Project resource intelligent scheduling and dynamic optimization method based on AI

Abstract

The invention discloses an AI-based project resource intelligent scheduling and dynamic optimization method, and relates to the technical field of project management and artificial intelligence crossing. According to the method, an improved demand increment time sequence prediction formula is adopted, the historical task progress data, the resource consumption data and the idle resource state are combined, accurate pre-judgment of project resource demand changes in the future 72 hours is achieved, the pre-judgment accuracy is improved to 89% or above, and meanwhile, the prediction efficiency is improved. For core resource demand abnormalities such as a computing power gap, a material consumption acceleration gap and a manpower gap, a differentiated linkage trigger rule is set, and a trigger condition is further optimized for a core project, so that the scheduling response speed is remarkably improved, and the scheduling response duration can be controlled within 10 minutes.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of project management and artificial intelligence, specifically to an AI-based intelligent scheduling and dynamic optimization method for project resources. Background Technology

[0002] As project management becomes increasingly complex, the effective scheduling and dynamic optimization of project resources have become key factors in ensuring that projects are completed on time, with quality, and within budget. Traditional project resource management methods often rely on human experience and static planning, which are difficult to cope with the rapid changes and uncertainties in resource requirements during project execution.

[0003] Traditional project resource scheduling methods rely primarily on human experience and pre-defined static plans. These methods fall short when faced with uncertainties and dynamic changes during project execution. Specifically, the shortcomings of traditional techniques are mainly reflected in the following aspects: First, the accuracy of resource demand forecasting is low, often based on historical data and experience estimates, making it difficult to accurately reflect resource consumption during actual project execution. Second, scheduling schemes lack flexibility; once formulated, they are difficult to adjust quickly according to actual conditions, leading to resource idleness or shortages. Finally, traditional methods lack the ability to adapt to long-term changes in resource demand and cannot continuously optimize scheduling strategies to cope with different stages and demand changes throughout the project lifecycle. These shortcomings limit the effectiveness of traditional project resource scheduling methods in complex and ever-changing project environments.

[0004] To address the problems of low prediction accuracy, lack of flexibility in scheduling schemes, and insufficient adaptability to long-term changes in resource demand in traditional project resource scheduling methods, this invention proposes an AI-based intelligent scheduling and dynamic optimization method for project resources, which is of particular importance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an AI-based intelligent scheduling and dynamic optimization method for project resources. This method can accurately predict changes in project resource demand within the next 72 hours by collecting all project operation data and using a time-series prediction AI model. It also establishes a linkage triggering rule between resource supply and demand time-series prediction and scheduling, achieving dynamic matching and efficient utilization of resource demand. This method not only improves the accuracy of resource demand prediction and the speed of scheduling response, but also ensures the continuous adaptability and accuracy of scheduling strategies through a long-term dynamic optimization mechanism, providing strong technical support for the effective management of project resources.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: an AI-based intelligent scheduling and dynamic optimization method for project resources, the specific steps of which are as follows: S1. Collect all project operation data, including project task progress data, real-time resource consumption data, and idle resource status data, and classify and preprocess the collected data. S2. Use a time-series prediction AI model to process the collected data, predict changes in project resource demand within the next 72 hours, identify abnormal types of resource demand, and verify and adjust the model prediction results. S3. Establish linkage triggering rules for resource supply and demand time-series prediction and scheduling, set differentiated triggering conditions for different types of abnormal resource demand, and classify and identify schedulable idle resources based on the predicted changes in resource demand and generate scheduling preparation plans. S4. Real-time monitoring of the actual resource demand status of the project. A demand deviation triggering formula is used to determine whether the deviation between actual and predicted demand reaches a preset threshold. If it does, a scheduling contingency plan is triggered to complete resource scheduling. After scheduling is completed, the scheduling effect is verified and the system is dynamically optimized over the long term to achieve dynamic matching of resource supply and demand. The formula is: ,in As a scheduling trigger identifier, This represents the actual demand for resources. Forecast resource demand values, This is the trigger threshold.

[0007] Furthermore, the project's full-scale operational data collection in step S1 involves the following steps: First, data is categorized into structured and unstructured operational data based on its attributes. Structured operational data is automatically collected via the API interface of the project management system at a frequency of 5 minutes per collection. The collected data includes the completion progress of each subtask, the specific values ​​of consumed computing power / materials / human resources, and the status information of registered idle resources. Unstructured operational data is collected through edge sensing devices and employee terminals at the project site. The collected data includes the real-time operating status of the equipment and the on-duty status of employees. After collection, preprocessing operations such as deduplication, outlier removal, and missing value completion are performed sequentially. Preprocessing can eliminate the interference of data noise on subsequent prediction and scheduling, increasing the data efficiency to over 96% and providing high-quality input data for subsequent steps.

[0008] Furthermore, the time-series forecasting AI model in step S2 employs an improved incremental demand time-series forecasting formula to achieve 72-hour resource demand prediction. The formula is as follows: ,in As the core output parameter, for The actual resource consumption requirements of the project at any given time. This is historical task progress data. This represents the average progress of historical tasks. This is historical data on resource consumption. This represents the historical average resource consumption. The number of steps for backtracking historical data. This represents the amount of idle resources available for use at the current moment. The base weight for current demand is derived from the average of resource demand stability data from 127 similar projects over the past three years, and is set at 0.72. The weight for the impact of task increments was determined to be 0.21 after 1000 iterations of training using a custom supply-demand matching loss function. The compensation weight for idle resources is derived from the average reuse rate of idle resources of enterprises over the past year, and is set at 0.07. This formula differs from existing time series forecasting algorithms that do not incorporate idle resource compensation logic, and can improve the accuracy of demand prediction to over 89%.

[0009] Furthermore, the verification and adjustment of the time-series prediction AI model in step S2 involves the following steps: First, the 72-hour resource demand prediction result output by the model is compared with the resource demand data of similar projects in the same period in history, and the relative deviation rate of the prediction result is calculated. If the relative deviation rate exceeds the preset threshold of 10%, the input feature dimension of the model is automatically adjusted to add features of project urgency and external environmental interference. After adjustment, the data is re-inputted for prediction until the relative deviation rate drops to within 10%. This step can avoid systematic deviations in the prediction results and ensure the reliability of subsequent scheduling preparation plans.

[0010] Furthermore, the resource supply and demand timing prediction and scheduling linkage triggering rules in step S3 are as follows: First, for the three types of core resource demand anomalies—computing power gap, accelerated material consumption, and manpower gap—independent triggering rules are established respectively; for the computing power gap, the triggering precondition is set as the predicted computing power gap accounting for ≥15% of the current available computing power; for the accelerated material consumption, the triggering precondition is set as the predicted material consumption rate exceeding the current consumption rate by ≥20%; for the manpower gap, the triggering precondition is set as the predicted manpower gap accounting for ≥12% of the current available manpower; at the same time, for core projects with priority level 1, the proportion of the above triggering preconditions is reduced by 5% to improve the scheduling response speed of core projects.

[0011] Furthermore, the idle resource scheduling preparation plan generated in step S3 involves the following steps: First, the idle resources within the enterprise and those of its partners are classified into three categories: computing power resources, material resources, and human resources for classified management. For predicted abnormal resource demand types, resources meeting the requirements are selected from the corresponding category of idle resources. The selection criteria include: available time of computing power resources ≥ 1.2 times the demand time; shelf life of material resources ≥ project usage cycle; and skill matching degree of human resources ≥ 85%. After selection, resource allocation paths and connection times are planned. For computing power resources, the transmission path with the optimal bandwidth is planned; for material resources, the path with the shortest transportation time is planned; and for human resources, the allocation path with the shortest commuting time is planned, ensuring the feasibility of the preparation plan.

[0012] Furthermore, the execution of the scheduling preparation plan in step S4 involves the following steps: First, after triggering the scheduling preparation plan, an allocation instruction is automatically sent to the idle resource management terminal. The instruction includes the resource allocation time, the project node to which the resource is allocated, and the resource usage requirements. Then, the resource allocation status is tracked in real time. If the resource does not reach the designated node within the preset buffer time, the allocation of alternative idle resources is automatically initiated. After the allocation is completed, the project resource usage status data is automatically updated to ensure the implementation of the scheduling plan. The buffer time for computing power resources is 2 hours, the buffer time for material resources is 4 hours, and the buffer time for human resources is 1 hour.

[0013] Furthermore, the scheduling effect verification in step S4 is specifically carried out as follows: First, within one hour after the scheduling plan is completed, collect actual resource usage data and project sub-task progress data; then calculate the resource supply and demand matching degree and task stagnation time, and compare them with the data before scheduling; if the improvement in resource supply and demand matching degree does not reach 30%, automatically adjust the threshold of the linkage triggering rule and the input features of the time series prediction model, regenerate the scheduling plan after adjustment and execute it; at the same time, store all the data of this scheduling in the project resource scheduling database for subsequent model training and rule updates, so as to realize the dynamic optimization of the scheduling system.

[0014] Furthermore, the long-term optimization of the dynamic matching of resource supply and demand in step S4 is specifically carried out as follows: First, the data in the project resource scheduling database is organized every 7 days, including all historical prediction data, trigger data, scheduling data, and effect data; then, the organized data is input into the time series prediction AI model to retrain the model's weight parameters and update the model's prediction logic; finally, the threshold of the resource supply and demand linkage trigger rules is updated synchronously to adapt the model and rules to the long-term changes in project resource demand, continuously improving the accuracy and adaptability of scheduling.

[0015] Compared with existing technologies, this AI-based intelligent scheduling and dynamic optimization method for project resources has the following advantages: I. This method, by adopting an improved incremental demand time-series prediction formula and combining historical task progress data, resource consumption data, and idle resource status, achieves accurate prediction of changes in project resource demand within the next 72 hours, with a prediction accuracy rate of over 89%. At the same time, it sets differentiated linkage trigger rules for core resource demand anomalies such as computing power gaps, accelerated material consumption, and manpower gaps, and further optimizes the trigger conditions for core projects, resulting in a significant improvement in scheduling response speed, with scheduling response time controllable within 10 minutes.

[0016] Second, this method monitors the actual demand status of project resources in real time and uses a demand deviation triggering judgment formula to accurately identify the scenarios that need to be scheduled, avoiding the problems of invalid triggering or untimely triggering. After scheduling is completed, the scheduling effect is verified by collecting actual resource usage data and project sub-task progress data, and the system is dynamically optimized in the long term. Every 7 days, historical data is sorted and the model is retrained to update the prediction logic and linkage triggering rule thresholds, so that the model and rules continuously adapt to the long-term changes in project resource demand, thereby continuously improving the accuracy and adaptability of scheduling and realizing dynamic matching and efficient utilization of resource supply and demand.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 A flowchart illustrating an AI-based intelligent scheduling and dynamic optimization method for project resources. Figure 2 A flowchart for determining the weights of indicators in an AI-based intelligent scheduling and dynamic optimization method for project resources; Figure 3 This is a detailed flowchart of the TOPSIS health assessment process, an AI-based intelligent scheduling and dynamic optimization method for project resources. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] Example 1 The core task of this project is deep learning model training and algorithm optimization. First, the operational data is divided into structured and unstructured data based on its attributes. Structured data is automatically collected through the project management system API interface, specifically including information such as the model training module completion rate of 80%, the algorithm optimization module completion rate of 65%, 5000 computing power units consumed, 20 man-days consumed, 3000 computing power units registered as idle, and 5 idle man-days. Unstructured data is collected through edge sensing devices at the project site to monitor the real-time server operating status, and through employee terminals to collect the on-duty status of 18 core R&D personnel. After collection, data preprocessing operations are performed sequentially: two duplicate computing power consumption records are removed, abnormal peak data of server CPU utilization caused by sensor malfunction are deleted, and information on one employee who failed to report their on-duty status in a timely manner is added.

[0022] A time-series predictive AI model is used to process the collected data, focusing on predicting changes in computing power and manpower demand over the next 72 hours, such as... Figure 2 As shown, the formula is: ,in As the core output parameter, for The actual resource consumption requirements of the project at any given time. This is historical task progress data. This represents the average progress of historical tasks. This is historical data on resource consumption. This represents the historical average resource consumption. The number of steps for backtracking historical data. This represents the amount of idle resources available for use at the current moment. As the basic weight for current needs, The weight of the impact of task increment. To compensate for idle resources, the model output indicated that the model training task would enter a computationally intensive phase within the next 48 hours, predicting a computational power shortage, which was identified as an abnormal computational demand type. The 72-hour resource demand prediction was then compared with historical data from similar AI R&D projects during the same period, resulting in a relative deviation rate of 12%, exceeding the preset 10% threshold. Based on this, the model's input feature dimensions were automatically adjusted, adding two new features: project urgency and external environmental interference. Preprocessed complete data was then re-inputted for prediction, ultimately reducing the relative deviation rate to 8%, meeting the preset requirements.

[0023] Establish linkage triggering rules for resource supply and demand time-series forecasting and scheduling, such as... Figure 1 As shown, for the identified computing power gap anomaly types, combined with the project's level 1 priority attribute, the triggering precondition was set as the predicted computing power gap accounting for ≥15% of the current available computing power. Calculations showed that the predicted computing power gap accounted for 18% of the current available computing power, thus meeting the triggering condition. Subsequently, idle resources within the enterprise and from partners were categorized into three types: computing power resources, human resources, and material resources. For computing power gap anomalies, eligible idle resources were selected: computing power resources needed to have an available duration ≥1.2 times the required duration; human resources needed to have a skill matching rate ≥85%. Finally, resource allocation paths were planned: computing power resources selected the cloud transmission path with the optimal bandwidth to ensure data transmission latency ≤5ms; human resources were allocated using the shortest commuting time path, ensuring that all three engineers could reach the project development node within 30 minutes. This resulted in a complete scheduling contingency plan.

[0024] Real-time monitoring of the actual resource demand status of the project, and continuous comparison of the difference between actual and predicted demand using a demand deviation triggering formula, such as... Figure 3 As shown, the formula is: ,in As a scheduling trigger identifier, This represents the actual demand for resources. Forecast resource demand values, To trigger the threshold, when the model training task progresses to the 10th hour, if the actual computing power consumption rate is 10% higher than the predicted value, and the deviation reaches the preset trigger threshold θ=10%, the scheduling preparation plan will be immediately triggered. The system will automatically send allocation instructions to the idle resource management terminal, specifying that computing power resources must be allocated to the model training node within 2 hours, with stable output computing power and fluctuation range ≤3%; human resources must be allocated to the algorithm optimization node within 4 hours to complete the model parameter tuning task according to the project schedule requirements, and the allocation status will be tracked in real time. All resources must arrive at the designated node within the preset buffer time. After the allocation is completed, the project resource usage status will be automatically updated. According to reports, within one hour of the scheduling completion, data on actual resource usage and project sub-task progress were collected. The results showed that the resource supply and demand matching rate improved from 65% before scheduling to 98%, an increase of 33% (≥30%). The task stagnation time decreased from one hour before scheduling to 0. All data from this scheduling was stored in the project resource scheduling database. Every seven days, the historical prediction data, trigger data, scheduling data, and effect data in the database were organized. The organized data was then input into the time series prediction AI model to retrain the weight parameters and update the model's prediction logic. At the same time, the threshold of the resource supply and demand linkage trigger rule was adjusted to adapt to the long-term changes in project resource demand.

[0025] Example 2 This project aims to automate the upgrade of three aging production lines, involving tasks such as equipment installation and data monitoring system deployment. Operational data is categorized into structured and unstructured data based on its attributes: Structured data is automatically collected via the project management system's API interface, including: 60% completion rate of equipment installation sub-tasks; 40% completion rate of data monitoring system deployment; 3000 computing units consumed; 500 consumed materials; 30 consumed manpower days; 2000 registered idle computing units; 300 idle materials; and 8 idle manpower days. Unstructured data is collected from edge sensors on the production lines to assess the operational status of existing equipment, and from employee terminals to track the on-duty status of 25 construction workers. After collection, preprocessing is performed: 15 duplicate attendance records are removed, 100 incorrectly recorded industrial sensor consumption data points are eliminated, and on-duty status information for 2 field construction workers is completed.

[0026] The collected data is processed using a time-series predictive AI model to predict changes in the demand for resources, computing power, and manpower over the next 72 hours. Figure 2 As shown, the model output results indicate that the consumption rate of industrial sensors will accelerate significantly in the next 36 hours, exceeding the current consumption rate by 25%, which is identified as an abnormal type of accelerated material consumption. The 72-hour resource demand forecast results are compared with the data of the same period of historical production line upgrade projects. The relative deviation rate is calculated to be 9%, which does not exceed the preset threshold of 10%. The model prediction results do not need to be adjusted and are directly verified.

[0027] Establish linkage triggering rules for resource supply and demand time-series forecasting and scheduling, such as... Figure 1 As shown, for abnormal material consumption rates, combined with the project's level 1 priority attribute, the triggering prerequisite is set as follows: the predicted material consumption rate exceeds the current consumption rate by ≥20%, and the current predicted rate of 25% meets the triggering requirement. Idle resources within the enterprise and from partners are categorized into computing resources, material resources, and human resources. Suitable resources are selected based on the following criteria: material resources must have a shelf life ≥ the project's usage cycle; computing resources must have an available duration ≥ 1.2 times the required duration; and human resources must have a skill matching rate ≥ 85%. Resource allocation paths are planned: for material resources, the shortest logistics route is selected, and local suppliers are coordinated to use expedited delivery to ensure delivery to the production line installation site within 4 hours; for computing resources, the optimal internal LAN transmission route is planned to ensure real-time synchronization of data monitoring signals; for human resources, the shortest commuting route is planned, ensuring that all 5 technicians can arrive at the work node within 20 minutes. Based on this, a scheduling contingency plan is generated.

[0028] Real-time monitoring of the actual demand status of project resources, and continuous judgment based on the determination formula triggered by demand deviation, such as... Figure 3As shown, when the equipment installation task progressed to the 8th hour, the deviation between the actual industrial sensor consumption demand and the predicted demand reached the trigger threshold θ=8%, triggering the scheduling backup plan. The system automatically sent allocation instructions to the idle resource management terminal: material resources needed to be allocated to the production line installation node within 4 hours, with usage requirements including compliance with production line equipment compatibility standards and sensitivity error ≤0.1%; computing resources needed to be allocated to the data monitoring node within 1 hour, ensuring real-time transmission of equipment operation data with a latency ≤10ms; and human resources needed to be allocated to the installation work node within 3 hours to complete equipment assembly and debugging according to construction specifications. During the tracking of allocation status, it was found that the first batch of 150 industrial sensors did not arrive within the preset buffer time due to logistical congestion. The system immediately activated the allocation of alternative idle resources, calling upon 100 industrial sensors reserved in the internal warehouse to ensure that the installation task was not interrupted. Ultimately, all resources were successfully delivered, and the resource usage status data was updated after allocation was completed. Within one hour of scheduling completion, data on actual resource usage and project sub-task progress were collected. The results showed that the resource supply and demand matching rate improved from 58% before scheduling to 92%, an increase of 34% (≥30%). The task stagnation time decreased from 2 hours before scheduling to 0.5 hours. All data from this scheduling was stored in the project resource scheduling database. The database data was systematically reorganized every 7 days. The reorganized historical prediction data, trigger data, scheduling data, and effect data were input into the time series prediction AI model to retrain the model weight parameters and update the prediction logic. At the same time, the threshold of the resource supply and demand linkage trigger rule was adjusted to achieve long-term dynamic optimization of the project resource scheduling system.

[0029] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An AI-based intelligent scheduling and dynamic optimization method for project resources, characterized in that, The specific steps of this method are as follows: S1. Collect all project operation data, including project task progress data, real-time resource consumption data, and idle resource status data, and classify and preprocess the collected data. S2. Use a time-series prediction AI model to process the collected data, predict changes in project resource demand within the next 72 hours, identify abnormal types of resource demand, and verify and adjust the model prediction results. S3. Establish linkage triggering rules for resource supply and demand time-series prediction and scheduling, set differentiated triggering conditions for different types of abnormal resource demand, and classify and identify schedulable idle resources based on the predicted changes in resource demand and generate scheduling preparation plans. S4. Monitor the actual resource demand status of the project in real time. Use a demand deviation trigger judgment formula to determine whether the deviation between actual demand and predicted demand reaches a preset threshold. If it does, trigger the scheduling preparation plan to complete resource scheduling. After scheduling is completed, verify the scheduling effect and perform long-term dynamic optimization of the system. The formula is: ,in As a scheduling trigger identifier, This represents the actual demand for resources. Forecast resource demand values, This is the trigger threshold.

2. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The specific steps for collecting full project operation data in step S1 are as follows: First, the data is divided into two categories according to its attributes: structured operation data and unstructured operation data. Structured operation data is automatically collected through the API interface of the project management system. The collected content includes the completion progress of each sub-task, the specific values ​​of consumed computing power / materials / human resources, and the status information of registered idle resources. Unstructured operation data is collected through edge sensing devices and employee terminals at the project site. The collected content includes the real-time operating status of the equipment and the on-duty status of employees. After the collection is completed, preprocessing operations such as deduplication, outlier removal, and missing value completion are performed in sequence.

3. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The time-series forecasting AI model in step S2 adopts an improved demand increment time-series forecasting formula, which is as follows: ,in As the core output parameter, for The actual resource consumption requirements of the project at any given time. This is historical task progress data. This represents the average progress of historical tasks. This is historical data on resource consumption. This represents the historical average resource consumption. The number of steps for backtracking historical data. This represents the amount of idle resources available for use at the current moment. As the basic weight for current needs, The weight of the impact of task increment. This is the compensation weight for idle resources.

4. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The verification and adjustment of the time-series prediction AI model in step S2 are as follows: First, the 72-hour resource demand prediction result output by the model is compared with the resource demand data of similar projects in the same period in history, and the relative deviation rate of the prediction result is calculated; if the relative deviation rate exceeds the preset threshold of 10%, the input feature dimension of the model is automatically adjusted to add the project urgency feature and the external environmental interference feature; after adjustment, the data is re-inputted for prediction until the relative deviation rate drops to within 10%.

5. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The resource supply and demand timing prediction and scheduling linkage triggering rules in step S3 are as follows: First, for the three types of core resource demand anomalies—computing power gap, accelerated material consumption, and manpower gap—independent triggering rules are established respectively. For the computing power gap, the triggering precondition is set as follows: the predicted computing power gap accounts for ≥15% of the current available computing power. For the accelerated material consumption, the triggering precondition is set as follows: the predicted material consumption rate exceeds the current consumption rate by ≥20%. For the manpower gap, the triggering precondition is set as follows: the predicted manpower gap accounts for ≥12% of the current available manpower. At the same time, for core projects with priority level 1, triggering rules are also established.

6. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The idle resource scheduling preparation plan generation in step S3 is as follows: First, the idle resources within the enterprise and its partners are classified into three categories: computing power resources, material resources, and human resources for classified management; for the predicted abnormal types of resource demand, resources that meet the requirements are screened from the corresponding category of idle resources. The screening conditions include that the available time of computing power resources is ≥ 1.2 times the demand time, the shelf life of material resources is ≥ the project usage cycle, and the skill matching degree of human resources is ≥ 85%; after screening, the resource allocation path and connection time are planned. The optimal transmission path with the best bandwidth is planned for computing power resources, the path with the shortest transportation time is planned for material resources, and the allocation path with the shortest commuting time is planned for human resources.

7. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The execution of the scheduling preparation plan in step S4 is as follows: First, after the scheduling preparation plan is triggered, an allocation instruction is automatically sent to the idle resource management terminal. The instruction includes the resource allocation time, the project node to which the resource is allocated, and the resource usage requirements. Then, the resource allocation status is tracked in real time. If the resource does not reach the designated node within the preset buffer time, the allocation of alternative idle resources is automatically started. After the allocation is completed, the project resource usage status data is automatically updated.

8. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The specific steps for verifying the scheduling effect in step S4 are as follows: First, within one hour after the scheduling plan is completed, collect actual resource usage data and project sub-task progress data; then calculate the resource supply and demand matching degree and task stagnation time, and compare them with the data before scheduling; if the improvement in resource supply and demand matching degree does not reach 30%, automatically adjust the threshold of the linkage triggering rule and the input features of the time series prediction model, regenerate the scheduling plan after adjustment and execute it; at the same time, store all the data of this scheduling in the project resource scheduling database.

9. The AI-based intelligent scheduling and dynamic optimization method for project resources according to claim 1, characterized in that, The long-term optimization of dynamic matching of resource supply and demand in step S4 is as follows: First, the data in the project resource scheduling database is organized every 7 days. The organized data includes all historical prediction data, trigger data, scheduling data, and effect data. Then, the organized data is input into the time series prediction AI model to retrain the model's weight parameters and update the model's prediction logic. Finally, the threshold of the resource supply and demand linkage trigger rules is updated synchronously to make the model and rules adapt to the long-term changes in project resource demand.

Images