A power transmission project cost analysis method based on big data

Abstract

The application relates to a power transmission engineering cost analysis method based on big data and belongs to the technical field of power transmission engineering construction management. A radial basis function neural network model is constructed by integrating multi-source historical data to realize linkage prediction and analysis of progress, safety and cost. A K-means clustering algorithm is used to automatically determine a Gaussian center, subjective setting is avoided, construction efficiency is quantified by combining dynamic resource input intensity, a plurality of construction schemes are generated to input the prediction model, the optimal scheme is evaluated and screened, a safety balance adjustment factor model is introduced, safety hard constraints and progress cost soft targets are dynamically balanced, model parameters are closed-loop optimized based on actual data, resource dynamic optimization configuration is realized, the management informatization level is improved, the cost prediction precision is effectively enhanced, construction safety and quality are guaranteed, and the intelligent transformation of engineering management is promoted.

Description

Technical Field

[0001] This invention relates to a method for cost analysis of power transmission projects based on big data, belonging to the field of power transmission project construction management technology. Background Technology

[0002] The continuous advancement of power engineering construction provides energy sources and infrastructure for industrial development and is a strong guarantee for the effective operation of local economies. However, for power transmission engineering construction, especially in scenarios involving cross-construction and power outage coordination, it is difficult to formulate construction plans due to the large number of coordinating departments and the narrow construction window. Therefore, the analysis and modeling technology of related projects has received increasing attention. Existing methods for analyzing transmission line project costs often rely heavily on manual experience and lack effective data integration mechanisms. Multi-source data is often fragmented, lacking spatiotemporal alignment and multi-dimensional fusion, leading to one-sided predictions that fail to fully reflect the actual project situation. Traditional methods, particularly the Gaussian center setting, depend on manual experience and lack adaptability, failing to dynamically respond to real-time changes during construction. This significantly reduces the accuracy and reliability of prediction models when facing complex and variable construction environments. Existing methods often use fixed weights for resource allocation and safety constraints, unable to dynamically adjust according to actual conditions during construction. This results in the contradiction between idle resources during peak periods and resource shortages during off-peak periods, and makes it difficult to effectively control safety risks. Transmission line project costs are significantly affected by seasonal factors, such as increased costs in winter due to low temperatures. However, existing methods often neglect the prediction and analysis of seasonal fluctuations, causing cost predictions to deviate from reality. Transmission line project cost analysis involves multiple objectives such as schedule, cost, and safety; existing methods often struggle to achieve a balance among these objectives, leading to poor overall efficiency of the construction plan. Summary of the Invention

[0003] This invention provides a big data-based method for analyzing the cost of power transmission projects, addressing the problems of insufficient data integration, poor adaptability of prediction models, lack of dynamic response to resource allocation and security constraints, and insufficient ability to predict seasonal fluctuations in existing technologies.

[0004] This invention provides a method for analyzing the cost of power transmission projects based on big data, which includes the following steps: Based on historical data of power transmission projects, we can identify the influencing factors related to the linkage between progress, safety, and cost, and construct a radial basis function neural network to provide more comprehensive and accurate predictions and analyses. The construction progress, safety and cost prediction model uses the K-means clustering algorithm to determine the Gaussian center and adaptively divides the input space, avoiding the subjectivity of manually setting the center; The actual resource input at each time point during the construction process is calculated by using dynamic resource input intensity calculation, which is used to quantify the efficiency and balance of resource input. Different construction plans are generated by adjusting the allocation of resources and safety measures. The candidate plans are then input into the prediction model to evaluate their feasibility and merits, and the optimal plan is retained. Dynamic resource input intensity calculation sets an early warning line for excessive resource concentration, and constructs a safety balance adjustment factor model to achieve a balance between hard indicators of safety constraints and soft optimization of schedule and cost objectives; By introducing dynamic adjustment factors, a dynamic balance is achieved between hard indicators of safety constraints and soft optimization of schedule and cost objectives. Based on actual data, model parameters are corrected to achieve closed-loop optimization.

[0005] Preferably, in step 1), the influencing factors related to the linkage between progress, safety, and cost are divided into different elements and element sets, and a spatiotemporal alignment matrix is ​​established. Discrete events are transformed into continuous influence coefficients, and multi-source data are fused using the entropy weight method to generate a three-dimensional influence matrix as the input feature of the neural network.

[0006] Preferably, in step 2), the model parameters are continuously optimized through 10-fold cross-validation and prediction accuracy evaluation to improve the accuracy and reliability of model prediction. Time series analysis of cost data can identify and extract seasonal components, thereby improving the sensitivity to seasonal fluctuations.

[0007] Preferably, in step 3), the predicted completed work volume and schedule deviation rate, safety score and accident probability prediction value, and predicted cost at each time point are obtained from the model. Combined with the resource input of the engineering dismantling plan and the actual resource input of material consumption data collected in real time, the actual resource input at each time node during the construction process is calculated.

[0008] Preferably, in step 4), a multi-objective optimization function including schedule, cost and safety is established, and construction period, resource input and safety threshold are considered to generate and evaluate different construction schemes.

[0009] Preferably, in step 5), the contradiction between idle resources during peak construction periods and resource shortages during off-peak periods is effectively avoided through a real-time penalty mechanism, the upper limit constraint on resource input is enforced through the projection method to reduce the risk of overspending, and the target weight is dynamically allocated through a safety balance adjustment factor to approach the Pareto optimal frontier.

[0010] Preferably, in step 6), the complexity of manpower, equipment, technical solutions, and work shifts are finely grouped. By tracking the location of manpower, the status of equipment, and the execution of shifts in real time and inputting them into the three-dimensional influence matrix, combined with dynamic adjustment factors and real-time data feedback, safety weights, resource allocation, and cost control strategies can be continuously adjusted.

[0011] Preferably, in step 6), new labels for water conservancy and transportation professional qualifications and cross-work collaboration experience are added. Equipment grouping and technical solution grouping are also integrated across fields to achieve resource sharing and optimized allocation. A hybrid weighting method combining improved entropy weighting method and cross-field expert weighting is adopted to ensure the comparability of data in different categories. The safety contribution model is corrected through joint review by experts from both fields to improve the accuracy and reliability of risk assessment.

[0012] Preferably, for common indicators, the entropy weight method is used for normalization to ensure the objectivity and accuracy of the evaluation; for industry-specific indicators, cross-domain expert scoring is introduced for correction, and the weight offset coefficient is determined by the Delphi method to improve the comprehensiveness and authority of the evaluation. The dynamic balance of safety constraints is strengthened by the coupling and expansion of multi-dimensional safety balance factors.

[0013] Preferably, the process of introducing cross-category process dependency in the schedule priority allocation, and prioritizing the allocation of manpower groups and shared equipment with cross-domain operation capabilities, optimizes the group combination through a three-dimensional particle swarm algorithm of time, cost, and safety, thereby improving the overall execution efficiency and schedule management level of the project.

[0014] The beneficial effects of this invention are: This invention provides a big data-based method for power transmission engineering cost analysis. It achieves spatiotemporal alignment and multi-dimensional fusion of multi-source data by constructing a radial basis function neural network and combining it with the entropy weight method, eliminating data fragmentation and improving the input quality of the prediction model. The K-means clustering algorithm adaptively determines the Gaussian center, avoiding the subjectivity of manually setting the center, enabling the model to better adapt to power transmission projects of different scales and types, thus improving the accuracy and reliability of predictions. For time series analysis of cost data, seasonal decomposition and seasonal factor extraction significantly improve the predictive ability for seasonal fluctuations, making the cost prediction results closer to reality. By calculating the actual resource input intensity at each time node during construction, the efficiency and balance of resource input are quantified, providing a scientific basis for optimizing resource allocation. Finally, a safety balance adjustment factor model is constructed to achieve a dynamic balance between hard indicators of safety constraints and soft optimization of schedule and cost objectives. By adjusting safety weights in real time, the system avoids uncontrolled safety risks caused by excessive pursuit of progress or cost reduction. It meticulously groups manpower, equipment, technical solutions, and work shifts based on complexity, and continuously adjusts safety weights, resource allocation, and cost control strategies based on dynamic adjustment factors and real-time data feedback. This improves the overall stability and controllability of the construction process. The introduction of professional qualifications from fields such as water conservancy and transportation, as well as cross-disciplinary collaboration experience tags, into the grouping of manpower, equipment, and technical solutions enables the sharing and optimized allocation of cross-domain resources. The establishment of cross-domain mapping relationships breaks down industry barriers and improves resource utilization efficiency. A hybrid weighting method combining an improved entropy weighting method and cross-domain expert weighting ensures the comparability of data across different categories. For common indicators and industry-specific indicators, entropy weight method and expert scoring method are used respectively to improve the comprehensiveness and authority of the evaluation. The water conservancy and transportation-specific monitoring data are integrated to construct a three-dimensional dynamic matrix of space, time and domain, realizing real-time positioning of cross-domain operation areas and labeling of safety risk levels, providing a more scientific basis for project management. A multi-objective optimization function including schedule, cost and safety is established, and constraints such as construction period, resource input and safety threshold are considered. It can generate and evaluate different construction plans, retain the optimal plan, and maximize the overall benefits of power transmission projects. By tracking data such as manpower location, equipment status and shift execution in real time and inputting it into the three-dimensional influence matrix for dynamic updates, combined with model parameter correction to achieve closed-loop optimization, the technical bottleneck of the existing system's decision-making lagging behind changes in on-site working conditions is solved. After each cross-domain construction stage is completed, water conservancy and transportation experts jointly review and correct the safety contribution model to ensure that the risk assessment meets the standards of both domains, improving the accuracy and reliability of the risk assessment. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a big data-based method for analyzing the cost of power transmission projects according to the present invention. Detailed Implementation

[0016] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0017] Example 1: This invention provides a method for analyzing the cost of power transmission projects based on big data, which includes: Based on historical data of power transmission projects, factors influencing the linkage between progress, safety, and cost are identified, and a radial basis function neural network is constructed based on these factors. Specifically, this includes dividing the influencing factors related to the linkage between progress, safety, and cost in historical data of power transmission projects into different elements and element sets, establishing a spatiotemporal alignment matrix, converting discrete events into continuous influence coefficients, and fusing them into a three-dimensional influence matrix using the entropy weight method, which serves as the input feature of a radial basis function neural network. In the radial basis function neural network, the first layer is the input layer. It receives input and independently normalizes the underlying sub-indices that constitute the three-dimensional influencing factors to eliminate dimensional differences. It then performs range standardization on the three-dimensional indicators to meet the non-negativity requirement of the entropy weight method, obtaining the weight information of key influencing factors. This weight information is input into the hidden layer, and the K-means clustering algorithm is used to determine the Gaussian centers, resulting in a multidimensional Gaussian neural network function. Ten-fold cross-validation is then performed based on this multidimensional Gaussian neural network function, and the prediction accuracy is evaluated. The specific multidimensional Gaussian neural network function is as follows: in: : The output of the j-th hidden layer neuron; The center of the j-th Gaussian function; : The width parameter of the j-th Gaussian function; Euclidean distance; The output of the hidden layer is fed into the output layer, and the final output of the radial basis function neural network is calculated using multivariate linear equations, as follows: in: : No. One output variable; Number of neurons in the hidden layer; : Output variable dimensions; : No. The hidden neuron to the first The weights of each output; The output includes values ​​for schedule, safety, and cost. Build a progress, safety and cost prediction model. The specific progress, safety and cost values ​​that have been obtained are integrated with the corresponding timestamps to form a complete dataset. The dataset is divided into three parts: 70% for training, 20% for validation, and 10% for testing. The training set is used for model training, the validation set is used to adjust model parameters, and the test set is used to finally evaluate the model performance. Using schedule-related influencing factors as input features, such as construction schedule and resource allocation, and with previously obtained schedule values ​​as the target output, an RBFNN schedule prediction sub-model is constructed. The input layer receives selected input features, the hidden layer uses K-means clustering to determine Gaussian centers, resulting in a multidimensional Gaussian neural network function, and the output layer uses multivariate linear equations to calculate the final schedule prediction value. The schedule prediction sub-model is trained using a training set, and the prediction error is minimized by continuously adjusting parameters such as the Gaussian function width and the number of hidden layer neurons. Similarly, using safety-related influencing factors as input features, such as safety training and the implementation of safety protection measures, and with existing safety values ​​as the target output, an RBFNN safety prediction sub-model is constructed. The sub-model takes cost-related factors as input features, such as material price fluctuations and labor cost changes, and uses existing cost as the target output to build an RBFNN cost prediction sub-model. The input layer receives selected input features, the hidden layer uses K-means clustering to determine Gaussian centers, and the output layer uses multivariate linear equations to calculate the final safety prediction value. The cost prediction sub-model is trained using a training set to optimize model parameters and improve prediction accuracy. Time series analysis is performed on the cost data to form a one-dimensional time series. If missing data exists, interpolation is used for prediction and filling to ensure the series is continuous and uninterrupted. By plotting the time series, it is observed whether the data exhibits periodic fluctuations, and the seasonality of the cost data is preliminarily identified. The seasonal difference unit root test is used to test the stationarity of the original series and the seasonally differencing series to determine whether the seasonality is significant. Based on the multiplicative model, seasonal decomposition is performed to decompose the cost data into trend components, seasonal components, and random components. The specific multiplicative model is as follows: in: : This represents the trend item; : This refers to seasonal items; : for random error term For the original sequence Taking the natural logarithm, we get The specific steps to transform the multiplication model into an addition model are as follows: right Perform a length of The moving average yields the trend-random term. ,right Smoothing yields the long-term trend. ,calculate This yields a sequence that includes seasonal effects and random noise; Grouping by period, calculating the average value of each group, yields the initial seasonal factor. Ensure that the sum of seasonal factors is L or the product is 1, specifically: The reconstructed decomposed sequence components are then subtracted from the original time series by the seasonal factor to obtain the deseasoned sequence, specifically: Verify the decomposition effect and plot the trend items after decomposition. Seasonal items and residuals Check whether the residuals are random (without obvious periodicity) to ensure that seasonality has been effectively extracted; For each time point Based on its position in the cycle, the corresponding seasonal factor is... As an additional input feature, the original input feature is combined with the seasonal feature in the RBFNN sub-model for cost prediction, which enhances the model's ability to fit seasonal fluctuations and improves its ability to predict seasonal fluctuations in cost data. The parameters of each sub-model are further adjusted and optimized using the validation set. The mean squared error is used to evaluate the performance of the model during the validation process. The trained progress, safety and cost prediction sub-models are tested using the test set. The values ​​of each evaluation index are calculated. The test results of each sub-model are combined to evaluate the entire progress, safety and cost prediction model as a whole. The project's predicted completed work volume and schedule deviation rate at each time point are obtained from the schedule sub-model; the safety score and accident probability prediction value are obtained from the safety sub-model; the predicted cost at each time point is obtained from the cost sub-model; and the actual resource input is collected in real time based on the material consumption data, combined with the project breakdown plan resource input. The actual resource input at each time point during construction is calculated using dynamic resource input intensity to quantify the efficiency and balance of resource input. Specifically: in: : ; : No. Planned resource allocation at specific points in time; Schedule weight (dynamically adjusted based on the tightness of the schedule); The first prediction of the progress sub-model The amount of work completed at a given time point; Plan No. The amount of work completed at a given time point; The cost-safety synergy coefficient is used to measure the synergy between safety level and cost, reflecting the marginal effect of unit safety investment on cost savings. Specifically: CSCC in: Safety measures investment costs; : Expected cost of safety accidents (predicting accident probability and loss through safety sub-models); : The security score output by the security sub-model; Establish a multi-objective optimization function Schedule objective: Minimize schedule deviation ; Cost objective: Minimize total construction cost ; Safety objective: Maximize the safety score ; Constraints: Construction period constraints: ( (Maximum allowable project delays); Resource input constraints: (Resource allocation at a single point in time shall not exceed the available limit); Safety threshold constraints: (Safety score not lower than industry standard); Different construction plans are generated by adjusting the allocation of resources and safety measures. The candidate plans are then input into the prediction model to obtain the predicted values ​​of schedule, safety, and cost. Based on the constraints and objective function, the feasibility and merits of the candidate plans are evaluated, and the optimal plan is retained to obtain the optimal resource input plan, safety measure configuration, and corresponding schedule, cost, and safety prediction results.

[0018] Historical data from power transmission projects is collected during operation to extract key influencing factors linked to progress, safety, and cost. These factors are then spatiotemporally aligned to construct a spatiotemporal alignment matrix, transforming discrete events into continuous influence coefficients and eliminating data fragmentation across time and space. Influence factors are fused using entropy weighting to generate a three-dimensional influence matrix, eliminating dimensional differences. Range standardization ensures comparability of multi-source data. These data serve as input features for a neural network, constructing a radial basis function neural network. The input layer receives the three-dimensional influence matrix, independently normalizing underlying sub-indices. The hidden layer uses K-means clustering to determine Gaussian centers, adaptively partitioning the input space to avoid the subjectivity of manually set centers. A multidimensional Gaussian function is constructed, and the output layer calculates the final output through multivariate linear equations. A nonlinear Gaussian kernel maps the input features to a high-dimensional space, capturing complex relationships and partitioning the dataset for further processing. Tenfold cross-validation and prediction accuracy assessment were conducted. The progress prediction sub-model was adjusted by adjusting the width of the Gaussian function and the number of hidden layer neurons to minimize prediction error. The safety prediction sub-model was adjusted to improve accuracy. The cost prediction sub-model combined the features after seasonal decomposition to improve the ability to predict seasonal fluctuations. For example, the periodic pattern of increased construction costs in winter due to low temperatures was identified. The trend term was extracted through moving average and smoothing, the seasonal factor was calculated, and the randomness of the residuals was verified. The seasonal factor was used as an additional input feature to enhance the seasonal fitting ability of the cost sub-model and enhance the model's sensitivity to periodic fluctuations. The resource input efficiency and cost-safety synergy coefficient were dynamically calculated. A multi-objective optimization function was established to constrain the upper limit of construction delay, the upper limit of resource input, and the lower limit of safety score. Candidate schemes were generated and input into the model for evaluation. The optimal resource plan, safety configuration, and prediction results were retained.

[0019] Compared with existing technologies, constructing a radial basis function neural network can simultaneously consider the linkage relationship, providing more comprehensive and accurate predictions and analyses. Using the entropy weight method to fuse multi-source data, a three-dimensional influence matrix is ​​generated as the input feature of the neural network. Through ten-fold cross-validation and prediction accuracy evaluation, model parameters are continuously optimized, significantly improving the accuracy and reliability of predictions. The K-means clustering algorithm is used to determine the Gaussian center, adaptively dividing the input space and avoiding the subjectivity of manually setting the center, enabling the model to better adapt to power transmission projects of different scales and types. Time series analysis of cost data can identify and extract seasonal components, using seasonal factors as additional input features, enhancing the sensitivity of the cost prediction sub-model to seasonal fluctuations and improving prediction accuracy. Dynamic calculation of resource input intensity and cost-safety synergy coefficients quantifies the efficiency and balance of resource input, as well as the marginal effect between safety input and cost savings, providing a scientific basis for optimizing resource allocation. A multi-objective optimization function including schedule, cost, and safety is established, considering constraints such as construction period, resource input, and safety thresholds. This allows for the generation and evaluation of different construction schemes, retaining the optimal scheme, thereby maximizing the overall benefits of power transmission projects.

[0020] Example 2: In Example 1, the efficiency and balance of resource input and the cost-safety synergy coefficient can be quantified by dynamically calculating the resource input intensity and the marginal effect between safety input and cost savings. Different construction schemes are generated and evaluated, and the optimal scheme is retained, thereby maximizing the overall benefits of the power transmission project. This example is a further improvement on the above example.

[0021] This embodiment also includes: during the schedule and cost optimization process, a balance is achieved between the hard indicators of safety constraints and the soft optimization of schedule and cost targets through dynamic adjustment factors, so as to avoid the loss of control of safety risks due to excessive pursuit of schedule or cost compression. The weights of schedule and cost targets are automatically adjusted according to the real-time safety status (safety score, accident probability, safety investment benefits) to form a safety-first and dynamically adapted optimization scheme. Specifically, a safety balance adjustment factor model is constructed, and the adjustment factor is calculated. in: The safety buffer zone (e.g., 10 points) represents the transition range from "qualified" to "good" in terms of safety status. , , High, medium, and low security weights, such as (Safety has a lower weight, allowing schedule cost to take precedence) (Balancing weights) (With extremely high safety weight, safety is forced to take priority) Embed the adjustment factor into the multi-objective function. Specifically: min(α·1 - ωsafe·Duration deviation + β·1 - ωsafe·Total cost + γ·ωsafe·Sscore) Where: α, β, γ: Original objective weight coefficients (the sum is 1); ωsafe: Dynamic adjustment factor, which adjusts the proportion of the safety objective in the total objective in real time; Optimize the generation of candidate solutions. Each particle represents a construction plan through particle coding. Constrained by the safety threshold: If Sscore < Smin, directly assign an infinite penalty value. Constrained by resources: Use the projection method to forcibly pull the Ractual exceeding the upper limit back to Rmax, and search for the optimal resource-cost combination that meets the safety threshold within the feasible solution space to avoid the defect of traditional algorithms ignoring safety constraints; Establish a fitness function. Specifically: Where: : Safety objective weight coefficient; : Schedule objective weight coefficient; : Cost objective weight coefficient; : Safety score output by the safety sub-model; : Actual duration; : Planned duration; : Sum of predicted costs at each time point; : Total project cost budget; Combined with the dynamic adjustment of the safety balance adjustment factor weight. Specifically When Sscore < Smin: ωs = 0.7, ωp = 0.2, ωc = 0.1 (Safety first); When Sscore ≥ Smin + 10: ωs = 0.3, ωp = 0.4, ωc = 0.3 (Schedule and cost first); According to Generate the resource input table for each process. If , Mark it as the resource emergency deployment plan Combined with And Generate a rush or slowdown plan, convert data such as resource input and safety measures of the candidate solutions into a three-dimensional impact matrix format, input the resource input and construction period plan into the schedule sub-model, and output and the construction period deviation, input the safety measure input into the safety sub-model, and output 、 , input the resource input, safety cost, and seasonal factor into the cost sub-model, and output Perform dynamic coefficient calculation and constraint verification. Specific dynamic resource input intensity calculation: If > 1.5 is marked as over-concentration of resources, and a penalty term is added; Cost-safety coordination coefficient: CSCC = Check and CSCC , if CSCC < 1: The safety input is uneconomical, directly eliminate this solution, and select the overall optimal solution of cost and construction period among the remaining solutions.

[0022] Define the safety score threshold, safety buffer interval, and weight coefficient during use, determine the upper limit of resource input, planned construction period, and total cost budget, achieve a smooth transition from "qualified" to "good", avoid sudden changes in weights. The schedule sub-model inputs the resource input and construction period plan, and outputs the predicted safety score and construction period deviation. The safety sub-model inputs the safety measure input, and outputs the real-time safety score and accident probability. The cost sub-model inputs the resource input, safety cost, and seasonal factor, and outputs the predicted cost sequence. Dynamically adjust the safety weight according to the safety score and cost-safety coordination coefficient, and dynamically adjust the weight ratio according to the real-time safety status to balance the safety, schedule, and cost objectives. Each particle represents a construction plan, including parameters such as resource input, construction period, and safety measures, to achieve multi-objective joint optimization. Through the dynamic adaptation of the objective weight and the plan parameters, approach the Pareto optimal frontier. If Sscore < Smin, directly assign an infinite penalty value and eliminate this solution. Use the projection method to forcibly pull the Ractual exceeding the upper limit back to Rmax to improve the algorithm convergence speed and ensure that the resource allocation conforms to the engineering reality. Perform dynamic resource input intensity and cost-safety coordination coefficient verification. If or CSCC < 1, directly eliminate it, leave the remaining solutions, and select the overall optimal solution of cost and construction period through the fitness function value among the solutions that meet the safety and coordination constraints.

[0023] Compared with existing designs, this approach sets an early warning line for excessive resource concentration through dynamic resource input intensity calculation, effectively avoids the contradiction between idle resources during peak construction periods and resource shortages during off-peak periods through a real-time penalty mechanism, enforces upper limit constraints on resource input through projection method to reduce the risk of cost overruns, and makes the actual resource input of the project closer to the theoretical value. It automatically eliminates schemes when the cost-safety synergy coefficient (CSCC) is less than 1, ensuring that each unit of safety input generates at least equivalent economic benefits. It integrates safety, schedule, and cost data through a three-dimensional influence matrix, shifting the decision-making basis from qualitative to quantitative. When generating candidate schemes through particle coding, safety threshold constraints improve the quality of the feasible solution space. It adopts a multi-objective dynamic adaptation function, simultaneously considering three dimensions: schedule deviation, total cost, and safety score. It achieves dynamic allocation of objective weights through a safety balance adjustment factor, approaching the Pareto optimal frontier. A safety balance adjustment factor model is constructed, dynamically adjusting the weights of schedule and cost objectives based on real-time safety status, achieving a balance between hard indicators of safety constraints and soft optimization of schedule and cost objectives.

[0024] Example 3: In Example 2, a balance is achieved between hard indicators of safety constraints and soft optimization of schedule and cost targets through dynamic adjustment factors, so as to avoid the loss of control of safety risks due to excessive pursuit of schedule or cost compression. This example is a further improvement on the above example. In this embodiment, establishing resource groups includes: Human Resources Grouping: Classified by job type, skill level, job role, and team size, with additional safety-related attributes such as safety training duration and historical accident rate; Equipment grouping: Classified by type, usage status, energy efficiency, and safety protection configuration; Complexity of technical solutions: classified according to construction technology, technical difficulty, risk level, and process complexity; Work shifts: Classified by time arrangement, continuous work duration, shift interval, and work intensity; Each group dimension is assigned a unique label, and the entropy weight method is used to normalize the data of each group to eliminate the difference in dimensions. Safety balance factors are incorporated into group attributes, defining safety contribution indicators for each group. In the human resources group, increased safety training time leads to higher safety scores, lower historical accident rates, and an increased safety coefficient. In the equipment group, improved safety protection configuration leads to higher safety scores, lower equipment availability, and an increased probability of accidents. In the technical solutions group, increased high-risk processes, increased safety officer cooperation, increased safety investment costs, and improved levels of safety technical measures lead to a decreased probability of accidents. In the shift group, increased continuous operation time leads to an increased probability of accidents, decreased shift intervals, and a lower safety score. In the safety balance adjustment factor model, a group safety coefficient is introduced as an input variable to dynamically adjust the safety weight and group priority. Prioritize the allocation of manpower groups, equipment groups, and day shift groups for key processes to ensure efficient construction. Each construction plan particle includes group combinations, input quantity of each group, and time allocation. Calculate the cost redundancy of each group. For example, if the actual input cost of a certain group exceeds the plan by 15% and does not bring about a schedule advancement of ≥5%, then add a penalty item of 0.2 × overspending ratio. Real-time tracking of manpower location, equipment status, and shift execution; input of a three-dimensional influence matrix; dynamic updating of group safety contribution, schedule efficiency coefficient, and cost consumption data; upon completion of each construction phase, correction of the group's safety contribution, schedule efficiency coefficient, and cost consumption model based on actual data; updating of RBFNN sub-model parameters through ten-fold cross-validation to form a closed-loop optimization.

[0025] When using the system, manpower is grouped according to job type, skill level, work role, and team size, with additional safety-related indicators such as safety training duration and historical accident rate. Equipment is grouped according to type, usage status, energy efficiency, and safety protection configuration. Technical solution complexity is grouped according to construction technology, technical difficulty, risk level, and process complexity. Work shifts are grouped according to time arrangement, continuous working hours, shift intervals, and work intensity. Entropy weighting is used to normalize the data for each group, eliminating dimensional differences, and a unique label is assigned to each group dimension. Based on the safety contribution index of each group's data, the group safety coefficient is input into the safety balance adjustment factor model to dynamically adjust safety weights and group priorities. Manpower groups, equipment groups, and daily shift groups are prioritized for key progress processes to ensure efficient construction. Each construction party... The case particles include: group combination, input quantity, and time allocation. It tracks the location of manpower, equipment status, and shift execution in real time, inputs them into the three-dimensional influence matrix, and dynamically updates the group safety contribution, schedule efficiency coefficient, and cost consumption data. This overcomes the shortcomings of traditional static weights that cannot reflect real-time changes in work status. After each construction phase is completed, the group safety contribution, schedule efficiency coefficient, and cost consumption model are corrected based on actual data. The radial basis function neural network is updated through ten-fold cross-validation, and the model parameters are updated to achieve closed-loop optimization, forming a closed-loop optimization of "data acquisition - model update - strategy adjustment". This solves the technical bottleneck of existing system decision-making lagging behind changes in on-site working conditions. Combined with dynamic adjustment factors and real-time data feedback, it continuously adjusts safety weights, resource allocation, and cost control strategies to ensure a balance between safety constraints and schedule and cost objectives.

[0026] Compared with existing designs, by introducing dynamic adjustment factors, a dynamic balance is achieved between hard safety constraints and soft optimization of schedule and cost targets. This avoids the loss of control over safety risks caused by excessive pursuit of schedule or cost reduction, ensuring the overall stability and controllability of the construction process. The complexity of manpower, equipment, and technical solutions, as well as work shifts, are finely grouped, and multiple dimensions such as safety training time, historical accident rate, safety protection configuration, construction technology, technical difficulty, risk level, process complexity, time arrangement, continuous working time, shift interval, and work intensity are considered. This helps to more accurately assess and manage resources, improve construction efficiency and safety, and dynamically update group safety contribution, schedule efficiency coefficient, and cost consumption data by tracking manpower location, equipment status, and shift execution in real time and inputting them into a three-dimensional influence matrix. After each construction phase is completed, the model parameters are adjusted based on actual data to achieve closed-loop optimization. This overcomes the shortcomings of traditional static weights that cannot reflect real-time changes in operational status, ensuring that decisions are synchronized with on-site conditions. Combined with dynamic adjustment factors and real-time data feedback, safety weights, resource allocation, and cost control strategies can be continuously adjusted, ensuring that various changes during construction can be addressed in a timely and effective manner, thus improving the stability and controllability of the entire construction process.

[0027] Example 4: In Example 3, dynamic resources are macroscopically grouped, and manpower, equipment, complexity of technical solutions, and work shifts are macroscopically grouped to optimize the safety balance and ultimately achieve the goal of accelerating progress and reducing costs. This example is a further improvement on the above examples. This embodiment also includes adding industry-specific attributes and establishing cross-domain mapping relationships on the basis of existing grouping dimensions. In the human resources group, new tags such as water conservancy and transportation professional qualifications and cross-job collaboration experience are added. The efficiency data of cross-category processes in historical projects are used to quantify cross-domain operation proficiency. In the equipment group, underwater operation equipment for water conservancy projects and road paving equipment for transportation projects are integrated and classified according to general functions and specific functions. In the technical solution group, a cross-domain technology coupling matrix is ​​established to mark the risk transmission path of cross-processes (such as the impact weight of river precipitation on bridge foundation settlement) and define safety collaboration indicators for cross-technology types, such as the mutual verification standards for water conservancy seepage prevention and transportation subgrade compaction. A hybrid weighting method combining an improved entropy weighting method and cross-domain expert weighting is adopted. Common indicators between water conservancy and transportation (such as safety training time and equipment integrity rate) are normalized using the entropy weighting method, while cross-domain expert scoring is introduced to correct industry-specific indicators (such as flood control level of water conservancy and tunnel ventilation efficiency of transportation). The weight offset coefficient is determined by the Delphi method to ensure the comparability of data of different categories. Multidimensional safety balance factors should be coupled and extended to strengthen the dynamic balance of safety constraints. Safety risks of cross-category work teams need to be superimposed (e.g., when a traffic team participates in water conservancy cofferdam construction, the risk amplification factor is caused by the lack of hydrological monitoring experience). Safety training time should include the time for learning cross-domain safety procedures. When using equipment in cross-scenario situations, the completeness of safety protection configuration should be dynamically adjusted according to the working environment. A dual-domain risk superposition model should be established for cross-processes. Safety investment costs should cover cross-domain monitoring equipment. In the schedule priority allocation, cross-category process dependency is introduced. For cross-processes on the critical path, priority is given to allocating manpower groups and shared equipment with cross-domain operation capabilities. The group combination is optimized by a three-dimensional particle swarm algorithm that considers time, cost, and safety to ensure that the time deviation of cross-category process overlap is not observed. By integrating water conservancy and transportation-specific monitoring data, a three-dimensional dynamic matrix of space, time, and domain is constructed. By merging the three-dimensional model of water conservancy hubs with the geographic information of transportation routes, cross-domain operation areas are located in real time, the safety risk level of each area is marked, and the hydrological cycle of water conservancy projects and the peak traffic period of transportation projects are synchronized. Cross-domain time window constraints are added to the shift schedule (such as prohibiting cross-river bridge hoisting operations during the high water level period of the flood season). Heterogeneous data such as water level monitoring of water conservancy and video surveillance of transportation are accessed in real time. The data platform performs format cleaning and feature fusion, and the grouping parameters are dynamically updated by inputting the safety balance factor model. The RBFNN sub-model is upgraded to a multi-input multi-output neural network. The input layer adds cross-domain feature variables, such as sediment content in water conservancy and load level in transportation. The output layer simultaneously optimizes the three objectives of safety, schedule, and cost. The model is pre-trained using historical data from water conservancy and transportation, and then fine-tuned using cross-project data. After each cross-domain construction phase is completed, the safety contribution model is revised through joint review by experts from both domains to ensure that the risk assessment meets both the scour resistance standards of water conservancy projects and the subgrade bearing requirements of transportation projects.

[0028] When using this system, the human resource grouping is expanded to include industry-specific tags for water conservancy professional qualifications, transportation professional qualifications, and cross-job collaboration experience. Efficiency data from cross-category processes in historical projects is used to quantify cross-domain operational proficiency. Equipment grouping integrates underwater operation equipment from water conservancy projects with road paving equipment from transportation projects. Technical solution grouping establishes a cross-domain technical coupling matrix to label the risk transmission path of cross-processes. Work shift grouping adds cross-domain time window constraints. Common cross-domain indicators such as safety training duration and equipment availability rate are normalized using the entropy weight method. For indicators such as flood control levels in water conservancy and tunnel ventilation efficiency in transportation, cross-domain expert scoring is introduced. The Delphi method is used to determine the weight offset coefficient, and a hybrid weighting method is used to eliminate domain differences, ensuring the comparability of data between water conservancy and transportation. A risk amplification factor for cross-category work teams is defined, and equipment safety configuration is dynamically adjusted according to the cross-domain operating environment, establishing a dual-domain risk superposition model. This approach defines composite risk levels for cross-functional processes, introduces cross-category process dependencies, identifies cross-functional processes on the critical path, prioritizes the allocation of manpower groups and shared equipment with cross-domain capabilities, and optimizes group combinations through a three-dimensional particle swarm optimization algorithm based on time, cost, and safety to ensure minimal deviations in the time and schedule of cross-category process overlaps. It integrates water conservancy and transportation-specific monitoring data to automatically avoid conflict periods, constructs a spatial-temporal-domain three-dimensional dynamic matrix, and merges three-dimensional models of water conservancy hubs with geographical information of transportation routes to mark the safety risk levels of cross-domain work areas in real time. A multi-input multi-output neural network is used, with the input layer expanded to add cross-domain feature variables and the output layer optimized to simultaneously optimize the three objectives of safety, schedule, and cost. The model is pre-trained using single-domain historical data and then fine-tuned using cross-project data. After each cross-domain construction phase is completed, water conservancy and transportation experts jointly revise the safety contribution model to ensure compliance with dual-domain standards.

[0029] Compared to existing technologies, the addition of tags such as water conservancy and transportation professional qualifications and cross-disciplinary collaboration experience improves the accuracy and efficiency of human resource grouping. Simultaneously, equipment grouping and technical solution grouping have also undergone corresponding cross-domain integration, enabling more efficient resource utilization, helping to break down industry barriers, and achieving resource sharing and optimized allocation. A hybrid weighting method combining an improved entropy weighting method and cross-domain expert weighting ensures the comparability of data across different categories. For common indicators, entropy weighting is used for normalization, guaranteeing the objectivity and accuracy of the evaluation; for industry-specific indicators, cross-domain expert scoring correction is introduced, and the weight offset coefficient is determined using the Delphi method, fully considering industry characteristics and expert opinions, improving the comprehensiveness and authority of the evaluation. The dynamic balance of safety constraints is strengthened through the coupling and expansion of multi-dimensional safety balance factors. Considering factors such as the overlapping safety risks of cross-category work teams, the cross-domain requirements for safety training time, and the dynamic adjustment of equipment safety protection configurations, it is possible to more effectively ensure the safety and stability of cross-domain operations. Cross-category process dependency is introduced into the schedule priority allocation, and priority is given to manpower groups and shared equipment with cross-domain operation capabilities. The group combination is optimized through a three-dimensional particle swarm optimization algorithm based on time, cost, and safety, which helps to minimize the time deviation of cross-category process overlap, improve the overall project execution efficiency and schedule management level, and integrate water conservancy and transportation-specific monitoring data to construct a three-dimensional dynamic matrix of space, time, and domain, realizing real-time positioning of cross-domain operation areas and labeling of safety risk levels. Meanwhile, by synchronizing information such as the hydrological cycle of water conservancy projects and the peak traffic periods of transportation projects, a more scientific basis is provided for shift scheduling, which helps to improve the informatization and intelligence level of project management. The RBFNN sub-model is upgraded to a multi-input multi-output neural network, and cross-domain feature variables are added as input layer variables. At the same time, the three objectives of safety, schedule and cost are optimized as output layer variables. It can make full use of historical data and cross-project data for model training and fine-tuning. The safety contribution model is corrected through joint review by experts from both domains, ensuring that the risk assessment meets both the scour resistance standards of water conservancy projects and the roadbed bearing requirements of transportation projects. This improves the accuracy and reliability of risk assessment and provides stronger support for project decision-making.

[0030] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for cost analysis of power transmission projects based on big data, characterized in that, Includes the following steps: Based on historical data of power transmission projects, we can identify the influencing factors related to the linkage between progress, safety, and cost, and construct a radial basis function neural network to provide more comprehensive and accurate predictions and analyses. The construction progress, safety and cost prediction model uses the K-means clustering algorithm to determine the Gaussian center and adaptively divides the input space, avoiding the subjectivity of manually setting the center; The actual resource input at each time point during the construction process is calculated by using dynamic resource input intensity calculation, which is used to quantify the efficiency and balance of resource input. Different construction plans are generated by adjusting the allocation of resources and safety measures. The candidate plans are then input into the prediction model to evaluate their feasibility and merits, and the optimal plan is retained. Dynamic resource input intensity calculation sets an early warning line for excessive resource concentration, and constructs a safety balance adjustment factor model to achieve a balance between hard indicators of safety constraints and soft optimization of schedule and cost objectives; By introducing dynamic adjustment factors, a dynamic balance is achieved between hard indicators of safety constraints and soft optimization of schedule and cost objectives. Based on actual data, model parameters are corrected to achieve closed-loop optimization.

2. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 1), the influencing factors related to the linkage between progress, safety, and cost are divided into different elements and element sets, and a spatiotemporal alignment matrix is ​​established. Discrete events are transformed into continuous influence coefficients, and multi-source data are fused using the entropy weight method to generate a three-dimensional influence matrix as the input feature of the neural network.

3. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 2), the model parameters are continuously optimized through 10-fold cross-validation and prediction accuracy evaluation to improve the accuracy and reliability of model prediction. Time series analysis of cost data can identify and extract seasonal components, thereby improving the sensitivity to seasonal fluctuations.

4. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 3), the predicted completed work volume and schedule deviation rate, safety score and accident probability prediction value, and predicted cost at each time point are obtained from the model. Combined with the resource input of the engineering dismantling plan and the real-time collection of material consumption data, the actual resource input at each time point during the construction process is calculated.

5. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 4), a multi-objective optimization function including schedule, cost and safety is established, and construction period, resource input and safety threshold are considered to generate and evaluate different construction schemes.

6. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 5), the contradiction between idle resources during peak construction periods and resource shortages during off-peak periods is effectively avoided through a real-time penalty mechanism. The upper limit of resource input is constrained by the projection method to reduce the risk of overspending. The target weight is dynamically allocated through the safety balance adjustment factor to approach the Pareto optimal frontier.

7. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 6), the complexity of manpower, equipment, technical solutions, and work shifts are finely grouped. By tracking the location of manpower, the status of equipment, and the execution of shifts in real time and inputting them into the three-dimensional influence matrix, combined with dynamic adjustment factors and real-time data feedback, safety weights, resource allocation, and cost control strategies can be continuously adjusted.

8. The method for analyzing the cost of power transmission projects based on big data according to claim 1, characterized in that: In step 6), new labels for water conservancy and transportation professional qualifications and cross-work collaboration experience were added. Equipment grouping and technical solution grouping were also integrated across fields to achieve resource sharing and optimized allocation. A hybrid weighting method combining improved entropy weighting method and cross-field expert weighting was adopted to ensure the comparability of data in different categories. The safety contribution model was revised through joint review by experts from both fields to improve the accuracy and reliability of risk assessment.

9. The method for analyzing the cost of power transmission projects based on big data according to claim 8, characterized in that: For common indicators, the entropy weight method is used for normalization to ensure the objectivity and accuracy of the evaluation. For industry-specific indicators, cross-domain expert scoring is introduced for correction, and the weight offset coefficient is determined by the Delphi method to improve the comprehensiveness and authority of the evaluation. The dynamic balance of safety constraints is strengthened by the coupling and expansion of multi-dimensional safety balance factors.

10. The method for analyzing the cost of power transmission projects based on big data according to claim 8, characterized in that: The proposed method introduces cross-category process dependency in schedule priority allocation and prioritizes the allocation of manpower groups and shared equipment with cross-domain operation capabilities. It also optimizes group combinations through a three-dimensional particle swarm algorithm that considers time, cost, and safety, thereby improving the overall execution efficiency and schedule management level of the project.

Images