Site hierarchical standby power control method and device, electronic equipment and storage medium

Abstract

The application provides a site hierarchical standby power control method and device, electronic equipment and storage medium, and relates to the technical field of communication power management. The method comprises the following steps: monitoring real-time state information of a plurality of power connection terminals, standby power units and a plurality of load devices in a target site; when the standby power unit meets a preset standby power control trigger condition, an input feature vector is constructed based on the real-time state information and a plurality of load devices corresponding to a preset priority label; the input feature vector is input into a neural network model to generate a hierarchical standby power control strategy for the plurality of load devices; and the corresponding load device is sequentially powered off or powered on according to the hierarchical standby power control strategy. The application can realize dynamic and fine hierarchical control of the load device without relying on additional hardware, thereby maximizing the running time of the core business device in an emergency power supply scene and significantly improving the reliability and resource utilization efficiency of site power supply guarantee.

Description

Technical Field

[0001] This invention relates to the field of communication power management technology, and in particular to a site-level backup power control method, device, electronic device, and storage medium. Background Technology

[0002] With the continuous evolution and commercial deployment of mobile communication network technology, the scale of communication network sites is rapidly expanding, and the number and types of equipment integrated within a single site are constantly increasing, leading to a significant increase in overall site energy consumption. The simultaneous growth in the number of sites and power consumption per point places higher demands on the power supply system of communication infrastructure in terms of deployment flexibility, operational economy, and energy utilization efficiency. Currently, common site backup power (referred to as "backup power") control schemes mainly rely on newly added hardware control units to realize the switching and connection of power supply circuits. These schemes are usually manifested as centralized backup power or backup power based on limited branch circuits. Their control logic is mostly based on preset fixed rules (such as pre-setting the power-down sequence according to equipment type) or simple threshold judgments based on parameters such as voltage and current. When an external main power interruption is detected, the entire site or a preset portion of the load circuits are automatically switched to backup power supply according to the above fixed rules or threshold conditions.

[0003] However, the above solutions have the following limitations: 1) Strong hardware dependence, limited cost and flexibility: The implementation of the solution requires the deployment of additional dedicated hardware control units, which not only increases the cost of system transformation and physical space occupation, but also restricts the rapid deployment and subsequent flexible expansion capabilities of the solution; 2) Rigid control strategy and weak adaptive capability: The backup power control is based on logic of fixed rules or simple thresholds. Under limited backup energy, it is difficult to adaptively adjust the power-down strategy to maximize the battery life of core business equipment; 3) Lack of global coordination and intelligent decision-making capabilities: There is a lack of a global equipment coordination management perspective within the site, and it is impossible to generate the optimal coordination control strategy. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a site-level backup power control method, device, electronic device and storage medium.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a site-level backup power control method, which adopts the following technical solution: A site-level backup power control method includes: Monitor the real-time status information of multiple power supply terminals, backup power units, and multiple load devices within the target site; When the backup power unit meets the preset backup power control triggering conditions, an input feature vector is constructed based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices. The input feature vector is fed into a neural network model to generate a hierarchical backup power control strategy for the multiple load devices; According to the hierarchical backup power control strategy, power-down or power-up operations are performed on the corresponding load devices in sequence.

[0006] The beneficial effects of this invention are as follows: By real-time sensing, fusion, and intelligent analysis of the status of multi-source devices within a communication site, an optimized hierarchical backup power control strategy can be dynamically generated and executed. This method can intelligently and automatically identify and prioritize the continuous operation of core business equipment within the limited time of backup power supply, thereby maximizing the continuous service duration of critical services in emergency situations and significantly improving the reliability and service assurance level of network services. Simultaneously, this method achieves refined and adaptive management of existing backup energy within the site, maximizing resource utilization efficiency through pure software algorithms, and improving the overall economy and intelligent operation and maintenance capabilities of the site's power system.

[0007] Based on the above technical solution, the present invention can be further improved as follows.

[0008] Furthermore, the step of sequentially performing power-down or power-saving operations on the corresponding load devices according to the hierarchical backup power control strategy includes: The following steps are repeated until the backup power unit no longer meets the preset backup power control trigger condition or all the load devices have been processed: Among the load devices that have not yet been processed, determine the device category with the highest current power-down priority, and designate it as the current category to be processed; For the current pending category, the following steps are executed repeatedly until the backup power unit no longer meets the preset backup power control trigger condition or all load devices under the current pending category have been processed: Obtain the real-time operating current of the power supply terminals connected to each load device that belongs to the current processing category and has not yet been powered down; The load devices are sorted from high to low based on their real-time operating current. Perform a power-down operation or a power-on operation on the load device that is ranked first. After performing a power-down operation or a power-on operation, it is re-evaluated whether the backup power unit still meets the preset backup power control trigger conditions.

[0009] The beneficial effects of adopting the above-mentioned further solution are as follows: First, it ensures that equipment is processed layer by layer from high to low according to the predetermined level of business importance, which strategically prioritizes the continuity of core businesses. Within each priority level, it further prioritizes and operates based on the real-time energy consumption of each device, ensuring that each action targets the device with the highest energy consumption in the current state and is most effective in alleviating system pressure, thereby relieving the overload or low voltage crisis of the backup power unit with the fastest speed and the fewest number of operations. Crucially, each operation is followed by real-time system status feedback and reassessment, which makes the entire control process adaptive. Once the system pressure is effectively relieved, subsequent operations can be stopped immediately, avoiding unnecessary business interruptions caused by excessive power-down. Finally, this solution achieves globally optimal dynamic allocation of limited backup energy at the logical level, maximizing the continuous operation time of critical loads in a highly structured and deterministic manner in emergency power supply scenarios.

[0010] Furthermore, the following steps are used to determine whether the backup power unit meets the preset backup power control triggering conditions: If the real-time output voltage of the backup power unit is lower than a preset voltage threshold and the real-time output current of the backup power unit is higher than a preset current threshold, then the backup power unit is determined to meet the preset backup power control triggering conditions.

[0011] The beneficial effect of adopting the above-mentioned further solution is that by combining the two conditions of the backup power unit's output voltage being lower than a preset voltage threshold and its output current being higher than a preset current threshold, it ensures that the system is only triggered in the most critical situation where the backup energy is about to be exhausted and the load is too heavy. This effectively avoids unnecessary control actions when the battery power is still sufficient or the system load is light, thereby improving the overall control accuracy and reliability.

[0012] Furthermore, the neural network model is a Long Short-Term Memory (LSTM) network, which is trained through the following steps: Acquire historical status datasets of multiple power supply terminals, backup power units, and multiple load devices within the target site within the historical time window, as well as historical power outage event records; The historical state dataset and the historical power outage event records are preprocessed, including data cleaning, data normalization and feature engineering, to construct a labeled sample set; The labeled sample set is divided into a training set, a validation set, and a test set; The training set is input into the initial long short-term memory network model, and iterative training is performed with the goal of minimizing the cross-entropy loss between the probability distribution of the model output of performing power-down operation on the load device and the corresponding actual power-down operation label. During training, the validation set is used to adjust the model hyperparameters and monitor model performance to prevent overfitting. The trained Long Short-Term Memory (LSTM) network model is evaluated using the test set. When the prediction accuracy of the trained LSM network model on the test set for the load device to be powered down reaches or exceeds a preset threshold, the LSM network model is obtained.

[0013] The beneficial effects of adopting the above-mentioned further approach are as follows: by utilizing the site's own historical operational data and decision records for training, the model can learn the optimal control mode in specific scenarios. Preprocessing through data cleaning, normalization, and feature engineering ensures input quality, while the standardized division of training, validation, and test sets effectively supports model learning, optimization, and final evaluation. In particular, using cross-entropy loss as the optimization objective and independent test set prediction accuracy as the acceptance criterion allows the model's output probability distribution to approximate the true optimal power-off decision to the greatest extent possible, resulting in a Long Short-Term Memory (LSTM) network model that not only fits historical data but also makes highly accurate intelligent recommendations in new scenarios.

[0014] Furthermore, the acquisition of historical status datasets of multiple power connection terminals, backup power units, and multiple load devices within the target site during the historical time window, as well as historical power outage event records, includes: Obtain the real-time operating voltage sequence, real-time operating current sequence, unique identifier, and rated load capacity of each power supply terminal within the historical time window; Obtain the real-time output voltage sequence and real-time output current sequence of the backup power unit within the historical time window; Obtain the real-time operating voltage sequence, real-time operating current sequence, device identifier, device category, unique identifier of the power supply terminal connected to the device, and its preset priority label for each load device within the historical time window; Obtain the device identifier sequence of each load device that was actually powered down at a historical moment when the preset backup power control trigger condition was met; Obtain the status change information of the backup power unit after the load device corresponding to each device identifier in the device identifier sequence is powered down.

[0015] The beneficial effects of adopting the above-mentioned further scheme are as follows: This scheme not only clarifies the need to collect complete time-series status information of power supply terminals, backup power units, and load equipment within the historical time window, but also specifically defines key data for recording the actual power-down operation sequence and its subsequent effects during historical power outage events. By precisely defining the data type and structure, such as obtaining the unique identifier, connection relationship, rated capacity, priority label of each device, as well as the specific device sequence of power-down operations and the status changes of backup power units after execution, this scheme ensures that the training data can comprehensively and accurately reproduce the actual operating status and decision-making process of the site under emergency conditions.

[0016] Furthermore, the preprocessing of the historical state dataset and the historical power outage event records, including data cleaning, data normalization, and feature engineering, to construct a labeled sample set, includes: Data cleaning is performed on the historical state dataset and the historical power outage event records; The numerical features in the cleaned historical state dataset are normalized, and the cleaned preset priority labels are encoded. For each power supply terminal, a real-time load rate sequence is calculated based on its normalized real-time operating current sequence and the rated load capacity. For the backup power unit, its real-time remaining range time sequence is estimated based on its normalized real-time output voltage sequence, real-time output current sequence, and battery capacity model. The numerical features, real-time load rate sequence, real-time remaining battery life time sequence, and encoding of each preset priority label in the normalized historical state dataset are time-aligned and associated with the device identification sequence and each state change information to construct the labeled sample set.

[0017] The advantages of adopting the above-mentioned further approach are as follows: This approach not only performs basic data processing such as data cleaning, normalization, and encoding, ensuring the quality and consistency of the original data, but also, crucially, extracts advanced time-series features from the original monitoring data by calculating the real-time load rate sequence of the power supply terminals and the estimated remaining power supply time sequence of the backup power units. These features directly reflect the system load pressure and energy sustainability, greatly enriching the feature dimensions that the model can learn. Finally, the processed multi-dimensional feature sequences are precisely aligned and correlated with historical actual power-down operation sequences and their effects on the timeline, thereby systematically constructing a clearly labeled sample set with clear feature-label correspondences.

[0018] Furthermore, the step of constructing an input feature vector based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices includes: The numerical features in the real-time status information are normalized. The preset priority tags corresponding to each of the multiple load devices are encoded. The real-time load rate of each power supply terminal and the estimated remaining battery life of the backup power unit are calculated based on the numerical features in the normalized real-time status information. The numerical features in the normalized real-time status information, the real-time load rate of each power supply terminal, the estimated remaining battery life of the backup power unit, and the encoding of the priority label are fused to form the input feature vector at the current moment.

[0019] The beneficial effects of adopting the above-mentioned further scheme are as follows: This scheme clarifies that the raw state information (such as voltage and current) collected in real time must be normalized, and the equipment importance labels (priority) must be encoded and converted. This ensures that data from different dimensions and types can be interpreted fairly and consistently by the model. More importantly, the scheme does not simply transmit raw data, but requires, based on normalization, the real-time load rate of the power supply terminals and the estimated remaining battery life of the backup power unit—two key high-dimensional derived features—to be calculated in real time. Finally, by fusing the processed basic features, the calculated high-level features, and the encoded priority labels, a structured, multi-dimensional input feature vector can be dynamically constructed. This input feature vector not only comprehensively and quantitatively represents the current operating pressure and energy status of the site, but also incorporates the business value weights of the equipment, thus providing high-quality, high-information-density real-time input for the subsequent intelligent decision-making of the neural network model.

[0020] Secondly, the present invention provides a site-level backup power control device, which adopts the following technical solution: A site-level backup power control device, comprising: The monitoring module is used to monitor the real-time status information of multiple power supply terminals, backup power units, and multiple load devices within the target site; The construction module is used to construct an input feature vector based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices when the backup power unit meets the preset backup power control trigger conditions; The generation module is used to input the input feature vector into the neural network model to generate a hierarchical backup power control strategy for the multiple load devices; The control module is used to sequentially perform power-down or power-up operations on the corresponding load devices according to the hierarchical backup power control strategy.

[0021] Thirdly, the present invention provides an electronic device that adopts the following technical solution: An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the site-level backup power control method as described in any of the first aspects.

[0022] Fourthly, the present invention provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the site-level backup power control method as described in any of the first aspects.

[0023] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of the invention. Attached Figure Description

[0024] Figure 1 A flowchart illustrating the site-level backup power control method provided by the present invention; Figure 2 A schematic diagram of the site hierarchical backup power control device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0026] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the site-level backup power control method provided by the present invention. Figure 1 As shown, the method may include the following steps S101-S104.

[0027] S101. Monitor the real-time status information of multiple power supply terminals, backup power units, and multiple load devices within the target site.

[0028] Specifically, a power supply terminal refers to a port on the power distribution equipment within the target site that supplies power to the load equipment. For each power supply terminal, the following information is monitored: 1) Real-time operating voltage: the current output voltage value of the terminal; 2) Real-time operating current: the total current flowing through the terminal; 3) Rated load capacity: the maximum safe current carrying capacity of the terminal (usually a static attribute, obtained during initialization).

[0029] A backup power unit refers to the backup power source for a target site, typically a battery bank, which supplies power to the site equipment when the main power supply is interrupted. For backup power units, the following information is monitored: 1) Real-time output voltage: The voltage currently output by the backup power unit is a key indicator for judging its remaining power and health status; 2) Real-time output current: The total current currently output by the backup power unit reflects the real-time total load demand of the entire site.

[0030] A load device refers to a communication device connected to a power supply terminal and performing a specific service, such as a wireless device, transmission device, data device, or access device. For each load device, the following information is monitored: 1) Real-time operating voltage: the voltage on the device's power supply circuit; 2) Real-time operating current: the current of the device, which is a direct indicator of its load capacity; 3) Device identification; 4) Device category, such as wireless device, data device, access device, or transmission device; 5) Identification of the connected power supply terminal.

[0031] S102. When the backup power unit meets the preset backup power control trigger conditions, an input feature vector is constructed based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices.

[0032] Specifically, when the backup power unit is detected to meet the preset backup power control triggering conditions, it signifies that the target site has entered an emergency state powered by the backup power source. In one embodiment, the following steps are used to determine whether the backup power unit meets the preset backup power control triggering conditions: if the real-time output voltage of the backup power unit is lower than a preset voltage threshold and the real-time output current of the backup power unit is higher than a preset current threshold, then the backup power unit is determined to meet the preset backup power control triggering conditions.

[0033] If the real-time output voltage of the backup power unit is lower than the preset voltage threshold, it indicates that the backup power supply's energy level is approaching the warning line. When a backup power supply (such as a battery) is operating, its output voltage gradually decreases as the remaining power decreases. When the real-time output voltage drops below the preset voltage threshold, it indicates that its energy reserves are insufficient to sustain the current load operation for an extended period, posing a direct threat to the system's power supply reliability. Without timely intervention, the output voltage may drop further, ultimately causing all equipment to shut down abnormally due to undervoltage, resulting in a complete business interruption.

[0034] If the real-time output current of the backup power unit exceeds the preset current threshold, it indicates that the total system load is too heavy, exceeding the healthy power supply capacity of the current backup power source. The backup power source has a rated maximum continuous output current capability. When the real-time output current consistently exceeds the preset safe current threshold, it means that the total power consumption of all equipment within the current site is excessive. This will not only accelerate battery consumption and shorten the overall backup power time, but may also lead to overheating and damage to the backup power source (battery) due to prolonged overload, and even cause safety issues.

[0035] Only when both of the above conditions are met simultaneously can it be considered that the system is in the most critical state of "backup power is about to be exhausted and the load is too heavy". At this time, the limited remaining power is being consumed rapidly, the system is on the verge of collapse, and intelligent hierarchical control must be activated immediately. The basis for this decision is to construct an input feature vector that the neural network model can understand.

[0036] In one embodiment, step S102, which involves constructing an input feature vector based on real-time status information and preset priority tags corresponding to each of the multiple load devices, includes: normalizing the numerical features in the real-time status information; encoding the preset priority tags corresponding to each of the multiple load devices; calculating the real-time load rate of each power connection terminal and the estimated remaining runtime of the backup power unit based on the numerical features in the normalized real-time status information; and fusing the numerical features in the normalized real-time status information, the real-time load rate of each power connection terminal, the estimated remaining runtime of the backup power unit, and the encoding of the priority tags to form the input feature vector at the current moment.

[0037] Specifically, firstly, numerical features such as voltage and current are normalized, and preset priority labels are encoded (e.g., one-hot encoding). Simultaneously, key derived features are calculated based on the normalized data, such as the real-time load rate (real-time operating current / rated current) of each power supply terminal and the estimated remaining operating time of the backup power unit (calculated based on the real-time output voltage, real-time output current, and battery capacity model of the backup power unit). Finally, these processed numerical features, derived features, and preset priority label codes are fused and concatenated at a unified time point to generate an input feature vector that comprehensively and structurally represents the current operating status and equipment importance of the target site.

[0038] S103. Input the input feature vector into the neural network model to generate a hierarchical backup power control strategy for multiple load devices.

[0039] Specifically, the input feature vector is fed into a neural network model, which dynamically generates a hierarchical backup power control strategy to guide the execution of power-down or power-up operations on each load device. This strategy clarifies the operation sequence of each load device, that is, under the current power shortage, which devices should be prioritized for power-down and which devices should be prioritized for power-up, thereby maximizing the extension of the core critical business operation time with limited backup power resources.

[0040] Considering that backup power control is essentially a time-series decision problem, the current optimal hierarchical backup power control strategy is highly dependent on the dynamic change trends of the status of each power supply terminal, backup power unit and load equipment over a period of time. In order to effectively capture and learn this time-series dependency and pattern, in one embodiment, the neural network model is a long short-term memory network, which is trained through the following steps S1031-S1036.

[0041] S1031. Obtain historical status datasets of multiple power supply terminals, backup power units, and multiple load devices within the target site during the historical time window, as well as historical power outage event records.

[0042] Optionally, step S1031 includes: Obtain the real-time operating voltage sequence, real-time operating current sequence, unique identifier (such as “terminal A-01”), and rated load capacity (such as 63A) for each power supply terminal within a historical time window (e.g., the past 10 minutes). Obtain the real-time output voltage sequence and real-time output current sequence of the backup power unit within a historical time window (e.g., the past 10 minutes); Obtain the real-time operating voltage sequence, real-time operating current sequence, device identifier (e.g., “RRU-1”), device category (e.g., wireless device, transmission device, data device, access device), unique identifier of the power supply terminal connected to the device, and its preset priority label (e.g., the number label “1” indicates the highest power-down priority) for each load device within a historical time window (e.g., the past 10 minutes). Obtain the device identifier sequence of each load device that was actually powered down at a historical moment when the preset backup power control trigger conditions were met (i.e., a list of device identifiers arranged in chronological order, indicating the execution order of the power down operation). Obtain the status change information of the backup power unit after powering down the load device corresponding to each device identifier in the device identifier sequence (e.g., the output voltage recovery value and / or output current decrease value of the backup power unit after the power-down operation, which are used to evaluate the effect of the power-down operation).

[0043] After obtaining the aforementioned historical state dataset and historical power outage event records, these raw data need to be preprocessed to construct a normalized sample set that can be used for model training.

[0044] S1032. Perform preprocessing on the historical state dataset and historical power outage event records, including data cleaning, data normalization, and feature engineering, to construct a labeled sample set.

[0045] Optionally, step S1032 includes: Perform data cleaning on historical state datasets and historical power outage event records (e.g., handle outliers such as missing values ​​and instantaneous voltage / current spikes). The numerical features (such as voltage and current) in the cleaned historical state dataset are normalized (e.g., the voltage and current values ​​are scaled to the [0,1] range), and the cleaned preset priority labels (e.g., wireless device-priority 1, access device-priority 2, data device-priority 3, transmission device-priority 4) are encoded (e.g., one-hot encoding is used to convert them into binary vectors that the model can process). For each power supply terminal, a real-time load rate sequence is calculated based on its normalized real-time operating current sequence and rated load capacity (i.e., real-time load rate = real-time operating current / rated load capacity, used to measure the degree of terminal load). For the backup power unit, its real-time remaining range time sequence is estimated based on its normalized real-time output voltage sequence, real-time output current sequence, and battery capacity model (i.e., the discharge model established based on battery physical characteristics and historical data). The numerical features, real-time load rate sequence, real-time remaining battery life time sequence, and the encoding of each preset priority label in the normalized historical state dataset are time-aligned and correlated with the device identification sequence and each state change information to construct a labeled sample set. In this set, the feature data serves as the samples, and the device identification sequence and state change information serve as the labels.

[0046] S1033. Divide the labeled sample set into a training set, a validation set, and a test set.

[0047] For example, the labeled sample set is divided into a training set, a validation set, and a test set in a ratio of 70%:15%:15%.

[0048] S1034. Input the training set into the initial long short-term memory network model and perform iterative training with the goal of minimizing the cross-entropy loss between the probability distribution of the model output of performing power-down operation on the load device and the corresponding actual power-down operation label.

[0049] Specifically, an initial Long Short-Term Memory (LSTM) network model is constructed, which includes: Input layer: It receives a time-series feature sequence of a time window (such as the past 10 minutes) as input. The feature vector of each time point includes normalized state features, derived features, encoded priority labels, etc. This is the sequence data format that the LSTM model is good at handling, enabling it to capture the dynamic change trend of the system state. Hidden layer: Employs one or more LSTM units. LSTM, through its unique gating structure (forget gate, input gate, output gate), can effectively learn long-term dependencies in a sequence. This is crucial in this embodiment because whether to trigger a power-down and which device to power down depends not only on the current instantaneous state but also on the evolution patterns of the power supply, load, and battery state over a previous period. Fully connected layer: Performs a linear transformation on the feature vector output by the last LSTM unit (which contains contextual information of the entire sequence); Output layer (Softmax activation): The Softmax function is applied to the output of the fully connected layer, which is converted into a probability distribution vector. The length of the vector is equal to the total number of online load devices in the current batch of training samples. Each element in this vector represents the probability predicted by the model to perform a power-off operation on the corresponding load device at the next decision time. The sum of the probabilities is 1.

[0050] This step uses the cross-entropy loss function to measure the accuracy of the model's predictions, and uses this as the target to drive the update of the model parameters. The specific training process is as follows: Model prediction: For each input training sample (containing a temporal feature sequence), the model outputs a probability distribution vector. For example, if there are currently 3 online load devices A, B, and C, the model may output [0.1, 0.7, 0.2], indicating that it believes the probabilities of powering down load devices A, B, and C are 10%, 70%, and 20%, respectively.

[0051] Actual Labels: The labels of the training samples are derived from the device identifier sequence, which is processed into a one-hot encoded vector. If the historical record shows that device B was actually powered down at that time, the corresponding label vector is [0, 1, 0].

[0052] Loss calculation: The cross-entropy loss function calculates the difference between the probability distribution predicted by the model (P_predicted) and the one-hot distribution of the true label (P_true).

[0053] Optimizer: The Adam optimizer can adaptively adjust the learning rate.

[0054] The training set is fed into the LSTM model in batches for forward propagation, ultimately outputting the power-down probability distribution for each input sample. Using the cross-entropy loss function, the average loss between the probability distributions of all model outputs in that batch and the actual power-down label corresponding to the sample is calculated. The gradient of the loss function with respect to all trainable parameters of the model (including the weights, biases, and fully connected layer parameters of the LSTM units) is calculated. This process is automatically completed by deep learning frameworks (such as TensorFlow and PyTorch). Using the Adam optimizer, the parameters of the LSTM model are iteratively updated based on the calculated gradients, with the goal of minimizing the cross-entropy loss. This process is repeated multiple times (epochs). Through repeated iterations, the model parameters are continuously adjusted, making its output probability distribution increasingly closer to historical, effective optimal decisions. Finally, the trained LSTM model will learn that when given a feature sequence representing the current site's critical state, it can output a probability distribution, where the load device with the highest probability is the load device that should be prioritized for power-down operation in the current state. This forms the decision basis for generating a tiered backup power control strategy.

[0055] S1035. During training, use a validation set to adjust the model hyperparameters and monitor model performance to prevent overfitting.

[0056] Specifically, while iteratively training the LSTM model using the training set, an independent validation set is used to evaluate the LSTM model's performance on unknown data. This serves to prevent overfitting, that is, to avoid the model mechanically memorizing noise or specific patterns in the training data, which could lead to a decline in its decision-making ability (generalization performance) in new scenarios.

[0057] S1036. The trained Long Short-Term Memory Network Model is evaluated using a test set. When the prediction accuracy of the trained Long Short-Term Memory Network Model on the test set for the load device to be powered down reaches or exceeds a preset threshold, the Long Short-Term Memory Network Model is obtained.

[0058] Specifically, after the model completes training and is fine-tuned on the validation set, a completely independent test set, never used during training and validation, is used to objectively and fairly evaluate the performance of the finally trained LSTM model. Its purpose is to verify the LSTM model's generalization ability in simulating entirely new and unknown scenarios, i.e., whether its intelligent decision-making strategy truly possesses universality and practicality. The key evaluation metric is the LSTM model's prediction accuracy for the load devices to be powered down on the test set. This accuracy is specifically defined as the proportion of samples in the test set where the model's recommended power-down device (i.e., the load device with the highest output probability) matches the load devices that were actually powered down in the corresponding historical state. Only when this prediction accuracy reaches or exceeds a pre-set performance threshold (e.g., 95%) is the LSTM model considered to have successfully learned an effective decision-making pattern, and its performance meets the requirements for actual deployment. At this point, the model training process is complete, and the model can be formally used for online intelligent decision-making in step S103.

[0059] S104. In accordance with the hierarchical backup power control strategy, perform power-down or power-protection operations on the corresponding load equipment in sequence.

[0060] In one embodiment, step S104 includes: repeatedly executing the following steps until the backup power unit does not meet the preset backup power control triggering conditions or all load devices have been processed: among the load devices that have not yet been processed, determine the device category with the highest current power-down priority as the current pending category; for the current pending category, repeatedly execute the following steps until the backup power unit does not meet the preset backup power control triggering conditions or all load devices under the current pending category have been processed: obtain the real-time operating current of the power supply terminals connected to each load device belonging to the current pending category that has not yet been powered down; sort the corresponding load devices from high to low based on each real-time operating current; perform a power-down operation or a power-saving operation on the load device with the highest sorting; after performing the power-down operation or power-saving operation, re-determine whether the backup power unit still meets the preset backup power control triggering conditions.

[0061] Specifically, this embodiment provides a specific cyclic control algorithm based on priority category nested current sorting for executing a hierarchical backup power control strategy. First, devices are processed from high to low importance (preset priority). Then, within devices of the same importance level, selective processing is performed based on current energy consumption (real-time operating current) from high to low. The system state is reassessed after each step.

[0062] The following is a detailed explanation using a specific example.

[0063] 1) Scene setting Assume that after the backup power unit is triggered, there are still 4 online load devices at a certain site, and their preset priority tags are as follows: Wireless device A: Priority 1 (highest); Wireless device B: Priority 1 (highest); Transmission device C: Priority 3; Data device D: Priority 4 (lowest).

[0064] The real-time operating currents of the power supply terminals connected to each load device are as follows: A: 8A, B: 5A, C: 10A, D: 2A.

[0065] 2) Specific flow of the loop control algorithm Step 1 (Start Outer Loop): The system determines that devices A, B, C, and D have not yet been processed. Among them, the device category with the highest power-down priority is the wireless device. Therefore, the wireless device is set as the current category to be processed, and the inner loop is entered.

[0066] Step 2 (Inner Loop - Processing Wireless Devices): a. Obtain devices in this category that have not yet been powered off: A and B.

[0067] b. Obtain the real-time operating current of the power supply terminals to which they are connected: A (8A), B (5A).

[0068] c. Sort by current from highest to lowest: A->B.

[0069] d. Perform a power-down operation on device A, which is ranked first (assuming the policy instruction is power-down).

[0070] e. After execution, reassess the status of the backup power unit. Assume the conditions are still met (voltage is still low, current is still high).

[0071] Step 3 (Return to the inner loop start point and process the remaining devices): Under the current pending category (wireless devices), only device B remains that has not been powered off.

[0072] Obtain the current (5A) of B.

[0073] After sorting, power off device B.

[0074] After execution, the status of the backup power unit is reassessed, assuming that the conditions are still met.

[0075] Step 4 (Inner Loop Termination and Category Switching): At this point, all devices (A and B) under the current pending category (wireless devices) have been processed, and the inner loop terminates.

[0076] The system exits the inner loop and returns to the starting point of the outer loop.

[0077] Step 5 (Outer loop - process the next priority category): Among the devices that have not yet been processed (currently C and D), the device category with the highest current power-down priority is determined to be the transmission device. This is set as the new current pending category, and the inner loop is entered again.

[0078] The processing procedure is the same as above: obtain the current of C (10A) and perform a power-off operation on device C.

[0079] After execution, the status of the backup power unit is reassessed. Assuming that after this operation, the output voltage of the backup power unit rises back above the threshold, the triggering condition is no longer met.

[0080] Step 6 (End of overall process): Since the backup power unit does not meet the preset backup power control triggering conditions, the inner loop terminates immediately, and since one of the termination conditions of the outer loop has also been met, the entire control process ends.

[0081] Final result: Device D (lowest priority) was never processed, and its state was implicitly considered as power preservation by the policy.

[0082] By sequentially powering down devices A, B, and C, the overload state of the backup power unit was relieved, and the continuous operation of device D, which had the lowest priority but had not yet been disconnected, was successfully ensured.

[0083] This invention provides a site-level backup power control method. Through real-time sensing, fusion, and intelligent analysis of the status of multiple source devices within a communication site, it can dynamically generate and execute optimized level-level backup power control strategies. This method can intelligently and automatically identify and prioritize the continuous operation of core service equipment within the limited time of backup power supply, thereby maximizing the continuous service duration of critical services in emergency situations and significantly improving network service reliability and service assurance levels. Simultaneously, this method achieves refined and adaptive management of existing backup energy within the site, maximizing resource utilization efficiency through pure software algorithms, and improving the overall economy and intelligent operation and maintenance capabilities of the site power system.

[0084] Please refer to Figure 2 , Figure 2 This is a schematic diagram of the site-level backup power control device provided by the present invention. Figure 2 As shown, the device may include: Monitoring module 201 is used to monitor the real-time status information of multiple power supply terminals, backup power units and multiple load devices within the target site; Module 202 is used to construct an input feature vector based on real-time status information and preset priority labels corresponding to multiple load devices when the backup power unit meets the preset backup power control trigger conditions. The generation module 203 is used to input the input feature vector into the neural network model to generate a hierarchical backup power control strategy for multiple load devices. The control module 204 is used to perform power-down or power-up operations on the corresponding load devices in sequence according to the hierarchical backup power control strategy.

[0085] In some embodiments, the site-level backup power control device of the present invention can be implemented in a combination of hardware and software. As an example, the site-level backup power control device of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the site-level backup power control method of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0086] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0087] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned site-level backup power control methods. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the site-level backup power control method shown in any embodiment of the present invention by calling the computer program.

[0088] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0089] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0090] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0091] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0092] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0093] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0094] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0095] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described site-level backup power control methods.

[0096] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0097] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned site-level backup power control method.

[0098] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0099] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0100] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0101] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0102] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0103] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0104] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0105] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A site-level hierarchical backup power control method, characterized in that, include: Monitor the real-time status information of multiple power supply terminals, backup power units, and multiple load devices within the target site; When the backup power unit meets the preset backup power control triggering conditions, an input feature vector is constructed based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices. The input feature vector is fed into a neural network model to generate a hierarchical backup power control strategy for the multiple load devices; According to the hierarchical backup power control strategy, power-down or power-up operations are performed on the corresponding load devices in sequence.

2. The site-level backup power control method according to claim 1, characterized in that, The step of sequentially performing power-down or power-saving operations on the corresponding load devices according to the hierarchical backup power control strategy includes: The following steps are repeated until the backup power unit no longer meets the preset backup power control trigger condition or all the load devices have been processed: Among the load devices that have not yet been processed, determine the device category with the highest current power-down priority, and designate it as the current category to be processed; For the current pending category, the following steps are executed repeatedly until the backup power unit no longer meets the preset backup power control trigger condition or all load devices under the current pending category have been processed: Obtain the real-time operating current of the power supply terminals connected to each load device that belongs to the current processing category and has not yet been powered down; The load devices are sorted from high to low based on their real-time operating current. Perform a power-down operation or a power-on operation on the load device that is ranked first. After performing a power-down operation or a power-on operation, it is re-evaluated whether the backup power unit still meets the preset backup power control trigger conditions.

3. The site-level backup power control method according to claim 1 or 2, characterized in that, The following steps are used to determine whether the backup power unit meets the preset backup power control triggering conditions: If the real-time output voltage of the backup power unit is lower than a preset voltage threshold and the real-time output current of the backup power unit is higher than a preset current threshold, then the backup power unit is determined to meet the preset backup power control triggering conditions.

4. The site-level backup power control method according to claim 1, characterized in that, The neural network model is a Long Short-Term Memory (LSTM) network, which is trained through the following steps: Acquire historical status datasets of multiple power supply terminals, backup power units, and multiple load devices within the target site within the historical time window, as well as historical power outage event records; The historical state dataset and the historical power outage event records are preprocessed, including data cleaning, data normalization and feature engineering, to construct a labeled sample set; The labeled sample set is divided into a training set, a validation set, and a test set; The training set is input into the initial long short-term memory network model, and iterative training is performed with the goal of minimizing the cross-entropy loss between the probability distribution of the model output of performing power-down operation on the load device and the corresponding actual power-down operation label. During training, the validation set is used to adjust the model hyperparameters and monitor model performance to prevent overfitting. The trained Long Short-Term Memory (LSTM) network model is evaluated using the test set. When the prediction accuracy of the trained LSM network model on the test set for the load device to be powered down reaches or exceeds a preset threshold, the LSM network model is obtained.

5. The site-level backup power control method according to claim 4, characterized in that, The acquisition of historical status datasets of multiple power connection terminals, backup power units, and multiple load devices within the target site during the historical time window, as well as historical power outage event records, includes: Obtain the real-time operating voltage sequence, real-time operating current sequence, unique identifier, and rated load capacity of each power supply terminal within the historical time window; Obtain the real-time output voltage sequence and real-time output current sequence of the backup power unit within the historical time window; Obtain the real-time operating voltage sequence, real-time operating current sequence, device identifier, device category, unique identifier of the power supply terminal connected to the device, and its preset priority label for each load device within the historical time window; Obtain the device identifier sequence of each load device that was actually powered down at a historical moment when the preset backup power control trigger condition was met; Obtain the status change information of the backup power unit after the load device corresponding to each device identifier in the device identifier sequence is powered down.

6. The site-level backup power control method according to claim 5, characterized in that, The preprocessing of the historical state dataset and the historical power outage event records, including data cleaning, data normalization, and feature engineering, to construct a labeled sample set, includes: Data cleaning is performed on the historical state dataset and the historical power outage event records; The numerical features in the cleaned historical state dataset are normalized, and the cleaned preset priority labels are encoded. For each power supply terminal, a real-time load rate sequence is calculated based on its normalized real-time operating current sequence and the rated load capacity. For the backup power unit, its real-time remaining range time sequence is estimated based on its normalized real-time output voltage sequence, real-time output current sequence, and battery capacity model. The numerical features, real-time load rate sequence, real-time remaining battery life time sequence, and encoding of each preset priority label in the normalized historical state dataset are time-aligned and associated with the device identification sequence and each state change information to construct the labeled sample set.

7. The site-level backup power control method according to claim 1, characterized in that, The step of constructing an input feature vector based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices includes: The numerical features in the real-time status information are normalized. The preset priority tags corresponding to each of the multiple load devices are encoded. The real-time load rate of each power supply terminal and the estimated remaining battery life of the backup power unit are calculated based on the numerical features in the normalized real-time status information. The numerical features in the normalized real-time status information, the real-time load rate of each power supply terminal, the estimated remaining battery life of the backup power unit, and the encoding of the priority label are fused to form the input feature vector at the current moment.

8. A site-level backup power control device, characterized in that, include: The monitoring module is used to monitor the real-time status information of multiple power supply terminals, backup power units, and multiple load devices within the target site. The construction module is used to construct an input feature vector based on the real-time status information and the preset priority labels corresponding to each of the multiple load devices when the backup power unit meets the preset backup power control trigger conditions; The generation module is used to input the input feature vector into the neural network model to generate a hierarchical backup power control strategy for the multiple load devices; The control module is used to sequentially perform power-down or power-up operations on the corresponding load devices according to the hierarchical backup power control strategy.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the site-level backup power control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the site-level backup power control method as described in any one of claims 1 to 7.

Images