Real-time management and control method, device and equipment for construction period and schedule of environmental engineering project and medium

Abstract

The invention relates to an environmental engineering project construction period schedule real-time management and control method, device and equipment and a medium. The method comprises the steps of collecting multi-source heterogeneous data of an environmental engineering project site and performing space-time alignment processing to generate a project real-time state data cube; based on the project real-time state data cube, predicting an environmental performance indicator track under a preset specific construction strategy through an environmental process dynamic prediction model; solving an optimization problem through a model prediction controller, and generating a control action adjustment sequence of resource configuration on the premise of meeting the environment quality hard constraint; and in combination with a preset action-instruction mapping table, generating a construction instruction and performing scheme rehearsal, and generating a construction process scheduling and resource scheduling updating scheme. By adopting the method, environment monitoring and construction period scheduling can be coupled in real time, the construction strategy is dynamically adjusted, and the problems of response lag, resource waste, environmental risk and the like in traditional static management are effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of engineering scheduling and management technology, and in particular relates to methods, devices, equipment and media for real-time management of the schedule of environmental engineering projects. Background Technology

[0002] With the rapid development of the environmental engineering field, especially the large-scale implementation of ecological restoration and pollution control projects, the complexity and dynamism of project management have become increasingly prominent. Currently, environmental engineering project management often draws on the management paradigms of traditional construction engineering, such as the Critical Path Method (CPM) and project management software (such as Microsoft Project). These technologies focus on linear programming of task schedule, resource allocation, and cost control, forming a static management approach with schedule and cost as the core objectives.

[0003] In traditional technologies, the management process for environmental engineering projects typically consists of three phases: The initial phase involves developing a detailed construction plan based on historical experience and environmental assessments, including work sequence scheduling, resource allocation, and quality standards; the mid-phase involves the construction team executing the plan and collecting environmental data (such as water quality and soil indicators) through regular inspections or fixed-point monitoring, but this data is primarily used for adjusting local process parameters (such as aeration rate and chemical dosage) or for post-project recording; the final phase involves environmental quality acceptance at project completion, with the final monitoring results determining project success. The entire process relies on the rigid execution of the pre-set plan, with environmental monitoring and progress management operating independently.

[0004] However, this current management approach has inherent flaws. Environmental engineering projects are essentially dynamic systems, and the environmental impact of construction activities changes in real time (such as fluctuations in pollutant concentrations and interference from climate factors). However, traditional methods treat environmental quality targets as static acceptance criteria at the project's end, rather than as dynamic control variables integrated into the construction process. This leads to a lag in management response when unexpected deviations occur in the environmental remediation process (such as delayed degradation rates), making it impossible to proactively adjust the project schedule, thereby causing overall project delays, resource waste, or environmental risks. Summary of the Invention

[0005] Therefore, it is necessary to provide an intelligent management and control method that can integrate environmental monitoring data and project scheduling in real time to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for real-time management of the scheduling of environmental engineering projects, including:

[0007] Collect multi-source heterogeneous data from environmental engineering project sites, and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data, and project plan data;

[0008] Based on the real-time status data cube of the project, the environmental performance index trajectory under the preset specific construction strategy is predicted by the environmental process dynamic prediction model, and the environmental performance index trajectory under the preset specific construction strategy is obtained; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on the attention mechanism.

[0009] Based on the predicted trajectory of environmental performance indicators and project schedule targets, with environmental quality compliance as a hard constraint, the objective function is solved in a rolling manner through a model predictive controller to generate a sequence of control actions for resource allocation; the objective function aims to minimize schedule deviation and cost deviation.

[0010] Based on the control action adjustment sequence and the preset action-instruction mapping table, construction instructions are generated, and a scheme pre-rehearsal is performed based on the construction instructions to generate a construction procedure schedule and resource scheduling update scheme.

[0011] Furthermore, the method also includes:

[0012] Real-time monitoring of environmental performance data and construction progress data after the resource scheduling update plan is implemented;

[0013] The project real-time status data cube is updated based on environmental performance data and construction progress data to obtain an updated project real-time status data cube. Based on the updated project real-time status data cube, the steps of generating environmental performance indicator trajectories and generating construction procedure schedules and resource scheduling update schemes are re-executed to obtain updated construction procedure schedules and updated resource scheduling update schemes.

[0014] Furthermore, based on the project's real-time status data cube, the environmental performance indicator trajectory under a preset specific construction strategy is predicted using an environmental process dynamic prediction model, thus obtaining the environmental performance indicator trajectory under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on an attention mechanism, including:

[0015] Extract historical data sequences for a preset past time period from the project's real-time status data cube; wherein, the input vector for each time step in the historical data sequence consists of environmental performance indicators, process control variables, and construction status variables;

[0016] Historical data sequences are encoded using a bidirectional long short-term memory neural network to generate context vector representations;

[0017] The attention mechanism is used to calculate the contribution weight of data at different time steps in the historical data sequence to the predicted value, and a weighted context vector is generated based on the contribution weight and context vector representation.

[0018] The preset specific construction strategy is input into a long short-term memory neural network, and combined with context vectors and weighted context vectors, multi-step prediction is performed to obtain the trajectory of environmental performance indicators under the preset specific construction strategy.

[0019] Furthermore, based on the predicted trajectory of environmental performance indicators and the project schedule target, and with environmental quality compliance as a hard constraint, a model predictive controller is used to solve the objective function in a rolling manner, generating a sequence of control actions for resource allocation adjustment. The objective function aims to minimize schedule and cost deviations, and includes:

[0020] Based on the predicted trajectory of environmental performance indicators and the project schedule target, an optimization function is constructed for each control cycle to minimize environmental performance tracking error, penalty for changes in control actions, and constraint slack. The optimization variables in this function are the adjustments made to future construction control actions.

[0021]

[0022] in, For time t, predict the environmental performance indicators for the next k steps. For the expected trajectory of environmental performance, Let Q be the weighted L2 norm, and let Q be the state weight matrix. Let t be the adjustment amount of the control action k steps in the future. To constrain the slack amount, To relax the penalty coefficient, For environmental performance tracking errors, Punishment for controlling changes in actions;

[0023] Set the output of the dynamic prediction model for environmental processes as the dynamic constraint of the optimization function;

[0024] Set the project schedule constraints, control variable constraints, and environmental quality constraints as hard constraints for the optimization problem;

[0025] Based on dynamic constraints and hard constraints, the optimization function is solved using a differential evolution algorithm to obtain the control action adjustment sequence for resource allocation.

[0026] Furthermore, based on the control action adjustment sequence and the preset action-instruction mapping table, construction instructions are generated, and based on the construction instructions, a scheme pre-simulation is performed to generate a construction procedure schedule and resource scheduling update scheme, including:

[0027] Based on the preset action-instruction mapping table, the adjustment amount in the control action adjustment sequence is converted into natural language instructions;

[0028] Natural language instructions are bound to a resource database to generate detailed scheduling commands; these detailed scheduling commands include the resource source task, target task, transfer time, and workload.

[0029] Based on detailed scheduling commands, the future construction process is simulated to generate a construction sequence schedule and resource scheduling plan.

[0030] Secondly, this application also provides a real-time scheduling control device for environmental engineering projects, including:

[0031] The data acquisition module is used to collect multi-source heterogeneous data from environmental engineering project sites and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data, and project plan data;

[0032] The trajectory prediction module is used to predict the trajectory of environmental performance indicators under a preset specific construction strategy based on the real-time status data cube of the project and through the environmental process dynamic prediction model. The environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on the attention mechanism.

[0033] The sequence generation module is used to predict the trajectory and project schedule target based on environmental performance indicators, with environmental quality compliance as a hard constraint. It generates a sequence of control actions for resource allocation by rolling the solution of the objective function through a model predictive controller. The objective function aims to minimize the schedule deviation and cost deviation.

[0034] The project schedule determination module is used to generate construction instructions based on the control action adjustment sequence and the preset action-instruction mapping table, and to perform scheme pre-simulation based on the construction instructions to generate construction procedure schedule and resource scheduling update scheme.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement any of the personalized learning path generation methods based on learning behavior profiles described in the embodiments of this application.

[0036] Fourthly, this application also provides a computer-readable storage medium storing at least one piece of program code, which is loaded and executed by a processor to implement the personalized learning path generation method based on learning behavior profiles as described in any of the embodiments of this application.

[0037] The aforementioned methods, devices, equipment, and media for real-time management of environmental engineering project schedules successfully construct a real-time project status data cube as the project data foundation by collecting and integrating multi-source heterogeneous data from the environmental engineering project site. A bidirectional LSTM model based on an attention mechanism is used to achieve accurate prediction of environmental performance. An optimization problem is solved through a model predictive controller, generating a sequence of control actions to adjust resource allocation while meeting hard environmental quality constraints. Through action-instruction mapping and digital twin pre-simulation, executable construction procedure scheduling and resource allocation update schemes are generated. This approach can couple environmental monitoring and project scheduling in real time, dynamically adjust construction strategies, and effectively solve problems such as response lag, resource waste, and environmental risks in traditional static management, thereby improving the intelligence and precision of environmental engineering project management. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating a real-time scheduling control method for environmental engineering projects in one embodiment.

[0040] Figure 2 This is a flowchart illustrating the steps of generating construction instructions based on a control action adjustment sequence and a preset action-instruction mapping table, and generating a construction procedure schedule and resource scheduling update plan based on the construction instructions for scheme pre-simulation.

[0041] Figure 3 This is a schematic diagram of the structure of a real-time management and control device for environmental engineering project scheduling in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] In one embodiment, a method for real-time management and control of the schedule of environmental engineering projects is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. Figure 1 As shown, in this embodiment, the method includes the following steps:

[0044] Step S101: Collect multi-source heterogeneous data from the environmental engineering project site, and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; wherein, the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data and project plan data.

[0045] Spatiotemporal alignment is a crucial step in integrating multi-source heterogeneous data in environmental engineering projects. Its core function is to eliminate inconsistencies in the temporal and spatial dimensions of data from different sources, providing a unified reference benchmark for all types of data. Ecological monitoring data originates from sensor networks deployed at the project site, such as water quality sensors, air quality sensors, and soil moisture sensors, reflecting the real-time status of the ecological environment within the project area. Construction machinery status data comes from the Internet of Things (IoT) modules integrated into the machinery, including information such as machine runtime, load rate, fuel consumption, and fault codes, characterizing the working status of the equipment. Project planning data comes from the project management system, covering schedule plans, work sequence arrangements, resource allocation plans, and project milestones, serving as the benchmark for construction execution. The IoT modules integrated into the construction machinery address the pain points of traditional construction machinery—difficulty in monitoring status, improving efficiency, and predicting maintenance—through a "sensing-transmission-interaction-application" link. The project management system (ProjectManagement)... A Project Management System (PMS) is a tool or platform used to plan, execute, monitor, and close a project throughout its entire lifecycle. Its core objective is to help teams collaborate efficiently, control project risks, and ensure that objectives (such as time, cost, and quality) are achieved. It is particularly suitable for complex project scenarios involving multi-task and multi-role collaboration. A Sensor Network is a distributed intelligent system composed of a large number of miniature sensor nodes, data transmission links (such as wireless / wired), and a data processing center. It is used to collaboratively sense, collect, transmit, and process various types of information in the physical world (such as water quality and air quality). A water quality sensor is a specialized detection device used to monitor key parameters of water bodies in real time and assess water quality. It can convert the physical / chemical / biological characteristics of water quality into readable electrical signals. An air quality sensor is an electronic device used to monitor and collect various pollutants and environmental parameters in the air in real time. It can convert chemical / physical signals in the air (such as the concentration of harmful gases and particulate matter content) into readable electrical signals. A soil moisture sensor is a core device used to monitor soil moisture content and related environmental parameters in real time.

[0046] For example, multi-source heterogeneous data is collected from environmental engineering project sites. Data cleaning is performed on this collected data. For ecological monitoring data, extreme values ​​and missing values ​​caused by sensor malfunctions are removed; interpolation can be used to supplement missing data and filter noisy data. For construction machinery status data, invalid data caused by machinery offline is removed, and data misalignment during transmission is corrected. For project plan data, logical consistency checks are performed to ensure that the sequence of procedures is not conflicting and resource allocation is reasonable. Subsequently, spatiotemporal alignment processing is performed to unify the timestamps (e.g., setting a sampling frequency of minutes or hours) and spatial coordinates of all data, enabling different types of data to be correlated within the same spatiotemporal dimension. The processed multi-source data is integrated into a real-time project status data cube. This cube has three dimensions: time, space, and indicators. The time dimension covers historical and real-time data sequences, the spatial dimension corresponds to the construction area and monitoring points, and the indicator dimension includes ecological, machinery, and plan-related indicators. Data cleaning refers to addressing problems in raw data to make it more standardized, accurate, and usable; interpolation is a mathematical method for estimating unknown values ​​between known data points, based on several known points, such as (x1, y1), (x2, y2)...(x...). n ,y n To construct a function that can "pass through" these known points, called an interpolation function, we then use this function to calculate the unknown function value at any position between the known points. Logical consistency verification is to check whether there are contradictions, conflicts or unreasonable relationships within the target object (such as data, plans, schemes) and between its components through preset rules or logical deduction, so as to ensure that it is logically sound and self-consistent.

[0047] Step S102: Based on the project's real-time status data cube, the environmental performance index trajectory under the preset specific construction strategy is predicted by the environmental process dynamic prediction model to obtain the environmental performance index trajectory under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on the attention mechanism.

[0048] The environmental process dynamic prediction model is a bidirectional long short-term memory (BiLSTM) neural network based on an attention mechanism. Its technical principle is as follows: Bidirectional LSTM extracts features from the forward and backward directions of the time series using forward LSTM and backward LSTM respectively, capturing the bidirectional temporal dependencies of historical data. This overcomes the limitation of traditional unidirectional LSTM, which can only process temporal information in a single direction. The attention mechanism refers to calculating the contribution weight of data at different time steps in the historical data sequence to the prediction results, strengthening the influence of key data, such as data during periods of sudden changes in environmental indicators after construction adjustments, and weakening interference from irrelevant data. The pre-set specific construction strategy is a set of candidate schemes containing clearly quantifiable construction control actions, pre-defined to verify the differences in the environmental impact of different construction period advancement plans. The core is a set of specific construction control schemes corresponding to the future planned control action sequence.

[0049] For example, historical data sequences from the past k time steps are extracted from the project's real-time status data cube. The input vector for each time step consists of environmental performance indicators (such as water quality chemical oxygen demand), process control variables (such as construction intensity, material input, and frequency of water spraying and dust suppression), and construction status variables (such as machinery uptime, personnel allocation, and equipment load rate). The historical data sequences are input into the environmental process dynamic prediction model to generate a context vector representation containing bidirectional time-series features. Then, an attention mechanism is used to calculate the contribution weight of the data at each time step, and a weighted context vector is generated by combining the context vector. Finally, a preset specific construction strategy (such as different construction intensities and resource allocation schemes) is input into the model. The model combines the context vector and the weighted context vector to perform multi-step predictions and outputs the trajectory of environmental performance indicators under the preset specific construction strategy. This trajectory is plotted with time on the horizontal axis and environmental performance indicator values ​​on the vertical axis to intuitively present the changing trend of future environmental indicators. Multi-step prediction refers to the task of predicting multiple consecutive time steps or multiple related results at once based on historical data or input information in fields such as time series analysis and machine learning. Its core is to break through the limitations of single-step prediction (predicting only the next result) and solve the need to predict trends over a period of time in the future.

[0050] Step S103: Based on the predicted trajectory of environmental performance indicators and the project schedule target, with environmental quality compliance as a hard constraint, the objective function is solved in a rolling manner through the model predictive controller to generate a sequence of control actions for resource allocation; wherein, the objective function aims to minimize the schedule deviation and cost deviation.

[0051] Rolling optimization is a core operation in the Model Predictive Control (MPC) framework, often used to handle real-time optimization problems of dynamic systems (such as schedule management in environmental engineering projects). Its core logic is to decompose long-cycle global optimization into multiple short-cycle local optimizations, and achieve dynamic adaptation and precise control through a cycle of "finite-time domain optimization - execute the current step - update the state - repeat optimization". The core principle of the model predictive controller is to predict the system behavior in the future finite time domain based on the current state in each control cycle, and determine the current control action by solving the optimization problem, thus achieving rolling optimization.

[0052] For example, based on the predicted trajectory of environmental performance indicators and project schedule targets (such as the upper limit of the total schedule and the completion time of key nodes), an objective function is constructed. This function aims to minimize the schedule deviation and cost deviation, while introducing environmental performance tracking error, control action change penalty, and constraint slack as objective terms. The optimization variables are set as the adjustment amounts of future construction control actions (such as the construction intensity adjustment rate and the change in machinery operating time). The output of the dynamic prediction model of the environmental process is used as the dynamic constraint of the optimization function to ensure that the optimization process conforms to the objective law of "construction control-environmental response". At the same time, hard constraints are set, including project schedule constraints, control variable constraints, and environmental quality constraints, to limit the feasible domain of the optimization solution. Based on the dynamic constraints and hard constraints, the objective function is solved in a rolling manner through the model predictive controller to obtain the resource allocation control action adjustment sequence that satisfies the constraints and minimizes the objective function, clarifying the specific adjustment direction and magnitude of construction control at each future time step.

[0053] Step S104: Based on the control action adjustment sequence and the preset action-instruction mapping table, generate construction instructions, and perform a scheme pre-rehearsal based on the construction instructions to generate a construction procedure schedule and resource scheduling update scheme.

[0054] Among them, the preset action-instruction mapping table is a standardized mapping tool in environmental engineering project schedule management that connects abstract control actions with concrete construction instructions. This table defines the correspondence between control action adjustment amounts and natural language instructions. For example, "the adjustment amount of machine operating time is +2 hours / day" corresponds to "starting tomorrow, the daily operating time of a single excavator will increase by 2 hours", and "the adjustment amount of water spraying and dust suppression frequency is -1 time / 2 hours" corresponds to "the water spraying frequency in high dust areas will be adjusted from once every 3 hours to once every 2 hours". This is used to ensure that abstract mathematical adjustment amounts can be transformed into understandable execution instructions. Scheme pre-drilling refers to the key step of verifying the feasibility of the optimized construction scheme in a virtual environment, which is usually achieved by relying on digital twin technology. Digital twin technology refers to the construction of a virtual environment that replicates the physical construction scene 1:1 through 3D modeling and Internet of Things (IoT) data synchronization. It can realistically simulate the entire construction process, including mechanical operation, personnel flow, process connection, and environmental response, and has functions such as conflict detection, process traceability, and data visualization.

[0055] For example, each adjustment amount in the control action adjustment sequence is matched with the action-instruction mapping table to generate preliminary natural language instructions. The project resource database is then invoked to bind the natural language instructions with resource information, supplementing the resource source task (e.g., the process to which idle machinery belongs), target task (e.g., the amount of work to be completed after adjustment), transfer time (e.g., the time consumed by machinery allocation), and workload (e.g., the amount of work expected to be completed with increased construction time), generating detailed scheduling commands. Based on these detailed scheduling commands, a scheme pre-simulation is conducted in a digital twin environment, a 1:1 digital replica of the project, containing a virtual model of elements such as the construction site, machinery, personnel, and environment. The virtual model is driven to simulate the execution process of construction instructions, recording progress changes, resource occupancy, and environmental response data in real time, detecting spatial conflicts (e.g., overlapping machinery movement lines) and resource contention conflicts (e.g., the same resource being called by multiple tasks). After the pre-simulation is conflict-free, a construction process schedule (clarifying the start and stop times and connections of each process) and a resource scheduling update plan (determining the allocation details of machinery, personnel, and materials) are generated. Among them, the project resource database is a structured data collection used to centrally store and manage various resource information related to project construction. The stored information includes machinery ledgers, personnel schedules, material inventory tables, construction area division maps, etc.; the digital twin environment is a "virtual-real mapping + interaction" system built on physical entities and through digital technology, realizing precise correspondence, real-time linkage and collaborative optimization between the physical world and digital space. In essence, it is to create a perceptible, analyzable and controllable digital mirror for physical objects or scenes; the virtual model is a digital simulation object that exists in virtual space and is built with the help of digital twin technology. It restores or creates digital substitutes for real objects, abstract concepts and even dynamic scenes through data and algorithms.

[0056] In this embodiment, multi-source heterogeneous data from environmental engineering project sites are collected and integrated to successfully construct a real-time project status data cube as the project data foundation. A bidirectional LSTM model based on an attention mechanism is used to achieve accurate prediction of environmental performance. An optimization problem is solved through a model predictive controller, generating a sequence of control actions to adjust resource allocation while meeting hard environmental quality constraints. Through action-instruction mapping and digital twin pre-simulation, an executable construction schedule and resource scheduling update plan are generated. This approach can couple environmental monitoring and project scheduling in real time, dynamically adjust construction strategies, and effectively solve problems such as response lag, resource waste, and environmental risks in traditional static management, thereby improving the intelligence and precision of environmental engineering project management.

[0057] In one exemplary embodiment, the method further includes:

[0058] Step S201: Monitor environmental performance data and construction progress data in real time after the resource scheduling update plan is implemented.

[0059] For example, environmental performance data and construction progress data are collected in real time during the execution of the resource scheduling update plan. Environmental performance data includes real-time monitoring values ​​of core indicators such as water quality, air quality, and soil condition; construction progress data includes information such as machinery operating status, personnel attendance, percentage of work completed, and amount of work completed. The collected raw data undergoes standardized preprocessing: for environmental performance data, outliers caused by temporary sensor malfunctions are removed, measurement noise is filtered using a data smoothing algorithm, and the data sampling frequency and units are standardized; for construction progress data, consistency verification of information from various data sources is performed, for example, cross-validating machinery operation time with the corresponding work completion amount, and correcting reporting delays or recording errors. Among them, data smoothing algorithms are a type of data analysis technique used to process random noise, abnormal fluctuations, or high-frequency interference in raw data. While preserving the core trends of the data (such as long-term growth and periodic changes), they can reduce the impact of irrelevant interference on data interpretation, making the data more stable and easier to reflect the true patterns. Consistency verification of information from various data sources refers to the process of checking the matching and rationality of data from different sources in key dimensions such as time, space, and logical relationships during the integration of multi-source data through preset rules and technical means, eliminating data contradictions or deviations, and ensuring that the data is unified and usable. Cross-validation (CV) is a commonly used model evaluation method in statistics. On a limited dataset, by dividing the data and repeating training and validation, it reduces the risk of model overfitting and more objectively evaluates the model's generalization ability.

[0060] Step S202: Update the project real-time status data cube based on environmental performance data and construction progress data to obtain an updated project real-time status data cube. Based on the updated project real-time status data cube, re-execute the steps from generating environmental performance indicator trajectories to generating construction procedure schedules and resource scheduling update schemes to obtain updated construction procedure schedules and updated resource scheduling update schemes.

[0061] For example, standardized environmental performance data and construction progress data are received, and the data is updated according to the spatiotemporal indexing rules of the project real-time status data cube. The update process follows the principle of "incremental replacement + dimensional supplementation": in the time dimension, newly collected real-time data replaces historical data in the same dimension of the cube according to the corresponding timestamp, while supplementing the dataset for new time nodes; in the spatial dimension, the spatial coverage dimension of the cube is improved for new data from different construction areas and monitoring points; in the indicator dimension, it is ensured that the environmental and progress-related indicator data are consistent with the indicator system of the original cube, maintaining the stability of the data structure. The updated project real-time status data cube is used to fully reflect the actual status of the project after the implementation of the plan. Based on the updated data cube, the environmental performance indicator trajectory prediction process is restarted. By mining the correlation patterns in the new data through a bidirectional long short-term memory neural network based on the attention mechanism, an environmental performance indicator trajectory that fits the current reality is generated; then, the objective function is solved in a rolling manner through the model prediction controller, and the constraints and objective weights are dynamically adjusted and optimized in combination with the latest data to generate a resource allocation control action adjustment sequence adapted to the current state; new construction instructions are generated based on the adjustment sequence and the plan is rehearsed to obtain the updated construction procedure schedule and resource scheduling update plan. Among them, data updates under the spatiotemporal indexing rules refer to ensuring that data query efficiency does not decrease significantly while maintaining the integrity of the "spatial location + time dimension" index structure; the "incremental replacement + dimension supplementation" principle is a methodology that leans towards practical optimization, achieving content iteration through local upgrades and perspective expansion without overturning the original foundation.

[0062] In this embodiment, by monitoring environmental and progress data after the implementation of the plan in real time, the deviation information between the actual execution effect and the expected goal is obtained. Based on this deviation, the core data carrier is dynamically updated. Then, a complete prediction-optimization-generation process is used to reconstruct the control plan to adapt to the current project status, forming a closed-loop control mechanism of "execution-monitoring-update-re-optimization". This can effectively solve the problems of disconnect between environmental monitoring and progress control and delayed response in traditional static management, ensuring that environmental performance indicators are always within a controllable range. At the same time, it dynamically optimizes the deviation between schedule and cost, improving the real-time performance, accuracy and adaptability of environmental engineering project schedule management.

[0063] In one embodiment, based on a real-time project status data cube, an environmental performance indicator trajectory under a preset specific construction strategy is predicted using an environmental process dynamic prediction model to obtain the environmental performance indicator trajectory under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on an attention mechanism, including:

[0064] Step S301: Extract a preset historical data sequence for past time periods from the project's real-time status data cube; wherein, the input vector for each time step in the historical data sequence consists of environmental performance indicators, process control variables, and construction status variables.

[0065] The preset past time period is a time interval set based on the project construction cycle, data sampling frequency, and the correlation between the environment and construction. Its purpose is to ensure that the extracted historical data can cover the key impact cycles and provide sufficient correlation information for model learning. The real-time status data cube is a multi-source data carrier after spatiotemporal alignment processing, including three dimensions: time, space, and indicators. Environmental performance indicators refer to monitorable environmental quality characterization parameters such as water quality chemical oxygen demand and air quality particulate matter concentration. Process control variables refer to human-adjustable construction parameters such as construction intensity, material input, and frequency of water spraying and dust suppression. Construction status variables refer to parameters reflecting the real-time construction status such as machinery uptime, personnel configuration, and equipment load rate.

[0066] For example, the real-time status data cube of the project is sliced ​​according to the time dimension to extract continuous time-series data within the preset time period. Three core variables are then selected from the indicator dimension: environmental performance indicators, process control variables, and construction status variables. The three types of variables at each time step are concatenated in a preset fixed order to form a vector with a unified dimension, ultimately integrating into a dimensionally regular historical data sequence. Here, the preset fixed order refers to a pre-defined arrangement of elements or the order of execution that cannot be arbitrarily changed.

[0067] Step S302: The historical data sequence is encoded using a bidirectional long short-term memory neural network to generate a context vector representation.

[0068] The encoding process refers to the process of extracting features and integrating contextual information from the extracted historical data sequence using a bidirectional long short-term memory neural network. The bidirectional long short-term memory neural network includes a forward long short-term memory neural network (Forward LSTM) and a backward long short-term memory neural network (Backward LSTM). The forward network processes the historical sequence in chronological order (from past to recent) to uncover forward temporal dependencies. The backward network processes the historical sequence in reverse chronological order (from recent to past) to capture backward temporal correlation features. Compared with the traditional unidirectional long short-term memory neural network (Long Short-Term Memory, or LSTM), it can extract contextual information from time-series data more comprehensively.

[0069] For example, using the input vector of each time step in the historical data sequence as the processing unit, the data is processed through the forget gate, input gate, and output gate of a bidirectional long short-term memory neural network to filter and retain key temporal features, generating the hidden state corresponding to each time step. Specifically, the forget gate uses the sigmoid function to judge and discard redundant information in the input vector that is irrelevant to the temporal pattern (such as abnormal fluctuation data); the input gate filters key information in the input vector that is valuable for subsequent prediction (such as the correlation data between construction intensity and environmental indicators) and updates the cell state using the tanh function; the output gate combines the cell state and the sigmoid function to output the hidden state of that time step, completing the retention of key temporal features for a single time step. The sigmoid function is a commonly used S-shaped curve function, frequently used in machine learning and neural networks, and its mathematical expression is: f(x) = 1 / (1+e^(-x) / (x-x)). -x This function maps input values ​​to outputs between 0 and 1, exhibiting smooth, continuously differentiable properties; the tanh function is a hyperbolic tangent function, commonly used in mathematics and neural networks, defined as tanh(x) = (e^(-1 / 2)). x -e -x ) / (e x +e -xThe output value ranges from -1 to 1. For example, a forward LSTM network (processing sequences chronologically from past to recent) and a reverse LSTM network (processing sequences chronologically in reverse order from recent to past) are run separately to obtain the hidden states of each network at each time step. Then, through vector concatenation, the forward hidden states reflecting the temporal trend from history to recent times and the reverse hidden states reflecting the temporal correlation from recent to past are integrated into a comprehensive hidden state sequence. Global feature extraction is then performed on this comprehensive hidden state sequence, such as through attention weight allocation, to fuse the bidirectional temporal features scattered across each time step, generating a context vector representation that can fully characterize the bidirectional temporal patterns of the historical data sequence. Global feature extraction refers to the process of mining key information that represents the overall attributes and global patterns of sequence data, such as time series, text, and images, rather than being limited to the processing of local fragment features.

[0070] Step S303: Through the attention mechanism, calculate the contribution weight of data at different time steps in the historical data sequence to the predicted value, and generate a weighted context vector based on the contribution weight and context vector representation.

[0071] For example, based on context vector representation, the contribution weights of data at different time steps in the historical data sequence to the predicted value (i.e., the prediction of future environmental performance indicators) are calculated: First, the context vector of each time step is linearly transformed and mapped to an attention score, which is used to quantify the degree of influence of the corresponding time step data on the prediction of future environmental performance indicators; then, the attention scores of all time steps are normalized to obtain the contribution weight of each time step, and the sum of the weights is 1. The higher the weight, the greater the influence of the data at that time step on the prediction result. The context vector of each time step is multiplied element-wise with the corresponding contribution weight to obtain a weighted time step vector, and then all weighted time step vectors are summed to generate a weighted context vector. Here, linear transformation is the core transformation form in linear algebra. The core is to satisfy linearity, that is, to maintain compatibility with vector addition and scalar multiplication operations. It can be understood as regularizing the vector without destroying its basic linear structure; normalization refers to mapping the original data of different magnitudes and units to a unified numerical range, most commonly [0,1] or [-1,1].

[0072] Step S304: Input the preset specific construction strategy into the long short-term memory neural network, combine the context vector and the weighted context vector, perform multi-step prediction, and obtain the trajectory of environmental performance indicators under the preset specific construction strategy.

[0073] Among them, the preset specific construction strategy is a pre-defined candidate construction scheme, which is represented by a sequence of process control variables over a future period of time, such as the sequence of control actions corresponding to different construction intensities and resource allocation schemes, used to provide clear construction scenario assumptions for prediction; Long Short-Term Memory Neural Network (LSTM) is a special type of Recurrent Neural Network (RNN) that includes three gating units (i.e., forget gate, input gate, and output gate) and a cell state. Its cell state is like an information highway, which can stably transmit key information in the sequence, while the three gating units are responsible for regulating the information of the cell state. This LSTM is used to solve the gradient vanishing / gradient explosion problem of traditional RNN when processing long sequence data, that is, traditional RNN has difficulty remembering the early key information in the sequence, while LSTM achieves selective memory and forgetting of information through a unique gating mechanism, which can effectively capture long-term dependencies in long sequences.

[0074] For example, a Long Short-Term Memory (LSTM) neural network is used as the decoder. The context vector and the weighted context vector are concatenated to obtain a concatenated vector, which serves as the initial hidden state and cell state of the decoder, ensuring that the decoder makes predictions based on historical core features. Simultaneously, a pre-defined specific construction strategy is sequentially input into the decoder at each time step. The decoder combines the initial state and the control action at the current time step to output the predicted value of the environmental performance index for the corresponding time step, and uses this predicted value as the input of the hidden state for the next time step, gradually completing multi-step predictions. The prediction results from all time steps are integrated to form the trajectory of the environmental performance index over a future period.

[0075] In this embodiment, a regularized historical data sequence is extracted from a unified data cube. A bidirectional long short-term memory neural network is used to comprehensively capture the bidirectional temporal dependence features of the historical data. An attention mechanism is combined to focus on key impact information, and a preset construction strategy is incorporated to achieve multi-step environmental performance prediction, forming an environmental performance indicator trajectory over a future period. This effectively solves the problem that traditional prediction models can only capture time-series information in a single direction and lack sufficient attention to key data, significantly improving the comprehensiveness, relevance, and accuracy of environmental performance indicator trajectory prediction.

[0076] In one embodiment, based on the predicted trajectory of environmental performance indicators and the project schedule target, and with environmental quality compliance as a hard constraint, a model predictive controller is used to solve the objective function in a rolling manner to generate a sequence of control actions for resource allocation. The objective function aims to minimize schedule and cost deviations, and includes:

[0077] Step S401: Based on the predicted trajectory of environmental performance indicators and the project schedule target, in each control cycle, construct an optimization function with the objective of minimizing environmental performance tracking error, control action change penalty, and constraint slack; wherein, the optimization variable of the optimization function is the adjustment amount of future construction control actions:

[0078]

[0079] in, For time t, predict the environmental performance indicators for the next k steps. For the expected trajectory of environmental performance, Let Q be the weighted L2 norm, and let Q be the state weight matrix. Let t be the adjustment amount of the control action k steps in the future. To constrain the slack amount, To relax the penalty coefficient, For environmental performance tracking errors, Punishment for controlling changes in actions.

[0080] Among them, the project schedule target includes preset requirements such as the upper limit of the total construction period and the completion time of key nodes; the environmental performance indicator prediction trajectory is the data on the future environmental quality change trend under a specific construction strategy; the optimization variables are the adjustment amount of future construction control actions, including adjustable parameters such as the construction intensity adjustment rate, the change in machinery operating time, and the adjustment value of material input.

[0081] For example, based on the predicted trajectory of environmental performance indicators and the project schedule target, a multi-objective optimization function is constructed in each control cycle. This function includes three core objective terms: an environmental performance tracking error term, which calculates the deviation between the predicted value and the expected trajectory using a weighted Euclidean Norm, with the state weight matrix Q used to reinforce the priority of core environmental indicators; a control action change penalty term, which addresses the fluctuation of control actions in adjacent time steps, used to avoid resource waste or operational risks caused by sudden changes in construction strategies; and a constraint relaxation amount ε, which is a non-negative variable, used in conjunction with the relaxation penalty coefficient ρ to provide a flexible buffer for constraints in extreme scenarios, ensuring that the optimization problem has a solution.

[0082] Step S402: Set the output of the environmental process dynamic prediction model as the dynamic constraint of the optimization function.

[0083] The technical principle of dynamic constraints is to transform the mapping relationship between "construction control actions and environmental performance" into mathematical constraints, ensuring that the optimization process conforms to the dynamic laws of the objective environment and avoiding ineffective solutions that are divorced from reality.

[0084] For example, the dynamic prediction model of environmental processes is transformed into a state-space equation form, including state equations and output equations. The state equation is: x(k+i+1|k)=f(x(k+i|k),u(k+i|k)), where x represents the internal state of the environment, such as core variables that cannot be directly monitored but affect environmental changes, such as pollutant accumulation and latent soil moisture; u represents construction control actions, such as adjustable parameters like construction intensity and material input; k is the current time; and i is the future time step. The nonlinear state transition function is generated by the encoding logic of a bidirectional long short-term memory neural network based on an attention mechanism. During the encoding process, the bidirectional LSTM simultaneously learns the positive and negative dependencies of historical time-series data, while the attention mechanism focuses on key time-series information affecting the environmental state. Ultimately, this learned pattern of "construction control actions driving changes in the internal state of the environment" is solidified into a function f, enabling the equation to accurately describe how the internal environmental state x(k+i|k) at time t=k+i evolves into the state x(k+i+1|k) at time t=k+i+1 under the action of construction control action u(k+i|k). The output equation is: y(k+i|k)=g(x(k+i|k),u(k+i|k)), where y is a directly monitorable environmental performance indicator (such as dust concentration). The decoding logic is solidified and generated by a bidirectional LSTM with an attention mechanism. The decoding process establishes a correspondence between the internal state and monitorable indicators based on the encoded environmental internal state features and construction control action information. This is achieved through a function... The intangible x(k+i|k) is transformed into a quantifiable and monitorable environmental performance indicator y(k+i|k), thereby enabling an explicit representation of environmental performance under construction control actions.

[0085] Step S403: Set the project schedule constraint, control variable constraint, and environmental quality constraint as hard constraints for the optimization problem.

[0086] For example, the project schedule, control variable limits, and environmental quality standards are obtained. The project schedule specifies the minimum progress completion rate required for each time step; the control variable limits are determined by the physical characteristics of construction resources, such as maximum machinery uptime and maximum daily material supply; and the environmental quality standards are the upper limits of environmental indicators stipulated by the state or industry. For example, these three types of data are transformed into a set of mathematical inequality constraints: the schedule constraint requires that the construction progress completion rate at each time step not be lower than a preset minimum value, ensuring that the project schedule target is not deviated from; the control variable constraint limits the adjustment range of construction control actions to within the upper and lower limits allowed by equipment and resources, avoiding unrealistic control instructions; and the environmental quality constraint specifies that the predicted values ​​of environmental performance indicators must not exceed the standard limits. A slack variable ε is introduced into the constraints along with a penalty coefficient ρ to allow slight fluctuations in extreme scenarios, but with corresponding penalties. Among them, the preset minimum value is a benchmark value set in advance based on business needs, rules and standards, or actual scenarios; the standard limit is usually clearly defined by industry or local regulations and standards, such as the "Ambient Air Quality Standard" and "Surface Water Environmental Quality Standard" in the environmental field, and the construction safety standard in the engineering field. Its value setting is based on scientific research, risk assessment and actual application needs, taking into account safety, compliance and feasibility.

[0087] Step S404: Based on dynamic constraints and hard constraints, the optimization function is solved using a differential evolution algorithm to obtain the control action adjustment sequence for resource allocation.

[0088] Among them, the differential evolution algorithm is a population-based stochastic search optimization algorithm that efficiently searches for the optimal solution in a constrained space by simulating the mutation, crossover, and selection operations in the biological evolution process. It is particularly suitable for solving complex nonlinear optimization problems in continuous space, such as function extrema and engineering parameter optimization, due to its simple structure.

[0089] For example, based on the optimization function, dynamic constraints, hard constraints, and preset differential evolution (DE) algorithm parameters, including population size, mutation factor, crossover probability, and maximum number of iterations. The optimization function is solved using a differential evolutionary algorithm: First, the population is initialized by generating multiple sets of candidate control action adjustment sequences based on the constraints of the construction control variables. Each set of sequences falls within the feasible region and covers different control amplitudes and combinations. A mutation operation generates a mutation vector based on individual differences. This involves selecting three different individuals from the population and generating a mutation vector according to the rule "baseline individual + mutation factor × (difference between different individuals)", introducing new control combinations based on inter-individual differences. A crossover operation randomly replaces the control parameters of the original individuals with the mutation vector according to a preset crossover probability, fusing the characteristics of both. A selection operation calculates the fitness of individuals, i.e., the optimization function value, and considers factors such as environmental performance error and control action smoothness, retaining individuals with better fitness that meet the schedule and environmental hard constraints, while eliminating inferior solutions. This process is repeated iteratively until the maximum number of iterations is reached or the fitness converges, ultimately selecting the globally optimal individual. This individual becomes the optimal control action adjustment sequence for the next K time steps, clearly defining the adjustment direction and amplitude of construction control at each time step.

[0090] In this embodiment, an optimization function that balances environmental compliance, construction stability, and constraint solvability is constructed. Dynamic constraints are used to correlate construction strategies with environmental response patterns. Hard constraints define the feasible region boundary, and a differential evolutionary algorithm is used to efficiently solve the nonlinear programming problem, resulting in a control action adjustment sequence for resource allocation. This effectively addresses the disconnect between environment, schedule, and resources in traditional static management, ensuring that the generated control action adjustment sequence conforms to the dynamic laws of the objective environment while meeting the rigid requirements of schedule, resources, and environment. Simultaneously, it avoids sudden changes in construction strategies, achieving precise optimization of the environmental engineering project's schedule and cost, and improving the scientific rigor, feasibility, and dynamic adaptability of the management and control scheme.

[0091] In one embodiment, such as Figure 2 As shown, based on the control action adjustment sequence and the preset action-instruction mapping table, construction instructions are generated. Based on these instructions, a scheme pre-simulation is performed, generating a construction sequence schedule and resource scheduling update scheme, including:

[0092] Step S501: Based on the preset action-instruction mapping table, convert the adjustment amount in the control action adjustment sequence into natural language instructions.

[0093] For example, each adjustment amount in the control action adjustment sequence is parsed to determine its corresponding control dimension (such as construction machinery, material delivery, dust suppression measures, etc.), adjustment direction (increase, decrease, maintain), and amplitude range; then the action-instruction mapping table is called, and the corresponding natural language expression template is matched according to the parsing results to transform "adjustment amount + control dimension" into a structured natural language instruction of "executor + action + time + requirement".

[0094] Step S502: Bind the natural language instructions and the resource database to generate detailed scheduling commands; wherein, the detailed scheduling commands include the resource source task, the target task, the transfer time, and the workload.

[0095] Among them, the resource database is a structured data carrier that stores all the resource information of the project, including core information such as machinery ledger (equipment number, operating status, and affiliated work group), personnel schedule (job configuration and attendance), material inventory table (material type, storage location, and available quantity), and construction area division map (area coordinates and process relationship).

[0096] For example, each natural language instruction is broken down into its elements, extracting three core elements: "task type," "execution object," and "adjustment requirements." Then, based on these elements, a resource database is searched to match suitable execution resources. For example, "excavator adjustment" is matched with the corresponding equipment number and operator. Execution details such as the resource source task (i.e., the process to which the idle resource belongs), the target task (i.e., the amount of work to be completed after adjustment), the transfer time (i.e., the time consumed by resource allocation), and the quantitative value of the workload (i.e., the expected amount to be completed after adjustment) are added. Finally, the elements, resources, and details are integrated into a standardized detailed scheduling command. The command format can be unified as "resource information + action requirements + time node + quantitative target."

[0097] Step S503: Based on detailed scheduling commands, simulate the future construction process and generate a construction sequence schedule and resource scheduling plan.

[0098] For example, a virtual construction environment (i.e., a digital twin environment) constructed using digital twin technology can be used as the simulation carrier. Detailed scheduling commands are converted into parameterized instructions that the virtual environment can recognize. For example, "machine scheduling" is converted into equipment coordinate movement and runtime parameters. The real-time status of the current project, such as the progress of each process and the current position of resources, is loaded and used as the initial conditions for simulation. Then, the virtual model in the virtual environment is driven to execute the construction process according to the scheduling commands. The progress of process completion, resource occupancy, and environmental indicator changes are recorded in real time. Potential problems such as spatial conflicts (such as equipment movement line intersections) and resource contention conflicts (such as multiple tasks calling the same resource) are detected simultaneously. After the simulation is conflict-free, the simulation data is integrated to generate a construction process schedule that clearly defines the start and stop times and connection relationships of each process, as well as a resource scheduling scheme that includes resource allocation paths and usage allocation.

[0099] In this embodiment, mathematical control actions are converted into natural language commands through a preset action-instruction mapping table. Detailed scheduling commands are generated by supplementing execution elements through binding a resource database. Digital twin technology is used to simulate the construction process and identify conflicts, ensuring the feasibility of the plan and generating a construction sequence schedule and resource scheduling scheme. This effectively solves the problems of vague construction instructions, disconnected resource allocation, and frequent on-site conflicts in traditional construction methods. It achieves precise matching of construction instructions with resources and procedures in environmental engineering projects, improving the scientific nature, accuracy, and feasibility of schedule planning and resource allocation, and providing a reliable guarantee for the efficient and orderly progress of projects.

[0100] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0101] Based on the same inventive concept, this application also provides an environmental engineering project schedule real-time management device for implementing the above-mentioned environmental engineering project schedule real-time management method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the environmental engineering project schedule real-time management device provided below can be found in the limitations of the environmental engineering project schedule real-time management method described above, and will not be repeated here.

[0102] In one exemplary embodiment, such as Figure 3 As shown, a real-time scheduling control device 300 for environmental engineering projects is provided, including:

[0103] The data acquisition module 301 is used to collect multi-source heterogeneous data from the environmental engineering project site and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data and project plan data;

[0104] The trajectory prediction module 302 is used to predict the trajectory of environmental performance indicators under a preset specific construction strategy based on the real-time status data cube of the project and through the environmental process dynamic prediction model, so as to obtain the trajectory of environmental performance indicators under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on the attention mechanism.

[0105] The sequence generation module 303 is used to generate a sequence of control actions for resource allocation by predicting the trajectory based on environmental performance indicators and project schedule targets, with environmental quality compliance as a hard constraint, and by using a model predictive controller to solve the objective function in a rolling manner; wherein, the objective function aims to minimize schedule deviation and cost deviation.

[0106] The schedule determination module 304 is used to generate construction instructions based on the control action adjustment sequence and the preset action-instruction mapping table, and to perform scheme pre-simulation based on the construction instructions to generate construction procedure schedule and resource scheduling update scheme.

[0107] In one exemplary embodiment, the system further includes:

[0108] The data monitoring module is used to monitor environmental performance data and construction progress data in real time after the resource scheduling update plan is implemented;

[0109] The project schedule update module is used to update the real-time status data cube of the project based on environmental performance data and construction progress data, so as to obtain an updated real-time status data cube of the project. Based on the updated real-time status data cube of the project, the steps of generating environmental performance indicator trajectories and generating construction procedure schedule and resource scheduling update plan are re-executed to obtain an updated construction procedure schedule and updated resource scheduling update plan.

[0110] In one embodiment, the trajectory prediction module 302 is further configured to:

[0111] Extract historical data sequences for a preset past time period from the project's real-time status data cube; wherein, the input vector for each time step in the historical data sequence consists of environmental performance indicators, process control variables, and construction status variables;

[0112] Historical data sequences are encoded using a bidirectional long short-term memory neural network to generate context vector representations;

[0113] The attention mechanism is used to calculate the contribution weight of data at different time steps in the historical data sequence to the predicted value, and a weighted context vector is generated based on the contribution weight and context vector representation.

[0114] The preset specific construction strategy is input into a long short-term memory neural network, and combined with context vectors and weighted context vectors, multi-step prediction is performed to obtain the trajectory of environmental performance indicators under the preset specific construction strategy.

[0115] In one embodiment, the sequence generation module 303 is further configured to:

[0116] Based on the predicted trajectory of environmental performance indicators and the project schedule target, an optimization function is constructed for each control cycle to minimize environmental performance tracking error, penalty for changes in control actions, and constraint slack. The optimization variables in this function are the adjustments made to future construction control actions.

[0117]

[0118] in, For time t, predict the environmental performance indicators for the next k steps. For the expected trajectory of environmental performance, Let Q be the weighted L2 norm, and let Q be the state weight matrix. Let t be the adjustment amount of the control action k steps in the future. To constrain the slack amount, To relax the penalty coefficient, For environmental performance tracking errors, Punishment for controlling changes in actions;

[0119] Set the output of the dynamic prediction model for environmental processes as the dynamic constraint of the optimization function;

[0120] Set the project schedule constraints, control variable constraints, and environmental quality constraints as hard constraints for the optimization problem;

[0121] Based on dynamic constraints and hard constraints, the optimization function is solved using a differential evolution algorithm to obtain the control action adjustment sequence for resource allocation.

[0122] In one embodiment, the schedule determination module 304 is further configured to:

[0123] Based on the preset action-instruction mapping table, the adjustment amount in the control action adjustment sequence is converted into natural language instructions;

[0124] Natural language instructions are bound to a resource database to generate detailed scheduling commands; these detailed scheduling commands include the resource source task, target task, transfer time, and workload.

[0125] Based on detailed scheduling commands, the future construction process is simulated to generate a construction sequence schedule and resource scheduling plan.

[0126] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the aforementioned method for real-time management and control of environmental engineering project scheduling.

[0127] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0128] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0129] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for real-time management and control of the schedule of environmental engineering projects, characterized in that, The method includes: Collect multi-source heterogeneous data from the environmental engineering project site, and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; wherein, the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data, and project plan data; Based on the project's real-time status data cube, the environmental performance index trajectory under a preset specific construction strategy is predicted by an environmental process dynamic prediction model, thereby obtaining the environmental performance index trajectory under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on an attention mechanism. Based on the predicted trajectory of the environmental performance indicators and the project schedule target, with environmental quality compliance as a hard constraint, the objective function is solved in a rolling manner through a model predictive controller to generate a sequence of control actions for resource allocation; wherein, the objective function aims to minimize the schedule deviation and cost deviation. Based on the control action adjustment sequence and the preset action-instruction mapping table, construction instructions are generated, and a scheme pre-rehearsal is performed based on the construction instructions to generate a construction procedure schedule and resource scheduling update scheme.

2. The method according to claim 1, characterized in that, The method further includes: Real-time monitoring of environmental performance data and construction progress data after the resource scheduling update scheme is implemented; The project real-time status data cube is updated based on the environmental performance data and the construction progress data to obtain an updated project real-time status data cube. Based on the updated project real-time status data cube, the steps of generating the environmental performance indicator trajectory to generating the construction procedure schedule and the resource scheduling update scheme are re-executed to obtain an updated construction procedure schedule and an updated resource scheduling update scheme.

3. The method according to claim 1, characterized in that, The method involves using the project's real-time status data cube and an environmental process dynamic prediction model to predict the trajectory of environmental performance indicators under a preset specific construction strategy, thereby obtaining the trajectory of environmental performance indicators under the preset specific construction strategy. The environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on an attention mechanism, comprising: Extract a preset historical data sequence for a past time period from the project's real-time status data cube; wherein, the input vector for each time step in the historical data sequence consists of environmental performance indicators, process control variables, and construction status variables; The historical data sequence is encoded using the bidirectional long short-term memory neural network to generate a context vector representation; The contribution weights of data at different time steps in the historical data sequence to the predicted value are calculated using an attention mechanism, and a weighted context vector is generated based on the contribution weights and the context vector representation. The preset specific construction strategy is input into a long short-term memory neural network, and combined with the context vector and the weighted context vector, multi-step prediction is performed to obtain the trajectory of the environmental performance index under the preset specific construction strategy.

4. The method according to claim 1, characterized in that, Based on the predicted trajectory of the environmental performance indicators and the project schedule target, and with environmental quality compliance as a hard constraint, the objective function is solved in a rolling manner through a model predictive controller to generate a sequence of control actions for resource allocation; wherein, the objective function aims to minimize schedule deviation and cost deviation, including: Based on the predicted trajectory of the environmental performance indicators and the project schedule target, an optimization function is constructed for each control cycle to minimize environmental performance tracking error, penalty for changes in control actions, and constraint slack; wherein the optimization variable of the optimization function is the adjustment amount of future construction control actions. in, For time t, predict the environmental performance indicators for the next k steps. For the expected trajectory of environmental performance, Let L be the weighted L2 norm, and Q be the state weight matrix. Let t be the adjustment amount of the control action k steps in the future. To constrain the slack amount, To relax the penalty coefficient, For environmental performance tracking errors, Punishment for controlling changes in actions; The output of the dynamic prediction model for the environmental process is set as the dynamic constraint of the optimization function; Set the project schedule constraint, control variable constraint, and environmental quality constraint as hard constraints for the optimization problem. Based on the dynamic constraints and the hard constraints, the optimization function is solved using a differential evolution algorithm to obtain the control action adjustment sequence for resource allocation.

5. The method according to claim 1, characterized in that, The process of generating construction instructions based on the control action adjustment sequence and a preset action-instruction mapping table, and performing a scheme pre-simulation based on the construction instructions to generate a construction procedure schedule and resource scheduling update scheme, includes: According to the preset action-instruction mapping table, the adjustment amount in the control action adjustment sequence is converted into natural language instructions; The natural language instructions and resource database are bound together to generate detailed scheduling commands; wherein, the detailed scheduling commands include resource source task, target task, transfer time and workload; Based on the detailed scheduling commands, the future construction process is simulated to generate a construction sequence schedule and resource scheduling plan.

6. A real-time scheduling control device for environmental engineering projects, characterized in that, The device includes: The data acquisition module is used to collect multi-source heterogeneous data from the environmental engineering project site and perform spatiotemporal alignment processing on the multi-source heterogeneous data to generate a real-time status data cube for the project; wherein, the multi-source heterogeneous data includes ecological monitoring data, construction machinery status data, and project plan data; The trajectory prediction module is used to predict the trajectory of environmental performance indicators under a preset specific construction strategy based on the real-time status data cube of the project and through an environmental process dynamic prediction model, so as to obtain the trajectory of the environmental performance indicators under the preset specific construction strategy; wherein, the environmental process dynamic prediction model is a bidirectional long short-term memory neural network based on an attention mechanism. The sequence generation module is used to predict the trajectory and project schedule target based on the environmental performance indicators, with environmental quality compliance as a hard constraint, and generate a sequence of control actions for resource allocation by rolling the solution of the objective function through the model prediction controller; wherein the objective function aims to minimize the schedule deviation and cost deviation. The construction schedule determination module is used to generate construction instructions based on the control action adjustment sequence and the preset action-instruction mapping table, and to perform scheme pre-simulation based on the construction instructions to generate construction procedure schedule and resource scheduling update scheme.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images