Optimization method and device for poor-quality cell problem, electronic equipment and storage medium

Abstract

The invention provides an optimization method and device for a poor-quality cell problem, electronic equipment and a computer readable storage medium, and relates to the technical field of network optimization. The optimization method comprises the following steps: identifying an initial intention task of a poor-quality cell problem based on a large language model and a poor-quality cell index problem rule base; based on a large language model and a poor-quality cell experience library, optimizing the initial intention task to obtain a target intention task of a poor-quality cell problem; and inputting the target intention task into a large language model, and outputting an optimization decision suggestion of the poor-quality cell problem. The problems that in the related technology, the content and form of an intention task are single and rigid, the adaptability to the actual situation is weak, and the capacity of a large language model cannot be fully utilized are at least solved. The method is suitable for poor quality identification and optimization scenes.

Description

Technical Field

[0001] The present invention relates to the field of network optimization technologies, and in particular, to an optimization method, device, electronic device, and computer-readable storage medium for poor-quality cell problems. Background Art

[0002] A poor-quality cell refers to a cell with one or more poor-quality problems in network communication quality. The types of cell poor-quality problems involve multiple aspects, such as voice service connection rate, call drop rate, packet loss rate, data service disconnection rate, access success rate, and user perceived rate.

[0003] Currently, the identification of the types of poor-quality cell problems mainly adopts the traditional network optimization (abbreviation: network optimization) method. Specifically: first, collect various key performance indicators (KPIs) of the cell; then, network optimization staff can set thresholds for each KPI relying on their own expert knowledge and experience, and then judge the types of poor-quality cell problems based on the KPI thresholds and the various KPIs of the cell; afterwards, rely on network optimization staff to analyze various data, manually check various network problems such as coverage, interference, and capacity, and conduct root cause analysis and formulate optimization solutions based on the experience of network optimization staff, so as to optimize network problems. However, using the traditional network optimization method to identify the types of poor-quality cell problems has low efficiency, poor accuracy, high requirements for the level of optimization personnel, but the user perception has not been greatly improved, and it is difficult to meet the current large-scale network optimization requirements.

[0004] In addition, the existing technologies for identifying and optimizing poor-quality cell problems using large language models (LLMs) usually only assemble intent tasks by matching the vectors of poor-quality cell problems. The content and form of the intent tasks are relatively single and rigid, and the adaptability to the actual situation is weak, and the existing technologies cannot make full use of the capabilities of large language models. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an optimization method, device, electronic device, and computer-readable storage medium for poor-quality cell problems in view of the above deficiencies of the prior art. This method can improve the flexibility, accuracy, and comprehensiveness of intent task generation, enhance the guiding effect of intent tasks on large language models, and further enhance the intent recognition ability of large language models, improve the efficiency and effect of optimizing poor-quality cell problems, and achieve rapid and effective optimization of poor-quality cell problems.

[0006] In a first aspect, the present invention provides an optimization method for poor-quality cell problems, including: identifying an initial intent task for poor-quality cell problems based on a large language model and a rule library for poor-quality cell index problems; optimizing the initial intent task based on a large language model and a poor-quality cell experience library to obtain a target intent task for poor-quality cell problems; and inputting the target intent task into the large language model to output an optimization decision suggestion for poor-quality cell problems.

[0007] Preferably, the step of identifying an initial intent task for poor-quality cell problems based on a large language model and a rule library for poor-quality cell index problems specifically includes: identifying a poor-quality cell ID and poor-quality cell province / city information from the poor-quality cell problems based on the large language model, and generating a first intent task for the poor-quality cell problems; obtaining problem judgment index data for the poor-quality cell problems based on the first intent task and the large language model, and obtaining rule judgment conditions for the poor-quality cell problems from the rule library for poor-quality cell index problems; and optimizing the first intent task based on the problem judgment index data and the rule judgment conditions to obtain an initial intent task for the poor-quality cell problems.

[0008] Preferably, the step of optimizing the first intent task based on the problem judgment index data and the rule judgment conditions to obtain an initial intent task for the poor-quality cell problems specifically includes: assembling the first intent task, the problem judgment index data, and the rule judgment conditions to obtain a second intent task for the poor-quality cell problems; inputting the second intent task into the large language model to identify the poor-quality type of the poor-quality cell problems; and assembling the poor-quality type and the second intent task to obtain an initial intent task for the poor-quality cell problems.

[0009] Preferably, the poor-quality cell experience library includes a poor-quality cell analysis experience library and a poor-quality cell decision suggestion library. The step of optimizing the initial intent task based on a large language model and a poor-quality cell experience library to obtain a target intent task for poor-quality cell problems specifically includes: obtaining root cause analysis experience for the poor-quality cell problems from the poor-quality cell analysis experience library based on the initial intent task and the large language model; optimizing the initial intent task based on the root cause analysis experience to obtain a third intent task for the poor-quality cell problems; obtaining decision suggestion cases for the poor-quality cell problems from the poor-quality cell decision suggestion library based on the third intent task and the large language model; and optimizing the third intent task based on the decision suggestion cases to obtain a target intent task for the poor-quality cell problems.

[0010] Preferably, based on the root cause analysis experience, optimizing the initial intent task to obtain a third intent task for the poor quality cell problem specifically includes: assembling the root cause analysis experience and the initial intent task to obtain a fourth intent task for the poor quality cell problem; inputting the fourth intent task into a large language model to determine the root cause localization tool and root cause analysis indicators for the poor quality cell problem; assembling the root cause localization tool, the root cause analysis indicators, and the fourth intent task to obtain a fifth intent task for the poor quality cell problem; inputting the fifth intent task into a large language model to locate the root cause of the poor quality cell problem; assembling the root cause and the fifth intent task to obtain a third intent task for the poor quality cell problem.

[0011] Preferably, based on the decision-making advice cases, optimizing the third intent task to obtain a target intent task for the poor quality cell problem specifically includes: assembling the decision-making advice cases and the third intent task to obtain a target intent task for the poor quality cell problem.

[0012] Preferably, the problem judgment index data includes at least one of the following: weak coverage ratio, proportion of sampling points with MRO - SINR ≥ 0 under measurement report offset, excellent rate of channel quality indicator CQI, average interference noise power of physical resource block PRB, number of days of high - load cells in cell capacity statistics within a preset period. The root cause localization tool includes at least one of the following: over - coverage diagnosis analysis tool, near - coverage diagnosis analysis tool, overlapping - coverage diagnosis analysis tool, MOD 30 interference diagnosis analysis tool, uplink interference type diagnosis analysis tool, uplink PRB utilization rate load diagnosis analysis tool, downlink PRB utilization rate load diagnosis analysis tool. The root cause analysis indicators include at least one of the following: over - coverage rate, near - coverage rate, overlapping - coverage ratio, MOD 30 interference, uplink interference type, uplink PRB utilization rate at the busiest time of the cell, and downlink PRB utilization rate at the busiest time of the cell.

[0013] In a second aspect, the present invention further provides an optimization device for poor quality cell problems, including an identification module, an optimization module, and an output module. The identification module is used to identify the initial intent task of the poor quality cell problem based on a large language model and a poor quality cell index problem rule base. The optimization module is connected to the identification module and is used to optimize the initial intent task based on a large language model and a poor quality cell experience base to obtain a target intent task for the poor quality cell problem. The output module is connected to the optimization module and is used to input the target intent task into the large language model and output an optimization decision - making advice for the poor quality cell problem.

[0014] In a third aspect, the present invention further provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to implement the optimization method for the poor-quality cell problem provided in the first aspect above.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the optimization method for the poor-quality cell problem provided in the first aspect above is implemented.

[0016] An optimization method, device, electronic device, and computer-readable storage medium for the poor-quality cell problem provided by the present invention utilize the rule library of the poor-quality cell index problem and the large speech model to identify the initial intention task, and then utilize the experience library of the poor-quality cell and the large language model to optimize the initial intention task to generate a target intention task. The target intention task for the poor-quality cell problem is flexibly and accurately generated from the poor-quality cell problem, enabling the target intention task to more effectively guide the large language model to generate optimization decision suggestions for the poor-quality cell problem. Therefore, the present invention can improve the flexibility, accuracy, and comprehensiveness of intention task generation, enhance the guiding effect of the intention task on the large language model, thereby enhancing the intention recognition ability of the large language model, improving the efficiency and effect of optimizing the poor-quality cell problem, and achieving rapid and effective optimization of the poor-quality cell problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of an optimization method for a poor-quality cell problem according to Embodiment 1 of the present invention;

[0018] Figure 2 It is a flowchart of another optimization method for a poor-quality cell problem according to Embodiment 2 of the present invention;

[0019] Figure 3 It is a flowchart of another optimization method for a poor-quality cell problem according to Embodiment 2 of the present invention;

[0020] Figure 4 It is a schematic structural diagram of an optimization device for a poor-quality cell problem according to Embodiment 3 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To enable those skilled in the art to better understand the technical solutions of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0022] It can be understood that the specific embodiments and the accompanying drawings described herein are only for explaining the present invention and are not intended to limit the present invention.

[0023] It can be understood that, without conflict, the various embodiments in the present invention and the various features in the embodiments can be combined with each other.

[0024] It is understood that for the convenience of description, only the parts related to the present invention are shown in the drawings of the present invention, and the parts unrelated to the present invention are not shown in the drawings.

[0025] It is understood that each unit and module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures. Alternatively, multiple units and modules may also be integrated into one entity structure.

[0026] It is understood that without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0027] It is understood that in the flowcharts and block diagrams of the present invention, the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the embodiments of the present invention are shown. Among them, each block in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified function. Moreover, each block or combination of blocks in the block diagram and flowchart may be implemented by a hardware-based system for implementing the specified function, or may be implemented by a combination of hardware and computer instructions.

[0028] It is understood that the units and modules involved in the embodiments of the present invention may be implemented in software or in hardware. For example, the units and modules may be located in a processor.

[0029] Embodiment 1:

[0030] As Figure 1 shown, this embodiment provides an optimization method for poor-quality cell problems. The optimization method for poor-quality cell problems includes:

[0031] S101, based on the large language model and the rule library of poor-quality cell index problems, identify the initial intent task of poor-quality cell problems.

[0032] In this embodiment, as shown in Table 1, the rule library for poor-quality cell indicators is used to store the rule judgment conditions for poor-quality cell problems in different communication scenarios. Among them, the communication scenarios include network, province, city, and poor-quality type. The network includes but is not limited to the 5G (5th Generation Mobile Communication Technology) network, and the poor-quality types include but are not limited to: coverage type, quality type, interference type, and capacity type. The rule judgment conditions include but are not limited to: the weak coverage ratio at the daily level is greater than 10%, the sampling point ratio of the daily-level MRO-SINR (Mobility Reference Signal Signal-to-Interference plus Noise Ratio) ≥ 0 is less than 90%, the excellent rate of the daily-level CQI (Channel Quality Indicator) is less than 95%, the average interference noise power (dBm) per PRB (Physical Resource Block) during busy hours is greater than -105 dBm (Decibel-milliwatts), the downlink PRB utilization rate during busy hours is greater than 80%, and the weak coverage ratio refers to the ratio of RSRP (Reference Signal Received Power) <= -105. The intention task refers to the prompt task (i.e., the prompt word) that is used to clarify what the large language model needs to execute.

[0033] Table 1 Rule Library for Poor-quality Cell Indicators

[0034]

[0035]

[0036]

[0037] Specifically, S101: Based on the large language model and the rule library for poor-quality cell indicators, identify the initial intention task of the poor-quality cell problem, including steps S1011 - S1013:

[0038] S1011, based on the large language model, identify the poor-quality cell ID from the poor-quality cell problem and generate the first intention task of the poor-quality cell problem.

[0039] In this embodiment, a large language model is used as an interaction entry point. The user inputs the problem of poor-quality cells in the network into the dialog box of the large language model through natural language interaction. Examples of poor-quality cell problems are as follows: 1) Please give suggestions for diagnosing and optimizing the poor-quality problem of the 5G cell with ID = 9208737793 in City a1, Province A on December 24th? 2) What is the analysis result of the poor-quality of the 5G cell with ID = 9208737793 in City a1, Province A on December 24th?

[0040] The large language model identifies the poor-quality cell ID and the province and city information of the poor-quality cell from the poor-quality cell problems input by the user, and at the same time assembles the poor-quality cell ID and the province and city information of the poor-quality cell to generate the first-intention task of the poor-quality cell problem, such as:

[0041] ① Cell ID = 9208737793, Province = Province A, City = City a1. In this embodiment, by using a large language model, the poor-quality cell ID and its corresponding province and city information can be effectively identified and extracted, ensuring higher accuracy in data processing and reducing errors caused by manual intervention; by generating the first-intention task of the poor-quality cell problem, it helps to automate the processing of poor-quality cell problems, improve work efficiency, and reduce labor costs.

[0042] S1012, based on the first-intention task and the large language model, obtain the problem judgment index data of the poor-quality cell problem, and obtain the rule judgment conditions of the poor-quality cell problem from the poor-quality cell index problem rule library.

[0043] Specifically, the problem judgment index data includes at least one of the following: weak coverage ratio, proportion of sampling points with signal-to-interference-plus-noise ratio MRO-SINR ≥ 0 under measurement report offset, excellent rate of channel quality indicator CQI, average interference noise power of physical resource block PRB, number of days with high-load cells in cell capacity statistics within a preset period.

[0044] In this embodiment, the problem judgment index data refers to the actual values stored in the operator's database and corresponding to the problem judgment index and the satisfaction method. The rule judgment condition refers to the judgment threshold corresponding to the problem judgment index and the satisfaction method. The first-intention task of the poor-quality cell problem is input into the large language model again. The large language model calls the API (Application Programming Interface) to obtain the actual values of the weak coverage ratio, the proportion of sampling points with MRO-SINR≥0, the excellent CQI rate, the average interference noise power of the PRB, and the number of days of high-load cells in the cell capacity statistics within a preset period corresponding to the first-intention task from the operator's database. For example: the weak coverage ratio is 25.98%, the proportion of sampling points with MRO-SINR≥0 is 95.4%, the excellent CQI rate is 98.63%, the average interference noise power of the PRB is -106 dBm, and the number of days of high-load cells in the cell capacity statistics within a preset period is 78%.

[0045] While obtaining the weak coverage ratio, the proportion of sampling points with MRO-SINR≥0, the excellent CQI rate, the average interference noise power of the PRB, and the number of days of high-load cells in the cell capacity statistics within a preset period corresponding to the first-intention task from the operator's database, the large language model also obtains from the poor-quality cell index problem rule library that: the daily weak coverage ratio is greater than 10%, the daily proportion of sampling points with MRO-SINR≥0 is less than 90%, the daily excellent CQI rate is less than 95%, the average interference noise power of each PRB during busy hours is greater than -105 dBm, and the downlink PRB utilization rate during busy hours is greater than 80%.

[0046] It should be noted that when the downlink PRB utilization rate during busy hours is high, it means that a large number of users are performing data transmission in the cell, and the network is operating at full load, which will cause the cell to reach the high-load standard, thereby increasing the number of days of high-load cells. On the contrary, when the downlink PRB utilization rate during busy hours remains at a high level, it will often be recorded as the situation of high-load cell appearance. Therefore, the actual value of the number of days of high-load cells in the cell capacity statistics within a preset period can be regarded as the actual value of the downlink PRB utilization rate.

[0047] S1013, optimize the first-intention task based on the problem judgment index data and the rule judgment condition to obtain the initial intention task of the poor-quality cell problem.

[0048] In this embodiment, by obtaining the problem judgment index data and rule judgment conditions corresponding to the poor quality cell problem, and combining the problem judgment index data and rule judgment conditions, a more comprehensive and accurate judgment basis can be obtained to ensure the correct identification and handling of the poor quality cell problem; by optimizing the first intention task with the problem judgment index data and rule judgment conditions, different poor quality cell conditions can be flexibly handled, improving the adaptability and response speed of the system. The optimized initial intention task is more rigorous in data processing logic and can effectively reduce the consequences of judgment errors.

[0049] Specifically, S1013: Optimize the first intention task based on the problem judgment index data and rule judgment conditions to obtain the initial intention task for the poor quality cell problem, including: Assembling the first intention task, the problem judgment index data, and the rule judgment conditions to obtain the second intention task for the poor quality cell problem; inputting the second intention task into the large language model to identify the poor quality type of the poor quality cell problem; assembling the poor quality type and the second intention task to obtain the initial intention task for the poor quality cell problem.

[0050] In this embodiment, the second intention task for the poor quality cell problem is as follows:

[0051] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0052] ② Problem judgment index data: The weak coverage ratio is 25.98%, the proportion of sampling points with MRO - SINR ≥ 0 is 95.4%, the excellent rate of CQI is 98.63%, the average interference noise power of PRB is -106 dBm, and the number of days with high - load cells in the cell capacity statistics within the preset period is 78%;

[0053] ③ Rule judgment conditions: The daily weak coverage ratio is greater than 10%, the daily proportion of sampling points with MRO - SINR ≥ 0 is less than 90%, the daily excellent rate of CQI is less than 95%, the average interference noise power per PRB during busy hours is greater than -105 dBm, and the downlink PRB utilization rate during busy hours is greater than 80%.

[0054] Input the second intention task into the large language model, and the large language model determines whether the problem judgment index data meets the rule judgment conditions to identify that the poor quality type of the poor quality cell problem is the coverage type. Assembling the poor quality type and the second intention task, the initial intention task for the poor quality cell problem is:

[0055] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0056] ② Problem judgment index data: The weak coverage ratio is 25.98%;

[0057] ③ Rule judgment condition: The weak coverage ratio at the day level is greater than 10%.

[0058] ④ Quality difference type: Coverage type. In this embodiment, by identifying the specific quality difference type of the quality difference cell problem, the quality difference problem can be classified more carefully, which helps for targeted analysis and processing; by combining the second intention task with the identified quality difference type, the context information of the task becomes more complete, which can provide richer background information, thereby enhancing the ability to understand the problem; integrating the quality difference type with the initial intention task helps to optimize the decision-making process and can formulate solutions for specific problems more quickly and accurately.

[0059] S102. Optimize the initial intention task based on the large language model and the quality difference cell experience library to obtain the target intention task of the quality difference cell problem.

[0060] Specifically, the quality difference cell experience library includes a quality difference cell analysis experience library and a quality difference cell decision-making suggestion library.

[0061] In this embodiment, as shown in Table 2, the quality difference cell analysis experience library is used to store the root cause analysis experiences of different quality types of cells (i.e., the root cause analysis experiences of coverage type, quality type, interference type, and capacity type). By analyzing the analysis process of the root cause analysis experiences of different quality types, this embodiment can further refine the quality difference cell analysis experience library shown in Table 2 to obtain the quality difference cell analysis experience library shown in Table 3.

[0062] Table 2 Quality difference cell analysis experience library

[0063]

[0064] Table 3 Quality difference cell analysis experience library

[0065]

[0066]

[0067] Specifically, S102: Optimize the initial intention task based on the large language model and the quality difference cell experience library to obtain the target intention task of the quality difference cell problem, including steps S1021 - S1024:

[0068] S1021. Based on the initial intention task and the large language model, obtain the root cause analysis experience of the quality difference cell problem from the quality difference cell analysis experience library.

[0069] In this embodiment, the initial intention task is input into the large language model, and the large language model obtains the root cause analysis experience of the quality difference cell problem from the quality difference cell analysis experience library shown in Table 1 as: determining the root cause result of the coverage problem according to the over-coverage and near-coverage analysis processes.

[0070] S1022. Optimize the initial intent task based on the root cause analysis experience to obtain the third intent task for the poor quality cell problem.

[0071] Specifically, S1022: Optimize the initial intent task based on the root cause analysis experience to obtain the third intent task for the poor quality cell problem, including: Assemble the root cause analysis experience and the initial intent task to obtain the fourth intent task for the poor quality cell problem; Input the fourth intent task into the large language model to determine the root cause location tool and root cause analysis indicators for the poor quality cell problem; Assemble the root cause location tool, root cause analysis indicators, and the fourth intent task to obtain the fifth intent task for the poor quality cell problem; Input the fifth intent task into the large language model to locate the root cause of the poor quality cell problem; Assemble the root cause and the fifth intent task to obtain the third intent task for the poor quality cell problem.

[0072] In this embodiment, assembling the root cause analysis experience and the initial intent task to obtain the fourth intent task for the poor quality cell problem is as follows:

[0073] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0074] ② Problem judgment index data: The weak coverage ratio is 25.98%;

[0075] ③ Rule judgment condition: The daily weak coverage ratio is greater than 10%;

[0076] ④ Poor quality type: Coverage type;

[0077] ⑤ Root cause analysis experience: Determine the root cause result of the coverage problem according to the over-coverage and near-coverage analysis processes.

[0078] Specifically, the root cause location tool includes at least one of the following: Over-coverage diagnosis analysis tool, near-coverage diagnosis analysis tool, overlapping coverage diagnosis analysis tool, MOD 30 interference diagnosis analysis tool, uplink interference type diagnosis analysis tool, uplink PRB utilization rate load diagnosis analysis tool, downlink PRB utilization rate load diagnosis analysis tool.

[0079] Specifically, the root cause analysis indicators include at least one of the following: Over-coverage rate, near-coverage rate, overlapping coverage ratio, MOD 30 interference, uplink interference type, uplink PRB utilization rate at the busiest time of the cell, and downlink PRB utilization rate at the busiest time of the cell.

[0080] In this embodiment, input the fourth intent task into the large language model. The large language model can determine the root cause location tools for the poor quality cell problem as the over-coverage diagnosis analysis tool and the near-coverage diagnosis analysis tool according to the over-coverage and near-coverage analysis processes, and the root cause analysis indicators for the poor quality cell problem are shown in Table 4.

[0081] Table 4 Root Cause Analysis Metrics for Poor Quality Cell Problems

[0082]

[0083] Assemble the root cause location tool, root cause analysis metrics, and the fourth intention task to obtain the fifth intention task for the poor quality cell problem as follows:

[0084] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0085] ② Problem judgment metric data: Weak coverage ratio is 25.98%;

[0086] ③ Rule judgment condition: The daily weak coverage ratio is greater than 10%;

[0087] ④ Poor quality type: Coverage type;

[0088] ⑤ Root cause analysis experience: Determine the root cause results of the coverage problem according to the over-coverage and near-coverage analysis processes;

[0089] ⑥ Root cause location tool: Over-coverage diagnosis analysis tool and near-coverage diagnosis analysis tool;

[0090] ⑦ Root cause analysis metrics: Over-coverage rate = 62.18%, near-coverage rate = 1.86%;

[0091] ⑧ Analysis judgment condition: The reasonable range of the over-coverage rate is 0 - 25%, and the reasonable range of the near-coverage rate is 0 - 20%.

[0092] Based on the fifth intention task, the large language model can locate the root cause of the poor quality cell problem as over-coverage, and assemble the root cause and the fifth intention task to obtain the third intention task for the poor quality cell problem as follows:

[0093] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0094] ② Problem judgment metric data: Weak coverage ratio is 25.98%;

[0095] ③ Rule judgment condition: The daily weak coverage ratio is greater than 10%;

[0096] ④ Poor quality type: Coverage type;

[0097] ⑤ Root cause analysis metrics: Over-coverage rate = 62.18%;

[0098] ⑥ Root cause: over-coverage. In this embodiment, through the root cause analysis experience extracted from the poor-quality cell analysis experience library, the root cause of the poor-quality cell problem can be more accurately located and understood, which helps to reduce the misjudgment of surface phenomena. Furthermore, by optimizing the initial intention task based on the feedback of the root cause analysis, the formed third intention task shows stronger pertinence and operability, making subsequent decisions more scientific and effective.

[0099] S1023. Based on the third intention task and the large language model, obtain decision-making suggestion cases for the poor-quality cell problem from the poor-quality cell decision-making suggestion library.

[0100] In this embodiment, the poor-quality cell decision-making suggestion library is used to store decision-making suggestions under different poor-quality types and root causes determined according to expert experience, as shown in Table 5.

[0101] Table 5 Poor-quality cell decision-making suggestion library

[0102]

[0103]

[0104] Input the third intention task into the large language model. The decision-making suggestion cases obtained by the large language model from Table 5 for the poor-quality cell problem are: 1) Check the downtilt angle; 2) Check the power configuration parameters; 3) Check the base station hanging height platform; 4) Check the azimuth angle.

[0105] S1024. Optimize the third intention task based on the decision-making suggestion cases to obtain the target intention task for the poor-quality cell problem.

[0106] Specifically, S1024: Optimize the third intention task based on the decision-making suggestion cases to obtain the target intention task for the poor-quality cell problem, including: Assemble the decision-making suggestion cases and the third intention task to obtain the target intention task for the poor-quality cell problem.

[0107] In this embodiment, the target intention task for the poor-quality cell problem obtained by assembling the decision-making suggestion cases and the third intention task is as follows:

[0108] ① Cell ID = 9208737793, Province = Province A, City = City a1;

[0109] ② Problem judgment index data: The weak coverage ratio is 25.98%;

[0110] ③ Rule judgment condition: The daily weak coverage ratio is greater than 10%;

[0111] ④ Poor-quality type: Coverage type;

[0112] ⑤ Root cause analysis index: Over-coverage rate = 62.18%;

[0113] ⑥Root cause: over-coverage;

[0114] ⑦Decision-making suggestion cases are as follows: 1) Checking the downtilt angle; 2) Checking the power configuration parameters; 3) Checking the base station hanging height platform; 4) Checking the azimuth angle.

[0115] S103: Input the target intention task into the large language model to output optimization decision-making suggestions for the poor-quality cell problem.

[0116] In this embodiment, the optimization decision-making suggestion refers to the optimization means corresponding to the poor-quality cell problem. Therefore, the optimization decision-making suggestion includes at least one of the decision-making suggestion cases. Input the target intention task into the large language model, and the large language model queries the cell engineering parameter data and outputs optimization decision-making suggestions for the poor-quality cell problem. Among them, the cell group engineering parameter data includes but is not limited to: station height, mechanical downtilt angle, electronic downtilt angle, and azimuth angle. By using the large language model to introduce actual decision-making suggestion cases from the poor-quality cell decision-making suggestion library, this embodiment provides rich reference materials and historical experience for decision-making, making the decision-making suggestions more systematic and practical, improving the effectiveness and feasibility of the decision-making suggestions, being able to better handle the poor-quality cell problem, reducing the risk of decision-making errors, making full use of the natural language processing ability of the large language model, enhancing the understanding and analysis ability of complex business scenarios, and improving the intelligent level.

[0117] It should be noted that the downtilt angle check refers to the optimization means of adjusting the downtilt angle configuration to improve the weak downlink coverage by checking the electronic downtilt angle and mechanical downtilt angle of the antenna feeder; the azimuth angle check refers to the optimization means of reasonably planning and adjusting the site azimuth angle in combination with the base station coverage range; the power configuration parameter check refers to the optimization means of evaluating the maximum transmit power supported in combination with the base station equipment capabilities and modifying the engineering parameter according to the actual situation; the base station hanging height platform check refers to the optimization means of evaluating the base station coverage range, checking the antenna hanging height, and improving the weak downlink coverage by adjusting the antenna feeder hanging height.

[0118] The planned site evaluation refers to the optimization means of evaluating the construction of a new site in the weak coverage area to improve the weak downlink coverage; effectively controlling the coverage rationality through RF optimization refers to the optimization means of adjusting the azimuth angle and downtilt angle to avoid the problem of the main direction of the antenna feeder being blocked; the on-site inspection refers to the optimization means of checking whether there are billboards or other physical obstructions such as beautifying antennas at the site and making rectifications if necessary; the site relocation refers to the optimization means of relocating the site for the problem of over-close coverage that cannot be improved through rectification due to serious obstruction; adding a new base station refers to the optimization means of suggesting to add a newly planned station in the surrounding area if the base station coverage cannot be optimized and improved; the neighboring cell optimization refers to the optimization means of optimizing the SINR poor quality caused by neighboring cell problems.

[0119] Combined with the specific parameters of different manufacturers, modify the CQI configuration parameters, specifically including: CQI reliability optimization switch, PUCCH (Physical Uplink Control Channel) channel optimization and improvement switch, optimization of the normalized PDSCH (Physical Downlink Shared Channel) power offset, optimization of the CQI misdetection algorithm, power offset optimization, PUCCH channel optimization to improve CQI reliability, enabling the periodic sub-band CQI reporting function, NPDSCH EPRE OFF (Non-Preemptive Dynamic Scheduled Preemption Off) optimization, and enabling CQI adaptation.

[0120] Adjusting the antenna downtilt angle is an optimization method to optimize the antenna coverage range, reduce signal interference, and improve spectrum efficiency; adjusting the antenna azimuth angle is an optimization method to enhance the signal directivity, improve the signal quality, enhance the system performance, and reduce the overlapping coverage rate; adjusting the antenna hanging height is an optimization method to improve the signal propagation effect, ensure that the coverage range matches the serving cell, and reduce the influence of occlusion; site rectification and relocation refer to the optimization method of rectifying the rooftop or relocating the site for interfering cells around the site due to unreasonable site location; replacing the site frequency band is an optimization method to avoid frequency interference, improve the data transmission rate, and optimize the network performance; adjusting the cell reference power is an optimization method to balance the coverage range, optimize the signal strength, and reduce interference.

[0121] PCI replanning refers to the optimization method of reasonably planning the cell PCI in combination with the whole network PCI planning data to ensure the PCI reuse distance and MOD value interference; antenna adjustment refers to the optimization method of adjusting the antennas of cells with the same MOD value to avoid interference between cells with the same MOD value; power modification refers to the optimization method of reducing the transmit power of interfering cells and reducing the interference signal level.

[0122] If the quality degradation type of the poor-quality cell problem is interference-related, then input the target intention task into the large language model. The large language model queries the hourly PRB-level waveform diagram of the cell and outputs optimization decision suggestions for the poor-quality cell problem based on the hourly PRB-level waveform diagram of the cell.

[0123] Increasing the isolation between two systems includes: raising the antenna height of the interfering source base station or the interfered base station to change the horizontal isolation to vertical isolation. Generally, the vertical isolation is more than 10 dB greater than the horizontal isolation. Checking the anti-interference ability of the equipment in the interfered cell refers to the optimization method of replacing the interfered equipment with equipment with stronger anti-blocking ability.

[0124] External interference troubleshooting, specifically including: Method 1: Direction determination, rendering the background noise value data into an intuitive heat map to guide the field engineer to troubleshoot the interference location and direction; Method 2: Using a drone to quickly and effectively locate the interference location and direction; Method 3: Conducting a manual carpet search to determine the final interference location; Method 4: The three-point positioning method.

[0125] The three-point positioning method specifically includes: Step 1: Determine several high points in the interference area. The first point is generally the building where the interfered base station is located. Use a spectrum analyzer to conduct a 360-degree test to find the direction with the strongest interference; Step 2: Move forward along the direction with the strongest interference to find the second point. Similarly, conduct a 360-degree frequency sweep and compare it with the interference intensity and direction of the first point to determine the approximate location of the interference source; Step 3: Conduct frequency sweeps at different high points in turn until the interference source is determined. Point selection principle: Select buildings with an open view and a relatively high position for frequency sweeping, which is convenient for observing the surrounding environment, checking whether there are devices such as antennas and amplifiers, and avoiding interference being blocked, which affects the judgment.

[0126] An optimization method for poor-quality cell problems provided in this embodiment, by making full use of the rule library of poor-quality cell index problems and the large voice model to identify the initial intention task, and then using the experience library of poor-quality cells and the large language model to optimize the initial intention task to generate the target intention task. Generate the target intention task for poor-quality cell problems flexibly and accurately from poor-quality cell problems, so that the target intention task can more effectively guide the large language model to generate optimization decision suggestions for poor-quality cell problems, improve the flexibility, accuracy and comprehensiveness of intention task generation, enhance the guiding effect of the intention task on the large language model, and then enhance the intention recognition ability of the large language model, improve the efficiency and effect of optimizing poor-quality cell problems, and achieve the rapid and effective optimization of poor-quality cell problems.

[0127] Embodiment 2:

[0128] As Figure 2 and Figure 3 shown, this embodiment provides an optimization method for poor-quality cell problems. The optimization method for poor-quality cell problems includes:

[0129] S201, based on the large language model, identify the poor-quality cell ID and the provincial, city and county information of the poor-quality cell from the poor-quality cell problem, and generate the first intention task of the poor-quality cell problem.

[0130] In this embodiment, the intention task is the Figure 2 prompt in, identifying the poor-quality cell ID and the provincial, city and county information of the poor-quality cell from the poor-quality cell problem, and generating the first intention task of the poor-quality cell problem is Figure 2 ① filling the user problem input in the prompt in.

[0131] S202. Based on the first-intention task and the large language model, obtain the problem judgment index data of the poor-quality cell problem, and obtain the rule judgment conditions of the poor-quality cell problem from the poor-quality cell index problem rule library.

[0132] In this embodiment, obtaining the problem judgment index data of the poor-quality cell problem is Figure 2 ② API interface call to query cell indexes in, and obtaining the rule judgment conditions of the poor-quality cell problem is Figure 2 ④ Cell province and city matching in.

[0133] S203. Assemble the first-intention task, the problem judgment index data, and the rule judgment conditions to obtain the second-intention task of the poor-quality cell problem.

[0134] In this embodiment, assembling the first-intention task, the problem judgment index data, and the rule judgment conditions is Figure 2 ⑤ Assembling context information in.

[0135] S204. Input the second-intention task into the large language model to identify the poor-quality type of the poor-quality cell problem; assemble the poor-quality type and the second-intention task to obtain the initial-intention task of the poor-quality cell problem.

[0136] In this embodiment, the large language model is Figure 2 LLM in, and the poor-quality type is Figure 2 The poor-quality cell type in.

[0137] S205. Based on the initial-intention task and the large language model, obtain the root cause analysis experience of the poor-quality cell problem from the poor-quality cell analysis experience library.

[0138] In this embodiment, obtaining the root cause analysis experience of the poor-quality cell problem is Figure 2 ⑦ Analysis experience recall in.

[0139] S206. Assemble the root cause analysis experience and the initial-intention task to obtain the fourth-intention task of the poor-quality cell problem.

[0140] In this embodiment, assembling the root cause analysis experience and the initial-intention task is Figure 2 ⑧ Assembling context information in.

[0141] S207. Input the fourth intention task into the large language model to determine the root cause location tool and root cause analysis indicators for the poor quality cell problem; assemble the root cause location tool, root cause analysis indicators, and the fourth intention task to obtain the fifth intention task for the poor quality cell problem; input the fifth intention task into the large language model to locate the root cause of the poor quality cell problem; assemble the root cause and the fifth intention task to obtain the third intention task for the poor quality cell problem; based on the third intention task and the large language model, obtain the decision-making suggestion cases for the poor quality cell problem from the poor quality cell decision-making suggestion library; assemble the decision-making suggestion cases and the third intention task to obtain the target intention task for the poor quality cell problem; input the target intention task into the large language model to output the optimized decision-making suggestion for the poor quality cell problem.

[0142] In this embodiment, inputting the fourth intention task into the large language model to determine the root cause location tool and root cause analysis indicators for the poor quality cell problem means Figure 2 in ⑨ the prompt guides the LLM to select tools and Figure 3 in ① call the API interface to query the cell root cause analysis indicators; assembling the root cause location tool, root cause analysis indicators, and the fourth intention task means Figure 3 in ② assemble the context information; obtaining the decision-making suggestion cases for the poor quality cell problem means Figure 3 in ③ decision-making suggestion recall; assembling the decision-making suggestion cases and the third intention task means Figure 3 in ④ assemble the context information.

[0143] An optimization method for the poor quality cell problem provided in this embodiment, by making full use of the poor quality cell index problem rule library and the large voice model to identify the initial intention task, and then using the poor quality cell experience library and the large language model to optimize the initial intention task to generate the target intention task. Flexibly and accurately generate the target intention task for the poor quality cell problem from the poor quality cell problem, so that the target intention task can more effectively guide the large language model to generate the optimized decision-making suggestion for the poor quality cell problem, improve the flexibility, accuracy, and comprehensiveness of the intention task generation, enhance the guiding effect of the intention task on the large language model, and further enhance the intention recognition ability of the large language model, improve the efficiency and effect of the poor quality cell problem optimization, and achieve the rapid and effective optimization of the poor quality cell problem.

[0144] Embodiment 3:

[0145] As Figure 4As shown in the figure, this embodiment provides an optimization device for poor-quality cell problems, including an identification module 31, an optimization module 32, and an output module 33. The identification module 31 is configured to identify the initial intention task of the poor-quality cell problem based on a large language model and a rule library for poor-quality cell index problems. The optimization module 32 is connected to the identification module 31 and is configured to optimize the initial intention task based on a large language model and an experience library for poor-quality cells to obtain the target intention task of the poor-quality cell problem. The output module 33 is connected to the optimization module 32 and is configured to input the target intention task into the large language model and output an optimization decision suggestion for the poor-quality cell problem.

[0146] Specifically, the identification module 31 includes: an identification unit 311, a first acquisition unit 312, and a first optimization unit 313. The identification unit 311 is configured to identify the poor-quality cell ID and the poor-quality cell province and city information from the poor-quality cell problem based on a large language model and generate the first intention task of the poor-quality cell problem. The first acquisition unit 312 is configured to obtain the problem judgment index data of the poor-quality cell problem based on the first intention task and the large language model, and obtain the rule judgment conditions of the poor-quality cell problem from the rule library for poor-quality cell index problems. The first optimization unit 313 is configured to optimize the first intention task based on the problem judgment index data and the rule judgment conditions to obtain the initial intention task of the poor-quality cell problem.

[0147] Specifically, the first optimization unit 313 includes: a first assembly subunit, an identification subunit, and a second assembly subunit. The first assembly subunit is configured to assemble the first intention task, the problem judgment index data, and the rule judgment conditions to obtain the second intention task of the poor-quality cell problem. The identification subunit is configured to input the second intention task into the large language model to identify the poor-quality type of the poor-quality cell problem. The second assembly subunit is configured to assemble the poor-quality type and the second intention task to obtain the initial intention task of the poor-quality cell problem.

[0148] Specifically, the optimization module 32 includes: a second acquisition unit 321, a second optimization unit 322, a third acquisition unit 323, and a third optimization unit 324. The second acquisition unit 321 is configured to obtain the root cause analysis experience of the poor-quality cell problem from the poor-quality cell analysis experience library based on the initial intention task and the large language model. The second optimization unit 322 is configured to optimize the initial intention task based on the root cause analysis experience to obtain the third intention task of the poor-quality cell problem. The third acquisition unit 323 is configured to obtain the decision-making suggestion cases of the poor-quality cell problem from the poor-quality cell decision-making suggestion library based on the third intention task and the large language model. The third optimization unit 324 is configured to optimize the third intention task based on the decision-making suggestion cases to obtain the target intention task of the poor-quality cell problem.

[0149] Specifically, the second optimization unit 322 includes: a third assembly subunit, a determination subunit, a fourth assembly subunit, a positioning subunit, and a fifth assembly subunit. The third assembly subunit is configured to assemble the root cause analysis experience and the initial intent task to obtain a fourth intent task for the poor quality cell problem. The determination subunit is configured to input the fourth intent task into a large language model to determine a root cause positioning tool and root cause analysis metrics for the poor quality cell problem. The fourth assembly subunit is configured to assemble the root cause positioning tool, the root cause analysis metrics, and the fourth intent task to obtain a fifth intent task for the poor quality cell problem. The positioning subunit is configured to input the fifth intent task into a large language model to locate the root cause of the poor quality cell problem. The fifth assembly subunit is configured to assemble the root cause and the fifth intent task to obtain a third intent task for the poor quality cell problem.

[0150] Specifically, the third optimization unit 324 includes: a sixth assembly subunit configured to assemble the decision-making suggestion case and the third intent task to obtain a target intent task for the poor quality cell problem.

[0151] It can be understood that the optimization device for the poor quality cell problem provided above executes the optimization method for the poor quality cell problem corresponding to Embodiment 1 provided above. Therefore, the beneficial effects it can achieve can refer to the beneficial effects of the solution corresponding to the optimization method for the poor quality cell problem in Embodiment 1 above, which will not be elaborated here.

[0152] Embodiment 4:

[0153] This embodiment provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to implement the optimization method for the poor quality cell problem in Embodiment 1 or Embodiment 2 above.

[0154] Embodiment 5:

[0155] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the optimization method for the poor quality cell problem in Embodiment 1 or Embodiment 2 above.

[0156] It can be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present invention. However, the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. An optimization method for poor-quality cell problems, characterized in that, Including: An initial intention task for identifying poor-quality cell problems based on a large language model and a rule library for poor-quality cell index problems; Optimizing the initial intention task based on a large language model and an experience library for poor-quality cells to obtain a target intention task for poor-quality cell problems; Inputting the target intention task into the large language model to output an optimized decision-making suggestion for poor-quality cell problems.

2. The optimization method for the problem of poor-quality cells according to claim 1, wherein The initial intention task for identifying poor-quality cell problems based on a large language model and a rule library for poor-quality cell index problems specifically includes: Based on the large language model, identifying the poor-quality cell ID and the information of the province, city, and district of the poor-quality cell from the poor-quality cell problem, and generating a first intention task for the poor-quality cell problem; Based on the first intention task and the large language model, obtaining the problem judgment index data for the poor-quality cell problem, and obtaining the rule judgment conditions for the poor-quality cell problem from the rule library for poor-quality cell index problems; Based on the problem judgment index data and the rule judgment conditions, optimizing the first intention task to obtain the initial intention task for the poor-quality cell problem.

3. The optimization method for the problem of poor-quality cells according to claim 2, wherein The optimizing the first intention task based on the problem judgment index data and the rule judgment conditions to obtain the initial intention task for the poor-quality cell problem specifically includes: Assembling the first intention task, the problem judgment index data, and the rule judgment conditions to obtain a second intention task for the poor-quality cell problem; Inputting the second intention task into the large language model to identify the poor-quality type of the poor-quality cell problem; Assembling the poor-quality type and the second intention task to obtain the initial intention task for the poor-quality cell problem.

4. The optimization method for the problem of poor-quality cells according to claim 1, characterized in that The experience library for poor-quality cells includes an experience library for poor-quality cell analysis and a decision-making suggestion library for poor-quality cells, The optimizing the initial intention task based on a large language model and an experience library for poor-quality cells to obtain a target intention task for poor-quality cell problems specifically includes: Based on the initial intention task and the large language model, obtaining the root cause analysis experience for the poor-quality cell problem from the experience library for poor-quality cell analysis; Based on the root cause analysis experience, optimizing the initial intention task to obtain a third intention task for the poor-quality cell problem; Based on the third intention task and the large language model, obtaining the decision-making suggestion cases for the poor-quality cell problem from the decision-making suggestion library for poor-quality cells; Based on the decision-making suggestion cases, optimizing the third intention task to obtain the target intention task for the poor-quality cell problem.

5. The optimization method for the problem of poor-quality cells according to claim 4, characterized in that The optimizing the initial intention task based on the root cause analysis experience to obtain a third intention task for the poor-quality cell problem specifically includes: Assembling the root cause analysis experience and the initial intention task to obtain a fourth intention task for the poor-quality cell problem; Inputting the fourth intention task into the large language model to determine the root cause location tool and root cause analysis index for the poor-quality cell problem; Assembling the root cause location tool, the root cause analysis index, and the fourth intention task to obtain a fifth intention task for the poor-quality cell problem; Inputting the fifth intention task into the large language model to locate the root cause of the poor-quality cell problem; Assembling the root cause and the fifth intention task to obtain the third intention task for the poor-quality cell problem.

6. The optimization method for the problem of poor-quality cells according to claim 4, characterized in that, Based on the decision-making suggestion case, optimize the third intent task to obtain the target intent task for the poor quality cell problem, specifically including: Assemble the decision-making suggestion case and the third intent task to obtain the target intent task for the poor quality cell problem.

7. The optimization method for the problem of poor-quality cells according to claim 5, characterized in that, The problem judgment index data includes at least one of the following: weak coverage ratio, proportion of sampling points with MRO-SINR≥0 under measurement report offset, excellent rate of channel quality indicator CQI, average interference noise power of physical resource block PRB, number of days with high-load cells in cell capacity statistics within a preset period. The root cause location tool includes at least one of the following: over-coverage diagnosis and analysis tool, near-coverage diagnosis and analysis tool, overlapping coverage diagnosis and analysis tool, MOD 30 interference diagnosis and analysis tool, uplink interference type diagnosis and analysis tool, uplink PRB utilization load diagnosis and analysis tool, downlink PRB utilization load diagnosis and analysis tool. The root cause analysis index includes at least one of the following: over-coverage rate, near-coverage rate, overlapping coverage ratio, MOD 30 interference, uplink interference type, uplink PRB utilization rate at the busiest time of the cell, and downlink PRB utilization rate at the busiest time of the cell.

8. An optimization device for the problem of poor-quality cells, characterized in that, It includes an identification module, an optimization module, and an output module. The identification module is used to identify the initial intent task for the poor quality cell problem based on the large language model and the poor quality cell index problem rule base. The optimization module is connected to the identification module and is used to optimize the initial intent task based on the large language model and the poor quality cell experience base to obtain the target intent task for the poor quality cell problem. The output module is connected to the optimization module and is used to input the target intent task into the large language model and output the optimized decision-making suggestion for the poor quality cell problem.

9. An electronic device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to implement an optimization method for a poor quality cell problem 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 an optimization method for a poor quality cell problem as described in any one of claims 1 to 7.

Images