Intelligent operator dumb resource management system and method based on artificial intelligence

Abstract

The invention relates to the technical field of communication network resource management, in particular to an operator dumb resource intelligent management system and method based on artificial intelligence, and the system comprises an intelligent data collection module, an edge intelligent processing module, a cloud AI core platform, a digital twinborn visualization module, and a natural language interaction module. The method has the beneficial effects that multi-dimensional data of dumb resources are collected through equipment such as an Internet of Things sensor and an industrial camera, the multi-dimensional data are uploaded to a cloud after being subjected to lightweight processing of edge nodes, intelligent decisions such as resource scheduling and fault diagnosis are realized by utilizing a knowledge graph and a machine learning model, and finally efficient management is realized through interaction of a digital twin interface and a natural language. The problems of inaccurate data, low scheduling efficiency, slow fault processing and the like in traditional dumb resource management are solved, the resource utilization rate and the operation and maintenance efficiency are remarkably improved, and the method is suitable for a large-scale dumb resource management scene of a communication operator.

Description

Technical Field

[0001] This invention relates to the field of communication network resource management technology, specifically to an intelligent management system and method for operator dumb resources based on artificial intelligence. Background Technology

[0002] The existing dumb resource management system of telecom operators has the following defects: Inefficient and inaccurate data collection: manual data entry error rate reaches 12%-18%, and the deviation rate between Excel ledgers and actual resource status exceeds 30% (data from a 2023 survey by a provincial mobile operator); RFID tags rely on handheld terminals for point-to-point scanning, and single-optical distribution box inventory takes 15-20 minutes, with severely delayed data updates. Resource scheduling relies on manual experience: service activation cycles are as long as 3-5 days, and cross-regional resource allocation response time exceeds 2 hours; the average utilization rate of optical fiber cores is only 55%, 20 percentage points lower than the international advanced level (ITU-T 2024 report). Time-consuming fault diagnosis and repair: fault location relies on "manual inspection + work order filling," with an average recovery time (MTTR) of 4.2 hours; 85% of faults require secondary on-site confirmation, and the duplicate work order rate reaches 25% (data from a certain operator's operations and maintenance). High management costs: the annual manpower cost for dumb resource management by operators nationwide exceeds 20 billion yuan, and the capital expenditure waste caused by resource idleness accounts for 15%-20%.

[0003] Therefore, there is a need to provide an end-to-end solution that integrates artificial intelligence technology, enabling automated management of the entire lifecycle of dumb resources through intelligent collection of multi-source data, edge-cloud collaborative computing, deep inference of machine learning models, and digital twin visualization interaction. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent management system and method for operator dumb resources based on artificial intelligence, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent management system for operator dumb resources based on artificial intelligence, comprising:

[0006] The intelligent data acquisition module is used to acquire multi-dimensional data of dumb resources through IoT sensors, visual recognition devices and positioning systems.

[0007] The edge intelligence processing module includes a lightweight AI model to enable real-time anomaly detection and local decision-making.

[0008] A cloud-based AI core platform that integrates knowledge graphs, machine learning models, and an intelligent scheduling engine;

[0009] The digital twin visualization module constructs a 3D resource model and maps its status in real time;

[0010] The natural language interaction module supports voice / text interaction for business queries and fault reporting.

[0011] Preferably, the edge intelligence processing module uses model distillation and dynamic quantization techniques to compress the parameter size of the deep learning model by more than 70% and control the inference latency to within 100ms.

[0012] Preferably, the cloud-based AI core platform includes resource prediction models and fault diagnosis models.

[0013] Preferably, the resource prediction model in the cloud-based AI core platform is based on the LSTM-Transformer hybrid architecture.

[0014] Preferably, the fault diagnosis model in the cloud-based AI core platform uses the XGBoost algorithm to fuse multi-source data.

[0015] A method for an AI-based intelligent management system for operator dumb resources includes the following steps:

[0016] Data acquisition steps: Utilize intelligent data acquisition modules to acquire multi-dimensional data of dumb resources through IoT sensors, visual recognition devices, and positioning systems;

[0017] Edge processing steps: With the help of the edge intelligent processing module, the lightweight AI model contained therein is used to perform real-time anomaly detection and local decision-making;

[0018] Cloud processing steps: Relying on the cloud AI core platform, knowledge graphs, machine learning models and intelligent scheduling engines are integrated to perform comprehensive analysis and processing of the collected and processed data;

[0019] Visualization steps: Construct a 3D resource model of the dumb resource using the digital twin visualization module and map its status in real time;

[0020] Interaction steps: Use the natural language interaction module to support voice / text interaction for business queries and fault reporting.

[0021] Preferably, in the edge processing step, the edge intelligent processing module uses model distillation and dynamic quantization techniques to compress the parameter size of the deep learning model by more than 70% and control the inference latency to within 100ms.

[0022] Preferably, in the cloud processing step, the cloud AI core platform includes a resource prediction model and a fault diagnosis model, which are used to perform in-depth analysis of dumb resource data to assist management decisions.

[0023] Preferably, in the cloud processing step, the resource prediction model is based on the LSTM-Transformer hybrid architecture to predict the future state of dummy resources, providing a basis for resource planning.

[0024] Preferably, in the cloud processing step, the fault diagnosis model uses the XGBoost algorithm to fuse multi-source data to accurately diagnose potential faults in dummy resources, so as to take timely maintenance measures.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] This invention proposes an AI-based intelligent management system and method for operator dumb resources. It collects multi-dimensional data on dumb resources through IoT sensors, industrial cameras, and other devices. After lightweight processing at edge nodes, the data is uploaded to the cloud. Knowledge graphs and machine learning models are used to achieve intelligent decision-making for resource scheduling and fault diagnosis. Finally, efficient management is achieved through a digital twin interface and natural language interaction. This invention solves the problems of inaccurate data, inefficient scheduling, and slow fault handling in traditional dumb resource management, significantly improving resource utilization and operational efficiency. It is suitable for large-scale dumb resource management scenarios in telecommunications operators. Attached Figure Description

[0027] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1, please refer to Figure 1 This invention provides a technical solution: an intelligent management system for operator dumb resources based on artificial intelligence, comprising:

[0030] The intelligent data acquisition module is used to acquire multi-dimensional data of dumb resources through IoT sensors, visual recognition devices and positioning systems.

[0031] The edge intelligence processing module includes a lightweight AI model to achieve real-time anomaly detection and local decision-making; it adopts model distillation and dynamic quantization techniques to compress the parameter size of deep learning models by more than 70% and control the inference latency to within 100ms.

[0032] The cloud-based AI core platform integrates knowledge graphs, machine learning models, and an intelligent scheduling engine. It includes resource prediction and fault diagnosis models. The resource prediction model is based on an LSTM-Transformer hybrid architecture. The fault diagnosis model uses the XGBoost algorithm to fuse multi-source data.

[0033] The digital twin visualization module constructs a 3D resource model and maps its status in real time.

[0034] The natural language interaction module supports voice / text interaction for business queries and fault reporting.

[0035] Example 2, based on Example 1, proposes a method for the intelligent management system of operator dumb resources based on artificial intelligence as described in claim 5, comprising the following steps:

[0036] (I) Multi-source data acquisition

[0037] 1. IoT Sensors: Fiber optic sensors (accuracy ±0.01dB) are deployed in the fiber optic junction box to monitor optical power, and pressure sensors (sensitivity ±0.1N) are installed on the patch panel to detect port insertion / removal status. Data is collected at a frequency of 1 time per minute.

[0038] 2. Visual recognition equipment: Industrial camera (resolution ≥ 12 megapixels) for real-time image capture of optical distribution box labels, integrated with OCR algorithm to recognize resource codes (accuracy ≥ 98%).

[0039] 3. Location Awareness: Interacts with BeiDou / GPS positioning systems (accuracy ≤ 5m) to acquire resource geographic coordinates and 3D spatial information.

[0040] (II) Data Preprocessing

[0041] 1. Outlier Filtering: Outliers in sensor data are removed using the 3σ rule.

[0042] 2. Multimodal alignment: Mapping image features (extracted by ResNet50), text features (encoded by BERT), and temporal data (encoded by LSTM) to a unified 256-dimensional feature space.

[0043] Example 3, based on Example 2, proposes the following:

[0044] Edge intelligent processing layer (edge ​​layer)

[0045] (I) Deployment of Lightweight Models:

[0046] 1. The model distillation technique (Teacher model: ResNet101 → Student model: MobileNetV3) is adopted, which reduces the parameter size by 75% and improves the inference speed by 4 times (single image recognition ≤80ms).

[0047] 2. Dynamic quantization technology (FP32→INT8) reduces memory usage by 60% and supports operation on edge nodes of ARM Cortex-A53 architecture (computing power ≥1TOPS).

[0048] (II) Real-time monitoring and local decision-making:

[0049] 1. Real-time status monitoring: Set multiple threshold levels (early warning / fault / serious fault), such as triggering a level one early warning when optical power drops by more than 3dB.

[0050] 2. Rapid response mechanism: Edge nodes store 200+ typical fault handling rules, supporting automatic work order generation for fiber breakage scenarios (delay ≤ 15 seconds).

[0051] Cloud-based AI core platform (cloud layer)

[0052] (I) Building Basic Capabilities

[0053] 1. Knowledge Graph Engine: Constructs a resource knowledge graph containing 8 types of entities (resources / ports / services / faults, etc.) and 12 types of relationships (attribution / connection / bearing, etc.), with a node scale of up to 10^8 levels, supporting complex relational queries (such as "query all optical distribution box ports with utilization > 80% in a certain area").

[0054] 2. Machine learning models

[0055] Resource forecasting model: Based on a hybrid LSTM+Transformer architecture, it takes 7 days of historical resource usage data (15-minute granularity) as input and outputs a resource demand forecast for the next 24 hours (error ≤10%).

[0056] Fault diagnosis model: Employing the XGBoost algorithm, which integrates sensor data, historical fault records, and work order text, it achieves fault type identification (accuracy ≥ 92%).

[0057] (II) Intelligent Decision Module

[0058] 1. Dynamic scheduling engine

[0059] Optimization objectives: Maximize resource utilization (weight 0.6), minimize business deployment latency (weight 0.3), and reduce energy consumption (weight 0.1).

[0060] Algorithm Implementation: Based on reinforcement learning (PPO algorithm), the state space includes resource load, service QoS requirements, and geographical location, while the action space consists of resource allocation / migration / expansion operations. It has been trained for over 100,000 rounds in a simulation environment.

[0061] 2. Work order automation system

[0062] NLP Analysis: Using the T5 model to parse unstructured chemical monogram text, 21 key entities (such as fault location and business impact) were extracted, with a field completeness of ≥95%.

[0063] Solution generation: By combining knowledge graphs and fault models, maintenance suggestions (including spare parts lists and operating procedures) are automatically generated.

[0064] Digital Twin and Interaction Layer

[0065] (I) 3D Visualization Platform:

[0066] 1. Utilize BIM + oblique photogrammetry technology to construct a 3D model of the equipment room / optical distribution box (with millimeter-level accuracy), and map the resource status in real time (green - normal / yellow - warning / red - fault).

[0067] 2. Service Link Visualization: Supports end-to-end path query and dynamically labels the resource load of each node (e.g., "Service A occupies ports 3-5 of optical distribution box J-01, with a current utilization rate of 85%").

[0068] (II) Natural Language Interaction:

[0069] 1. Intelligent Customer Service: Based on Dialogflow, it implements multi-turn dialogues, supports 100+ business intents (such as "query idle optical cables in Lixia District" and "locate port faults"), and has a response latency of ≤1 second.

[0070] 2. Voice control: Integrates automatic speech recognition (ASR) and text-to-speech (TTS), supporting resource scheduling or fault reporting triggered by voice commands.

[0071] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent management system for operator dumb resources based on artificial intelligence, characterized in that: include: The intelligent data acquisition module is used to acquire multi-dimensional data of dumb resources through IoT sensors, visual recognition devices and positioning systems. The edge intelligence processing module includes a lightweight AI model to enable real-time anomaly detection and local decision-making. A cloud-based AI core platform that integrates knowledge graphs, machine learning models, and an intelligent scheduling engine; The digital twin visualization module constructs a 3D resource model and maps its status in real time; The natural language interaction module supports voice / text interaction for business queries and fault reporting.

2. The intelligent management system for operator dumb resources based on artificial intelligence according to claim 1, characterized in that: The edge intelligence processing module uses model distillation and dynamic quantization techniques to compress the parameter size of deep learning models by more than 70% and control the inference latency to within 100ms.

3. The intelligent management system for operator dumb resources based on artificial intelligence according to claim 2, characterized in that: The cloud-based AI core platform includes resource prediction models and fault diagnosis models.

4. The intelligent management system for operator dumb resources based on artificial intelligence according to claim 3, characterized in that: The resource prediction model in the cloud AI core platform is based on the LSTM-Transformer hybrid architecture.

5. The intelligent management system for operator dumb resources based on artificial intelligence according to claim 4, characterized in that: The fault diagnosis model in the cloud-based AI core platform uses the XGBoost algorithm to fuse multi-source data.

6. A method for an AI-based intelligent management system for operator dumb resources as described in claim 5, characterized in that: Includes the following steps: Data acquisition steps: Utilize intelligent data acquisition modules to acquire multi-dimensional data of dumb resources through IoT sensors, visual recognition devices, and positioning systems; Edge processing steps: With the help of the edge intelligent processing module, the lightweight AI model contained therein is used to perform real-time anomaly detection and local decision-making; Cloud processing steps: Relying on the cloud AI core platform, knowledge graphs, machine learning models and intelligent scheduling engines are integrated to perform comprehensive analysis and processing of the collected and processed data; Visualization steps: Construct a 3D resource model of the dumb resource using the digital twin visualization module and map its status in real time; Interaction steps: Use the natural language interaction module to support voice / text interaction for business queries and fault reporting.

7. A method according to claim 6, characterized in that: In the edge processing step, the edge intelligent processing module uses model distillation and dynamic quantization techniques to compress the parameter size of the deep learning model by more than 70% and control the inference latency to within 100ms.

8. A method according to claim 7, characterized in that: In the cloud processing steps, the cloud-based AI core platform includes resource prediction models and fault diagnosis models, which are used to perform in-depth analysis of dumb resource data to assist management decisions.

9. A method according to claim 8, characterized in that: In the cloud processing steps, the resource prediction model, based on the LSTM-Transformer hybrid architecture, predicts the future state of dummy resources, providing a basis for resource planning.

10. A method according to claim 9, characterized in that: In the cloud processing steps, the fault diagnosis model uses the XGBoost algorithm to fuse multi-source data to accurately diagnose potential faults in dummy resources, so as to take timely maintenance measures.

Images