Intelligent storehouse system for secondary equipment of transformer substation
By introducing intelligent management systems and real-time monitoring technology into the secondary equipment warehouse of the substation, the shortcomings of equipment management and flame detection are solved, the safe and reliable operation and rapid failure recovery of the warehouse are achieved, and the safety and reliability of the power grid and users are improved. .
Patent Information
- Application Number
- CN202411078247.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-08-07
AI Technical Summary
Traditional substation secondary equipment warehouses cannot effectively manage and trace equipment, and cannot realize intelligent spare parts management and replenishment of equipment, resulting in the inability to resume operation quickly in the event of sudden failures, affecting the safety and reliability of the power grid and users.
A smart warehouse system for secondary equipment of substations was designed, including sensors, cameras, RFID devices and face recognition scanners, combined with YOLO flame detection model and dung beetle optimization algorithm, real-time monitoring and automated management are realized to ensure inventory tracking and environmental monitoring. and equipment security.
The normal operation of the warehouse and the safety and reliability of the equipment are achieved, and the required equipment and accessories can be quickly provided in the event of a sudden failure, reducing the impact of accidents on the power grid and users, and improving the accuracy and response speed of inventory management and flame detection.
Smart Images

Figure CN119047971A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of substation engineering, and in particular relates to an intelligent warehouse system for secondary equipment of a substation. Background Art
[0002] The safe and stable operation of the secondary equipment of the substation is directly related to the safety and stability of the entire power grid. As an important place for storing and placing secondary equipment, the secondary equipment warehouse directly determines the performance of the stored secondary equipment. Therefore, the substation secondary equipment warehouse is of great significance to ensure the normal operation of the substation and improve the reliability and safety of the power grid. The substation secondary equipment warehouse usually stores spare parts and replacement parts to deal with sudden failures or equipment damage. The timely supply and management of these spare parts directly affects the maintenance and recovery time of the substation equipment, and then affects the reliability and stability of the power grid. In the event of a sudden failure or emergency, if the warehouse cannot quickly provide the required equipment and accessories, it will not be able to quickly restore the substation to operation, resulting in the accident affecting the power grid and users. The traditional substation secondary equipment warehouse does not take into account the storage needs and equipment safety protection requirements of special equipment, and cannot achieve equipment management and traceability, as well as intelligent spare parts management and replenishment. Summary of the invention
[0003] In order to solve the deficiencies in the prior art, the present invention provides a substation secondary equipment intelligent warehouse system to solve the technical problems of normal operation of the secondary equipment warehouse and safety and reliability of the equipment.
[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions.
[0005] The present invention first discloses a substation secondary equipment intelligent warehouse system, including a secondary equipment warehouse and a secondary equipment intelligent management system. The secondary equipment warehouse is provided with a management area, a common tool area, a test instrument area, an accessories area, a relay protection spare parts area, an automation spare parts area and a network security spare parts area; the secondary equipment warehouse is deployed with sensors and cameras for temperature and humidity detection, smoke detection, personnel detection and flame detection; the secondary equipment warehouse is installed with RFID equipment to record storage and outbound information by capturing RFID tag data; the secondary equipment warehouse is installed with a face recognition scanner for authentication and entry and exit management of authorized personnel;
[0006] The secondary equipment intelligent management system is used to query spare parts, display spare parts information, record and manage the entry and exit of spare parts, issue inventory alarms based on the number of spare parts, use inventory data for analysis, monitor and query warehouse environment data, and perform user management for warehouse managers;
[0007] The flame detection adopts the YOLOv8 architecture and the YOLO loss function. First, the YOLO model is trained on the flame detection dataset, and the early stopping technique is used to prevent overfitting. The model is evaluated on the test set, and mAP is used as the performance indicator. According to the performance on the test set, the optimal detection threshold is determined, and the trained model is deployed in the real-time monitoring system.
[0008] To perform real-time flame detection on the images captured by the camera, image processing is first performed, including:
[0009] Use cameras to collect video streams in the warehouse, use images within a specific wavelength range to enhance the characteristics of the flame, and image preprocessing includes grayscale and histogram equalization;
[0010] I'(x,y)=T·I(x,y)
[0011] Where I'(x,y) is the processed image intensity, I(x,y) is the original image intensity, and T is the transformation matrix based on the flame spectral characteristics;
[0012] The processed image is input into the trained YOLO model for flame detection.
[0013] The present invention further includes the following preferred embodiments:
[0014] Each storage area in the management area, common tool area, test instrument area, accessories area, relay protection spare parts area, automation spare parts area and network security spare parts area is marked with category names, different areas are marked with different colors, and physical dividing lines are used to divide different categories; office computers and servers are set up in the management area, and the secondary equipment intelligent management system is deployed.
[0015] The secondary equipment intelligent management system will pre-register the facial data of authorized personnel. The facial recognition scanner is integrated with the electronic door lock and uses facial recognition technology for detection. Only authorized personnel can enter the warehouse. Access control software is set to manage and monitor access permissions and define access levels for different groups or individuals.
[0016] The secondary equipment intelligent management system includes:
[0017] The spare parts search and query module is used to query spare parts according to name information and display spare parts information and the shelf to which they belong according to the query results;
[0018] The in-and-out record module is used to implement inventory management and check the equipment in-and-out time;
[0019] The spare parts inventory alarm module is used to generate equipment inventory alarm records based on the equipment inventory quantity; define the minimum inventory level replenishment threshold for each spare part. When the inventory is lower than the corresponding threshold, the replenishment alarm is triggered and sent to the warehouse manager terminal via SMS;
[0020] Data analysis and reporting modules for using inventory data for analysis, identifying equipment usage trends, and predicting future demand.
[0021] The in-and-out record module and the spare parts inventory alarm module use RFID smart tags for real-time inventory tracking and automatic replenishment notifications. All spare parts are equipped with RFID tags. Each tag stores spare parts related data, including spare parts type, manufacturer, model, shelf, and project. The RFID reader is connected to the secondary equipment intelligent management system to process data from the tag and update inventory changes in real time. The secondary equipment intelligent management system verifies the shelf in the information recorded on the spare parts tag to ensure that the shelf and spare parts type are consistent. If they are inconsistent, an error is reported, indicating that the RFID data is wrong. When the item is placed in the wrong position, an alarm is issued to the warehouse manager, and the message is sent to the warehouse manager terminal via text message.
[0022] The secondary equipment intelligent management system includes:
[0023] The environmental monitoring status and alarm recording module is used to monitor the real-time temperature, humidity value, camera surveillance screen of the warehouse, and query the sensor, camera history record and alarm record; the sensors include temperature sensors, humidity sensors, and smoke detectors. All sensors are connected to the environmental monitoring status and alarm recording module to continuously collect and analyze environmental data, define safety thresholds for environmental temperature, humidity and smoke status, and trigger an alarm when the corresponding threshold is exceeded, automatically notifying the warehouse manager.
[0024] The Dung Beetle Optimization Algorithm (DBO) is used to dynamically adjust the safety thresholds of temperature, humidity and smoke, including:
[0025] 1.1: Determine the search space of the alarm threshold, that is, the possible value range of the threshold;
[0026] 1.2: Select as the initial population size;
[0027] 1.3: For each dung beetle individual, calculate the false positive rate and false positive rate according to the threshold it represents, and then calculate the fitness function:
[0028] Fitness = missed alarm rate (missed alarm number / total alarm number) + false alarm rate (false alarm number / total alarm number)
[0029] The goal is to minimize the fitness function, i.e., reduce false negatives and false positives;
[0030] 1.4: Simulate the social behavior of dung beetles, that is, individuals tend to follow individuals with lower fitness;
[0031] 1.5: Simulate the foraging behavior of dung beetles and search the solution space through random perturbations in the hope of finding a better threshold solution;
[0032] 1.6: Select dung beetle individuals for reproduction based on fitness to generate a new offspring population, using Gaussian perturbation, and the perturbation size is based on the percentage of the threshold;
[0033] 1.7: Combine the offspring after reproduction with the current population to update the population;
[0034] 1.8: Repeat 1.3 to 1.7, set the maximum number of iterations to 100, and stop the algorithm if the improvement of the fitness function is less than 0.01% in 10 consecutive iterations;
[0035] 1.9: Output the threshold represented by the dung beetle individual with the lowest fitness as the optimal solution. After every 5 correct alarms, false alarms, and missed alarms, re-test the threshold to optimize the threshold.
[0036] The camera is equipped with an image recognition module and is used for uninterrupted real-time monitoring. The monitoring range covers the entire area of the warehouse. The camera executes an AI video image recognition algorithm and is connected to the environmental monitoring status and alarm recording module.
[0037] The image recognition module includes a flame detection module and a personnel detection module. The flame detection module is used to immediately issue an alarm when signs of flame are found, link intelligent video capture, and automatically notify the warehouse manager by phone. Flame detection uses the YOLO target detection algorithm, which has high recognition accuracy and stability. The personnel detection module is used to link and turn off all lights in the warehouse when it is detected that there is no one in each area of the warehouse for a duration of 10 minutes.
[0038] The secondary equipment warehouse uses constant temperature and humidity precision air conditioning, and is equipped with a static eliminator and an ion fan device to reduce the generation of static electricity in the warehouse.
[0039] The secondary equipment warehouse is equipped with water retaining dams, drainage ditches and floor drains around the ground to prevent the transfer and infiltration of underground water. Water-sensitive detection instruments or components are installed around the warehouse leakage risk areas, including under the warehouse floor and around the air conditioners, to perform waterproof detection and alarm for the warehouse.
[0040] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides an intelligent warehouse system for secondary equipment of a substation, which can ensure the normal operation of the warehouse and the safety and reliability of the equipment. In the event of a sudden failure or emergency, the warehouse can quickly provide the required equipment and accessories to help the substation quickly resume operation and reduce the impact of the accident on the power grid and users.
[0041] The alarm strategy is optimized based on the Deep Q Network (DQN). DQN can handle high-dimensional and continuous state spaces, which makes it very suitable for complex environments such as smart warehouse monitoring systems, involving a large amount of sensor readings and historical data. The DQN model can continuously learn from new data and continuously improve the alarm strategy to adapt to the environmental changes of the secondary equipment warehouse, thereby improving the practicality and accuracy of the alarm strategy.
[0042] According to the historical data and real-time changes of environmental parameters in the warehouse, the dung beetle optimization algorithm is used to dynamically adjust the safety thresholds of temperature, humidity and smoke. The dung beetle optimization algorithm has a strong global search capability by simulating the behavior of dung beetles, which helps to find a better threshold setting and improve the accuracy of the alarm system. DBO can still maintain good search efficiency and stability in the face of complex and changing monitoring environments, reducing the occurrence of false alarms and missed alarms; the DBO algorithm has a simple principle and few parameters, and is easy to implement and apply in the existing alarm system. It does not require complex mathematical models and a large amount of computing resources. By simulating the random walk mechanism of dung beetles, DBO can effectively jump out of the local optimal solution, avoid the threshold adjustment from falling into the local extreme value, and improve the comprehensiveness and accuracy of the threshold adjustment.
[0043] Intelligent flame detection based on the YOLO algorithm. The YOLO model can process images at an extremely fast speed to achieve real-time detection. This is especially important for flame detection, because flames may appear suddenly and require a quick response. The YOLO model has many pre-trained models that can be downloaded and applied for free, which speeds up the development process and reduces the need for data collection. After proper training and optimization, the YOLO model can achieve high detection accuracy and ensure accurate identification of flames. The YOLO model can adapt to different environmental conditions, including different lighting conditions, background noise, etc., which makes it suitable for secondary equipment warehouse application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a schematic diagram of the area division result of the intelligent storehouse of the substation secondary equipment in the present invention.
[0045] Figure 2 It is a schematic diagram of the function page of the secondary equipment intelligent management system in the present invention.
[0046] Figure 3 It is a structural diagram of the environment monitoring and intelligent alarm system in the present invention.
[0047] Figure 4 This is a flow chart of optimizing the safety threshold using the dung beetle optimization algorithm in the present invention. DETAILED DESCRIPTION
[0048] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0049] The embodiments described in this application are only some embodiments of the present invention, not all embodiments. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.
[0050] In view of the shortcomings of the prior art, the present invention proposes an intelligent warehouse system for secondary equipment of a substation, so that when a sudden failure or emergency occurs, the warehouse can quickly provide the required equipment and accessories to help the substation quickly resume operation and reduce the impact of the accident on the power grid and users. The secondary equipment warehouse has its own characteristics and requirements in terms of equipment storage, safety protection, management and maintenance, and a series of measures need to be taken to ensure the normal operation of the warehouse and the safety and reliability of the equipment. The secondary equipment warehouse usually stores various types of electrical equipment, such as relays, protection devices, switchgear, etc. These devices have special requirements for storage and management, and need to be kept dry and ventilated, and avoid being affected by environmental factors such as humidity and dust. Secondary equipment is usually a key equipment for the operation of a substation, so its safety and integrity are crucial. The secondary equipment warehouse needs to be equipped with safety protection facilities, such as surveillance cameras, intrusion alarm systems, etc., to prevent unauthorized personnel from entering or damaging the equipment. Due to the importance of secondary equipment, strict management and traceability are necessary. The warehouse needs to establish a complete equipment management system, including equipment entry registration, outbound records, equipment status tracking and other functions, to ensure that the use and maintenance of the equipment can be effectively managed and monitored. To ensure the normal operation and safety of secondary equipment, the warehouse needs to be inspected and maintained regularly. This includes regular inspection, cleaning, maintenance, and necessary repairs and replacements of the equipment to ensure the performance and reliability of the equipment. In order to deal with sudden failures and equipment damage, the secondary equipment warehouse needs to establish a spare parts management system and formulate corresponding replenishment plans. Warehouse managers need to reasonably arrange the storage and replenishment of spare parts according to the use and maintenance needs of the equipment to ensure that the equipment can be replaced and repaired in time when needed. Based on the above needs, an intelligent warehouse system for substation secondary equipment is developed.
[0051] The regional division method of the substation secondary equipment intelligent warehouse disclosed in the present invention is as follows: Figure 1As shown. The warehouse is divided into eight areas, namely the management area, common tools area, test instrument area, accessories area, relay protection spare parts area, automation spare parts area and network security spare parts area. Each storage area is marked with a category name, including visual aids such as symbols and color coding, and different areas are marked with different colors. Physical dividing lines are used between different categories to divide the end of one category and the beginning of another category. The management area is equipped with a warehouse management system and safety precautions, where office computers and servers are placed, and a secondary equipment intelligent management system is deployed. The warehouse management system includes: personnel responsibilities and obligations, spare parts in and out management system, warehouse safety management system, warehouse emergency response plan, daily patrol system, etc. Arrows are posted on the ground to guide the patrol route map.
[0052] Secondary equipment is usually the key equipment for substation operation, so its safety and integrity are of vital importance. The secondary equipment warehouse needs to be equipped with security protection facilities to prevent unauthorized personnel from entering or damaging the equipment.
[0053] First, pre-register authorized personnel into the secondary equipment intelligent management system to store the authorized personnel's facial data. Install a facial recognition scanner at the warehouse entrance point, integrate it with the electronic door lock, and use facial recognition technology for detection. Only authorized personnel can enter the warehouse. Managers manage and monitor access rights by setting up access control software. Define access levels for different groups or individuals. Record the entry and exit of personnel.
[0054] Secondary equipment intelligent management system function page Figure 2 As shown, it includes a spare parts search and query module, an in-and-out storage record module, a spare parts inventory alarm record module, an environmental monitoring status and alarm record module, a user management module and a warehouse management system query module.
[0055] The spare parts search query module is used to query spare parts according to information such as name, and display spare parts information and the shelf to which they belong according to the query results. The in-and-out storage record module is used to implement inventory management and view the equipment in-and-out storage time. The spare parts inventory alarm module is used to generate equipment inventory alarm records based on the equipment inventory quantity. The data analysis and reporting module is used to use inventory data for analysis, such as identifying equipment usage trends, predicting future demand and optimizing inventory levels. The environmental monitoring status and alarm record module is used to monitor the real-time temperature, humidity value, and camera monitoring screen of the warehouse, and can query sensors, camera history records and alarm records. The user management module is required by the warehouse manager to log in to the secondary equipment intelligent management system. The terminal number of the warehouse manager and security personnel and other information. The warehouse management system query includes personnel responsibilities and obligations, spare parts in-and-out storage management system, warehouse safety management system, warehouse emergency response plan, daily inspection system and other inquiries.
[0056] In the inventory management process of the in-and-out recording module, smart tags (RFID) are used for real-time inventory tracking and automatic replenishment notification. Specifically, all spare parts are equipped with RFID tags. Each tag stores spare part-related data, including spare part type, manufacturer, model, shelf, and project. An RFID reader is installed at the door of the warehouse to capture data from the tag, record the in-and-out situation, and record the storage. The RFID reader is connected to the secondary equipment intelligent management system. The system processes the data from the tag and updates the inventory changes in real time. The management system verifies the shelf described in the information recorded by the spare part tag to ensure that the shelf and spare part type are consistent. If they are inconsistent, an error is reported and the RFID data error is displayed. An alarm is issued to the warehouse manager when the item is placed in the wrong position. And it is sent to the warehouse manager terminal by SMS. Define the minimum inventory level replenishment threshold for each spare part. When the inventory is lower than this threshold, the system will trigger a replenishment alarm. And it is sent to the warehouse manager terminal by SMS. The system will improve inventory accuracy, reduce manual work, and ensure timely replenishment of spare parts.
[0057] The secondary equipment warehouse deploys sensors and cameras to detect temperature and humidity, smoke, personnel, and flames, and transmits information to the environmental monitoring status and alarm recording module. When an abnormal situation is detected, it automatically notifies the warehouse manager or security personnel. Figure 3 shown.
[0058] The sensors include temperature sensors, humidity sensors, and smoke detectors. All sensors are connected to the environmental monitoring status and alarm recording module to continuously collect and analyze environmental data. Safety thresholds are defined for environmental temperature, humidity, and smoke status. When the corresponding threshold is exceeded, the system will trigger an alarm and automatically notify the warehouse manager or security personnel by phone.
[0059] In a preferred embodiment, the safety thresholds of temperature, humidity and smoke are dynamically adjusted according to the historical data and real-time changes of the environmental parameters in the warehouse. The Dung Beetle Optimization (DBO) algorithm simulates the behavior of dung beetles and has a strong global search capability, which helps to find a better threshold setting and improve the accuracy of the alarm system. When facing a complex and changeable monitoring environment, DBO can still maintain good search efficiency and stability, reducing the occurrence of false alarms and missed alarms; the DBO algorithm has a simple principle, few parameters, and is easy to implement and apply in the existing alarm system. It does not require complex mathematical models and a large amount of computing resources. By simulating the random walk mechanism of dung beetles, DBO can effectively jump out of the local optimal solution, avoid the threshold adjustment from falling into the local extreme value, and improve the comprehensiveness and accuracy of the threshold adjustment.
[0060] Use the dung beetle optimization algorithm to optimize the alarm threshold. Follow the steps below: Figure 4 shown.
[0061] 1.1. Determine the search space of the alarm threshold, that is, the possible value range of the threshold.
[0062] 1.2. Initialize the population and select 30 as the initial population size to balance the breadth and depth of the search.
[0063] 1.3. For each dung beetle individual, calculate the missed alarm rate and false alarm rate according to the threshold it represents, and then calculate the fitness function:
[0064] Fitness = missed alarm rate (missed alarm number / total alarm number) + false alarm rate (false alarm number / total alarm number)
[0065] The goal is to minimize the fitness function, i.e., reduce false negatives and false positives.
[0066] 1.4. Simulate the social behavior of dung beetles, that is, individuals tend to follow individuals with lower fitness (i.e., the sum of missed alarm rate and false alarm rate).
[0067] 1.5. Simulate the foraging behavior of dung beetles and search the solution space through random perturbations in the hope of finding a better threshold solution.
[0068] 1.6: Select dung beetle individuals for reproduction based on fitness to generate a new offspring population. Use fitness proportional selection to ensure that excellent threshold solutions have a higher chance of reproduction. Implement an elite strategy to ensure that the top 10% of the optimal solutions are directly inherited to the next generation. Use Gaussian perturbation, and the perturbation size is based on the percentage of the threshold, such as 5% of the threshold, to avoid falling into the local optimal solution.
[0069] 1.7: Combine the offspring after reproduction with the current population and select as needed to update the population.
[0070] 1.8: Repeat 1.3 to 1.7, set the maximum number of iterations to 100, and stop the algorithm if the improvement of the fitness function is less than 0.01% in 10 consecutive iterations to prevent the algorithm from converging prematurely or over-iteration.
[0071] 1.9: Output the threshold represented by the dung beetle individual with the lowest fitness (i.e., the sum of the missed alarm rate and the false alarm rate is the smallest) as the optimal solution. After every 5 related events (including correct alarms, false alarms, and missed alarms), re-test the threshold to optimize the threshold and ensure the rationality of the threshold and the effectiveness of the alarm system.
[0072] The environmental monitoring status and alarm recording module further uses reinforcement learning algorithms to optimize the alarm strategy. The deep Q network (DQN) has significant advantages in optimizing the alarm strategy, especially in dealing with complex and dynamically changing environments and decision-making scenarios that need to consider long-term impacts. The deep Q network (DQN) is used to optimize the alarm strategy. The specific implementation process includes:
[0073] 2.1: Collect and preprocess data related to warehouse monitoring, including sensor readings (temperature, humidity, smoke level, etc.), equipment status, historical alarm records, etc. Convert this data into feature vectors as the input state s of DQN.
[0074] 2.2: List all possible actions a that the agent can take, including sounding an alarm, not sounding an alarm, turning off the power, adjusting sensor thresholds, etc. Assign a unique index to each action.
[0075] 2.3: Design a reward function based on business needs. The reward function contains multiple aspects to ensure the security and efficiency of the system. Consider the following factors when designing the reward function:
[0076] 1) Alarm accuracy: reduce false alarms and missed alarms.
[0077] R accuracy =-(w fp FP+w fn ·FN)
[0078] Where FP is the number of false positives, FN is the number of missed negatives, and w fp and w fn is the corresponding weight coefficient.
[0079] 2) Response time: Real alarms that encourage quick response.
[0080] R response =w time (Alarm time - response time)
[0081] w time is the corresponding weight coefficient.
[0082] 3) Alarm processing efficiency: Effective alarm processing measures are encouraged. The processing success rate is the ratio of successfully processed alarms to total alarms.
[0083] R efficiency =w eff Processing success rate
[0084] where w eff is the corresponding weight coefficient.
[0085] 4) User satisfaction: Consider the warehouse managers’ satisfaction with the alarm system, where the satisfaction score is 1-3 points, with 3 points being the most satisfactory.
[0086] R satisfaction =w sat Satisfaction rating
[0087] where w sat is the corresponding weight coefficient.
[0088] The final reward function is the weighted sum of the above rewards:
[0089] r=R accuracy +R response +R efficiency +R satisfaction
[0090] According to the expert experience and the actual situation of the warehouse, the weight value w fp 、w fn 、w time 、w eff 、w sat Take 0.3, 0.3, 0.1, 0.2, and 0.1 respectively, and can be adjusted according to actual conditions.
[0091] 2.4: Use a deep neural network to approximate the Q value function. The network has an input layer, two hidden layers (RELU as activation function), and an output layer. The Adam optimizer is used, the learning rate α is 0.001, and the decay rate is 0.99; the mean square error loss (MSE Loss), L2 regularization, and the weight decay coefficient is 0.001 to prevent overfitting; the ε greedy strategy is used, the initial ε value is 1, and it decreases by 0.01 every 1000 training steps until the minimum value is 0.1. The network accepts the state as input and outputs the Q value of each possible action.
[0092] 2.5: During training, all experienced states, actions, rewards, and next states are stored in an experience pool. In each training step, a mini-batch of samples is randomly drawn from the experience pool for training.
[0093] 2.6: Use a separate target network to calculate the target Q value, which copies the weights from the online network at regular intervals. The target network is used to calculate the target Q value to reduce the variance during training.
[0094] 2.7: Define the mean squared error loss function (MSE) as
[0095]
[0096] Where r is the immediate reward, γ is the discount factor, which is used to weigh the importance of immediate rewards and future rewards and is set to 0.95. Q'(s',a') is the Q value of the next state s' and action a' output by the target network, and Q(s,a) is the Q value of the current state s and action a output by the online network.
[0097] 2.8: Use the gradient descent algorithm (Adam optimizer) to update the network weights to minimize the loss function.
[0098] At each training step, the weights of the online network are updated and periodically copied to the target network.
[0099] Update Q value:
[0100] Q(s,a)←Q(s,a)+α[r+γmax a' Q'(s',a')-Q(s,a)]
[0101] Where α is the learning rate and γ is the discount factor.
[0102] 2.9: Use the ε greedy strategy to balance exploration (randomly selecting actions) and exploitation (selecting the best action based on the current network). As training progresses, gradually reduce the ε value to reduce random exploration.
[0103] 2.10: Use an independent test set to evaluate the performance of the model. If the model performance meets the requirements, deploy it in the actual environment. Through the above steps, optimize the alarm strategy and improve the efficiency and security of the system.
[0104] The camera is installed in the warehouse. In a preferred embodiment, 6 cameras equipped with image recognition modules are installed to conduct 24-hour uninterrupted real-time monitoring, and the monitoring range covers the entire area of the warehouse. With the help of advanced AI video image recognition algorithm, the image recognition module is divided into a flame detection module and a personnel detection module. The camera is connected to the environmental monitoring status and alarm recording module.
[0105] The flame detection module uses the YOLO model to realize intelligent flame detection.
[0106] The flame detection module uses a public fire flame detection dataset. All images are resized to 416×416 pixels and standardized. The YOLOv8 architecture is selected, which strikes a good balance between speed and accuracy. The YOLO loss function is used, the learning rate is set to 0.001, and the SGD optimizer is used. The model is trained on the flame detection dataset, and the early stopping technique is used to prevent overfitting. The model is evaluated on the test set, and mAP (mean Average Precision) is used as the performance indicator. Based on the performance on the test set, the optimal detection threshold is determined. The trained model is deployed to the real-time monitoring system.
[0107] The trained YOLO model is integrated into the real-time monitoring system to perform real-time flame detection on the images captured by the camera. First, image processing is performed, including:
[0108] Use a highly sensitive camera to capture the video stream inside the warehouse. The camera should have good low-light performance to ensure reliable detection of flames in different lighting conditions.
[0109] Due to the spectral characteristics of flames, images within a specific wavelength range can be used to enhance the characteristics of flames. Image preprocessing includes grayscale, histogram equalization and other steps.
[0110] I'(x,y)=T·I(x,y)
[0111] Where I'(x, y) is the processed image intensity, I(x, y) is the original image intensity, and T is the transformation matrix designed according to the flame spectrum characteristics.
[0112] After image processing, the image is input into the YOLO model for flame recognition to ensure the effectiveness of flame recognition.
[0113] For the flame detection module, once signs of flame are detected, the system will immediately issue an alarm after 3 seconds, link up with the intelligent video recording and automatically notify the warehouse management personnel and security personnel by phone, reminding relevant personnel to take timely action, thereby effectively preventing the occurrence of fire.
[0114] For the personnel detection module, once it detects that there is no one in each area of the warehouse for 10 minutes, all the lights in the warehouse will be turned off to achieve energy saving effect.
[0115] In terms of physical security protection, a dedicated constant temperature and humidity precision air conditioner is used in the warehouse to ensure constant temperature and humidity in each area. The warehouse is equipped with static eliminators, ion fans and other devices to reduce the generation of static electricity in the warehouse. Anti-static wristbands are placed at the entrance of the warehouse for staff to wear when operating electronic equipment to reduce the risks caused by static electricity. Anti-static shoe covers are set at the entrance of the warehouse for staff to wear anti-static shoe covers after entering the door to reduce the risks caused by static electricity. Water dams, drainage ditches and floor drains are set around the ground in the warehouse leakage risk area to prevent the transfer and infiltration of underground water caused by the rupture of the air conditioning drainage pipe. Water-sensitive detection instruments or components, such as leakage ropes, are set around the warehouse leakage risk area, such as under the warehouse floor and around the air conditioner, to conduct waterproof detection and alarm for the warehouse.
[0116] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides an intelligent warehouse system for secondary equipment of a substation, which can ensure the normal operation of the warehouse and the safety and reliability of the equipment. In the event of a sudden failure or emergency, the warehouse can quickly provide the required equipment and accessories to help the substation quickly resume operation and reduce the impact of the accident on the power grid and users.
[0117] The alarm strategy is optimized based on the Deep Q Network (DQN). DQN can handle high-dimensional and continuous state spaces, which makes it very suitable for complex environments such as smart warehouse monitoring systems, involving a large amount of sensor readings and historical data. The DQN model can continuously learn from new data and continuously improve the alarm strategy to adapt to the environmental changes of the secondary equipment warehouse, thereby improving the practicality and accuracy of the alarm strategy.
[0118] According to the historical data and real-time changes of environmental parameters in the warehouse, the dung beetle optimization algorithm is used to dynamically adjust the safety thresholds of temperature, humidity and smoke. The dung beetle optimization algorithm has a strong global search capability by simulating the behavior of dung beetles, which helps to find a better threshold setting and improve the accuracy of the alarm system. DBO can still maintain good search efficiency and stability in the face of complex and changing monitoring environments, reducing the occurrence of false alarms and missed alarms; the DBO algorithm has a simple principle and few parameters, and is easy to implement and apply in the existing alarm system. It does not require complex mathematical models and a large amount of computing resources. By simulating the random walk mechanism of dung beetles, DBO can effectively jump out of the local optimal solution, avoid the threshold adjustment from falling into the local extreme value, and improve the comprehensiveness and accuracy of the threshold adjustment.
[0119] Intelligent flame detection based on the YOLO algorithm. The YOLO model can process images at an extremely fast speed to achieve real-time detection. This is especially important for flame detection, because flames may appear suddenly and require a quick response. The YOLO model has many pre-trained models that can be downloaded and applied for free, which speeds up the development process and reduces the need for data collection. After proper training and optimization, the YOLO model can achieve high detection accuracy and ensure accurate identification of flames. The YOLO model can adapt to different environmental conditions, including different lighting conditions, background noise, etc., which makes it suitable for secondary equipment warehouse application scenarios.
[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A substation secondary equipment intelligent warehouse system, including a secondary equipment warehouse and a secondary equipment intelligent management system, Features: The secondary equipment warehouse is equipped with a management area, a common tool area, a test instrument area, an accessories area, a relay protection spare parts area, an automation spare parts area, and a network security spare parts area; the secondary equipment warehouse is equipped with sensors and cameras for temperature and humidity detection, smoke detection, personnel detection, and flame detection; the secondary equipment warehouse is equipped with RFID equipment to record the inbound and outbound information by capturing RFID tag data; the secondary equipment warehouse is equipped with a face recognition scanner for authentication and entry and exit management of authorized personnel; The secondary equipment intelligent management system is used to query spare parts, display spare parts information, record and manage the entry and exit of spare parts, issue inventory alarms based on the number of spare parts, use inventory data for analysis, monitor and query warehouse environment data, and perform user management for warehouse managers; The flame detection adopts the YOLOv8 architecture and the YOLO loss function. First, the YOLO model is trained on the flame detection dataset, and the early stopping technique is used to prevent overfitting. The model is evaluated on the test set, and mAP is used as the performance indicator. According to the performance on the test set, the optimal detection threshold is determined, and the trained model is deployed in the real-time monitoring system. To perform real-time flame detection on the images captured by the camera, image processing is first performed, including: Use cameras to collect video streams in the warehouse, use images within a specific wavelength range to enhance the characteristics of the flame, and image preprocessing includes grayscale and histogram equalization; I'(x,y)=T·I(x,y) Where I'(x,y) is the processed image intensity, I(x,y) is the original image intensity, and T is the transformation matrix based on the flame spectral characteristics; The processed image is input into the trained YOLO model for flame detection.
2. The substation secondary equipment intelligent warehouse system according to claim 1, It is characterized in that Each storage area in the management area, common tool area, test instrument area, accessories area, relay protection spare parts area, automation spare parts area and network security spare parts area is marked with category names, different areas are marked with different colors, and physical dividing lines are used to divide different categories; office computers and servers are set up in the management area, and the secondary equipment intelligent management system is deployed.
3. The substation secondary equipment intelligent warehouse system according to claim 2, It is characterized in that The secondary equipment intelligent management system will pre-register the facial data of authorized personnel. The facial recognition scanner is integrated with the electronic door lock and uses facial recognition technology for detection. Only authorized personnel can enter the warehouse. Access control software is set to manage and monitor access permissions and define access levels for different groups or individuals.
4. The substation secondary equipment intelligent warehouse system according to claim 3, It is characterized in that The secondary equipment intelligent management system includes: The spare parts search and query module is used to query spare parts according to name information and display spare parts information and the shelf to which they belong according to the query results; The in-and-out record module is used to implement inventory management and check the equipment in-and-out time; The spare parts inventory alarm module is used to generate equipment inventory alarm records based on the equipment inventory quantity; define the minimum inventory level replenishment threshold for each spare part. When the inventory is lower than the corresponding threshold, the replenishment alarm is triggered and sent to the warehouse manager terminal via SMS; Data analysis and reporting modules for using inventory data for analysis, identifying equipment usage trends, and predicting future demand.
5. The substation secondary equipment intelligent warehouse system according to claim 4, It is characterized in that The in-and-out record module and the spare parts inventory alarm module use RFID smart tags for real-time inventory tracking and automatic replenishment notifications. All spare parts are equipped with RFID tags. Each tag stores spare parts related data, including spare parts type, manufacturer, model, shelf, and project. The RFID reader is connected to the secondary equipment intelligent management system to process data from the tag and update inventory changes in real time. The secondary equipment intelligent management system verifies the shelf in the information recorded on the spare parts tag to ensure that the shelf and spare parts type are consistent. If they are inconsistent, an error is reported, indicating that the RFID data is wrong. When the item is placed in the wrong position, an alarm is issued to the warehouse manager, and the message is sent to the warehouse manager terminal via text message.
6. The substation secondary equipment intelligent warehouse system according to claim 5, It is characterized in that The secondary equipment intelligent management system includes: The environment monitoring status and alarm recording module is used to monitor the real-time temperature, humidity value, camera monitoring screen of the warehouse, and query the sensor, camera history record and alarm record; the sensors include temperature sensors, humidity sensors, and smoke detectors. All sensors are connected to the environment monitoring status and alarm recording module to continuously collect and analyze environmental data, define safety thresholds for environmental temperature, humidity and smoke status, and trigger an alarm when the corresponding threshold is exceeded, automatically notifying the warehouse management personnel; The Dung Beetle Optimization Algorithm (DBO) is used to dynamically adjust the safety thresholds of temperature, humidity and smoke, including: 1.1: Determine the search space of the alarm threshold, that is, the possible value range of the threshold; 1.2: Select as the initial population size; 1.3: For each dung beetle individual, calculate the false positive rate and false negative rate according to the threshold it represents, and then calculate the fitness function: Fitness = missed alarm rate (missed alarm number / total alarm number) + false alarm rate (false alarm number / total alarm number) The goal is to minimize the fitness function, i.e., reduce false negatives and false positives; 1.4: Simulate the social behavior of dung beetles, that is, individuals tend to follow individuals with lower fitness; 1.5: Simulate the foraging behavior of dung beetles and search the solution space through random perturbations in the hope of finding a better threshold solution; 1.6: Select dung beetle individuals for reproduction based on fitness to generate a new offspring population, using Gaussian perturbation, and the perturbation size is based on the percentage of the threshold; 1.7: Combine the offspring after reproduction with the current population to update the population; 1.8: Repeat 1.3 to 1.7, set the maximum number of iterations to 100, and stop the algorithm if the improvement of the fitness function is less than 0.01% in 10 consecutive iterations; 1.9: Output the threshold represented by the dung beetle individual with the lowest fitness as the optimal solution. After every 5 correct alarms, false alarms, and missed alarms, re-test the threshold to optimize the threshold.
7. The substation secondary equipment intelligent warehouse system according to claim 6, It is characterized in that The camera is equipped with an image recognition module and is used for uninterrupted real-time monitoring. The monitoring range covers the entire area of the warehouse. The camera executes an AI video image recognition algorithm and is connected to the environmental monitoring status and alarm recording module.
8. The substation secondary equipment intelligent warehouse system according to claim 7, It is characterized in that The image recognition module includes a flame detection module and a personnel detection module. The flame detection module is used to immediately issue an alarm when signs of flame are found, link intelligent video recording and capture, and automatically notify the warehouse management personnel by phone. The personnel detection module is used to link and turn off all lights in the warehouse when it is detected that there is no one in each area of the warehouse and the duration lasts for 10 minutes.
9. The substation secondary equipment intelligent warehouse system according to claim 8, It is characterized in that The secondary equipment warehouse uses constant temperature and humidity precision air conditioning, and is equipped with a static eliminator and an ion fan device to reduce the generation of static electricity in the warehouse.
10. The substation secondary equipment intelligent warehouse system according to claim 9, It is characterized in that The secondary equipment warehouse is equipped with water retaining dams, drainage ditches and floor drains around the ground to prevent the transfer and infiltration of underground water. Water-sensitive detection instruments or components are installed around the warehouse leakage risk areas, including under the warehouse floor and around the air conditioners, to perform waterproof detection and alarm for the warehouse.
Citation Information
Patent Citations
Secondary spare part inventory management system for power system enterprise
CN112508486A
Storehouse monitoring management system for guaranteeing safety of electric power overhaul
CN115512476A
Intelligent storage safety management system based on Internet of Things
CN116993265A
Semi-supervised video fire detection method based on consistency regularization and distribution alignment
CN117274881A
