Digital project management method and system based on artificial intelligence

Abstract

The present invention provides a digital project management method and system based on artificial intelligence, which relates to the field of artificial intelligence technology, including collecting power grid engineering construction resource data; constructing a multi-level resource constraint matrix; using a deep neural network to generate a resource collaborative optimization configuration model; calculating a real-time scheduling weight coefficient; constructing a resource scheduling objective function; using an improved ant colony algorithm to solve the optimal resource scheduling solution; inputting the optimal solution into an intelligent collaborative controller for dynamic adjustment; and realizing intelligent collaborative scheduling of power grid engineering construction resources. The present invention improves the efficiency and accuracy of resource scheduling.

Description

Technical Field

[0001] The present invention relates to artificial intelligence technology, and in particular to a digital project management method and system based on artificial intelligence. Background Art

[0002] Existing resource management methods lack comprehensive consideration of multi-dimensional resource constraints. They usually only focus on single-dimensional resource constraints, such as time or space, while ignoring the comprehensive impact of multi-dimensional constraints such as resource type and quantity, resulting in inaccurate resource allocation and difficulty in achieving optimal resource utilization.

[0003] Secondly, traditional resource scheduling methods often adopt static scheduling strategies, which cannot respond to dynamic changes in the construction process in a timely manner. This fixed scheduling scheme is difficult to adapt to various emergencies in actual construction and real-time changes in resource status, which can easily lead to resource waste or project delays.

[0004] Finally, the existing resource collaborative management system lacks intelligence and adaptive capabilities. During the resource scheduling process, it is impossible to monitor the resource status in real time, and it is difficult to promptly discover and resolve conflicts in resource scheduling. At the same time, the system lacks self-optimization and learning capabilities, and is unable to continuously improve scheduling strategies based on historical data and real-time feedback, resulting in low resource scheduling efficiency and difficulty in achieving truly intelligent collaborative management. Summary of the invention

[0005] The embodiments of the present invention provide a digital project management method and system based on artificial intelligence, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention,

[0007] Provides digital project management methods based on artificial intelligence, including:

[0008] Collecting power grid project construction resource data, the power grid project construction resource data including material resource data, equipment resource data and construction site resource data; constructing a multi-level resource constraint matrix based on the power grid project construction resource data, the multi-level resource constraint matrix including a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; inputting the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model;

[0009] Based on the collaborative optimization configuration model of power grid engineering construction resources, the real-time dispatch weight coefficients of various types of resources are calculated, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; a resource dispatch objective function is constructed according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; an improved ant colony algorithm is used to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch plan, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm;

[0010] The optimal resource scheduling plan is input into a preset intelligent collaborative controller, which includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; the resource status monitoring module monitors the scheduling execution status of various types of resources in real time, the conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base, and the scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent collaborative scheduling of power grid project construction resources.

[0011] Constructing a multi-level resource constraint matrix based on the power grid engineering construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix, and a resource type constraint sub-matrix; inputting the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid engineering construction resource collaborative optimization configuration model includes:

[0012] The construction period of the project is divided into multiple time units according to the construction process, and a resource availability vector is established for each time unit. The resource availability vector includes equipment operation time parameters, material supply time parameters and site occupancy time parameters. According to the time sequence mapping relationship between the time unit and the resource availability vector, a time dimension constraint sub-matrix is ​​generated;

[0013] The construction site is divided into a plurality of regional units according to the construction area. The division of the regional units is based on the connection relationship of the construction process and the resource allocation radius. A resource capacity vector is established for each of the regional units. The resource capacity vector includes operator capacity parameters, mechanical equipment capacity parameters, material storage capacity parameters and site carrying capacity parameters. Based on the corresponding relationship between the spatial distribution of the regional units and the resource capacity vectors, a spatial dimension constraint sub-matrix is ​​constructed.

[0014] Construct a resource feature recognition model for different resource types, extract a resource attribute feature vector based on the resource feature recognition model, the resource attribute feature vector includes a resource use priority feature, a resource allocation flexibility feature, a resource use cost feature and a resource substitutability feature, establish a resource type classifier, the resource type classifier uses a support vector machine algorithm to classify resources, and generates a resource type constraint submatrix according to the association between the resource type and the resource attribute feature vector;

[0015] Inputting the time dimension constraint submatrix, the space dimension constraint submatrix and the resource type constraint submatrix into a deep neural network model, wherein the deep neural network model includes an input layer, three hidden layers and an output layer, and a ReLU activation function is set between adjacent layers;

[0016] The deep neural network model is trained based on a composite loss function, wherein the composite loss function includes a resource utilization loss term, a spatiotemporal constraint violation loss term, and a resource synergy efficiency loss term, and each loss term is set with a corresponding weight coefficient; a dropout layer is set after each hidden layer of the deep neural network model, and the dropout rate of the dropout layer is 0.3, and an L2 regularization weight decay term is added;

[0017] The Adam optimizer is used to optimize the deep neural network model, set the initial learning rate, and dynamically adjust the learning rate according to the training rounds to generate a collaborative optimization configuration model for power grid engineering construction resources.

[0018] Based on the collaborative optimization configuration model of power grid engineering construction resources, the real-time dispatch weight coefficients of various types of resources are calculated, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; a resource dispatch objective function is constructed according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function, including:

[0019] Construct a resource evaluation index system, which includes timeliness index, criticality index, substitutability index and coordination index, among which:

[0020] The timeliness index is calculated by task deadline, current time and planned duration to obtain the first-level timeliness index, and the second-level timeliness index is calculated by associating process weight and resource process association coefficient; the criticality index is calculated based on the ratio of the longest process path length to the critical path length in the process network topology structure and the resource usage intensity; the substitutability index is calculated based on the resource feature similarity matrix and the substitution cost coefficient matrix; the coordination index is calculated based on the resource node centrality calculated based on the resource coordination efficiency matrix;

[0021] Establish an indicator importance evaluation matrix, the matrix elements of which represent the relative importance of different indicators in the resource evaluation indicator system, calculate an initial weight vector based on the indicator importance evaluation matrix, construct a time-varying adjustment factor, the time-varying adjustment factor is related to the total construction period and the adjustment coefficient, and generate a resource scheduling dynamic weight based on the initial weight vector, the time-varying adjustment factor and the normalized score of each indicator in the resource evaluation indicator;

[0022] The resource scheduling objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function. The resource utilization sub-function is calculated based on the actual utilization, the planned utilization and the resource efficiency coefficient. The project progress sub-function is calculated based on the actual progress, the planned progress and the progress influencing factor. The cost control sub-function is calculated based on the actual cost, the planned cost and the cost sensitivity coefficient.

[0023] The fuzzy hierarchical analysis method is used to determine the resource scheduling dynamic weights corresponding to the resource utilization sub-function, the project progress sub-function and the cost control sub-function, and the resource scheduling objective function is obtained by linearly weighted combination of each sub-function based on the resource scheduling dynamic weights.

[0024] The resource scheduling objective function is iteratively optimized and solved by using an improved ant colony algorithm to generate an optimal resource scheduling solution including:

[0025] Establish a pheromone matrix, wherein the row vectors of the pheromone matrix represent available resources, the column vectors represent task requirements, and the matrix elements are determined according to the product of the pheromone attenuation coefficient and the pheromone matrix at the previous moment and the pheromone increment, and the pheromone increment is calculated by the target value corresponding to the resource scheduling objective function of the path passed by the ants;

[0026] Calculating a resource scheduling decision distance, wherein the resource scheduling decision distance is determined based on a matching degree between the available resources and the task requirements, and calculating a dynamic heuristic factor according to the resource scheduling decision distance, wherein the dynamic heuristic factor is obtained by exponentially calculating the inverse of the resource scheduling decision distance and a dynamic adjustment coefficient;

[0027] The state transition probability is calculated based on the pheromone matrix and the dynamic heuristic factor, the ant colony search path is initialized, and the resource scheduling scheme of the initial ant population is randomly generated under the condition of satisfying the preset constraint condition list according to the state transition probability, and the target value corresponding to the resource scheduling objective function corresponding to each resource scheduling scheme is calculated, and the resource scheduling scheme with the smallest target value is recorded as the current optimal solution, wherein the preset constraint condition list includes resource layer constraints, task layer constraints and system layer constraints;

[0028] Performing ant colony iterative search, determining a new resource scheduling decision plan for each ant based on the state transition probability; optimizing the current optimal solution by using a local search mechanism in each iteration, calculating the movement increment by defining a neighborhood structure centered on the current optimal solution, and determining whether to accept the candidate solution in the neighborhood based on the movement increment and the Metropolis criterion;

[0029] The global optimal solution is updated according to the local search results. When a global optimal solution that is better than the current optimal solution appears, the pheromone matrix is ​​updated and the number of local searches is reset, and the new global optimal solution is recorded. When the maximum number of iterations is reached, the iteration is stopped and the optimal resource scheduling plan is output.

[0030] The scheduling optimization module dynamically adjusts the optimal resource scheduling scheme according to the conflict event, including:

[0031] Acquire a task state vector, a resource state vector and a conflict state vector, and combine the task state vector, the resource state vector and the conflict state vector to construct a state space vector; and establish a constraint condition set based on the state space vector;

[0032] Obtain the duration of all tasks, determine the maximum task duration, and use the product of the maximum task duration and a preset time coefficient as the reference window size; calculate the rolling step size according to the reference window size, the rolling step size is the product of the reference window size and the step coefficient, and the step coefficient has a value range of 0.2 to 0.5; determine the prediction interval according to the reference window size, the prediction interval is the product of the reference window size and the prediction coefficient, and the prediction coefficient is greater than or equal to 2; within the prediction interval, use a rolling optimization method based on the constraint condition set to generate multiple candidate adjustment schemes;

[0033] Constructing a task dependency graph, wherein nodes of the task dependency graph represent tasks, and edges represent dependencies between tasks; using a critical path algorithm to identify a critical path from the task dependency graph, wherein the critical path represents a task chain that has the greatest impact on the total construction period; calculating the priority of each task based on the critical path, wherein the priority is positively correlated with the position of the task on the critical path and the number of its successor tasks; decomposing and reorganizing the tasks according to the task priorities, and generating a reorganization plan;

[0034] Construct a resource capability matrix, wherein the rows of the resource capability matrix represent resources, the columns represent capability indicators, and the matrix elements represent specific capability values ​​of resources; construct a task requirement vector, wherein each component of the task requirement vector represents the requirement value of the capability indicator corresponding to the task; calculate the matching degree between tasks and resources using cosine similarity based on the resource capability matrix and the task requirement vector; perform resource allocation for the tasks in the reorganization plan according to the matching degree, and generate a resource allocation plan;

[0035] The degree of conflict impact is calculated in real time, and the degree of conflict impact is positively correlated with the number of conflicts, the intensity of conflicts and the scope of conflict impact; the adjustment cost of the adjustment plan is calculated, and the adjustment cost is positively correlated with the number of resource reallocations, the task delay time and the system performance loss; when the degree of conflict impact is less than a first preset threshold and the adjustment cost is less than a second preset threshold, the plan with the largest target value of the resource scheduling objective function is selected from the candidate adjustment plans as the output plan, otherwise the resource allocation plan is adopted as the output plan.

[0036] The task state vector includes task completion, task urgency and task dependency, the resource state vector includes resource occupancy, resource remaining capacity and resource available time period, and the conflict state vector includes conflict type, conflict impact range and conflict duration;

[0037] The constraint condition set includes timing constraints, resource constraints and space constraints, wherein the timing constraints define the order of task execution, the resource constraints define the upper and lower limits of resource capabilities and their mutually exclusive relationships, and the space constraints define the spatial distribution requirements of task execution.

[0038] According to a second aspect of the embodiments of the present invention,

[0039] Provides an AI-based digital project management system, including:

[0040] The first unit is used to collect power grid project construction resource data, wherein the power grid project construction resource data includes material resource data, equipment resource data and construction site resource data; construct a multi-level resource constraint matrix based on the power grid project construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; input the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model;

[0041] The second unit is used to calculate the real-time dispatch weight coefficients of various types of resources based on the collaborative optimization configuration model of power grid engineering construction resources, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; construct a resource dispatch objective function according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; adopt an improved ant colony algorithm to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch plan, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm;

[0042] The third unit is used to input the optimal resource scheduling plan into a preset intelligent collaborative controller, and the intelligent collaborative controller includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; the resource status monitoring module monitors the scheduling execution status of various types of resources in real time, the conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base, and the scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent collaborative scheduling of power grid project construction resources.

[0043] According to a third aspect of the embodiments of the present invention,

[0044] An electronic device is provided, comprising:

[0045] processor;

[0046] a memory for storing processor-executable instructions;

[0047] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0048] A fourth aspect of the embodiments of the present invention is:

[0049] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0050] The beneficial effects of this application are as follows:

[0051] 1. Improve the efficiency and accuracy of resource allocation:

[0052] The present invention collects multi-dimensional resource data of power grid engineering construction, constructs a multi-level resource constraint matrix, and uses a deep neural network model to generate a resource collaborative optimization configuration model. This method can comprehensively consider the constraints of multiple dimensions such as time, space and resource type, realize the precise configuration of power grid engineering construction resources, and significantly improve the efficiency and accuracy of resource configuration.

[0053] 2. Optimize resource scheduling scheme:

[0054] The present invention introduces a real-time scheduling weight coefficient to construct a resource scheduling objective function including multiple sub-functions such as resource utilization, project progress and cost control. Through iterative optimization and solving by the improved ant colony algorithm, a more reasonable and efficient resource scheduling solution can be generated. The improved ant colony algorithm introduces dynamic heuristic factors, local search mechanisms and global update rules, which further improves the convergence speed and solution quality of the algorithm.

[0055] 3. Realize intelligent collaborative scheduling:

[0056] The present invention adopts an intelligent collaborative controller, including modules such as resource status monitoring, conflict detection and scheduling optimization, which can monitor the resource scheduling execution status in real time, timely identify and handle conflict events in the scheduling process, and dynamically adjust the scheduling plan. This intelligent collaborative scheduling mechanism greatly improves the flexibility and adaptability of resource scheduling in power grid engineering construction, effectively copes with complex and changeable engineering environments, and ensures the continuity and efficiency of resource scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flowchart of a digital project management method based on artificial intelligence according to an embodiment of the present invention;

[0058] Figure 2 It is a structural diagram of a digital project management system based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0061] Figure 1 FIG. 1 is a flow chart of a digital project management method based on artificial intelligence according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0062] S101. Collecting power grid project construction resource data, the power grid project construction resource data includes material resource data, equipment resource data and construction site resource data; constructing a multi-level resource constraint matrix based on the power grid project construction resource data, the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; inputting the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model;

[0063] S102. Based on the collaborative optimization configuration model of power grid engineering construction resources, calculate the real-time dispatch weight coefficients of various types of resources, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; construct a resource dispatch objective function according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; use an improved ant colony algorithm to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch solution, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm;

[0064] S103. Input the optimal resource scheduling plan into a preset intelligent collaborative controller, which includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; the resource status monitoring module monitors the scheduling execution status of various types of resources in real time, the conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base, and the scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent collaborative scheduling of power grid project construction resources.

[0065] The specific implementation methods of the digital project management method based on artificial intelligence are as follows:

[0066] First, collect the resource data of power grid project construction. This step includes collecting material resource data, equipment resource data and construction site resource data. Material resource data includes the inventory, supplier information, procurement cycle, etc. of various construction materials. Equipment resource data includes the model, quantity, usage status, maintenance plan, etc. of various mechanical equipment. Construction site resource data includes the area, geographical location, site conditions, etc. of each construction area. These data are collected in real time through multiple channels such as IoT sensors, ERP systems, and construction site management systems.

[0067] Next, a multi-level resource constraint matrix is ​​constructed based on the collected power grid engineering construction resource data. The matrix contains three sub-matrices: the time dimension constraint sub-matrix, the space dimension constraint sub-matrix, and the resource type constraint sub-matrix. The time dimension constraint sub-matrix reflects the availability and usage restrictions of various resources in different time periods. The space dimension constraint sub-matrix represents the distribution and deployment restrictions of various resources in different construction areas. The resource type constraint sub-matrix describes the complementary and substitution relationship between different types of resources. For example, if a certain equipment is occupied within a specific time period, the corresponding time dimension constraint sub-matrix element is set to 0; if a certain material can only be used in a specific construction area, the corresponding space dimension constraint sub-matrix element is set to 1.

[0068] Then, the constructed multi-level resource constraint matrix is ​​input into the preset deep neural network model. The model adopts a multi-layer convolutional neural network structure, and generates a resource collaborative optimization configuration model for power grid engineering construction by extracting features and performing nonlinear mapping on the constraint matrix. The model can automatically learn the implicit rules of resource configuration and predict future resource demand.

[0069] Based on the generated collaborative optimization configuration model of power grid project construction resources, the real-time dispatch weight coefficients of various resources are calculated. This coefficient reflects the dispatch priority of different resource types at the current moment, and the larger the value, the higher the priority. The calculation process takes into account multiple dimensions such as resource scarcity, the degree of impact on project progress, and cost factors. For example, for a key equipment, if its current utilization rate is low and it has an impact on multiple construction links, its dispatch weight coefficient will be increased accordingly.

[0070] Next, the resource scheduling objective function is constructed based on the calculated real-time scheduling weight coefficient. This function consists of three sub-functions: resource utilization sub-function, project progress sub-function and cost control sub-function. The resource utilization sub-function is used to maximize the efficiency of the use of various resources. The project progress sub-function aims to ensure that key construction links are completed on schedule. The cost control sub-function is committed to minimizing the resource scheduling cost while meeting the project requirements. These three sub-functions are combined into the final objective function by weighted summation.

[0071] Subsequently, the improved ant colony algorithm is used to iteratively optimize and solve the constructed resource scheduling objective function to generate the optimal resource scheduling solution. The improved ant colony algorithm introduces dynamic heuristic factors, local search mechanisms and global update rules on the basis of the standard ant colony algorithm. The dynamic heuristic factor is adaptively adjusted according to the search situation during the iteration to balance the exploration and utilization capabilities of the algorithm. The local search mechanism performs local optimization on the current optimal solution after each iteration to improve the quality of the solution. The global update rule updates the global pheromone after the iteration to avoid falling into the local optimum. Through these improvements, the algorithm can find a resource scheduling solution close to the global optimum more quickly and efficiently.

[0072] The generated optimal resource scheduling plan is input into the preset intelligent collaborative controller. The controller includes a resource status monitoring module, a conflict detection module and a scheduling optimization module. The resource status monitoring module continuously tracks the scheduling execution status of various resources through real-time data collection and analysis. The conflict detection module automatically identifies possible conflict events that may occur in the resource scheduling process based on a pre-defined rule base, such as duplicate resource allocation and scheduling time conflicts. The scheduling optimization module dynamically adjusts the optimal resource scheduling plan based on the detected conflict events and the real-time resource status. For example, when a device failure is detected, the scheduling optimization module will automatically find a replacement device and rearrange the scheduling plan for the related construction tasks.

[0073] Finally, the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent coordinated scheduling of power grid project construction resources. The execution terminal includes mobile device apps, smart wearable devices, on-site electronic display screens, etc., to ensure that various resources can operate efficiently and collaboratively according to the optimized scheduling plan.

[0074] The beneficial effects of this method are mainly reflected in the following three aspects:

[0075] 1. Improve resource utilization efficiency: Through the multi-level resource constraint matrix and deep neural network model, accurate configuration and optimal scheduling of various resources are achieved, which significantly improves resource utilization and reduces resource waste. 2. Optimize project progress management: The improved ant colony algorithm is used to optimize and solve the resource scheduling objective function, which can accelerate the construction progress and shorten the construction period to the greatest extent while ensuring the quality of the project. 3. Enhance the intelligence and adaptability of the scheduling process: The introduction of the intelligent collaborative controller enables the resource scheduling process to have the ability of autonomous monitoring, conflict detection and dynamic optimization, which can quickly respond to on-site emergencies and ensure the reliability and flexibility of the scheduling plan.

[0076] In an optional implementation, a multi-level resource constraint matrix is ​​constructed based on the power grid engineering construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix, and a resource type constraint sub-matrix; the multi-level resource constraint matrix is ​​input into a preset deep neural network model to generate a power grid engineering construction resource collaborative optimization configuration model, including:

[0077] The construction period of the project is divided into multiple time units according to the construction process, and a resource availability vector is established for each time unit. The resource availability vector includes equipment operation time parameters, material supply time parameters and site occupancy time parameters. According to the time sequence mapping relationship between the time unit and the resource availability vector, a time dimension constraint sub-matrix is ​​generated;

[0078] The construction site is divided into a plurality of regional units according to the construction area. The division of the regional units is based on the connection relationship of the construction process and the resource allocation radius. A resource capacity vector is established for each of the regional units. The resource capacity vector includes operator capacity parameters, mechanical equipment capacity parameters, material storage capacity parameters and site carrying capacity parameters. Based on the corresponding relationship between the spatial distribution of the regional units and the resource capacity vectors, a spatial dimension constraint sub-matrix is ​​constructed.

[0079] Construct a resource feature recognition model for different resource types, extract a resource attribute feature vector based on the resource feature recognition model, the resource attribute feature vector includes a resource use priority feature, a resource allocation flexibility feature, a resource use cost feature and a resource substitutability feature, establish a resource type classifier, the resource type classifier uses a support vector machine algorithm to classify resources, and generates a resource type constraint submatrix according to the association between the resource type and the resource attribute feature vector;

[0080] Inputting the time dimension constraint submatrix, the space dimension constraint submatrix and the resource type constraint submatrix into a deep neural network model, wherein the deep neural network model includes an input layer, three hidden layers and an output layer, and a ReLU activation function is set between adjacent layers;

[0081] The deep neural network model is trained based on a composite loss function, wherein the composite loss function includes a resource utilization loss term, a spatiotemporal constraint violation loss term, and a resource synergy efficiency loss term, and each loss term is set with a corresponding weight coefficient; a dropout layer is set after each hidden layer of the deep neural network model, and the dropout rate of the dropout layer is 0.3, and an L2 regularization weight decay term is added;

[0082] The Adam optimizer is used to optimize the deep neural network model, set the initial learning rate, and dynamically adjust the learning rate according to the training rounds to generate a collaborative optimization configuration model for power grid engineering construction resources.

[0083] The present invention provides a method for collaborative optimization configuration of power grid engineering construction resources based on a multi-level resource constraint matrix. The method first constructs a multi-level resource constraint matrix, including three sub-matrices: time dimension, space dimension, and resource type. Then, the matrix is ​​input into a preset deep neural network model to generate a collaborative optimization configuration model for power grid engineering construction resources. The specific implementation method is as follows:

[0084] First, collect and organize the resource data of power grid project construction, including human resources, equipment resources, material resources, site resources, etc. Analyze and preprocess these resource data, remove outliers, fill in missing values, and perform standardization.

[0085] Next, construct the time dimension constraint submatrix. Divide the entire construction cycle into multiple time units according to the construction process, for example, divide it into 24 time units by month. For each time unit, establish a resource availability vector, which includes parameters such as equipment operation time, material supply time, and site occupancy time. For example, the resource availability vector of a time unit may be [180, 15, 30], indicating that the equipment can run for 180 hours, the material supply takes 15 days, and the site can be occupied for 30 days. According to the correspondence between the time unit and the resource availability vector, generate a 24×3 time dimension constraint submatrix.

[0086] Then, a spatial dimension constraint sub-matrix is ​​constructed. The construction site is divided into multiple regional units according to the construction area, and the division basis includes the connection relationship of the construction process and the radius of resource allocation. For example, it can be divided into 10 regional units. A resource capacity vector is established for each regional unit, including parameters such as operator capacity, machinery and equipment capacity, material storage capacity, and site carrying capacity. The resource capacity vector of a regional unit may be [50, 10, 1000, 5000], indicating that the area can accommodate 50 operators, 10 machinery and equipment, 1000 cubic meters of material storage space, and 5000 square meters of site area. Based on the correspondence between the spatial distribution of regional units and the resource capacity vector, a 10×4 spatial dimension constraint sub-matrix is ​​constructed.

[0087] Then, we construct the resource type constraint submatrix. We build a resource feature recognition model for different resource types (such as manpower, equipment, materials, etc.), and use deep learning methods to extract resource attribute feature vectors, including usage priority, deployment flexibility, usage cost, substitutability and other features. For example, the feature vector of a certain type of resource may be [0.8, 0.6, 0.7, 0.5]. Then, we build a resource type classifier based on the support vector machine algorithm to classify resources. According to the relationship between resource type and feature vector, we generate a resource type constraint submatrix. Assuming there are 5 types of resources, the matrix dimension is 5×4.

[0088] The above three sub-matrices are used as input and sent to the preset deep neural network model. The model includes an input layer, three hidden layers and an output layer, and a ReLU activation function is set between adjacent layers. The number of nodes in the input layer is the sum of the elements of the three sub-matrices of time, space, and resource type. The number of nodes in the three hidden layers is 128, 64, and 32 respectively. The number of nodes in the output layer is the number of resources to be optimized.

[0089] During the model training process, a composite loss function is used, including resource utilization loss term, space-time constraint violation loss term, and resource synergy efficiency loss term. Each loss term is set with a corresponding weight coefficient, for example, it can be set to 0.4, 0.3, and 0.3 respectively. A dropout layer is set after each hidden layer, with a dropout rate of 0.3, and an L2 regularization weight decay term is added, with a coefficient of 0.01 to prevent overfitting.

[0090] The Adam optimizer is used to optimize the model, and the initial learning rate is set to 0.001. The learning rate is dynamically adjusted according to the training rounds, for example, the learning rate is reduced to 0.1 times the original every 50 epochs. Through repeated iterative training, a collaborative optimization configuration model for power grid engineering construction resources is finally generated.

[0091] In practical applications, by inputting the resource data of a specific project into the model, an optimized resource allocation plan can be obtained, including the allocation quantity of various resources at different times and in different regions. For example, 20 construction workers, 2 excavators, and 100 tons of steel are allocated to a certain area in a certain month.

[0092] The beneficial effects of this method are mainly reflected in the following three aspects:

[0093] 1. Improved resource utilization efficiency. Through the multi-level constraint matrix, factors such as time, space and resource type are fully considered, and accurate resource allocation is achieved, which avoids resource waste and improves overall utilization.

[0094] 2. Enhanced flexibility in resource allocation. The deep learning-based optimization model can quickly adapt to the dynamic changes of the construction site, adjust the resource allocation plan in time according to the actual situation, and ensure the construction progress.

[0095] 3. Optimized resource synergy. By considering the correlation and substitutability between different resources, the synergy optimization of various resources is achieved, which improves the overall construction efficiency, shortens the construction period and reduces the cost.

[0096] In an optional implementation, based on the collaborative optimization configuration model of power grid engineering construction resources, the real-time dispatch weight coefficients of various types of resources are calculated, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; a resource dispatch objective function is constructed according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function, including:

[0097] Construct a resource evaluation index system, which includes timeliness index, criticality index, substitutability index and coordination index, among which:

[0098] The timeliness index is calculated by task deadline, current time and planned duration to obtain the first-level timeliness index, and the second-level timeliness index is calculated by associating process weight and resource process association coefficient; the criticality index is calculated based on the ratio of the longest process path length to the critical path length in the process network topology structure and the resource usage intensity; the substitutability index is calculated based on the resource feature similarity matrix and the substitution cost coefficient matrix; the coordination index is calculated based on the resource node centrality calculated based on the resource coordination efficiency matrix;

[0099] Establish an indicator importance evaluation matrix, the matrix elements of which represent the relative importance of different indicators in the resource evaluation indicator system, calculate an initial weight vector based on the indicator importance evaluation matrix, construct a time-varying adjustment factor, the time-varying adjustment factor is related to the total construction period and the adjustment coefficient, and generate a resource scheduling dynamic weight based on the initial weight vector, the time-varying adjustment factor and the normalized score of each indicator in the resource evaluation indicator;

[0100] The resource scheduling objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function. The resource utilization sub-function is calculated based on the actual utilization, the planned utilization and the resource efficiency coefficient. The project progress sub-function is calculated based on the actual progress, the planned progress and the progress influencing factor. The cost control sub-function is calculated based on the actual cost, the planned cost and the cost sensitivity coefficient.

[0101] The fuzzy hierarchical analysis method is used to determine the resource scheduling dynamic weights corresponding to the resource utilization sub-function, the project progress sub-function and the cost control sub-function, and the resource scheduling objective function is obtained by linearly weighted combination of each sub-function based on the resource scheduling dynamic weights.

[0102] According to the resource collaborative optimization configuration model for power grid engineering construction, a resource evaluation index system is first constructed, which includes timeliness index, criticality index, substitutability index and coordination index.

[0103] For timeliness index, the first-level timeliness index is calculated by the task deadline, current time and planned duration. For example, if the deadline of a task is December 31, 2023, the current time is June 1, 2023, and the planned duration is 180 days, the first-level timeliness index can be expressed by the ratio of the remaining time to the planned duration. The second-level timeliness index is calculated by associating the process weight with the resource-process correlation coefficient. Assuming that a resource is related to three processes, the correlation coefficients are 0.8, 0.6 and 0.4 respectively, and the corresponding process weights are 0.5, 0.3 and 0.2, the second-level timeliness index can be obtained by weighted average.

[0104] The criticality index is calculated based on the ratio of the longest path length of the process to the critical path length in the process network topology and the resource usage intensity. For example, if the longest path length of a process is 10 days, the critical path length is 15 days, and the resource usage intensity is 0.8, then these factors can be considered together to get the criticality index.

[0105] The substitutability index is calculated based on the resource feature similarity matrix and the substitution cost coefficient matrix. Assuming there are three resources, the feature similarity matrix between them is a 3x3 matrix, and the substitution cost coefficient matrix is ​​also a 3x3 matrix. The substitutability index can be obtained by calculating these two matrices.

[0106] The synergy index is obtained by calculating the centrality of resource nodes based on the resource synergy efficiency matrix. For example, there are 5 types of resources, and the synergy efficiency between them can be represented by a 5x5 matrix. By calculating the degree of association between each resource node and other resource nodes, the resource synergy index can be obtained.

[0107] Next, we build an indicator importance evaluation matrix. The matrix elements represent the relative importance of different indicators in the resource evaluation indicator system. Assuming there are 4 indicators, we can build a 4x4 evaluation matrix. Based on this matrix, we can calculate the initial weight vector. For example, we may get an initial weight of [0.3, 0.25, 0.2, 0.25].

[0108] Construct a time-varying adjustment factor, which is related to the total duration of the project and the adjustment coefficient. For example, the ratio of the current time to the total duration can be used as the basis, and then the adjustment coefficient can be introduced to obtain the time-varying adjustment factor. Assuming the total duration is 365 days, the current day is the 100th day, and the adjustment coefficient is 0.8, the time-varying adjustment factor may be 0.22.

[0109] The resource scheduling dynamic weight is generated based on the initial weight vector, the time-varying adjustment factor, and the normalized score of each indicator in the resource evaluation index. For example, if the normalized scores of the four indicators are 0.8, 0.7, 0.9, and 0.6, respectively, the final dynamic weight can be calculated by combining the initial weight and the time-varying adjustment factor.

[0110] Then, the resource scheduling objective function is constructed, including the resource utilization sub-function, the project progress sub-function and the cost control sub-function. The resource utilization sub-function is calculated based on the actual utilization, the planned utilization and the resource efficiency coefficient. For example, if the actual utilization of a resource is 80%, the planned utilization is 75%, and the resource efficiency coefficient is 1.2, the value of the resource utilization sub-function can be calculated through these parameters.

[0111] The engineering progress sub-function is calculated based on the actual progress, planned progress and progress impact factor. Assuming the actual progress is 60%, the planned progress is 65%, and the progress impact factor is 1.1, the value of the engineering progress sub-function can be calculated using these parameters.

[0112] The cost control sub-function is calculated based on the actual cost, planned cost and cost sensitivity coefficient. For example, if the actual cost of a project is 5 million yuan, the planned cost is 4.8 million yuan, and the cost sensitivity coefficient is 0.9, the value of the cost control sub-function can be calculated based on these parameters.

[0113] Finally, the fuzzy hierarchical analysis method is used to determine the dynamic weights of resource scheduling corresponding to the resource utilization sub-function, project progress sub-function and cost control sub-function. Assume that the weights of the three sub-functions obtained by fuzzy hierarchical analysis are 0.4, 0.35 and 0.25 respectively. Based on these dynamic weights of resource scheduling, each sub-function is linearly weighted combined to finally obtain the resource scheduling objective function.

[0114] The beneficial effects of this method include the following three aspects:

[0115] 1. By building a comprehensive resource evaluation index system, taking into account multiple dimensions such as timeliness, criticality, substitutability and synergy, it is possible to more comprehensively and accurately evaluate the importance and scheduling priority of various resources. 2. The introduction of time-varying adjustment factors and dynamic weight mechanisms enables the resource scheduling strategy to be dynamically adjusted according to the progress of the project, improving the flexibility and adaptability of resource scheduling. 3. The multi-objective optimization method is adopted, taking into account multiple objectives such as resource utilization, project progress and cost control at the same time, which can achieve a balance between multiple key performance indicators and realize the coordinated optimization configuration of power grid project construction resources.

[0116] In an optional implementation, using an improved ant colony algorithm to iteratively optimize and solve the resource scheduling objective function to generate an optimal resource scheduling solution includes:

[0117] Establish a pheromone matrix, wherein the row vectors of the pheromone matrix represent available resources, the column vectors represent task requirements, and the matrix elements are determined according to the product of the pheromone attenuation coefficient and the pheromone matrix at the previous moment and the pheromone increment, and the pheromone increment is calculated by the target value corresponding to the resource scheduling objective function of the path passed by the ants;

[0118] Calculating a resource scheduling decision distance, wherein the resource scheduling decision distance is determined based on a matching degree between the available resources and the task requirements, and calculating a dynamic heuristic factor according to the resource scheduling decision distance, wherein the dynamic heuristic factor is obtained by exponentially calculating the inverse of the resource scheduling decision distance and a dynamic adjustment coefficient;

[0119] The state transition probability is calculated based on the pheromone matrix and the dynamic heuristic factor, the ant colony search path is initialized, and the resource scheduling scheme of the initial ant population is randomly generated under the condition of satisfying the preset constraint condition list according to the state transition probability, and the target value corresponding to the resource scheduling objective function corresponding to each resource scheduling scheme is calculated, and the resource scheduling scheme with the smallest target value is recorded as the current optimal solution, wherein the preset constraint condition list includes resource layer constraints, task layer constraints and system layer constraints;

[0120] Performing ant colony iterative search, determining a new resource scheduling decision plan for each ant based on the state transition probability; optimizing the current optimal solution by using a local search mechanism in each iteration, calculating the movement increment by defining a neighborhood structure centered on the current optimal solution, and determining whether to accept the candidate solution in the neighborhood based on the movement increment and the Metropolis criterion;

[0121] The global optimal solution is updated according to the local search results. When a global optimal solution that is better than the current optimal solution appears, the pheromone matrix is ​​updated and the number of local searches is reset, and the new global optimal solution is recorded. When the maximum number of iterations is reached, the iteration is stopped and the optimal resource scheduling plan is output. DETAILED DESCRIPTION

[0123] The improved ant colony algorithm is used to iteratively optimize and solve the resource scheduling objective function. The specific steps to generate the optimal resource scheduling solution are as follows:

[0124] First, a pheromone matrix is ​​established. The row vectors of the matrix represent available resources, and the column vectors represent task requirements. Each element in the matrix is ​​determined by the product of the pheromone decay coefficient and the pheromone matrix at the previous moment and the pheromone increment. The pheromone increment is calculated by the target value corresponding to the resource scheduling objective function of the path passed by the ant. For example, assuming there are 3 available resources and 4 task requirements, the pheromone matrix at the initial moment can be set to a 3x4 all-1 matrix. The pheromone decay coefficient is set to 0.5, and the corresponding objective function value of the path passed by an ant is 10, then the pheromone increment on the path can be set to 0.1.

[0125] Next, calculate the resource scheduling decision distance. This distance is determined based on the matching degree between available resources and task requirements. For example, it can be represented by calculating the Euclidean distance between the resource capability vector and the task requirement vector. The dynamic heuristic factor is calculated based on the resource scheduling decision distance, which can be obtained by the power operation of the inverse of the resource scheduling decision distance and the dynamic adjustment coefficient. Assuming that the decision distance between a resource and a task is 2 and the dynamic adjustment coefficient is 0.8, the corresponding dynamic heuristic factor can be calculated as (1 / 2)^0.8.

[0126] Then, the state transition probability is calculated based on the pheromone matrix and the dynamic heuristic factor. The ant colony search path is initialized, and the resource scheduling scheme of the initial ant population is randomly generated under the condition of satisfying the preset constraint list according to the state transition probability. The preset constraint list includes resource layer constraints, task layer constraints and system layer constraints. For example, resource layer constraints may include resource quantity restrictions; task layer constraints may include task completion time requirements; system layer constraints may include total energy consumption restrictions, etc. Calculate the resource scheduling objective function value corresponding to each resource scheduling scheme, and record the resource scheduling scheme with the smallest objective value as the current optimal solution.

[0127] Then, the ant colony iterative search is performed to determine a new resource scheduling decision for each ant based on the state transition probability. In each iteration, a local search mechanism is used to optimize the current optimal solution. Specifically, the moving increment can be calculated by defining a neighborhood structure centered on the current optimal solution, and whether to accept the candidate solution in the neighborhood is determined based on the moving increment and the Metropolis criterion. For example, the neighborhood structure can be defined as the allocation of two tasks randomly exchanged in the current optimal solution, and the difference in the objective function value before and after the exchange is calculated as the moving increment. If the moving increment is negative, the new solution is accepted; if it is positive, the new solution is accepted with a certain probability.

[0128] Finally, the global optimal solution is updated according to the local search results. When a global optimal solution that is better than the current optimal solution appears, the pheromone matrix is ​​updated and the number of local searches is reset, and the new global optimal solution is recorded. For example, the pheromone on the path corresponding to the new global optimal solution can be increased by a fixed value. When the maximum number of iterations is reached, the iteration is stopped and the optimal resource scheduling solution is output.

[0129] The use of the above improved ant colony algorithm for resource scheduling optimization has the following beneficial effects:

[0130] By introducing dynamic heuristic factors, the algorithm's ability to perceive problem features is improved, and the search is more targeted and effective. The local search mechanism is adopted to further optimize the current optimal solution in each iteration, effectively avoiding the algorithm from falling into the local optimum. Combining the constraints of the resource layer, task layer and system layer, the feasibility and practicality of the generated resource scheduling solution are guaranteed, and the practical application value of the algorithm is improved.

[0131] In an optional implementation, the scheduling optimization module dynamically adjusts the optimal resource scheduling solution according to the conflict event, including:

[0132] Acquire a task state vector, a resource state vector and a conflict state vector, and combine the task state vector, the resource state vector and the conflict state vector to construct a state space vector; and establish a constraint condition set based on the state space vector;

[0133] Obtain the duration of all tasks, determine the maximum task duration, and use the product of the maximum task duration and a preset time coefficient as the reference window size; calculate the rolling step size according to the reference window size, the rolling step size is the product of the reference window size and the step coefficient, and the step coefficient has a value range of 0.2 to 0.5; determine the prediction interval according to the reference window size, the prediction interval is the product of the reference window size and the prediction coefficient, and the prediction coefficient is greater than or equal to 2; within the prediction interval, use a rolling optimization method based on the constraint condition set to generate multiple candidate adjustment schemes;

[0134] Constructing a task dependency graph, wherein nodes of the task dependency graph represent tasks, and edges represent dependencies between tasks; using a critical path algorithm to identify a critical path from the task dependency graph, wherein the critical path represents a task chain that has the greatest impact on the total construction period; calculating the priority of each task based on the critical path, wherein the priority is positively correlated with the position of the task on the critical path and the number of its successor tasks; decomposing and reorganizing the tasks according to the task priorities, and generating a reorganization plan;

[0135] Construct a resource capability matrix, wherein the rows of the resource capability matrix represent resources, the columns represent capability indicators, and the matrix elements represent specific capability values ​​of resources; construct a task requirement vector, wherein each component of the task requirement vector represents the requirement value of the capability indicator corresponding to the task; calculate the matching degree between tasks and resources using cosine similarity based on the resource capability matrix and the task requirement vector; perform resource allocation for the tasks in the reorganization plan according to the matching degree, and generate a resource allocation plan;

[0136] The degree of conflict impact is calculated in real time, and the degree of conflict impact is positively correlated with the number of conflicts, the intensity of conflicts and the scope of conflict impact; the adjustment cost of the adjustment plan is calculated, and the adjustment cost is positively correlated with the number of resource reallocations, the task delay time and the system performance loss; when the degree of conflict impact is less than a first preset threshold and the adjustment cost is less than a second preset threshold, the plan with the largest target value of the resource scheduling objective function is selected from the candidate adjustment plans as the output plan, otherwise the resource allocation plan is adopted as the output plan.

[0137] The specific implementation method of the scheduling optimization module dynamically adjusting the optimal resource scheduling scheme according to the conflict event is as follows:

[0138] First, we obtain the task state vector, resource state vector, and conflict state vector. The task state vector contains the current execution status of each task, such as not started, executing, completed, etc. The resource state vector contains the current available quantity of each resource. The conflict state vector contains the occurrence of various conflict events. These three vectors are combined to form a state space vector, which is used to represent the overall state of the system.

[0139] A set of constraints is established based on the state space vector. Constraints include total resource constraints, task dependencies, task execution time constraints, etc. For example, if a task requires 2 devices A and 3 workers B, then the constraints are that the usage of device A during the execution of the task does not exceed 2, and the usage of worker B does not exceed 3.

[0140] Next, obtain the duration of all tasks and find the maximum task duration. Assuming that the maximum task duration is 10 days and the preset time coefficient is 1.5, the benchmark window size is 15 days. Calculate the rolling step size based on the benchmark window size, and take the step coefficient as 0.3, so the rolling step size is 4.5 days. Determine the prediction interval, take the prediction coefficient as 2.5, and the prediction interval is 37.5 days. Within the 37.5-day prediction interval, use the rolling optimization method with a step size of 4.5 days to generate multiple candidate adjustment plans.

[0141] Then, a task dependency graph is constructed, where nodes represent tasks and edges represent dependencies between tasks. For example, Task 2 and Task 3 can only be started after Task 1 is completed, so there are edges connecting Task 1 with Task 2 and Task 3. The critical path algorithm is used to identify the task chain that has the greatest impact on the total construction period. Assume that the identified critical path is Task 1-Task 3-Task 5-Task 7. The priority of each task is calculated based on the critical path. The later the task is on the critical path, the higher the priority. At the same time, the number of successor tasks of the task is considered. The more successor tasks there are, the higher the priority. The tasks are decomposed and reorganized according to the calculated task priorities, and a reorganization plan is generated.

[0142] Construct a resource capability matrix, where rows represent different types of resources, columns represent various capability indicators, and matrix elements represent specific capability values. For example, the capability values ​​of a certain device in terms of accuracy, efficiency, and stability are 0.8, 0.9, and 0.7, respectively. Construct a task requirement vector, where each component represents the required value of the capability indicator corresponding to the task. Based on the resource capability matrix and the task requirement vector, the cosine similarity is used to calculate the matching degree between tasks and resources. Allocate resources to the tasks in the reorganization plan according to the matching degree, and generate a resource allocation plan.

[0143] Calculate the impact of conflicts in real time, taking into account the number of conflicts, conflict intensity, and conflict scope. For example, if three conflict events occur with an average intensity of 0.6, affecting five tasks, the impact of the conflict can be calculated as 3*0.6*5=9. Calculate the adjustment cost of the adjustment plan, taking into account the number of resource reallocations, task delay time, and system performance loss. For example, if 10 resources need to be reallocated, resulting in a delay of 2 days for three tasks and a 5% drop in system performance, the adjustment cost can be calculated as 10+3*2+5=21.

[0144] When the impact of the conflict is less than the preset threshold of 10 and the adjustment cost is less than the preset threshold of 25, the solution with the largest target value of the resource scheduling objective function is selected from the candidate adjustment solutions as the output solution. Otherwise, the resource allocation solution generated previously is used as the output solution. Through the above steps, the optimal resource scheduling solution can be dynamically adjusted according to the conflict event.

[0145] This technical solution has the following beneficial effects:

[0146] 1. By constructing the state space vector and constraint condition set, the system state and various constraints are fully considered, and the feasibility and rationality of the scheduling scheme are improved. 2. The rolling optimization and critical path analysis are adopted to achieve the prediction optimization for a period of time in the future, while ensuring the priority scheduling of important tasks, improving the scheduling efficiency and quality. 3. The resource capacity matrix and task demand vector are introduced to achieve the precise matching of tasks and resources. Combined with the evaluation of the impact degree of conflict and the adjustment cost, the dynamic adjustment and optimization of the scheduling scheme is guaranteed, and the robustness and adaptability of the system are improved.

[0147] In an optional implementation, the task state vector includes task completion, task urgency and task dependency, the resource state vector includes resource occupancy, resource remaining capacity and resource available time period, and the conflict state vector includes conflict type, conflict impact range and conflict duration;

[0148] The constraint condition set includes timing constraints, resource constraints and space constraints, wherein the timing constraints define the order of task execution, the resource constraints define the upper and lower limits of resource capabilities and their mutually exclusive relationships, and the space constraints define the spatial distribution requirements of task execution.

[0149] In a specific implementation, a task state vector, a resource state vector and a conflict state vector are first constructed. The task state vector contains three dimensions: task completion, task urgency and task dependency. Task completion uses 0-100% to represent the completion progress of the current task, and task urgency uses an integer from 1 to 5 to represent the urgency of the task, with 1 representing the least urgent and 5 representing the most urgent. Task dependency is represented by a binary matrix, and the element aij in the matrix represents whether task i depends on task j, 1 represents dependency and 0 represents non-dependency.

[0150] The resource status vector contains three dimensions: resource occupancy, resource remaining capacity, and resource available time period. Resource occupancy uses 0-100% to represent the current resource usage, resource remaining capacity uses actual values ​​to represent the amount of work that can be undertaken, and resource available time period uses time intervals to represent the available time of the resource. For example, the resource status vector of a device can be expressed as [80%, 20 working hours, 9:00-17:00].

[0151] The conflict state vector contains three dimensions: conflict type, conflict impact range, and conflict duration. The conflict type is coded with integers to represent different types of conflicts, such as 1 for resource conflict and 2 for timing conflict. The conflict impact range is represented by the number of tasks affected. The conflict duration is represented by the number of hours. For example, the state vector of a conflict can be represented as [1, 3, 2.5h].

[0152] Next, define the constraint set, including timing constraints, resource constraints, and space constraints. Timing constraints define the order of task execution, which can be represented by a directed acyclic graph. The nodes in the graph represent tasks, and the edges represent the order of tasks. For example, A->B means that task A must be completed before task B. Resource constraints define the upper and lower limits of resource capabilities and their mutually exclusive relationships. The upper and lower limits of resource capabilities are represented by intervals to represent the minimum and maximum workload of the resource, such as [10, 50], which means that the resource works at least 10 hours a day and at most 50 hours. The mutually exclusive relationship of resources is represented by a binary matrix, and a matrix element of 1 means that two resources cannot be used at the same time. Spatial constraints define the spatial distribution requirements for task execution, and coordinates or areas can be used to represent the spatial location restrictions of tasks.

[0153] In practical applications, we first collect the status information of each task, resource, and conflict, and construct the corresponding state vector. Then, we define various constraints according to the actual situation. Next, we use this information to optimize the scheduling. During the optimization process, we first check whether the constraints are met. For scheduling schemes that do not meet the constraints, we optimize them by adjusting the task order, allocating different resources, or changing the execution time until a feasible solution that meets all the constraints is found. Finally, we select the best solution from the feasible solutions according to the established optimization goals (such as minimizing the total execution time, maximizing resource utilization, etc.).

[0154] During the scheduling process, each state vector needs to be updated dynamically. For example, when a task is 50% complete, its task completion degree is updated from 0% to 50%. When a resource is occupied for 2 hours, its resource occupancy rate and remaining capacity are adjusted accordingly. If a new conflict occurs, a new conflict state vector is added. Constraints may also change over time, such as adding new task dependencies, adjusting resource available time, etc. Therefore, it is necessary to monitor the system status in real time and update various information in a timely manner to ensure the timeliness and accuracy of the scheduling plan.

[0155] In specific implementation, heuristic algorithms or intelligent optimization algorithms can be used to solve scheduling problems. For example, a genetic algorithm can be used to encode each scheduling solution as a chromosome, generate new solutions through operations such as selection, crossover, and mutation, and gradually optimize to obtain a high-quality scheduling solution that meets the constraints. An ant colony algorithm can also be used to regard the task sequence as the path of ants, guide the search process through the accumulation and volatilization of pheromones, and finally find a near-optimal scheduling solution.

[0156] The beneficial effects of this method are mainly reflected in three aspects:

[0157] First, by constructing a multi-dimensional state vector, the characteristics of tasks, resources and conflicts are fully characterized, providing a rich information basis for accurate scheduling. Secondly, by adopting multiple types of constraints, the constraints of timing, resources and space are fully considered to ensure the feasibility and rationality of the scheduling scheme. Finally, by using the dynamic update mechanism, the changes in the system state can be reflected in a timely manner, adapting to the complex and changing actual environment, and improving the flexibility and adaptability of scheduling.

[0158] Figure 2 FIG. 1 is a schematic diagram of the structure of a digital project management system based on artificial intelligence according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0159] The first unit is used to collect power grid project construction resource data, wherein the power grid project construction resource data includes material resource data, equipment resource data and construction site resource data; construct a multi-level resource constraint matrix based on the power grid project construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; input the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model;

[0160] The second unit is used to calculate the real-time dispatch weight coefficients of various types of resources based on the collaborative optimization configuration model of power grid engineering construction resources, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; construct a resource dispatch objective function according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; adopt an improved ant colony algorithm to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch plan, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm;

[0161] The third unit is used to input the optimal resource scheduling plan into a preset intelligent collaborative controller, and the intelligent collaborative controller includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; the resource status monitoring module monitors the scheduling execution status of various types of resources in real time, the conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base, and the scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent collaborative scheduling of power grid project construction resources.

[0162] According to a third aspect of the embodiments of the present invention,

[0163] An electronic device is provided, comprising:

[0164] processor;

[0165] a memory for storing processor-executable instructions;

[0166] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0167] A fourth aspect of the embodiments of the present invention is:

[0168] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0169] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital project management method based on artificial intelligence, characterized by: include: Collecting power grid project construction resource data, the power grid project construction resource data including material resource data, equipment resource data and construction site resource data; constructing a multi-level resource constraint matrix based on the power grid project construction resource data, the multi-level resource constraint matrix including a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; inputting the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model; Based on the collaborative optimization configuration model of power grid engineering construction resources, the real-time dispatch weight coefficients of various types of resources are calculated, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; a resource dispatch objective function is constructed according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; an improved ant colony algorithm is used to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch plan, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm; Inputting the optimal resource scheduling solution into a preset intelligent collaborative controller, wherein the intelligent collaborative controller includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; The resource status monitoring module monitors the scheduling execution status of various types of resources in real time. The conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base. The scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent coordinated scheduling of power grid project construction resources.

2. The method according to claim 1, characterized in that: Constructing a multi-level resource constraint matrix based on the power grid engineering construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix, and a resource type constraint sub-matrix; inputting the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid engineering construction resource collaborative optimization configuration model includes: The construction period of the project is divided into multiple time units according to the construction process, and a resource availability vector is established for each time unit. The resource availability vector includes equipment operation time parameters, material supply time parameters and site occupancy time parameters. According to the time sequence mapping relationship between the time unit and the resource availability vector, a time dimension constraint sub-matrix is ​​generated; The construction site is divided into a plurality of regional units according to the construction area. The division of the regional units is based on the connection relationship of the construction process and the resource allocation radius. A resource capacity vector is established for each of the regional units. The resource capacity vector includes operator capacity parameters, mechanical equipment capacity parameters, material storage capacity parameters and site carrying capacity parameters. Based on the corresponding relationship between the spatial distribution of the regional units and the resource capacity vectors, a spatial dimension constraint sub-matrix is ​​constructed. Construct a resource feature recognition model for different resource types, extract a resource attribute feature vector based on the resource feature recognition model, the resource attribute feature vector includes a resource use priority feature, a resource allocation flexibility feature, a resource use cost feature and a resource substitutability feature, establish a resource type classifier, the resource type classifier uses a support vector machine algorithm to classify resources, and generates a resource type constraint submatrix according to the association between the resource type and the resource attribute feature vector; Inputting the time dimension constraint submatrix, the space dimension constraint submatrix and the resource type constraint submatrix into a deep neural network model, wherein the deep neural network model includes an input layer, three hidden layers and an output layer, and a ReLU activation function is set between adjacent layers; The deep neural network model is trained based on a composite loss function, wherein the composite loss function includes a resource utilization loss term, a spatiotemporal constraint violation loss term, and a resource synergy efficiency loss term, and each loss term is set with a corresponding weight coefficient; a dropout layer is set after each hidden layer of the deep neural network model, and the dropout rate of the dropout layer is 0.3, and an L2 regularization weight decay term is added; The Adam optimizer is used to optimize the deep neural network model, set the initial learning rate, and dynamically adjust the learning rate according to the training rounds to generate a collaborative optimization configuration model for power grid engineering construction resources.

3. The method according to claim 1, characterized in that Based on the collaborative optimization configuration model of power grid engineering construction resources, the real-time dispatch weight coefficients of various types of resources are calculated, and the real-time dispatch weight coefficients are used to characterize the dispatch priority of different resource types; A resource scheduling objective function is constructed according to the real-time scheduling weight coefficient, and the resource scheduling objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function, including: Construct a resource evaluation index system, which includes timeliness index, criticality index, substitutability index and coordination index, among which: The timeliness index is calculated by task deadline, current time and planned duration to obtain the first-level timeliness index, and the second-level timeliness index is calculated by associating process weight and resource process association coefficient; the criticality index is calculated based on the ratio of the longest process path length to the critical path length in the process network topology structure and the resource usage intensity; the substitutability index is calculated based on the resource feature similarity matrix and the substitution cost coefficient matrix; the coordination index is calculated based on the resource node centrality calculated based on the resource coordination efficiency matrix; Establish an indicator importance evaluation matrix, the matrix elements of which represent the relative importance of different indicators in the resource evaluation indicator system, calculate an initial weight vector based on the indicator importance evaluation matrix, construct a time-varying adjustment factor, the time-varying adjustment factor is related to the total construction period and the adjustment coefficient, and generate a resource scheduling dynamic weight based on the initial weight vector, the time-varying adjustment factor and the normalized score of each indicator in the resource evaluation indicator; The resource scheduling objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function. The resource utilization sub-function is calculated based on the actual utilization, the planned utilization and the resource efficiency coefficient. The project progress sub-function is calculated based on the actual progress, the planned progress and the progress influencing factor. The cost control sub-function is calculated based on the actual cost, the planned cost and the cost sensitivity coefficient. The fuzzy hierarchical analysis method is used to determine the resource scheduling dynamic weights corresponding to the resource utilization sub-function, the project progress sub-function and the cost control sub-function, and the resource scheduling objective function is obtained by linearly weighted combination of each sub-function based on the resource scheduling dynamic weights.

4. The method according to claim 1, characterized in that The resource scheduling objective function is iteratively optimized and solved by using an improved ant colony algorithm to generate an optimal resource scheduling solution including: Establish a pheromone matrix, wherein the row vectors of the pheromone matrix represent available resources, the column vectors represent task requirements, and the matrix elements are determined according to the product of the pheromone attenuation coefficient and the pheromone matrix at the previous moment and the pheromone increment, and the pheromone increment is calculated by the target value corresponding to the resource scheduling objective function of the path passed by the ants; Calculating a resource scheduling decision distance, wherein the resource scheduling decision distance is determined based on a matching degree between the available resources and the task requirements, and calculating a dynamic heuristic factor according to the resource scheduling decision distance, wherein the dynamic heuristic factor is obtained by exponentially calculating the inverse of the resource scheduling decision distance and a dynamic adjustment coefficient; The state transition probability is calculated based on the pheromone matrix and the dynamic heuristic factor, the ant colony search path is initialized, and the resource scheduling scheme of the initial ant population is randomly generated under the condition of satisfying the preset constraint condition list according to the state transition probability, and the target value corresponding to the resource scheduling objective function corresponding to each resource scheduling scheme is calculated, and the resource scheduling scheme with the smallest target value is recorded as the current optimal solution, wherein the preset constraint condition list includes resource layer constraints, task layer constraints and system layer constraints; Performing ant colony iterative search, determining a new resource scheduling decision plan for each ant based on the state transition probability; optimizing the current optimal solution by using a local search mechanism in each iteration, calculating the movement increment by defining a neighborhood structure centered on the current optimal solution, and determining whether to accept the candidate solution in the neighborhood based on the movement increment and the Metropolis criterion; The global optimal solution is updated according to the local search results. When a global optimal solution that is better than the current optimal solution appears, the pheromone matrix is ​​updated and the number of local searches is reset, and the new global optimal solution is recorded. When the maximum number of iterations is reached, the iteration is stopped and the optimal resource scheduling plan is output.

5. The method according to claim 1, characterized in that The scheduling optimization module dynamically adjusts the optimal resource scheduling scheme according to the conflict event, including: Acquire a task state vector, a resource state vector and a conflict state vector, and combine the task state vector, the resource state vector and the conflict state vector to construct a state space vector; and establish a constraint condition set based on the state space vector; Obtain the duration of all tasks, determine the maximum task duration, and use the product of the maximum task duration and a preset time coefficient as the reference window size; calculate the rolling step size according to the reference window size, the rolling step size is the product of the reference window size and the step coefficient, and the step coefficient has a value range of 0.2 to 0.5; determine the prediction interval according to the reference window size, the prediction interval is the product of the reference window size and the prediction coefficient, and the prediction coefficient is greater than or equal to 2; within the prediction interval, use a rolling optimization method based on the constraint condition set to generate multiple candidate adjustment schemes; Constructing a task dependency graph, wherein nodes of the task dependency graph represent tasks, and edges represent dependencies between tasks; using a critical path algorithm to identify a critical path from the task dependency graph, wherein the critical path represents a task chain that has the greatest impact on the total construction period; calculating the priority of each task based on the critical path, wherein the priority is positively correlated with the position of the task on the critical path and the number of its successor tasks; decomposing and reorganizing the tasks according to the task priorities, and generating a reorganization plan; Construct a resource capability matrix, wherein the rows of the resource capability matrix represent resources, the columns represent capability indicators, and the matrix elements represent specific capability values ​​of resources; construct a task requirement vector, wherein each component of the task requirement vector represents the requirement value of the capability indicator corresponding to the task; calculate the matching degree between tasks and resources using cosine similarity based on the resource capability matrix and the task requirement vector; perform resource allocation for the tasks in the reorganization plan according to the matching degree, and generate a resource allocation plan; The degree of conflict impact is calculated in real time, and the degree of conflict impact is positively correlated with the number of conflicts, the intensity of conflicts and the scope of conflict impact; the adjustment cost of the adjustment plan is calculated, and the adjustment cost is positively correlated with the number of resource reallocations, the task delay time and the system performance loss; when the degree of conflict impact is less than a first preset threshold and the adjustment cost is less than a second preset threshold, the plan with the largest target value of the resource scheduling objective function is selected from the candidate adjustment plans as the output plan, otherwise the resource allocation plan is adopted as the output plan.

6. The method according to claim 5, characterized in that The task state vector includes task completion, task urgency and task dependency, the resource state vector includes resource occupancy, resource remaining capacity and resource available time period, and the conflict state vector includes conflict type, conflict impact range and conflict duration; The constraint condition set includes timing constraints, resource constraints and space constraints, wherein the timing constraints define the order of task execution, the resource constraints define the upper and lower limits of resource capabilities and their mutually exclusive relationships, and the space constraints define the spatial distribution requirements of task execution.

7. A digital project management system based on artificial intelligence, used to implement the method according to any one of claims 1 to 6, characterized in that: include: The first unit is used to collect power grid project construction resource data, wherein the power grid project construction resource data includes material resource data, equipment resource data and construction site resource data; construct a multi-level resource constraint matrix based on the power grid project construction resource data, wherein the multi-level resource constraint matrix includes a time dimension constraint sub-matrix, a space dimension constraint sub-matrix and a resource type constraint sub-matrix; input the multi-level resource constraint matrix into a preset deep neural network model to generate a power grid project construction resource collaborative optimization configuration model; The second unit is used to calculate the real-time dispatch weight coefficients of various types of resources based on the collaborative optimization configuration model of power grid engineering construction resources, and the real-time dispatch weight coefficients are used to characterize the dispatch priorities of different resource types; construct a resource dispatch objective function according to the real-time dispatch weight coefficients, and the resource dispatch objective function includes a resource utilization sub-function, a project progress sub-function and a cost control sub-function; adopt an improved ant colony algorithm to iteratively optimize and solve the resource dispatch objective function to generate an optimal resource dispatch plan, wherein the improved ant colony algorithm introduces a dynamic heuristic factor, a local search mechanism and a global update rule on the basis of the ant colony algorithm; The third unit is used to input the optimal resource scheduling plan into a preset intelligent collaborative controller, and the intelligent collaborative controller includes a resource status monitoring module, a conflict detection module and a scheduling optimization module; the resource status monitoring module monitors the scheduling execution status of various types of resources in real time, the conflict detection module identifies conflict events in the resource scheduling process based on a preset rule base, and the scheduling optimization module dynamically adjusts the optimal resource scheduling plan according to the conflict events; the adjusted optimal resource scheduling plan is sent to each resource scheduling execution terminal to realize the intelligent collaborative scheduling of power grid project construction resources.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images