CN-121998338-A - Task matching method and system for wind power industry

CN121998338ACN 121998338 ACN121998338 ACN 121998338ACN-121998338-A

Abstract

The invention discloses a task matching method and a task matching system for wind power industry, which are applied to the technical field of intelligent employment and comprise the steps of splitting and marking task exclusive demands, calculating multi-dimensional similarity based on a deployed task matching model, combining cosine distance, manhattan distance and Jaccard distance, dynamically adjusting weights, outputting optimal technical talents, matching confidence, alternative talents sequencing and sub-task parallel execution suggestions corresponding to each sub-task, and realizing accurate matching of task demands and technical talents. Compared with the prior art, the method has the advantages that aiming at the problem of matching intelligent labor and task talents of a wind power enterprise, the subtask is endowed with split specific wind power scene attributes, a subtask and technical talents accurate matching method is established, the intelligent labor is modeled as a problem of matching a demand pool with a talent pool, simple system design is realized, system expansion and function update are facilitated, and flexible application of the intelligent labor method in the wind power enterprise is greatly promoted.

Inventors

CUI MENGYU
YANG CONGXI
XIAO YING
CHEN XIANMING
Chen Chuangying
YANG WU
Hao Xiaoya
LI JINXIU

Assignees

运达智服新能源技术(浙江)有限公司

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (10)

1. The task matching method for the wind power industry is characterized by comprising the following steps of: S1, a wind power exclusive demand pool is established based on task demands and wind power exclusive information, task demands are collected according to operation and maintenance classes, technical improvement classes, rush repair classes and pre-grinding classes, the wind power exclusive information corresponding to each task is recorded, the wind power exclusive information comprises a fan model, an operation environment, an emergency degree, equipment operation years and fault codes, each task in the wind power exclusive demand pool is configured with a state monitoring module and a task quality detection module, the state monitoring module is used for detecting whether the task is matched with a technical talent, and the task quality detection module is used for checking and evaluating the task completion condition according to wind power task acceptance criteria; S2, splitting the task based on the physical space dimension, the technical difficulty dimension and the aging requirement dimension of the wind power task to obtain subtasks, checking the dependency relationship of the subtasks and avoiding resource conflict, labeling each subtask with a category label, a technical requirement label and a wind power exclusive label according to a preset updatable correction label library, wherein the wind power exclusive label comprises a fan model label, a working environment label, an authentication requirement label and a task association label, and dynamically adjusting label weight through a reinforcement learning algorithm based on history matching data; S3, acquiring talent information of enterprise technical talents and social technical talents, wherein the talent information comprises identity information, education information, technical capability information, working experience information, wind power qualification authentication information, model experience information and operation limiting information; S4, counting the talent information in the S3, marking the technical capability, working experience, wind power exclusive skills and qualification authentication conditions of talents, establishing a technical talent portrait comprising a general skill vector and a wind power exclusive skill vector, and updating parameters of the technical talent portrait according to talent skill learning and project expression change according to a preset time period; S5, using the working experience mark and the special skill mark of wind power in the step S4 as input data, using the technical capability mark as tag data, adopting a neural network of a double-branch attention mechanism, and adjusting a general worker data pre-training model by using small sample data of the wind power industry through a transfer learning mode until model loss is not reduced any more, and deploying a trained task matching model to terminal equipment, wherein a loss function of the neural network is a set loss function and comprises a smooth L1 loss item and a wind power task fault tolerance penalty item; S6, acquiring task exclusive demands in an electric exclusive demand pool in S1, splitting and marking the task exclusive demands in S2, calculating multi-dimensional similarity based on a task matching model deployed in S5, combining cosine distance, manhattan distance and Jack-card distance, dynamically adjusting weights, and outputting optimal technical talents, matching confidence, alternative talent sequencing and sub-task parallel execution suggestions corresponding to each sub-task.
2. The method for matching tasks in wind power industry according to claim 1, wherein the dependency relationship of the subtasks in S2 includes checking the sequential logic order of execution of the subtasks.
3. The method for matching tasks in wind power industry according to claim 1, wherein the task association tag in S2 comprises dependency relationship between subtasks and other tasks and required tool equipment information.
4. The method for matching tasks in wind power industry according to claim 1, wherein the state monitoring module in S3 further comprises reducing the corresponding matching weight of talents which do not engage in corresponding models or skill related tasks beyond a preset period through a talent skill attenuation model, and recovering after re-authentication.
5. The wind power industry task matching method according to claim 1, wherein the neural network of the dual-branch attention mechanism in S5 includes processing common skill matching and strengthening wind power exclusive skill matching, and the priorities of the two branches are dynamically allocated through attention weights.
6. The wind power industry task matching method according to claim 1, wherein a calculation formula of a smooth L1 loss in S5 is , wherein, Representing the predicted value and the true value of the model for the ith input sample respectively, The loss of the smooth L1 is represented, N is the number of training samples, and the wind power task fault tolerance penalty term is , For the fault tolerance coefficient of the ith subtask, The penalty factor is 2.0.
7. The method for matching tasks in the wind power industry according to claim 1, wherein the weight adjustment rule corresponding to the multi-dimensional similarity in S6 is that the cosine distance weight of the technical intensive subtask is 0.6, and the Manhattan distance weight of the empirical intensive subtask is 0.7.
8. The method for matching tasks in a wind power industry according to claim 1, further comprising a subtask cooperative scheduling step of scheduling a plurality of subtask personnel, equipment and time resources based on a constraint satisfaction algorithm of a wind power operation flow.
9. The method for matching tasks in a wind power industry according to claim 1, further comprising a step of integrating a wind power knowledge base, wherein the wind power knowledge base comprises a wind power equipment fault base, a maintenance manual and industry standards, the wind power knowledge base is automatically associated when subtasks are matched, and the technical talents with relevant experiences and consulted with the corresponding wind power knowledge base are preferentially matched.
10. A wind power industry task matching system, comprising: the system comprises a demand pool construction module, a task quality detection module, a wind power demand management module and a wind power demand management module, wherein the demand pool construction module is used for executing S1, establishing a wind power exclusive demand pool based on task demands and wind power exclusive information, collecting task demands according to operation and maintenance classes, technical improvement classes, rush repair classes and pre-grinding classes, recording the wind power exclusive information corresponding to each task, wherein the wind power exclusive information comprises a fan model, an operation environment, an emergency degree, an equipment operation age and a fault code; The sub-task splitting module is used for executing S2, splitting the task based on the physical space dimension, the technical difficulty dimension and the aging requirement dimension of the wind power task to obtain sub-tasks, checking the dependency relationship of the sub-tasks and avoiding resource conflict; The talent pool construction module is used for executing S3, acquiring talent information of enterprise technical talents and social technical talents, wherein the talent information comprises identity information, education information, technical capability information, working experience information, wind power qualification authentication information, model experience information and operation limiting information; The talent representation marking module is used for executing S4, counting the talent information in S3, marking the technical capability, working experience, wind power exclusive skills and qualification authentication conditions of talents, establishing a technical talent representation comprising a general skill vector and a wind power exclusive skill vector, and updating parameters of the technical talent representation according to talent skill learning and project expression change according to a preset time period; The system comprises a demand talent modeling module, a neural network, a wind power task fault tolerance penalty item, a training task matching model, a terminal device and a power generation task fault tolerance penalty item, wherein the demand talent modeling module is used for executing S5, taking a working experience mark and a wind power exclusive skill mark in the step S4 as input data, taking a technical capability mark as tag data, adopting a neural network of a double-branch attention mechanism, and adjusting a universal working data pre-training model by utilizing small sample data of the wind power industry through a transfer learning mode until model loss is not reduced any more, wherein a loss function of the neural network is a set loss function and comprises a smooth L1 loss item and a wind power task fault tolerance penalty item; the matching execution module is used for executing S6, acquiring task exclusive demands in the power exclusive demand pool in S1, splitting and marking the task exclusive demands in S2, calculating multi-dimensional similarity based on a task matching model deployed in S5, combining cosine distance, manhattan distance and Jaccard distance, dynamically adjusting weights, and outputting optimal technical talents, matching confidence, alternative talent sequencing and sub-task parallel execution suggestions corresponding to each sub-task.

Description

Task matching method and system for wind power industry Technical Field The invention relates to the field of intelligent labor, in particular to a task matching method and system in the wind power industry. Background Enterprises can flexibly use and manage talents according to own personnel requirements, and can be used as a supplement for all-day talents. The method is more flexible in form and simpler and more convenient in flow, can effectively support the expansion of the enterprise scale and save the cost, can fully exert the talent value, and is an important labor form for the rapid development of the enterprise. Prior Art The prior art cannot realize the accurate matching of task demands and talents. Disclosure of Invention In view of the above, the invention aims to provide a task matching method and a task matching system in the wind power industry, which solve the technical problem that tasks and talents cannot be matched accurately in the prior art. In order to solve the technical problems, the invention provides a task matching method for wind power industry, which comprises the following steps: S1, a wind power exclusive demand pool is established based on task demands and wind power exclusive information, task demands are collected according to operation and maintenance classes, technical improvement classes, rush repair classes and pre-grinding classes, the wind power exclusive information corresponding to each task is recorded, the wind power exclusive information comprises a fan model, an operation environment, an emergency degree, equipment operation years and fault codes, each task in the wind power exclusive demand pool is configured with a state monitoring module and a task quality detection module, the state monitoring module is used for detecting whether the task is matched with a technical talent, and the task quality detection module is used for checking and evaluating the task completion condition according to wind power task acceptance criteria; S2, splitting the task based on the physical space dimension, the technical difficulty dimension and the aging requirement dimension of the wind power task to obtain subtasks, checking the dependency relationship of the subtasks and avoiding resource conflict, labeling each subtask with a category label, a technical requirement label and a wind power exclusive label according to a preset updatable correction label library, wherein the wind power exclusive label comprises a fan model label, a working environment label, an authentication requirement label and a task association label, and dynamically adjusting label weight through a reinforcement learning algorithm based on history matching data; S3, acquiring talent information of enterprise technical talents and social technical talents, wherein the talent information comprises identity information, education information, technical capability information, working experience information, wind power qualification authentication information, model experience information and operation limiting information; S4, counting the talent information in the S3, marking the technical capability, working experience, wind power exclusive skills and qualification authentication conditions of talents, establishing a technical talent portrait comprising a general skill vector and a wind power exclusive skill vector, and updating parameters of the technical talent portrait according to talent skill learning and project expression change according to a preset time period; S5, using the working experience mark and the special skill mark of wind power in the step S4 as input data, using the technical capability mark as tag data, adopting a neural network of a double-branch attention mechanism, and adjusting a general worker data pre-training model by using small sample data of the wind power industry through a transfer learning mode until model loss is not reduced any more, and deploying a trained task matching model to terminal equipment, wherein a loss function of the neural network is a set loss function and comprises a smooth L1 loss item and a wind power task fault tolerance penalty item; S6, acquiring task exclusive demands in an electric exclusive demand pool in S1, splitting and marking the task exclusive demands in S2, calculating multi-dimensional similarity based on a task matching model deployed in S5, combining cosine distance, manhattan distance and Jack-card distance, dynamically adjusting weights, and outputting optimal technical talents, matching confidence, alternative talent sequencing and sub-task parallel execution suggestions corresponding to each sub-task. Optionally, the dependency relationship of the subtasks in S2 includes checking the sequential logic order of execution of the subtasks. Optionally, the task association tag in S2 includes the dependency relationship between the subtasks and other tasks, and the required tool equipment information. Optionally, the state monitoring modu