabinet, and based on the industrial computer, the energy storage cabinet control strategy is executed to start the remote control of the control device corresponding to the energy storage cabinet, including:

[0160] S2031: At the edge computing gateway, the energy storage cabinet control strategy is transmitted to the industrial computer of the corresponding energy storage cabinet;

[0161] S2032: At the industrial computer, the energy storage cabinet control strategy is converted into a remote control instruction of the corresponding control device, and the remote control instruction is sent to the corresponding control device;

[0162] For example, the strategy "power reduction operation" can be converted into an instruction "adjust the power setting value of the power conversion system to 50% of the current value";

[0163] In this embodiment, the industrial computer sends these remote control instructions to the corresponding control devices in the energy storage cabinet, such as the power conversion system, the battery management system, the air conditioning system, the fire relay, etc.

[0164] S2033: At the control device, the remote control instruction is executed to realize the remote control of the energy storage cabinet;

[0165] In this embodiment, after the control device receives the instruction, it immediately performs the corresponding action, such as adjusting the charge and discharge power, starting the fan, cutting off the circuit, etc., thereby completing the precise and remote control of the energy storage cabinet.

[0166] The embodiment of the present application also provides an energy storage cabinet remote control device based on the Internet of Things, which is described with reference to Figure 3 , shows a functional unit diagram of an energy storage cabinet remote control device 300 based on the Internet of Things of the present application, which can include the following units:

[0167] The model deployment unit 301 is used to use a federated learning mechanism to collaboratively train an energy storage cabinet abnormality identification model, an Internet of Things abnormal communication detection model and an energy storage cabinet control strategy generation model on a cloud server, and deploy them to an edge computing gateway.

[0168] The automatic function unit 302 is used for identifying the abnormality of the energy storage cabinet, detecting the abnormal communication of the Internet of Things, and generating the control strategy of the energy storage cabinet using the energy storage cabinet abnormality identification model, the Internet of Things abnormal communication detection model, and the energy storage cabinet control strategy generation model in the edge computing gateway, and uploading the obtained energy storage cabinet abnormality identification result, the Internet of Things abnormal communication detection result, and the energy storage cabinet control strategy to the cloud server through the state-instruction Internet of Things double-channel.

[0169] The remote control unit 303 is used for transmitting the energy storage cabinet control strategy to the industrial computer of the corresponding energy storage cabinet in the edge computing gateway, and executing the energy storage cabinet control strategy based on the industrial computer to start the control equipment corresponding to the energy storage cabinet to realize remote control.

[0170] Based on the same inventive concept, another embodiment of the present application provides an electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete communication with each other through the communication bus,

[0171] The memory is used for storing a computer program.

[0172] The processor is used for executing the program stored on the memory to realize the remote control method of the energy storage cabinet based on the Internet of Things.

[0173] The communication bus mentioned above can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used for communication between the terminal and other devices. The memory can include a random access memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. Optionally, the memory can also be at least one storage device located away from the aforementioned processor.

[0174] The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0175] In addition, to achieve the above object, the embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the method for remote control of the energy storage cabinet based on the Internet of Things.

[0176] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device, or computer program product. Therefore, the embodiments of the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can adopt a form of a computer program product implemented on one or more computer usable vehicles (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer usable program codes.

[0177] The embodiments of the present application are described with reference to flowcharts and / or block diagrams according to the method, terminal device (apparatus), and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal device produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device for implementing the functions specified in one flow or multiple flows and / or blocks.

[0178] These computer program instructions can also be stored in a computer readable memory capable of guiding a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one flow or multiple flows and / or blocks.Figure 1 the function specified in the one or more blocks.

[0179] These computer program instructions can also be loaded into computer or other programmable data processing terminal devices, so that a series of operation steps are performed on the computer or other programmable terminal devices to generate a computer-implemented process, so that the instructions executed on the computer or other programmable terminal devices provide a process for implementing the flowchart Figure 1 the flowchart or flowcharts and / or a block Figure 1 the steps of the function specified in the one or more blocks.

[0180] Finally, it should be noted that the relational terms herein, such as first and second, and the like, are used solely to distinguish one from another entity or action, without necessarily requiring or implying any actual relationship or order between such entities or actions. "And / or" means that either one of the two, or both, can be selected. Also, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, so that a process, method, article, or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or other elements inherent in such process, method, article, or terminal device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or terminal device including the element.

[0181] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A remote control method for an energy storage cabinet based on the Internet of Things, characterized in that, The method includes: On a cloud server, a federated learning mechanism is used to collaboratively train an energy storage cabinet anomaly identification model, an IoT anomaly communication detection model, and an energy storage cabinet control strategy generation model, and then deploys them to an edge computing gateway. At the edge computing gateway, an energy storage cabinet anomaly identification model, an IoT anomaly communication detection model, and an energy storage cabinet control strategy generation model are used to identify energy storage cabinet anomalies, detect IoT anomalies, and generate energy storage cabinet control strategies. The obtained energy storage cabinet anomaly identification results, IoT anomaly communication detection results, and energy storage cabinet control strategies are then uploaded to the cloud server through the status-command IoT dual channel. At the edge computing gateway, the energy storage cabinet control policy is transmitted to the industrial control computer of the corresponding energy storage cabinet, and based on the industrial control computer, the energy storage cabinet control policy is executed to start the corresponding control equipment of the energy storage cabinet to achieve remote control. On a cloud server, a federated learning mechanism is used to collaboratively train an energy storage cabinet anomaly identification model, an IoT anomaly communication detection model, and an energy storage cabinet control policy generation model, which are then deployed to an edge computing gateway, including: On a cloud server, artificial intelligence algorithms are used to build an initial model; the initial model includes an initial energy storage cabinet anomaly identification model, an initial Internet of Things anomaly communication detection model, and an initial energy storage cabinet control strategy generation model. The initial model metadata is distributed to several training servers within the federated learning scope, and the initial model is reconstructed on each training server. On the training server, the initial model is iteratively trained locally based on the first local dataset of the local energy storage cabinet in different scenarios to obtain the local model; the local model includes a local energy storage cabinet anomaly identification model, a local IoT anomaly communication detection model, and a local energy storage cabinet control strategy generation model. During each local iteration of training, the improved optimization algorithm is used to simultaneously optimize the hyperparameters of the local model, and the optimized hyperparameters are applied to the next local iteration of training. If the local iterative training ends or the first comprehensive loss value of the local model meets the requirements, the local model is output, the corresponding model parameter update amount and data amount are extracted, and uploaded to the cloud server. On the cloud server, based on the model parameter update volume and data volume of all training servers, a federated aggregation weight dynamic adjustment mechanism is used to safely aggregate and adjust the model parameters of the initial model to obtain a global model; the global model includes a global energy storage cabinet anomaly identification model, a global IoT abnormal communication detection model, and a global energy storage cabinet control strategy generation model. The second model metadata of the global model is distributed to all edge computing gateways within the remote control scope, and the global model is reconstructed on each edge computing gateway; At the edge computing gateway, the global model is transferred and fine-tuned based on the second local dataset of the local energy storage cabinet in the current scenario to obtain the final model; the final model includes the final energy storage cabinet anomaly identification model, the final IoT abnormal communication detection model, and the final energy storage cabinet control strategy generation model. If the migration fine-tuning is completed or the second comprehensive loss value of the final model meets the requirements, the final model deployment in the edge computing gateway is completed. The improved optimization algorithm is the IFPA algorithm; During each local training iteration, the IFPA algorithm is used to synchronously optimize the hyperparameters of the local model, and the optimized hyperparameters are applied to the next local training iteration, including: Based on the hyperparameter synchronization optimization problem, a multi-objective optimization function is set and used as the fitness function of the IFPA algorithm; Encode the hyperparameters of the local model into individual vectors for the IFPA algorithm, and set the IFPA population parameters and the maximum number of iterations; Based on the IFPA population parameters, the initial IFPA population is obtained by initializing using the Tent chaotic mapping sequence; each IFPA individual in the IFPA population corresponds to a candidate adapter parameter. Based on the first local dataset, the fitness function is used to obtain the fitness value of each initial IFPA individual, and the initial IFPA individual with the best fitness value is taken as the optimal solution. Transition probability q By selecting cross-pollination and self-pollination, the initial IFPA population is iteratively updated to obtain an updated IFPA population. Introducing a dynamic reversal mechanism, using reversal probability p The updated IFPA individuals are dynamically reversed to obtain the reversed IFPA population. Based on the first local dataset, the fitness function is used to obtain the fitness value of each updated IFPA individual and the reversed IFPA individual, and the updated IFPA individual or the reversed IFPA individual with the best fitness value is updated as the optimal solution. When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, the iterative update of the IFPA population is terminated, and the optimal solution of the current iteration is output. The individual vectors of the IFPA individuals corresponding to the optimal solution are decoded to obtain the optimal hyperparameters, and the optimized hyperparameters are used as the initial hyperparameters for the next local iteration training.

2. The remote control method for energy storage cabinet based on the Internet of Things according to claim 1, characterized in that, On the cloud server, based on the model parameter update volume and data volume of all training servers, a federated aggregation weight dynamic adjustment mechanism is used to safely aggregate and adjust the model parameters of the initial model, resulting in a global model, including: On the cloud server, a federated aggregation weight dynamic adjustment mechanism is used to generate corresponding federated aggregation weights based on the amount of data on each training server. Based on the federated aggregation weights of all training servers and the corresponding model parameter update amounts, the model parameters of the initial model are safely aggregated to obtain the global model parameters. The initial model is adjusted based on the global model parameters to obtain the global model.

3. The remote control method for energy storage cabinet based on the Internet of Things according to claim 2, characterized in that, The state-command IoT dual channel includes a critical state and command IoT channel and a regular state and data IoT channel. The critical state and command IoT channel is equipped with a first IoT communication protocol with high reliability and low latency, while the regular state and data IoT channel is equipped with a second IoT communication protocol with low overhead and high efficiency.

4. The remote control method for energy storage cabinet based on the Internet of Things according to claim 3, characterized in that, At the edge computing gateway, an energy storage cabinet anomaly identification model, an IoT anomaly communication detection model, and an energy storage cabinet control strategy generation model are used to perform energy storage cabinet anomaly identification, IoT anomaly communication detection, and energy storage cabinet control strategy generation. The obtained energy storage cabinet anomaly identification results, IoT anomaly communication detection results, and energy storage cabinet control strategies are then uploaded to the cloud server via a dual-channel status-command IoT system, including: The energy storage cabinet monitoring device collects multi-dimensional monitoring data of the energy storage cabinet and uploads the multi-dimensional monitoring data to the edge computing gateway within the communication range through the underlying local area network. At the edge computing gateway, multi-dimensional monitoring data is input into the energy storage cabinet anomaly identification model to identify energy storage cabinet anomalies and obtain the energy storage cabinet anomaly identification results. The abnormal identification results of the energy storage cabinet are input into the energy storage cabinet control strategy generation model to generate the energy storage cabinet control strategy. At the edge computing gateway, communication data of the IoT communication channel between the gateway and the cloud server is captured and input into the IoT abnormal communication detection system to perform IoT abnormal communication detection and obtain the IoT abnormal communication detection results. If the energy storage cabinet anomaly identification result indicates that there is an energy storage cabinet anomaly, or the IoT anomaly communication detection result indicates that there is an IoT anomaly communication, then switch to the critical state and command IoT channel in the state-command IoT dual channel; otherwise, maintain the default state and data IoT channel in the state-command IoT dual channel. Multidimensional monitoring data, energy storage cabinet anomaly identification results, IoT anomaly communication detection results, and energy storage cabinet control strategies are uploaded to the cloud server through the key status and command IoT channel or the normal status and data IoT channel in the status-command IoT dual channel.

5. The remote control method for energy storage cabinet based on the Internet of Things according to claim 4, characterized in that, The energy storage cabinet anomaly identification model is constructed based on the Convolutional Neural Network (CNN), Long Short-Term Memory Network (LSTM), Attention, Multilayer Perceptron (MLP) algorithm. The IoT abnormal communication detection model is constructed based on the autoencoder AE algorithm; The energy storage cabinet control strategy generation model is constructed based on the multi-agent deep deterministic policy gradient (MADDPG) algorithm.

6. The remote control method for energy storage cabinet based on the Internet of Things according to claim 5, characterized in that, At the edge computing gateway, the energy storage cabinet control policy is transmitted to the corresponding industrial control computer of the energy storage cabinet. Based on the industrial control computer, the energy storage cabinet control policy is executed, and the corresponding control equipment of the energy storage cabinet is started to achieve remote control, including: At the edge computing gateway, the control strategy for the energy storage cabinet is transmitted to the industrial control computer of the corresponding energy storage cabinet; In the industrial control computer, the energy storage cabinet control strategy is converted into remote control commands for the corresponding control devices, and the remote control commands are sent to the corresponding control devices. The control equipment executes remote control commands to achieve remote control of the energy storage cabinet.

7. A remote control device for an energy storage cabinet based on the Internet of Things, used to implement the remote control method for an energy storage cabinet as described in any one of claims 1-6, characterized in that, The device includes: The model deployment unit is used to collaboratively train the energy storage cabinet anomaly identification model, the IoT anomaly communication detection model, and the energy storage cabinet control strategy generation model on a cloud server using a federated learning mechanism, and then deploy them to the edge computing gateway. The automation function unit is used at the edge computing gateway to perform energy storage cabinet anomaly identification, IoT anomaly communication detection, and energy storage cabinet control strategy generation by using the energy storage cabinet anomaly identification model, IoT anomaly communication detection model, and energy storage cabinet control strategy generation model. The obtained energy storage cabinet anomaly identification results, IoT anomaly communication detection results, and energy storage cabinet control strategies are then uploaded to the cloud server through the status-command IoT dual channel. The remote control unit is used at the edge computing gateway to transmit the energy storage cabinet control strategy to the corresponding industrial control computer of the energy storage cabinet, and based on the industrial control computer, execute the energy storage cabinet control strategy and start the corresponding control equipment of the energy storage cabinet to achieve remote control.

Images