A power distribution room multi-device integrated control system and method based on an industrial internet of things architecture

Abstract

The application relates to the technical field of integrated control of power distribution room equipment, and discloses a power distribution room multi-equipment integrated control system and method based on an industrial Internet of Things architecture, the method comprising: collecting operation data of power distribution equipment and performing regularization, dividing into slowly varying operation data and rapidly varying operation data, using an improved LSTM network and a sliding differential mode combined with an attention mechanism to respectively extract the time sequence characteristics of the two; using an improved weighted polling mode to sequentially receive equipment characteristic information of the power distribution equipment, performing mapping processing on the equipment characteristic information based on fuzzy logic rules, and generating control parameters. The application deploys an edge collection module, a protocol conversion module and an edge gateway to construct an industrial Internet of Things architecture, realizes classified collection of rapidly varying and slowly varying operation data, combines an improved LSTM with a sliding differential mode combined with an attention mechanism to respectively extract time sequence characteristics, maps control parameters through fuzzy logic rules, and realizes unified scheduling and intelligent control of multiple devices.

Description

Technical Field

[0001] This invention relates to an Internet of Things (IoT) control system for industrial big data, and more particularly to the technical field of integrated control of power distribution room equipment. Specifically, it relates to a multi-device integrated control system and method for power distribution rooms based on an industrial IoT architecture. Background Technology

[0002] With the continuous improvement of energy digitalization and intelligence, substations, as key nodes in the power system, are increasingly equipped with a wide variety of devices, including smart meters, protection and control devices, circuit breakers, power quality analyzers, fault recorders, and environmental monitoring devices. The growing demand for collaborative operation of multiple devices places higher demands on substations in terms of monitoring, protection, control, and information exchange. To improve operational efficiency, ensure power quality, and guarantee operational safety, there is an urgent need to construct a highly efficient and intelligent multi-device integrated control system.

[0003] The development of the Industrial Internet of Things (IIoT) has provided crucial support for the intelligentization of power distribution rooms. Through devices such as edge acquisition modules, protocol conversion modules, and edge gateways, real-time acquisition, protocol parsing, and standardized data processing of various power distribution equipment can be achieved, providing a data foundation for subsequent intelligent control. However, efficiently integrating operational data with different sampling granularities (milliseconds, minutes) to construct a unified data processing and control logic remains a significant technical challenge.

[0004] Some research has already attempted integrated control of large-scale equipment. For example, CN111651332B proposes a large-scale heterogeneous equipment integrated control method based on message middleware. This method abstracts a unified device access model and utilizes a publish / subscribe mechanism to achieve attribute read / write control of devices. This method is applicable to general industrial control systems and has certain advantages in improving the flexibility of equipment integration and the scalability of the control system. However, it still has the following limitations: it does not differentiate between fast-changing and slow-changing operating data in power distribution equipment, which can easily lead to the loss of critical data or response delays; the control strategy lacks time sensitivity and cannot dynamically adjust the access frequency and control methods according to the operating characteristics of different devices.

[0005] To address this issue, this invention proposes a multi-device integrated control system and method for power distribution rooms based on an industrial Internet of Things (IoT) architecture. It constructs a unified industrial IoT architecture, connects data interfaces between multi-source heterogeneous devices, forms a structured access system, and realizes integrated access and interconnection of multiple devices, as well as adjustment of control parameters. Summary of the Invention

[0006] In view of this, the present invention provides a multi-device integrated control system and method for power distribution rooms based on an industrial Internet of Things (IIoT) architecture. Due to the diverse types of equipment and varying communication protocols in power distribution rooms, information collection standards are inconsistent, data interoperability is difficult, and efficient integrated control and unified management are challenging. This invention deploys edge acquisition modules, protocol conversion modules, and edge gateways in the power distribution room to construct a unified industrial Internet of Things architecture, opening up data interfaces between multi-source heterogeneous devices and forming a structured access system, enabling integrated access and interconnection of multiple devices. The operating data of power distribution equipment includes two feature types: slow-changing and fast-changing. Traditional acquisition methods often mix these, leading to the masking or failure of important state features in analysis. This application introduces an improved LSTM network and a sliding difference method combined with an attention mechanism to extract refined temporal features from both slow-changing and fast-changing operating data, significantly improving the accuracy and semantic expression capability of device feature modeling, thereby enhancing the foundation for state recognition and dynamic control. Furthermore, by improving the weighted polling mechanism and fuzzy logic rule mapping method, intelligent fusion processing of the feature information of each device is performed to dynamically generate highly adaptable and real-time responsive control parameters, achieving collaborative control and efficient scheduling of multiple devices in the power distribution room.

[0007] To achieve the above objectives, the present invention provides a multi-device integrated control method for a power distribution room based on an industrial Internet of Things (IoT) architecture, comprising the following steps:

[0008] S1: Deploy edge acquisition modules, protocol conversion modules, and edge gateways in the power distribution room. Power distribution equipment can access the multi-device integrated control system of the power distribution room through the edge acquisition modules.

[0009] The edge acquisition module, protocol conversion module, edge gateway, and connected power distribution equipment are considered as industrial IoT in the power distribution room scenario;

[0010] S2: Use the edge acquisition module to collect the operating data of the power distribution equipment, and use the protocol conversion module to normalize the collected operating data, and divide the normalized operating data into slow change operating data and fast change operating data;

[0011] S3: The improved LSTM network and the sliding difference method combined with the attention mechanism are used to extract the time series features of slow-changing operation data and fast-changing operation data respectively, which are used as the equipment feature information of the power distribution equipment associated with the slow-changing operation data and fast-changing operation data.

[0012] S4: The improved weighted polling method is used to receive the equipment feature information of the power distribution equipment in sequence, and the equipment feature information is processed by mapping based on fuzzy logic rules to generate the control parameters of the power distribution equipment, and to perform integrated control of multiple power distribution equipment in the power distribution room.

[0013] As a further improvement of the present invention:

[0014] Optionally, the edge acquisition module is installed inside the power distribution equipment, the protocol conversion module and the edge gateway are integrated into the control cabinet, and the control cabinet is deployed in the power distribution room, including:

[0015] The power distribution equipment in the power distribution room is registered by connecting to the multi-device integrated control system of the power distribution room through the edge acquisition module. After successful registration, the multi-device integrated control system of the power distribution room sends a unique device ID to the control cabinet.

[0016] After receiving the unique ID of the device, the edge gateway in the control cabinet calls the device identification service to automatically match the device model, operating data sampling frequency, proprietary communication protocol and data sampling template of the device received by the edge gateway. It then sends the proprietary communication protocol to the protocol conversion module and the operating data sampling frequency and data sampling template to the edge acquisition module.

[0017] Optionally, the edge acquisition module is used to collect operating data of the power distribution equipment, including:

[0018] The edge acquisition module collects the operating data of the power distribution equipment according to the operating data sampling frequency, and preprocesses the operating data, including data caching, filtering and noise reduction. The preprocessed operating data is sent to the edge gateway, and the edge gateway transmits the preprocessed operating data to the protocol conversion module. The operating data includes the unique ID of the power distribution equipment.

[0019] The protocol conversion module identifies the power distribution equipment according to its unique ID, obtains the proprietary communication protocol and data frame structure of the power distribution equipment, and extracts key fields from the preprocessed running data. The key fields include register number, value and status bit. The standard sequential structure mapping unit organizes the key fields into a unified data format as the organized running data. The data packet encapsulation unit encapsulates the organized running data into a data packet and transmits it to the multi-device integrated control system of the power distribution room through the edge gateway.

[0020] The integrated control system for multiple devices in the power distribution room parses data packets to obtain normalized operating data.

[0021] Optionally, the normalized operating data is divided into slow-changing operating data and fast-changing operating data according to the sampling frequency of the operating data. The sampling frequency range of the slow-changing operating data is below 2 Hz, and the sampling frequency range of the fast-changing operating data is not less than 2 Hz. The data lengths of the slow-changing operating data and the fast-changing operating data are the same.

[0022] Optionally, the temporal features of slow-changing and fast-changing running data are extracted using an improved LSTM network and a sliding difference method incorporating an attention mechanism, respectively, including:

[0023] Acquire slow-changing operational data x and fast-changing operational data y. Use an improved LSTM network to extract the temporal features of the slow-changing operational data x. The improved LSTM network employs a multi-gated residual connection. The temporal feature extraction process includes:

[0024] Calculate the current LSTM hidden state at any time step n in slowly varying running data x using an LSTM network. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation ,in N represents the length of the slowly changing running data. This represents the nth data value in the slowly changing running data x. By nonlinearly fusing the current LSTM hidden state and the input data, the joint representation of the two is extracted, enhancing the discriminative power of temporal features. The current LSTM hidden state... It reflects the trend of slow-changing running data x at time step t;

[0025] The current LSTM hidden state is calculated using a multi-gating mechanism. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation The gating weights, for the current LSTM hidden state Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Perform weighted fusion to obtain the current LSTM hidden state. The multi-gated residual connection state, and its relationship to the current LSTM hidden state. Update:

[0026] ;

[0027] in, These represent the current LSTM hidden states, respectively. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Gating weights;

[0028] Indicates the current LSTM hidden state The multi-gated residual connection state, and the current LSTM hidden state. Updated to multi-gated residual connection state. Indicates the current LSTM hidden state The updated results;

[0029] Extract the current LSTM hidden state after each time step update in the slowly changing running data x to form the temporal features of the slowly changing running data x.

[0030] Optionally, the temporal features of the rapidly changing running data y are extracted using the sliding difference method combined with the attention mechanism, including:

[0031] Performing a sliding window difference on the rapidly varying operational data y yields a difference sequence of y, where the difference sequence is represented as follows: ,in k represents the length of the sliding window. ,in This represents the i-th data value in the rapidly changing running data y. This represents the first element in the rapidly changing running data y. One data value;

[0032] Represents the difference sequence The difference value in;

[0033] A nonlinear transformation method is used to perform feature mapping on the difference values ​​in the difference sequence to obtain the nonlinear transformation characteristics of the difference values;

[0034] The time attention of the nonlinear transformation features is calculated, and the nonlinear transformation features of the difference values ​​are weighted according to the time attention to obtain the time series features of the fast-changing running data y:

[0035] ;

[0036] ;

[0037] in, This represents the time-series characteristics of rapidly changing running data y. Represents the difference value Temporal attention, where T denotes transpose and v denotes a learnable mapping coefficient vector used to adjust the length of temporal features. Represents the learnable attention weights. This represents the learnable attentional bias. Represents the difference value The nonlinear transformation characteristics.

[0038] Optionally, an improved weighted polling method is used to sequentially receive equipment characteristic information of the power distribution equipment, including:

[0039] Extract the normalized operating data of the power distribution equipment, and calculate the operational risk level of the normalized operating data. The higher the operational risk level, the more abnormal the operating status of the power distribution equipment. The calculation method for the operational risk level is as follows:

[0040] ;

[0041] in, This indicates the level of operational risk of the normalized operational data. This represents the nth data value in the normalized running data. , This represents the mean of the normalized operating data. This represents the standard deviation of the normalized operating data. This represents the j-th data value in the normalized running data. This represents the (j-1)th data value in the normalized running data;

[0042] Based on the level of operational risk, calculate the number of times the equipment characteristic information of the power distribution equipment is received within a scheduling cycle. :

[0043] ;

[0044] in, This indicates the preset minimum number of times. Indicates the preset maximum number of times, set. =1, It is 5. This represents the sum of the operational risk levels of all power distribution equipment.

[0045] Optionally, the equipment feature information is mapped based on fuzzy logic rules to generate control parameters for the power distribution equipment, including:

[0046] The mean and standard deviation of the equipment feature information are extracted and input into the trigonometric membership function and the Gaussian membership function, respectively. The trigonometric membership function is used to output the membership degree of the equipment feature information in low, medium and high operating states. The operating state with the highest membership degree is extracted. The operating states include high operating state, medium operating state and low operating state. The Gaussian membership function is used to output the fluctuation membership degree value of the equipment feature information. The higher the fluctuation membership degree value, the smaller the operating fluctuation of the equipment feature information. The clustering algorithm is used to cluster the fluctuation membership degree values ​​of the received equipment feature information. The real-time received equipment feature information is divided into the closest clusters according to the fluctuation membership degree value. The clusters include high fluctuation, medium fluctuation and low fluctuation.

[0047] The operating state with the highest membership degree and its corresponding cluster are extracted according to fuzzy logic rules to extract corresponding control parameters, which are used as control parameters for the power distribution equipment to control and adjust the operating state of the power distribution equipment. The fuzzy logic rules are the correspondence rules between different types of operating states, clusters and control parameters.

[0048] To address the aforementioned problems, this invention provides a multi-device integrated control system for power distribution rooms based on an industrial Internet of Things (IIoT) architecture. This control system includes an auxiliary module, a timing feature extraction module, and a control module.

[0049] The auxiliary module includes a data acquisition module, a protocol conversion module, and an edge gateway. The power distribution equipment is connected to the multi-device integrated control system of the power distribution room through the edge acquisition module. The edge acquisition module collects the operating data of the power distribution equipment, and the protocol conversion module normalizes the collected operating data. The normalized operating data is then uploaded to the multi-device integrated control system of the power distribution room through the edge gateway.

[0050] The time-series feature extraction module is used to extract the time-series features of slow-changing operation data and fast-changing operation data respectively using an improved LSTM network and a sliding difference method combined with an attention mechanism, as equipment feature information of the associated power distribution equipment.

[0051] The control module is used to sequentially receive equipment feature information of power distribution equipment using an improved weighted polling method, perform mapping processing on the equipment feature information based on fuzzy logic rules, generate control parameters for the power distribution equipment, and perform integrated control of multiple power distribution equipment in the power distribution room.

[0052] This is to achieve the above-described method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things (IoT) architecture.

[0053] To address the above problems, the present invention provides an electronic device, the electronic device comprising:

[0054] Memory, storing at least one instruction;

[0055] Communication interfaces enable communication between electronic devices; and

[0056] The processor executes the instructions stored in the memory to implement the above-described method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture.

[0057] To address the aforementioned issues, the present invention also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the aforementioned multi-device integrated control method for a power distribution room based on an industrial Internet of Things architecture.

[0058] Compared with existing technologies, this invention proposes a multi-device integrated control system and method for power distribution rooms based on an industrial Internet of Things (IoT) architecture. This technology has the following beneficial effects:

[0059] Firstly, in the extraction of time-series features from slowly varying operational data, traditional gated residual connections typically use only a single gate weight to balance the current LSTM hidden state and the current input data. This fails to flexibly capture the simultaneous presence of multiple feature components (such as trends, fluctuations, and noise) in complex time-series data. When it is necessary to utilize hidden state, input, historical state, or composite transformation information simultaneously, a single gate weight cannot guarantee the full utilization of various types of information, potentially leading to the neglect or weakening of important information. Multi-gated weights support parallel fusion of multiple information streams, flexibly capturing multi-dimensional features such as trends, short-term fluctuations, and historical dependencies, enriching the feature representation space. The information streams include the current LSTM hidden state, the current input data, the LSTM hidden state from the previous time step, and a nonlinear mapping combination representation. Multi-gated weights are adaptively generated based on the context, dynamically adjusting the fusion ratio for different time steps and input data, improving adaptability and generalization ability. At different operating times, the characteristics of power distribution equipment... The importance of features can change, and multi-gated weights can adapt to such changes. For example, during transient faults, the current input data becomes more critical. During stable operation, the LSTM hidden state of the previous time step and the current LSTM hidden state, which represent the trend of change, become more critical. Under complex coupled conditions, nonlinear mapping combination representations can capture more detailed features. At the same time, gating mechanisms can reduce the weights of noise, anomalies, or irrelevant information flow, reduce their interference on the model output, and improve robustness. For example, when abnormal input signals or sudden interference are caused by occasional sensor failures, gating weights will automatically reduce their dependence on abnormal inputs to prevent misjudgment by the control strategy. They can also flexibly adjust the influence ratio of the current LSTM hidden state and the LSTM hidden state of the previous time step to strengthen the balance between long-term dependence and short-term dynamics. For example, in a power distribution room, the state of power distribution equipment is often affected by the past. Strengthening historical dependence through gating helps to accurately capture slowly changing features such as equipment aging and gradual load changes.

[0060] Meanwhile, in the process of extracting time-series features from rapidly changing operational data, the sliding window differential operation effectively amplifies the local change information in the data, especially the instantaneous signal jumps caused by sudden anomalies, enhancing the sensitivity to anomaly patterns and preventing the stationary components of the data from masking anomaly detection. Furthermore, the feature encoding through nonlinear mapping improves the expressive power of the difference values ​​in the difference sequence, capturing richer temporal nonlinear features and helping to reveal complex dynamic changes. Crucially, the time attention mechanism, through adaptive weight allocation, focuses on key anomaly moments in the time series, strengthening the ability to identify sudden events and avoiding the dilution of important information caused by averaging. In addition, the fixed-dimensional aggregated features output by this method are concise and expressive, facilitating subsequent control strategy formulation or anomaly identification. Overall, this approach combines local change capture with global temporal attention, significantly improving the detection and response efficiency for millisecond-level sudden anomalies in power distribution rooms. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a multi-device integrated control method for a power distribution room based on an industrial Internet of Things (IoT) architecture, as provided in an embodiment of the present invention.

[0062] Figure 2 This is an industrial Internet of Things (IoT) architecture diagram provided in one embodiment of the present invention.

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0065] This application provides a multi-device integrated control system and method for power distribution rooms based on an Industrial Internet of Things (IIoT) architecture. The execution entity of the multi-device integrated control method for power distribution rooms based on the IIoT architecture includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the multi-device integrated control method for power distribution rooms based on the IIoT architecture can be executed by software or hardware installed on terminal devices or server devices, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0066] Reference Figure 1 as well as Figure 2 Embodiment 1 of the present invention is as follows:

[0067] A method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things (IoT) architecture includes the following steps:

[0068] S1: Deploy edge acquisition modules, protocol conversion modules, and edge gateways in the power distribution room. Power distribution equipment can access the multi-device integrated control system of the power distribution room through the edge acquisition modules.

[0069] The edge acquisition module is installed inside the power distribution equipment, and the protocol conversion module and edge gateway are integrated into the control cabinet, which is deployed in the power distribution room.

[0070] The edge acquisition module includes a multi-channel signal acquisition interface, a multi-source sensor, a communication interface, and a sampling buffer. The multi-channel signal acquisition interface is used to connect to the multi-source sensor to collect the operating data of the power distribution equipment. The communication interface supports multiple communication protocols and is used to send the registration information and operating data of the power distribution equipment to the multi-device integrated control system of the power distribution room and the edge gateway in the control cabinet, respectively. The sampling buffer is used to perform data caching, filtering, and noise reduction processing on the collected operating data.

[0071] The protocol conversion module includes a multi-protocol parsing engine, a standard data structure mapping unit, and a data packet encapsulation unit. The multi-protocol parsing engine is used to parse the proprietary communication protocols and data frame structures of various power distribution equipment. The standard sequential structure mapping unit is used to regularize the operating data of the power distribution equipment into the same data format. The data packet encapsulation unit is used to encapsulate the regularized operating data into data packets and transmit them to the multi-device integrated control system of the power distribution room through the edge gateway.

[0072] The edge gateway includes a caching unit and a communication module. The caching unit is used to compress and cache the data packets encapsulated by the protocol conversion module, and the communication module is used to receive the operating data of the power distribution equipment and transmit the data packets encapsulated by the protocol conversion module to the multi-device integrated control system of the power distribution room.

[0073] The power distribution equipment in the power distribution room is connected to the multi-device integrated control system of the power distribution room through the edge acquisition module. When the power distribution equipment is connected, it automatically sends registration information to the multi-device integrated control system of the power distribution room. After successful registration, the multi-device integrated control system of the power distribution room sends a unique ID of the equipment to the control cabinet. The registration information includes MAC address, equipment type, manufacturer and model.

[0074] After receiving the unique ID of the device, the edge gateway in the control cabinet calls the device identification service to automatically match the device model, operating data sampling frequency, proprietary communication protocol and data sampling template of the device received by the edge gateway. It then sends the proprietary communication protocol to the protocol conversion module and the operating data sampling frequency and data sampling template to the edge acquisition module.

[0075] like Figure 2 The diagram shown is an industrial Internet of Things (IoT) architecture provided in an embodiment of the present invention.

[0076] In this embodiment, the power distribution equipment includes smart meters, protection and control devices (such as microcomputer protection units), fault recorders, circuit breakers, environmental monitoring equipment, power quality analyzers, monitoring terminals, and communication equipment (such as edge gateways). The operating data of the smart meters includes voltage, current, active / reactive power, frequency, and harmonics. The operating data of the protection and control devices is tripping signals. The operating data of the fault recorders is voltage fluctuation data. The operating data of the circuit breakers is the circuit breaker's action state sequence. The operating data of the environmental monitoring equipment is the temperature and humidity data and smoke concentration data of the power distribution room. The operating data of the power quality analyzers is voltage deviation data and three-phase imbalance data. The monitoring terminal provides access control opening status data. The operating data of the communication equipment is link quality data.

[0077] It should be noted that all running data is in the form of a data sequence.

[0078] S2: Use the edge acquisition module to collect the operating data of the power distribution equipment, and use the protocol conversion module to normalize the collected operating data, dividing the normalized operating data into slow-change operating data and fast-change operating data.

[0079] The edge acquisition module is used to collect operational data from power distribution equipment, including:

[0080] The edge acquisition module collects the operating data of the power distribution equipment according to the operating data sampling frequency, and preprocesses the operating data, including data caching, filtering and noise reduction. The preprocessed operating data is sent to the edge gateway, and the edge gateway transmits the preprocessed operating data to the protocol conversion module. The operating data includes the unique ID of the power distribution equipment.

[0081] The protocol conversion module identifies the power distribution equipment according to its unique ID, obtains the proprietary communication protocol and data frame structure of the power distribution equipment, and extracts key fields from the preprocessed running data. The key fields include register number, value and status bit. The standard sequential structure mapping unit organizes the key fields into a unified data format as the organized running data. The data packet encapsulation unit encapsulates the organized running data into a data packet and transmits it to the multi-device integrated control system of the power distribution room through the edge gateway.

[0082] In this embodiment, the proprietary communication protocols of the power distribution equipment include Modbus, IEC104, IEC61850, DNP3, PROFIBUS, etc. For example, register number 0x0010 represents the active power register, register value 315.2 indicates that the current power is 315.2kW, and status bit 0x04 indicates that the power distribution equipment is in an abnormal state.

[0083] The integrated control system for multiple devices in the power distribution room parses data packets to obtain normalized operating data.

[0084] It should be noted that the data length of the normalized running data is consistent, therefore the higher the sampling frequency of the running data, the lower the data collection time.

[0085] Based on the sampling frequency of the operating data, the normalized operating data is divided into slow-changing operating data and fast-changing operating data. The sampling frequency range of the slow-changing operating data is below 2 Hz, and the sampling frequency range of the fast-changing operating data is not less than 2 Hz. The data length of the slow-changing operating data and the fast-changing operating data is the same.

[0086] In this embodiment, the fast-transformer operation data includes the normalized operation data of smart meters, protection and control devices (such as microcomputer protection units), fault recorders, and circuit breakers;

[0087] Slow-change operation data includes the standardized operation data of environmental monitoring equipment, power quality analyzers, monitoring terminals, and communication equipment.

[0088] S3: Utilize an improved LSTM network and a sliding differential method incorporating an attention mechanism to extract the temporal features of slow-changing and fast-changing operating data, respectively, as equipment feature information of the power distribution equipment associated with the slow-changing and fast-changing operating data.

[0089] Temporal features of slow-changing and fast-changing running data are extracted using an improved LSTM network and a sliding difference method incorporating an attention mechanism, including:

[0090] Acquire slow-changing operational data x and fast-changing operational data y. Use an improved LSTM network to extract the temporal features of the slow-changing operational data x. The improved LSTM network employs a multi-gated residual connection. The temporal feature extraction process includes:

[0091] Calculate the current LSTM hidden state at any time step n in slowly varying running data x using an LSTM network. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation ,in N represents the length of the slowly changing running data. This represents the nth data value in the slowly changing running data x. By nonlinearly fusing the current LSTM hidden state and the input data, the joint representation of the two is extracted, enhancing the discriminative power of temporal features. The current LSTM hidden state... It reflects the trend of slow-changing running data x at time step t;

[0092] Specifically, ,in This represents the activation function (the activation function is a non-linear function, such as the ReLU function). This represents the learnable weight matrix parameters. This represents the learnable bias parameters. This represents the sampling frequency of the slowly varying running data x.

[0093] It should be noted that the current LSTM hidden state This is a deep representation of the slowly changing operating data x at time step t, reflecting the load trend and operating pattern of the power distribution equipment. It includes the LSTM hidden state from the previous time step. Used to determine the current LSTM hidden state Perform time-series smoothing to enhance time dependence. Used to improve the ability to recognize complex patterns;

[0094] The current LSTM hidden state is calculated using a multi-gating mechanism. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation The gating weights, for the current LSTM hidden state Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Perform weighted fusion to obtain the current LSTM hidden state. The multi-gated residual connection state, and its relationship to the current LSTM hidden state. Update:

[0095] ;

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] in, These represent the current LSTM hidden states, respectively. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Gating weights;

[0101] Indicates the current LSTM hidden state The multi-gated residual connection state, and the current LSTM hidden state. Updated to multi-gated residual connection state. Indicates the current LSTM hidden state The updated results;

[0102] In this embodiment, This represents an exponential function with the natural constant as its base. These represent the current LSTM hidden state, the current input data, the LSTM hidden state at the previous time step, and the learnable gated scoring matrix represented by the nonlinear mapping combination, respectively.

[0103] Extract the updated LSTM hidden state at each time step in the slowly changing running data x to construct the temporal features of the slowly changing running data x. Specifically, the temporal features of the slowly changing running data x are:

[0104] ;

[0105] in, This represents the time-series characteristics of slowly varying data x. These represent the current LSTM hidden states of the slowly changing running data x after the updates at time steps 1 to N, respectively.

[0106] Extracting the temporal features of rapidly changing running data y using the aforementioned sliding difference method incorporating an attention mechanism includes:

[0107] Performing a sliding window difference on the rapidly varying operational data y yields a difference sequence of y, where the difference sequence is represented as follows: ,in k represents the length of the sliding window. ,in This represents the i-th data value in the rapidly changing running data y. This represents the first element in the rapidly changing running data y. One data value;

[0108] Represents the difference sequence The difference value in;

[0109] A nonlinear transformation method is used to perform feature mapping on the difference values ​​in the difference sequence to obtain the nonlinear transformation characteristics of the difference values;

[0110] The time attention of the nonlinear transformation features is calculated, and the nonlinear transformation features of the difference values ​​are weighted according to the time attention to obtain the time series features of the fast-changing running data y:

[0111] ;

[0112] ;

[0113] in, This represents the time-series characteristics of rapidly changing running data y. Represents the difference value The time attention is defined as T denotes transpose, and v denotes the learnable mapping coefficient vector. Represents the learnable attention weights. This represents the learnable attentional bias. Represents the difference value The nonlinear transformation features are extracted; specifically, the ReLU function is used to extract the nonlinear transformation features of the difference values. It should be noted that by restricting the rows and columns of the attention weights, the temporal features of both slowly changing and rapidly changing data are made into feature vectors of the same length.

[0114] S4: The improved weighted polling method is used to receive the equipment feature information of the power distribution equipment in sequence, and the equipment feature information is processed by mapping based on fuzzy logic rules to generate the control parameters of the power distribution equipment, and to perform integrated control of multiple power distribution equipment in the power distribution room.

[0115] An improved weighted polling method is used to sequentially receive equipment characteristic information of power distribution equipment, including:

[0116] Extract the normalized operating data of the power distribution equipment, and calculate the operational risk level of the normalized operating data. The higher the operational risk level, the more abnormal the operating status of the power distribution equipment. The calculation method for the operational risk level is as follows:

[0117] ;

[0118] in, This indicates the level of operational risk of the normalized operational data. This represents the nth data value in the normalized running data. , This represents the mean of the normalized operating data. This represents the standard deviation of the normalized operating data. This represents the j-th data value in the normalized running data. This represents the (j-1)th data value in the normalized running data;

[0119] It should be noted that, This is used to measure the standardized cumulative deviation of normalized operational data. The larger the standardized cumulative deviation, the greater the deviation of the data values ​​from the normal range. Used to reflect the instantaneous fluctuations of the normalized operating data;

[0120] Based on the level of operational risk, calculate the number of times the equipment characteristic information of the power distribution equipment is received within a scheduling cycle. :

[0121] ;

[0122] in, This indicates the preset minimum number of times. Indicates the preset maximum number of times, set. =1, It is 5. This represents the sum of the operational risk levels of all power distribution equipment; weighted polling of equipment characteristic information is performed based on the operational risk level to improve the real-time performance and sensitivity of control, and power distribution equipment with potential risks is prioritized for processing. The scheduling cycle is 20 minutes.

[0123] The equipment feature information is mapped using fuzzy logic rules to generate control parameters for the power distribution equipment, including:

[0124] The mean and standard deviation of the equipment feature information are extracted and input into the trigonometric membership function and the Gaussian membership function, respectively. The trigonometric membership function is used to output the membership degree of the equipment feature information in low, medium and high operating states. The operating state with the highest membership degree is extracted. The operating states include high operating state, medium operating state and low operating state. The Gaussian membership function is used to output the fluctuation membership degree value of the equipment feature information. The higher the fluctuation membership degree value, the smaller the operating fluctuation of the equipment feature information. The clustering algorithm is used to cluster the fluctuation membership degree values ​​of the received equipment feature information. The real-time received equipment feature information is divided into the closest clusters according to the fluctuation membership degree value. The clusters include high fluctuation, medium fluctuation and low fluctuation.

[0125] The operating state with the highest membership degree and its corresponding cluster are extracted according to fuzzy logic rules to extract the corresponding control parameters, which are then used as control parameters for the power distribution equipment to control and regulate the operating state of the power distribution equipment.

[0126] In this embodiment, the mean, as a central trend indicator representing the operating status of power distribution equipment, can accurately classify different operating levels such as high, medium, and low operating states using a triangular membership function, thereby enabling fine-grained classification of equipment load or operating intensity. The standard deviation reflects the fluctuation amplitude and stability of the data; a Gaussian membership function can be used to smoothly model the degree of fluctuation, capturing abnormal fluctuations or unstable states in the operation of power distribution equipment. The combination of these two aspects allows the multi-equipment integrated control system in the power distribution room to not only perceive the load status of power distribution equipment in real time but also dynamically monitor fluctuation risks, enhancing comprehensive control over the equipment's operating status.

[0127] The load of the power distribution equipment decreases sequentially from high operating state to medium operating state to low operating state;

[0128] It should be noted that, based on the generated control parameters and the optimal control parameters set by manual adjustment, a mean squared error loss function is constructed to train and optimize the learnable parameters.

[0129] Based on these two types of membership results, the constructed fuzzy logic rules can achieve flexible and practical control decisions. By using rules such as "if the mean is high and the fluctuation is large, then reduce the load" or "if the mean is medium and the fluctuation is low, then maintain the current operation," the fuzzy logic rules can automatically adjust the control strategy for different operating states, improving the accuracy and robustness of control parameter responses. This method effectively avoids the rigidity of traditional hard threshold judgments, adapts to the continuous changes in the operating state of power distribution equipment, enhances the adaptability and fault tolerance of the multi-device integrated control system in the power distribution room to complex operating environments, and significantly improves the intelligence level and safe and stable operation capability of the multi-device integrated control system in the power distribution room.

[0130] Specifically, examples of the fuzzy logic rules are as follows:

[0131]

[0132] Example 2:

[0133] An integrated control system for multiple devices in a power distribution room based on an industrial Internet of Things (IIoT) architecture, to implement the integrated control method for multiple devices in a power distribution room based on an IIoT architecture as described in Example 1, includes an auxiliary module, a timing feature extraction module, and a control module:

[0134] The auxiliary module includes a data acquisition module, a protocol conversion module, and an edge gateway. The power distribution equipment is connected to the multi-device integrated control system of the power distribution room through the edge acquisition module. The edge acquisition module collects the operating data of the power distribution equipment, and the protocol conversion module normalizes the collected operating data. The normalized operating data is then uploaded to the multi-device integrated control system of the power distribution room through the edge gateway.

[0135] The time-series feature extraction module is used to extract the time-series features of slow-changing operation data and fast-changing operation data respectively using an improved LSTM network and a sliding difference method combined with an attention mechanism, as equipment feature information of the associated power distribution equipment.

[0136] The control module is used to sequentially receive equipment feature information of power distribution equipment using an improved weighted polling method, perform mapping processing on the equipment feature information based on fuzzy logic rules, generate control parameters for the power distribution equipment, and perform integrated control of multiple power distribution equipment in the power distribution room.

[0137] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0138] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0139] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0140] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things (IoT) architecture, characterized in that, The method includes: S1: Deploy edge acquisition modules, protocol conversion modules, and edge gateways in the power distribution room. Power distribution equipment can access the multi-device integrated control system of the power distribution room through the edge acquisition modules. S2: Use the edge acquisition module to collect the operating data of the power distribution equipment, and use the protocol conversion module to normalize the collected operating data, and divide the normalized operating data into slow change operating data and fast change operating data; S3: The improved LSTM network and the sliding difference method combined with the attention mechanism are used to extract the time series features of slow-changing operation data and fast-changing operation data respectively, which are used as the equipment feature information of the power distribution equipment associated with the slow-changing operation data and fast-changing operation data. Acquire slow-changing operational data x and fast-changing operational data y. Use an improved LSTM network to extract the temporal features of the slow-changing operational data x. The improved LSTM network employs a multi-gated residual connection. The temporal feature extraction process includes: Calculate the current LSTM hidden state at any time step n in slowly varying running data x using an LSTM network. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation ,in N represents the length of the slowly changing running data. This represents the nth data value in the slowly changing running data x. By nonlinearly fusing the current LSTM hidden state and the input data, the joint representation of the two is extracted, enhancing the discriminative power of temporal features. The current LSTM hidden state... It reflects the trend of slow-changing running data x at time step t; The current LSTM hidden state is calculated using a multi-gating mechanism. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation The gating weights, for the current LSTM hidden state Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Perform weighted fusion to obtain the current LSTM hidden state. The multi-gated residual connection state, and its relationship to the current LSTM hidden state. Update: ； ； ； ； ； in, These represent the current LSTM hidden states, respectively. Current input data LSTM hidden state at the previous time step and nonlinear mapping combination representation Gating weights; Indicates the current LSTM hidden state The multi-gated residual connection state, and the current LSTM hidden state. Updated to multi-gated residual connection state. Indicates the current LSTM hidden state The updated results; This represents an exponential function with the natural constant as its base. These represent the current LSTM hidden state, the current input data, the LSTM hidden state at the previous time step, and the learnable gated scoring matrix represented by the nonlinear mapping combination, respectively. Extract the current LSTM hidden state after each time step update in the slowly changing running data x to form the temporal features of the slowly changing running data x; Extracting the temporal features of rapidly changing running data y using the aforementioned sliding difference method incorporating an attention mechanism includes: Performing a sliding window difference on the rapidly varying operational data y yields a difference sequence of y, where the difference sequence is represented as follows: ,in k represents the length of the sliding window. ,in This represents the i-th data value in the rapidly changing running data y. This represents the first element in the rapidly changing running data y. One data value; Represents the difference sequence The difference value in; A nonlinear transformation method is used to perform feature mapping on the difference values ​​in the difference sequence to obtain the nonlinear transformation characteristics of the difference values; The time attention of the nonlinear transformation features is calculated, and the nonlinear transformation features of the difference values ​​are weighted according to the time attention to obtain the time series features of the fast-changing running data y: ； ； in, This represents the time-series characteristics of rapidly changing running data y. Represents the difference value The time attention, where T denotes transpose and v denotes the learnable mapping coefficient vector. Represents the learnable attention weights. This represents the learnable attentional bias. Represents the difference value The nonlinear transformation features are extracted using the ReLU function; the nonlinear transformation features of the difference values ​​are extracted using the ReLU function; by restricting the rows and columns of the attention weights, the time-series features of slow-changing and fast-changing data are made into feature vectors of the same length. S4: The improved weighted polling method is used to receive the equipment feature information of the power distribution equipment in sequence, and the equipment feature information is processed by mapping based on fuzzy logic rules to generate the control parameters of the power distribution equipment, and to perform integrated control of multiple power distribution equipment in the power distribution room.

2. The method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in claim 1, characterized in that, The edge acquisition module is installed inside the power distribution equipment, and the protocol conversion module and edge gateway are integrated into the control cabinet, which is deployed in the power distribution room. The power distribution equipment in the power distribution room is registered by connecting to the multi-device integrated control system of the power distribution room through the edge acquisition module. After successful registration, the multi-device integrated control system of the power distribution room sends a unique device ID to the control cabinet. After receiving the unique ID of the device, the edge gateway in the control cabinet calls the device identification service to automatically match the device model, operating data sampling frequency, proprietary communication protocol and data sampling template of the device received by the edge gateway. It then sends the proprietary communication protocol to the protocol conversion module and the operating data sampling frequency and data sampling template to the edge acquisition module.

3. The method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in claim 2, characterized in that, The edge acquisition module is used to collect operational data from power distribution equipment, including: The edge acquisition module collects the operating data of the power distribution equipment according to the operating data sampling frequency, and preprocesses the operating data, including data caching, filtering and noise reduction. The preprocessed operating data is sent to the edge gateway, and the edge gateway transmits the preprocessed operating data to the protocol conversion module. The operating data includes the unique ID of the power distribution equipment. The protocol conversion module identifies the power distribution equipment according to its unique ID, obtains the proprietary communication protocol and data frame structure of the power distribution equipment, and extracts key fields from the preprocessed running data. The key fields include register number, value and status bit. The standard sequential structure mapping unit organizes the key fields into a unified data format as the organized running data. The data packet encapsulation unit encapsulates the organized running data into a data packet and transmits it to the multi-device integrated control system of the power distribution room through the edge gateway. The integrated control system for multiple devices in the power distribution room parses data packets to obtain normalized operating data.

4. The method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in claim 3, characterized in that, Based on the sampling frequency of the operating data, the normalized operating data is divided into slow-changing operating data and fast-changing operating data. The sampling frequency range of the slow-changing operating data is below 2 Hz, and the sampling frequency range of the fast-changing operating data is not less than 2 Hz. The data length of the slow-changing operating data and the fast-changing operating data is the same.

5. The method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in claim 1, characterized in that, An improved weighted polling method is used to sequentially receive equipment characteristic information of power distribution equipment, including: Extract the normalized operating data of the power distribution equipment, and calculate the operational risk level of the normalized operating data. The higher the operational risk level, the more abnormal the operating status of the power distribution equipment. The calculation method for the operational risk level is as follows: ； in, This indicates the level of operational risk of the normalized operational data. This represents the nth data value in the normalized running data. , This represents the mean of the normalized operating data. This represents the standard deviation of the normalized operating data. This represents the j-th data value in the normalized running data. This represents the (j-1)th data value in the normalized running data; Based on the level of operational risk, calculate the number of times the equipment characteristic information of the power distribution equipment is received within a scheduling cycle. : ； in, This indicates the preset minimum number of times. Indicates the preset maximum number of times, set. =1, It is 5. This represents the sum of the operational risk levels of all power distribution equipment.

6. The method for integrated control of multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in claim 5, characterized in that, The equipment feature information is mapped using fuzzy logic rules to generate control parameters for the power distribution equipment, including: The mean and standard deviation of the equipment feature information are extracted and input into the trigonometric membership function and the Gaussian membership function, respectively. The trigonometric membership function is used to output the membership degree of the equipment feature information in low, medium and high operating states. The operating state with the highest membership degree is extracted. The operating states include high operating state, medium operating state and low operating state. The Gaussian membership function is used to output the fluctuation membership degree value of the equipment feature information. The higher the fluctuation membership degree value, the smaller the operating fluctuation of the equipment feature information. The clustering algorithm is used to cluster the fluctuation membership degree values ​​of the received equipment feature information. The real-time received equipment feature information is divided into the closest clusters according to the fluctuation membership degree value. The clusters include high fluctuation, medium fluctuation and low fluctuation. The operating state with the highest membership degree and its corresponding cluster are extracted according to fuzzy logic rules to extract corresponding control parameters, which are used as control parameters for the power distribution equipment to control and adjust the operating state of the power distribution equipment. The fuzzy logic rules are the correspondence rules between different types of operating states, clusters and control parameters.

7. A multi-device integrated control system for a power distribution room based on an industrial Internet of Things (IoT) architecture, characterized in that, The integrated control system for power distribution rooms based on the industrial Internet of Things architecture includes an auxiliary module, a time-series feature extraction module, and a control module: The auxiliary module includes a data acquisition module, a protocol conversion module, and an edge gateway. The power distribution equipment is connected to the multi-device integrated control system of the power distribution room through the edge acquisition module. The edge acquisition module collects the operating data of the power distribution equipment, and the protocol conversion module normalizes the collected operating data. The normalized operating data is then uploaded to the multi-device integrated control system of the power distribution room through the edge gateway. The time-series feature extraction module is used to extract the time-series features of slow-changing operation data and fast-changing operation data respectively using an improved LSTM network and a sliding difference method combined with an attention mechanism, as equipment feature information of the associated power distribution equipment. The control module is used to sequentially receive equipment feature information of power distribution equipment using an improved weighted polling method, perform mapping processing on the equipment feature information based on fuzzy logic rules, generate control parameters for the power distribution equipment, and perform integrated control of multiple power distribution equipment in the power distribution room. To achieve the integrated control method for multiple devices in a power distribution room based on an industrial Internet of Things architecture as described in any one of claims 1-6.

Images