Method and system for scheduling regional power grid ring network cabinet in data space

Abstract

The application discloses a production scheduling decision method and system for regional power grid ring network cabinets in a data space, relates to the technical field of production scheduling decision analysis, and comprises the following steps: taking power grid side related parameters as core inputs, constructing predicted demand and delivery time limit, so that production scheduling can face real power grid demand instead of single order signals; designing a nonlinear coupling function, so that device reliability risk indexes and supply chain states jointly act on priority calculation, high-risk tasks are highlighted through mean deviation mode, and supply chain limited tasks are inhibited through exponential decay, so that dynamic balance of risk response and executability is realized; adopting a standard deviation driven capacity adjustment algorithm, so that production capacity distribution is tilted when priority difference is large to guarantee key tasks, and is balanced when the difference is small to stabilize production; and combining process resource constraints to generate a landable production scheduling plan. Therefore, the production scheduling decision result meets the reliability demand of the power grid side and meets the supply chain and production execution conditions.

Description

Technical Field

[0001] This application relates to the field of production scheduling decision analysis technology, and in particular to a production scheduling decision method and system for regional power grid ring network cabinets in the data space. Background Technology

[0002] Ring main units (RMUs) are common high-voltage switchgear in regional power grids. They connect multiple distribution transformers into a closed loop via a ring network structure, enabling flexible power distribution and rapid fault isolation, making them indispensable equipment in urban power distribution networks. With the advancement of urban power grid upgrades and distribution automation, the demand for RMUs in regional power grids is constantly growing, with significant differences in product types and configurations. Production scheduling has become a crucial interface between manufacturers and power grid companies.

[0003] Traditional production scheduling strategies often focus on internal order quantities, delivery dates, and capacity constraints, paying little attention to external power grid operation status and equipment health information. This type of production scheduling disconnects enterprises from power grid demand, making it difficult to support the reliable operation of smart grids.

[0004] Specifically, traditional production scheduling methods are often based on fixed order demands and single production capacities, aiming to maximize capacity utilization or minimize delivery times. However, as power equipment, ring main units (RMS) production plans are closely linked to grid load, equipment status, and operation and maintenance plans. For example, the construction and renovation plans of distribution networks are highly phased and uncertain, and traditional production scheduling, relying on order demands, cannot accurately reflect future real-world demand. Furthermore, as a critical node in power supply, a failure of a RMS can lead to widespread power outages. Ignoring information such as equipment aging and load factors in the grid during production scheduling may result in low-risk equipment being produced prematurely while high-risk equipment is delayed, impacting power supply security. Moreover, RMS manufacturing involves multiple stages, including material procurement, component processing, assembly, and testing. Supply chain data is scattered across different enterprise systems, lacking unified data sharing and collaboration, causing production scheduling decisions to lag behind changes in market and grid demand. Summary of the Invention

[0005] This application provides a production scheduling decision-making method, system, storage medium, computer program product, and electronic device for regional power grid ring network cabinets in a data space, which can at least solve the problems of planning mismatch and power supply safety hazards caused by the disconnect between power grid demand and manufacturing scheduling, the lack of visibility of risks, and insufficient supply chain coordination in the current related technologies.

[0006] In a first aspect, embodiments of this application provide a production scheduling decision method for ring main units in a regional power grid within a data space. The method includes: acquiring grid-side correlation parameters, supply chain status parameters, and enterprise-side production parameters corresponding to each regional production scheduling task; the grid-side correlation parameters include the equipment reliability risk index of the ring main unit corresponding to the task in the current regional power grid, regional load forecast data, and regional power grid planning data; the enterprise-side production parameters include the available production capacity of the production line and process resource constraints; based on the grid-side correlation data corresponding to each regional production scheduling task, generating the predicted demand and predicted delivery time for ring main units for the corresponding regional production scheduling task; and for each regional production scheduling task, based on the predicted demand, equipment reliability risk index, and supply chain status parameters corresponding to the regional production scheduling task... The comprehensive priority index is calculated using a nonlinear coupling function, which includes a risk enhancement function and a supply chain resilience function. The risk enhancement function is used to nonlinearly correct the priority in response to the mean deviation of the equipment reliability risk index, thereby increasing the priority of risky tasks. The supply chain resilience function is used to correct the priority based on an exponential function of supply chain smoothness, thereby reducing the priority of tasks when the supply chain is not smooth. A standard deviation-driven capacity adjustment algorithm is used to process the comprehensive priority index of production scheduling tasks in each region to generate capacity allocation results for production scheduling tasks in each region. Based on the capacity allocation results and the process resource constraints, a production schedule is generated, which includes the production time and corresponding production capacity for each region's production scheduling tasks.

[0007] Secondly, embodiments of this application provide a production scheduling decision system for ring main units in a regional power grid within a data space. The system includes: a data acquisition unit, used to acquire grid-side correlation parameters, supply chain status parameters, and enterprise-side production parameters corresponding to each regional production scheduling task; the grid-side correlation parameters include the equipment reliability risk index of the ring main unit corresponding to the task in the current regional power grid, regional load forecast data, and regional power grid planning data; the enterprise-side production parameters include the available capacity of the production line and process resource constraints; a demand forecasting unit, used to generate the predicted demand and predicted delivery time of the ring main units for each regional production scheduling task based on the grid-side correlation data corresponding to each regional production scheduling task; and a priority calculation unit, used to calculate the priority of each regional production scheduling task based on the predicted demand, equipment reliability risk index, and supply chain status parameters corresponding to the regional production scheduling task. The system calculates a comprehensive priority index based on supply chain status and delivery deadlines using a nonlinear coupling function. This nonlinear coupling function includes a risk enhancement function and a supply chain resilience function. The risk enhancement function is used to nonlinearly correct priorities in response to deviations from the mean of the equipment reliability risk index, thereby increasing the priority of risky tasks. The supply chain resilience function is used to correct priorities based on an exponential function of supply chain smoothness, thereby reducing the priority of tasks when the supply chain is disrupted. A capacity allocation unit processes the comprehensive priority index of production scheduling tasks in each region using a standard deviation-driven capacity adjustment algorithm to generate capacity allocation results for each region's production scheduling tasks. A production scheduling plan generation unit generates a production scheduling plan based on the capacity allocation results and the process resource constraints. The production scheduling plan includes the scheduling time and corresponding production capacity for each region's production scheduling tasks.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the scheduling decision method for regional power grid ring main units in the data space of any embodiment of this application.

[0009] Fourthly, embodiments of this application provide a storage medium storing a computer program, characterized in that, when the program is executed by a processor, it implements the steps of the scheduling decision method for regional power grid ring network cabinets in the data space of any embodiment of this application.

[0010] Fifthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implements the steps of the scheduling decision method for regional power grid ring network cabinets in the data space of any embodiment of this application.

[0011] The scheduling decision-making method and system for regional power grid ring network cabinets in a data space provided in this application can produce at least the following technical effects:

[0012] (1) By introducing equipment reliability risk index, regional load forecast, and power grid planning data, the production tasks of ring main units are directly linked to the power grid operation status. The risk enhancement function is used to nonlinearly enhance the demand for high-risk equipment, thereby ensuring its priority in production scheduling. The production scheduling demand is transformed from being driven by a single order to being driven by dynamic demand forecast for power grid operation and construction. Furthermore, the risk enhancement function is used to nonlinearly enhance the priority of high-risk equipment tasks. As a result, the production schedule of high-risk nodes is effectively ensured not to be delayed due to simple delivery time or capacity constraints. Thus, even with limited capacity, priority can still be given to ensuring the safe operation of the power grid and the power supply reliability of high-load areas, thereby improving the overall safety resilience of the power grid.

[0013] (2) By constructing a supply chain resilience function in conjunction with supply chain status parameters, the priority of tasks with restricted materials or supply disruptions is suppressed and corrected, thereby avoiding situations where high-priority tasks are difficult to implement due to supply chain shortages. In addition, through a standard deviation-driven capacity adjustment algorithm, the degree of capacity tilting or balancing is dynamically determined based on the comprehensive priority distribution, enabling rapid response to urgent and high-risk tasks and smooth production scheduling under normal conditions. Thus, through a dual optimization mechanism, the impact of material constraints on production scheduling is reduced, and the concentrated congestion of bottleneck processes is also reduced, ensuring the feasibility and flexibility of priority allocation, and enabling the production plan to form a dynamic balance between risk response and feasibility, thereby significantly improving the overall capacity utilization rate and the execution stability of production scheduling.

[0014] This technical solution establishes a production scheduling decision model based on the dual constraints of power grid risk and supply chain feasibility. By combining the nonlinear coupling of risk enhancement and elasticity suppression with an adaptive capacity allocation method, a closed-loop linkage from "actual power grid demand - supply chain carrying capacity - production execution constraints" is achieved. This ensures that the production scheduling plan can guarantee the reliability of key power grid nodes and also has the feasibility of supply chain and production, thus forming a highly reliable production scheduling system for smart grid environments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1A flowchart illustrating an example of a scheduling decision method for regional power grid ring network cabinets in a data space according to an embodiment of this application is shown.

[0017] Figure 2 A flowchart illustrating an example of obtaining a device reliability risk index according to an embodiment of this application is shown.

[0018] Figure 3 A flowchart illustrating an example of regional production demand forecasting according to an embodiment of this application is shown.

[0019] Figure 4 A flowchart illustrating an example of generating capacity allocation results for production scheduling tasks in various regions according to an embodiment of this application is provided.

[0020] Figure 5 A flowchart illustrating an example of generating a production schedule based on capacity allocation results according to an embodiment of this application is shown.

[0021] Figure 6 This diagram illustrates the simulation comparison of the monthly on-time completion rate.

[0022] Figure 7 A schematic diagram illustrating the simulation comparison of cumulative risk-weighted hourly output is shown;

[0023] Figure 8 A structural block diagram of an example of a scheduling decision system for a regional power grid ring main unit in a data space, according to an embodiment of this application, is shown. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In recent years, by establishing energy data aggregation centers and adopting a "physical + logical" data aggregation model and technologies such as "data usable but invisible, data usage controllable and measurable," the secure flow and privacy protection of data at multiple stages have been resolved. The concept of data space provides a new approach to ring main unit production scheduling: comprehensively considering supply and demand, equipment reliability, and policy guidance to make more precise production decisions.

[0026] It should be noted that, among the current related technologies, some experts and scholars have proposed some novel technical directions for ring main unit monitoring or production scheduling decision-making technologies.

[0027] The literature "Design of Remote Online Monitoring System for Cable Joint Temperature of Ring Main Unit" points out that overheating of power cable joints is an important cause of ring main unit failure. By remotely monitoring the temperature of cable joints, online monitoring and alarm can be achieved for problems such as joint aging, overload, and insufficient quality in ring main units. However, it does not involve production scheduling decisions and focuses on individual equipment rather than the overall regional power grid.

[0028] Currently, the manufacturing industry widely uses Advanced Planning and Scheduling (APS) systems for production optimization. APS systems typically integrate Master Production Schedule (MPS) and Material Requirements Planning (MRP) modules, determining the production quantity and required materials for each product in different time periods based on forecasts and contracts. The flexible shop floor scheduling problem is considered a classic combinatorial optimization problem. APS systems often employ tabu search and Genetic Algorithms (GA) to solve job sequencing and resource allocation, satisfying constraints such as only one workpiece being processed per machine at a time and uninterrupted processes. Furthermore, network planning techniques are commonly used for project-based scheduling, adjusting plans through critical path analysis. Although APS systems are widely used in industrial manufacturing, these algorithms primarily focus on internal indicators such as processing time, machine load, and delivery time, lacking comprehensive consideration of external power grid load, equipment reliability, and data security, and failing to utilize the cross-domain data resources provided by the energy data space. Ring main unit manufacturers face a unique scenario in power equipment manufacturing: relatively small production batches, many varieties, and high technical requirements. Algorithms such as GA, TS (Tabu Search), and network planning are general optimization algorithms that lack specificity for the special equipment of ring main units and neglect the importance of equipment reliability and power grid safety.

[0029] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0030] The technical solutions in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of users' personal information, comply with relevant laws and regulations and do not violate public order and good morals.

[0031] Figure 1A flowchart illustrating an example of a scheduling decision method for regional power grid ring main units in a data space according to an embodiment of this application is shown.

[0032] Regarding the execution entity of the method in this application embodiment, it can be any controller or processor with computing or processing capabilities, such as a ring main unit production scheduling management platform. It deploys a nonlinear priority coupling mechanism oriented towards grid risk and supply chain resilience, supplemented by a standard deviation-driven capacity adjustment strategy, thereby breaking away from the single logic of relying solely on orders and capacity in traditional production scheduling. Thus, it not only ensures the priority production of high-risk equipment but also maintains the stability and implementation of the production schedule under supply chain fluctuations, achieving a balance between grid security, delivery reliability, and production efficiency.

[0033] like Figure 1 As shown, in step S110, the grid-side associated parameters, supply chain status parameters, and enterprise-side production parameters corresponding to the production scheduling tasks of each region are obtained.

[0034] In production scheduling decisions, the power grid operating status, supply chain conditions, and enterprise production capacity are the three key factors influencing task priority and feasibility. Therefore, it is necessary to first construct a complete set of parameter inputs to support subsequent prediction and optimization. The power grid-side parameters include the equipment reliability risk index of the ring main unit corresponding to the task in the current regional power grid, regional load forecast data, and regional power grid planning data. The enterprise-side production parameters include the available capacity of the production line and process resource constraints.

[0035] Equipment reliability risk index can be obtained by utilizing equipment reliability risk monitoring technologies from various related fields, such as calculating based on historical failure rate, service life, and load factor of the equipment. Regional load forecast data can be obtained based on historical load curves and modeling of climate and electricity demand fluctuations, and can also employ various unrestricted regional load modeling methods. Regional power grid planning data can be obtained by combining official power grid announcements, which can describe the construction and renovation plans of the regional power grid to determine phased task windows, such as planned maintenance, capacity expansion, and renovation node information. Enterprise-side production parameters include real-time available production capacity (calculated per shift), equipment and manpower constraints for key processes (such as the number of assembly workstations, parallel testing capabilities, and assembly bottleneck time), ensuring that subsequent production scheduling results are implemented at the execution level.

[0036] Therefore, by collecting parameters from multiple sources, the production scheduling input data not only reflects the urgency of the power grid demand, but also incorporates supply chain feasibility and production line execution constraints, achieving panoramic coverage of the information layer and realizing a multi-dimensional characterization of the production scheduling task.

[0037] In step S120, based on the grid-side associated data corresponding to the production tasks in each region, the predicted demand and delivery deadline for ring main units for the corresponding regional production tasks are generated.

[0038] Here, the operation and planning data of the power grid side are transformed into real demand signals for ring main units, rather than relying solely on external order data. By predicting demand and delivery deadlines, a shift from "passive order acceptance" to "proactive planning" is achieved.

[0039] In some implementations, regional load forecasting data is combined with power grid planning to identify the need for insufficient grid capacity or node upgrades within certain time windows, and to estimate the number of ring main units that need to be added or replaced in the corresponding areas. Additionally, by incorporating equipment reliability risk indices, if the risk value of a node is significantly higher than the average, it will be included in potential demand even if no orders have been triggered yet, to avoid delays in replacing high-risk equipment. Regarding delivery date determination, in addition to considering the construction timelines in the power grid plan, differentiated delivery dates should be set based on risk levels. For example, high-risk nodes can have shorter delivery dates locked in advance, while low-risk nodes are allowed to be completed over a longer period.

[0040] By adopting a demand forecasting mechanism, the production scheduling system obtains dynamic demand and delivery constraints for the future, which can accurately reflect the real urgency and construction pace in power grid operation, avoid the disconnect between production and actual power grid demand, and improve the foresight and pertinence of the production scheduling results.

[0041] It should be noted that by collecting and integrating multi-source data from the power grid side, the supply chain side, and the enterprise side, external orders are no longer used as the sole input. Instead, the actual demand and delivery time of future ring main units are calculated based on power grid load forecasts, equipment risk indices, and planning data.

[0042] Specifically, grid demand is often phased and uncertain. Relying solely on order data cannot fully reflect potential replacement or new tasks. However, multi-source data forecasting can identify high-risk equipment and potential gaps in advance, thus making production scheduling decisions more forward-looking and accurate. By using multi-source data forecasting as a prerequisite for production scheduling, a shift from order-driven to data-driven approaches is achieved. This allows production scheduling to be based on real future demand, effectively avoiding the lag and distortion problems caused by relying solely on order data, thereby improving the scientific nature and feasibility of production planning.

[0043] In step S130, for each regional production scheduling task, a comprehensive priority index is calculated using a nonlinear coupling function based on the predicted demand, equipment reliability risk index, supply chain status, and delivery deadline corresponding to the regional production scheduling task.

[0044] Here, when facing production scheduling tasks from multiple regions, it is necessary to establish a unified measurement standard that can balance safety risks, delivery deadlines, and supply chain constraints. The nonlinear coupling function includes a risk reinforcement function and a supply chain resilience function. The risk reinforcement function is used to nonlinearly adjust priorities in response to deviations from the mean of the equipment reliability risk index, thereby increasing the priority of risky tasks. The supply chain resilience function is used to adjust priorities based on an exponential function of supply chain smoothness, thereby reducing the priority of tasks when the supply chain is disrupted.

[0045] By using nonlinear functions to synthesize multi-dimensional parameters, the problems of averaging and dilution in linear weighted models are avoided, thus forming priority results that better reflect key differences.

[0046] In some implementations, the risk enhancement function uses the mean deviation of the equipment reliability risk index as the independent variable. It amplifies risk tasks with a risk level higher than the mean through a nonlinear function (such as an exponential function or a sigmoid function), thereby highlighting the priority of tasks in high-risk areas, while tasks with a risk level close to or below the mean are not overemphasized.

[0047] In some implementations, the supply chain resilience function can be used to process the supply chain smoothness index through an exponential decay model. When materials are readily available and supply is stable, the function approaches 1 and has no significant inhibition on priority. However, when there are shortages or fluctuations in the supply chain, the function rapidly decays the priority to prevent tasks from being queued as unexecutable tasks.

[0048] Furthermore, by coupling risk enhancement with elasticity functions, the predicted demand, delivery urgency, equipment reliability risk index, and supply chain status are integrated into a single priority indicator, which ensures both the response to power grid risks and the feasibility of supply chain and execution.

[0049] Therefore, by dynamically reflecting the interaction of multiple factors through priority indicators, high-risk tasks are guaranteed in advance, while tasks limited by material supply are prevented from being mistakenly scheduled ahead of time, thereby improving the rationality and robustness of production scheduling.

[0050] In step S140, a standard deviation-driven capacity adjustment algorithm is used to process the comprehensive priority index of production scheduling tasks in each region to generate capacity allocation results for production scheduling tasks in each region.

[0051] It should be understood that even with defined priority indicators, some processes may still be overloaded or production lines may remain idle if capacity distribution is not optimized for balance. Therefore, the degree of capacity allocation should be dynamically adjusted based on the overall distribution of task priorities to achieve flexible resource scheduling.

[0052] In some implementations, the overall standard deviation σ is calculated for the comprehensive priority index of all regional tasks. A high overall standard deviation σ indicates a significant difference in priority between tasks, in which case the system will concentrate more capacity on the highest priority tasks to quickly resolve critical risk points. Conversely, a low overall standard deviation σ indicates that the differences between tasks are not significant, allowing for a balanced allocation strategy to maintain smooth production line operation and avoid bottlenecks and fluctuations caused by excessive resource concentration. Furthermore, the capacity adjustment algorithm can be combined with a dynamic sliding window mechanism to update in real time within the scheduling cycle, ensuring the adaptability of capacity allocation.

[0053] By employing an adaptive capacity allocation method, the system effectively enhances rapid response capabilities during high-risk periods while maintaining stability during normal periods. This reduces fluctuations and changeover losses in the production process, and improves capacity utilization and scheduling flexibility. Consequently, it effectively reduces the volatility of capacity utilization, minimizes congestion and excessive overtime at bottleneck processes, and enhances the overall balance of the production line.

[0054] In step S150, a production schedule is generated based on the capacity allocation results and process resource constraints. The production schedule includes the production time and corresponding production capacity of each region's production task.

[0055] It should be noted that production scheduling plans need to further constrain process resources on top of priorities and capacity allocation to ensure feasibility of execution. In some implementations, based on the capacity allocation results and combined with process resource constraints (such as the number of workstations, testing stations, and manpower shifts for a certain process), a constraint solving model is used to generate a production schedule containing specific production times and capacity allocations. This production schedule not only covers the production sequence of tasks in each area but also details it down to the production line and shift level.

[0056] More preferably, process-level resource constraints can be overlaid on the existing capacity allocation results, such as the workstation capacity of the assembly process, the test bench time of the inspection process, and the lead time of key components. The constraint satisfaction solver generates specific time sequences, clarifying the start and end times of production and capacity allocation for each task. The scheduling results are output in the form of a plan table or Gantt chart and can be written into the enterprise's MES (Manufacturing Execution System) or scheduling platform to drive subsequent production execution.

[0057] As a result, the final production schedule not only reflects the urgency of the power grid demand and the feasibility of the supply chain, but also strictly meets the process and resource constraints of enterprise production, ensuring the feasibility and delivery stability of the plan, achieving coordination and consistency between the power grid side, the supply chain side and the production side, and solving the coordination inconsistency problem that may be caused by system data silos.

[0058] Regarding the nonlinear coupling function in step S130, in some examples of embodiments of this application, the system platform uses "regional production scheduling task" as the decision granularity and comprehensively calculates the overall priority of the task by integrating four types of factors: predicted demand intensity, equipment reliability risk, delivery urgency, and supply chain status. Demand and risk act as amplification factors, delivery urgency as a suppression factor (the tighter the delivery date, the weaker the suppression), and supply chain status as an availability correction factor. After dimensionless / standardized processing, each factor is coupled nonlinearly to ensure that tasks with "strong demand, high risk, tight delivery date, and smooth supply chain" receive higher priority.

[0059] Specifically, it can be expressed by the following formula:

[0060] Equation (1)

[0061] In the formula, Indicates the first The comprehensive priority index for production scheduling tasks in each region. Indicates the first The equipment reliability risk index corresponding to the production scheduling task in each region ranges from (0,1). The larger the value, the older the equipment or the heavier the load, and the higher the risk. Indicates the first The supply chain status index for production scheduling tasks in each region ranges from (0,1) and reflects the raw material arrival rate and logistics support level. The larger the value, the smoother the supply chain. Indicates the first The predicted demand intensity corresponding to the production scheduling tasks in each region Indicates the first The predicted demand corresponding to the production scheduling tasks in each region. Indicates the baseline demand; Indicates the first The urgency of delivery deadlines for production tasks in a given region is defined as the ratio of remaining delivery time to the standard production cycle. For the first The risk enhancement function value corresponding to the production scheduling task in each region. For the first The supply chain elasticity function value corresponding to the production scheduling tasks in each region.

[0062] To adjust parameters, the strength of their impact on priority is configurable and can be adjusted according to enterprise strategy and power grid status. Specifically, historical data is used as a monitoring signal, and multi-target calibration is performed by comprehensively considering indicators such as "high-risk satisfaction rate, delay, plan hit rate, and reordering rate." In addition, to prevent short-term fluctuations, sliding windows and exponential smoothing can be used to perform temporal denoising and gradual control of priority.

[0063] It should be noted that high-risk areas (equipment aging, abnormal temperature rise, high load) should be prioritized, but linear processing may not highlight boundary risks. Therefore, nonlinear enhancement is used to amplify risks "above the overall level" and moderately weaken risks "below the overall level." Furthermore, tasks with material shortages or high logistical risks may remain idle even if scheduled, so nonlinear suppression is needed to quickly downgrade priorities while preserving recovery capabilities after supply chain improvements, reducing idle waiting and back-scheduling, and avoiding ineffective capacity utilization. Additionally, the more urgent the delivery, the more "resistance" needs to be reduced in the priority system. Urgency is described by the ratio of remaining delivery time to standard production cycle time; the smaller the ratio, the tighter the urgency.

[0064] Here, the actual demand for ring main units is driven by the load evolution, equipment health, and planned events on the grid side, while the manufacturing side's ability to deliver on schedule is also constrained by both supply chain and production line resources. Specifically, the three types of grid-side data are first mapped in the data space as calculable elements of regional production scheduling tasks (predicted demand and delivery time, equipment reliability risk index, and supply chain smoothness), and then transformed into decision variables on the same scale through dimensionless transformation. Subsequently, they are coupled in a non-linear manner: S-shaped reinforcement is used for equipment risks higher than the overall level to highlight the differences, and exponential suppression is used for supply chain disruptions to avoid infeasible loading. At the same time, delivery urgency is used as a suppression term and predicted demand intensity is used as an amplification term, thereby forming a comprehensive priority that takes into account both safety and feasibility.

[0065] Therefore, without changing the total capacity, the planning objective has shifted from passive orders to proactive tasks oriented towards power grid events: high-risk areas enter the production line earlier due to enhanced mapping, and urgent delivery tasks are prioritized due to suppression of urgency; tasks with supply chain disruptions are automatically downgraded or frozen, reducing waiting and rescheduling caused by material shortages.

[0066] In some examples of embodiments of this application, the risk enhancement function is expressed by the following formula:

[0067] Equation (2)

[0068] In the formula, This represents the average equipment reliability risk index for all regional production tasks awaiting scheduling. For steepness parameter, As the risk amplification coefficient, It is the hyperbolic tangent function.

[0069] It should be noted that the intensity of equipment risk varies with time and location, and the definition may also differ. Here, the risk amplification function is set as a hyperbolic tangent function centered on the mean risk of all tasks. This maps the "deviation from the mean" to a bounded, monotonic, and smoothly adjustable gain: when the risk of a task is higher than the mean, it receives an amplification of >1; when it is lower than the mean, it receives suppression of <1. The gain range is controlled by the risk amplification coefficient (upper and lower bounds are...). The slope is controlled by the kurtosis parameter (which determines the sensitivity to small deviations near the mean) and naturally saturates at extremely high / low risk levels to avoid unbounded amplification.

[0070] In this way, without relying on absolute thresholds, it can adaptively align with the risk baseline of the current task set. It is consistent with the multiplicative framework of priorities, can work in conjunction with factors such as supply chain and delivery time, and is easy to update smoothly online because the function is continuously differentiable.

[0071] After adopting this risk enhancement approach, tasks with risk higher than the overall level for the current period will have their overall priority steadily increased; tasks with risk significantly lower than the average will have their priority moderately decreased. This is because the gain is strictly limited to... Internally, the system saturates at extreme points, and priorities will not experience uncontrolled jumps due to individual anomalies. As a result, high-risk tasks receive relatively more available capacity and earlier start times, while the overall allocation of resources remains stable and controllable.

[0072] In some examples of embodiments of this application, the supply chain resilience function is expressed by the following formula:

[0073] Equation (3)

[0074] In the formula, These are parameters for adjusting supply chain sensitivity.

[0075] Here, whether regional production scheduling can be implemented on schedule depends primarily on the "feasibility" of material availability and logistical support. A multiplicative correction is applied to the supply chain status index, making it monotonically bounded and... The value is equal to 1; in addition, rapid compression is formed in the low smoothness range (by... (Controlling the slope) maintains a mild response in the medium to high smoothness range, avoiding ranking jumps caused by slight fluctuations; in addition, the function is continuously differentiable, which facilitates coordination with "capacity allocation - scheduling constraints" (capacity, delivery date, unique assignment), so that priority is only suppressed when there are constraints in the supply chain, and can smoothly rebound when the supply chain recovers.

[0076] Therefore, after adopting this flexible adjustment, the overall priority of tasks with supply chain disruptions will be promptly and controllably lowered, thus reducing the number of tasks entering the infeasible start-up queue during capacity allocation and loading. When the supply chain is smooth, the adjustment factor is close to 1, without affecting the original sorting logic. This reduces the probability of waiting and temporary rescheduling due to material shortages, improves the feasibility and stability of the plan, and concentrates limited capacity on tasks that can be implemented in the short term. Simultaneously, because the adjustment range is influenced by... Since the constraints and functions are bounded, the overall allocation will not fluctuate excessively, which is conducive to maintaining scheduling rhythm and resource balance.

[0077] Figure 2 A flowchart illustrating an example of obtaining a device reliability risk index according to an embodiment of this application is shown.

[0078] like Figure 2 As shown, in step S210, the task instruction area information of the regional production scheduling task is parsed, and the set of ring network cabinets corresponding to the task instruction area information is determined, and the ring network cabinet monitoring data of each ring network cabinet is obtained.

[0079] Here, the ring main unit monitoring data includes the cable joint hotspot temperature and ambient temperature at different sampling times. Risk assessments must correspond one-to-one with the actual power supply areas of the production schedule, and the overall status of multiple ring main units within that area should be used to represent the regional risk, avoiding distortion from single-point data.

[0080] Specifically, the regional identifiers of the production scheduling tasks are spatially linked with the power grid asset ledger to obtain a set of ring main units within that region (only ledger entries with online temperature monitoring or periodic inspection data are retained). A minimum sample size rule is set (e.g., at least two units with valid data), and the set size is recorded for subsequent quality assessment and reporting. If the same feeder spans multiple regions, it is deduplicated based on the task's regional boundary or by the primary user's affiliation.

[0081] In some implementations, two types of time series are synchronized from the data lake / monitoring platform: cable joint hotspot temperature and ambient temperature (sampling intervals, e.g., 1-15 minutes). Resampling is performed using a fixed-time raster; interpolation with a window upper limit is used for short-term measurements, and long-term missing measurements are directly removed from that time period.

[0082] In step S220, the thermal risk of each ring network cabinet in the ring network cabinet set is calculated, and the equipment reliability risk index corresponding to the regional production scheduling task is obtained by arithmetic averaging.

[0083] Regional production scheduling priorities should not be determined by individual anomalies; cross-counter aggregation should be used to obtain a robust regional risk representation.

[0084] Equation (4)

[0085] Equation (5)

[0086] In the formula, Indicates the first The set of ring network cabinets with available monitoring data within the geographical area indicated by the regional production scheduling task. For set The number of Zhonghuan network cabinets, Indicates the first The thermal risk of a single unit in the Taiwan Ring power distribution cabinet. These represent the weighting coefficients for the temperature difference term and the temperature rise rate term, respectively. ; and These represent the number of times the data was monitored at the sampling time. The hot spot temperature of the cable joints in the Taihuan network cabinet and the ambient temperature. Indicates the temperature difference reference threshold. Indicates the number of monitored The rate of temperature rise of the cable joint hotspots in the Taiwan Ring main unit, and This represents the baseline threshold for the rate of temperature rise.

[0087] It should be noted that the true risk at the regional level should be characterized by the combined operating thermal state of multiple ring main units within that region. The relative temperature difference between the hotspot temperature and the ambient temperature is used to characterize long-term thermal stress, while the rate of increase in hotspot temperature characterizes the failure evolution speed. Then, these two risk factors are weighted and combined to obtain the thermal risk of a single unit. Finally, the risk index is obtained by aggregating all monitorable ring main units in the same region, thereby mitigating the noise and occasional fluctuations of individual sensors through cross-unit averaging.

[0088] Here, for sets All valid cabinets inside Get the arithmetic mean To avoid frequent rearrangements caused by short-term fluctuations, one can also... Perform lightweight time smoothing (such as exponential smoothing between the previous and current periods). For regions with insufficient samples or substandard data quality, regress to conservative estimates of neighboring regions / historical means and label them with a "low confidence" tag to limit their amplification effect on priority. This results in... It is robust and interpretable, reflecting both the overall regional thermal risk and suppressing individual anomalies.

[0089] The regional risk index obtained through the embodiments of this application can stably reflect whether there is a persistently high temperature or rapid temperature rise in the ring main units within the region. This allows tasks in high-risk areas to be prioritized, while low-risk areas do not excessively consume resources, enabling production capacity to be more focused on tasks in areas that have greater mitigation value for grid security. Furthermore, since the index is derived from multi-unit aggregation and undergoes threshold normalization and boundary processing, the production scheduling is not overly sensitive to outliers, helping to reduce unnecessary plan fluctuations.

[0090] Figure 3 A flowchart illustrating an example of regional production demand forecasting according to an embodiment of this application is shown.

[0091] like Figure 3 As shown, in step S310, the multimodal features corresponding to the grid-side associated data are extracted. The multimodal features include equipment risk features, load time series features, and planning text features.

[0092] Here, the equipment risk characteristics include the equipment reliability risk index of the ring main unit corresponding to the task and its regional distribution standard deviation; the load time series characteristics include the time series characteristics of regional load forecast data in multi-scale time windows; and the planning text characteristics include the semantic characteristics of demand events corresponding to the regional power grid planning data.

[0093] Specifically, the risk characteristics of the equipment can be based on the risk index sequence of multiple ring network cabinets in the mission area, and statistical quantities such as mean, quantile, regional distribution standard deviation, rise / fall rate, and duration of exceeding threshold can be constructed. Time series labels such as the time interval from the most recent alarm to the present can also be retained to characterize the risk level and evolution speed.

[0094] Load time series characteristics can be obtained by multi-scale decomposition of regional load forecasts (such as hourly / daily / weekly windows) to extract features such as trends, seasonality, volatility, peak-to-valley difference, probability of peak occurrence, and month-on-month / year-on-year load growth. At the same time, the evolution trajectory of the load-existing capacity ratio can be generated to reflect whether the gap is approaching.

[0095] Planning text features can be extracted from regional power grid planning documents to identify event types (addition, renovation, decommissioning), expected commissioning / decommissioning times, target capacity, and constraints. Events are encoded as sparse sequences on a timeline, and a text encoder is used to obtain event semantic vectors, along with metadata such as "time elapsed" and "event confidence level." Furthermore, the three types of features are unified into a task time window grid (e.g., daily or weekly).

[0096] In step S320, the multimodal features are input into the demand prediction neural network model to determine the predicted demand and delivery time of the corresponding ring main unit.

[0097] Here, the demand forecasting neural network model includes: a cross-modal interactive attention layer, used to perform cross-attention among equipment risk features, load time-series features and planning text features to complete time alignment and weight allocation, and obtain a fused feature representation; and a multi-task forecasting output layer, used to output the predicted demand vector and predicted delivery deadline of the ring main unit based on the fused feature representation.

[0098] It should be noted that demand and delivery dates are often triggered by the "coincidence of events across multiple modalities in time," such as when the load is about to exceed capacity and equipment risk increases, or when a specific month is planned for commissioning. Modeling a single modality alone is insufficient to capture this coupling relationship; it is necessary to "mutually influence" the modalities on the timeline to achieve alignment and weight allocation.

[0099] Specifically, the cross-modal interactive attention layer uses the load sequence as the main time axis, introducing key / value sequences from equipment risk and planning semantics. It employs cross-attention to calculate the attention weight of each time step to other modalities. Masking ensures that attention is only focused on "effective windows before and after a given time point," avoiding the leakage of future information. When the confidence of a planning event is high, higher weight is given to windows with adjacent times. When the standard deviation of the risk distribution is large (significant differentiation within the region), the proportion of risk modalities in the attention is increased, emphasizing the impact of "local hotspots" on demand. The output of the cross-modal interactive attention layer is a fused feature representation, containing interpretable weights for which type of event is most critical to demand / delivery time at what time, which can be echoed back to the operations / planning side for verification.

[0100] As a result, the model can learn the nonlinear relationship of "event co-occurrence" and is no longer driven by single-modal noise; under the same training data, the fusion representation is more sensitive to demand peaks and delivery nodes, while maintaining smoothness and interpretability.

[0101] Furthermore, while demand and delivery deadlines are related, they have different objectives. A multi-task forecasting output layer can simultaneously predict and share an understanding of the same event, improving sample efficiency, but mutual interference must be avoided. Specifically, on top of the shared fusion representation, two lightweight branches are connected: the demand branch outputs a "demand vector for future rolling cycles" (by model or total quantity), employing a more robust regression head for skewed distributions and providing quantiles or confidence intervals as uncertainty indicators. The delivery deadline branch outputs suggested time points or time windows for demand fulfillment, such as regressing to the earliest fulfillment time. Preferably, an interpretable trigger factor label (such as "planned commissioning approaching," "load-capacity ratio exceeding limits," or "risk continuously exceeding thresholds") can also be output simultaneously.

[0102] It should also be noted that, compared to simple splicing or single-modal models, cross-modal interactive attention can establish time alignment and causal relationships between asynchronous and heterogeneous data sources, avoiding information dilution caused by missing data and inconsistent granularity. After sharing the backbone, setting up multi-task branches based on "demand / delivery date" allows for the reuse of understanding of the same event to improve sample efficiency and robustness, while also preventing mutual constraints through task decoupling. Simultaneously, attention weights can reflect key time periods and dominant factors, facilitating verification by operations and planning teams and providing interpretable and auditable evidence.

[0103] For training the demand forecasting neural network model, the demand quantity is supervised by summarizing historical installation / replacement / procurement records; the delivery date is aligned with the commissioning / substation project milestones and the actual material arrival / start-up time; and confidence weights are set for outlier samples to reduce the impact of noise. An adaptive loss weight mechanism is adopted, which balances the "demand quantity error" and "delivery date error" through uncertainty weighting or dynamic weights, avoiding the suppression of the gradient of one task by another.

[0104] In this embodiment, the reliability risks on the equipment side, the temporal evolution on the load side, and the construction milestones on the planning side are three types of data that are inherently inconsistent in granularity, form, and time alignment, but they collectively drive and influence the actual demand for ring main units. First, the risk, load, and planning texts are structured into comparable features. Then, cross-modal interactive attention is used to complete alignment and weight allocation on a unified time axis, enabling the model to identify when and by which factors trigger demand growth and the earliest feasible time to meet that demand. Subsequently, multi-task output simultaneously provides the predicted demand and predicted delivery deadline, forming an interpretable mapping from external power grid signals to production-schedulable indicators.

[0105] Figure 4 A flowchart illustrating an example of generating capacity allocation results for production scheduling tasks in various regions according to an embodiment of this application is shown.

[0106] like Figure 4 As shown, in step S410, the statistical characteristics of the comprehensive priority index of all regional production scheduling tasks are calculated.

[0107] Here, we select three types of distribution information corresponding to the comprehensive priority index: center, dispersion and skewness. The mean determines the "relative baseline", the standard deviation provides the "scale" and the skewness describes "whether the high-low end is unbalanced".

[0108] Equation (6)

[0109] Equation (7)

[0110] Equation (8)

[0111] In the formula, This indicates the total number of regional production tasks allocated within the current scheduling period. , and These represent the mean, standard deviation, and skewness of the comprehensive priority index for production scheduling tasks in all regions, respectively.

[0112] In step S420, a capacity allocation benchmark is established based on priority distribution characteristics.

[0113] Equation (9)

[0114] In the formula, Indicates the first The baseline value for capacity allocation of production tasks in each region. Indicates the first The comprehensive priority index for production scheduling tasks in each region. This represents the sum of the priorities of all tasks; For the available capacity of the production line, The tilt coefficient; For a sign function, when the independent variable is positive, zero, or negative, the corresponding sign function value takes the values ​​of +1, 0, or -1, respectively.

[0115] Here, the allocation consists of two parts: first, a fair baseline based on share ( First, the proportion of all task priorities; second, a directional tilt around the mean (those above the mean, and the further away from it, receive more allocation), the tilt strength is adjusted by distribution skewness and carries capacity units, consistent with the baseline dimension. Thus, under the premise of constant total available capacity, more capacity is stably allocated to tasks significantly above the average level; furthermore, when all... When the angles are very close, the tilt automatically weakens, resulting in smoother distribution and less fluctuation.

[0116] In step S430, a supply chain status index is introduced to constrain capacity elasticity.

[0117] It should be noted that even if the priority is very high, if the materials / logistics are not available in the short term, the machine will still be idle and rearranged. Therefore, feasibility filtering and flexible callback are introduced after allocation.

[0118] Equation (10)

[0119] In the formula, Indicates the first The production capacity of each region is subject to flexible constraints after supply chain limitations. The threshold for the supply chain status index. This is the supply chain state elasticity coefficient. This indicates taking the smaller of the two values.

[0120] As shown in equation (10), the supply chain smoothness is read for each task. .when Below the threshold At that time, compress proportionally ( Until frozen; within the feasible range, a mild elasticity coefficient (such as...) is then added. ,in (This is an exponential adjustment), meaning that the smoother the supply chain, the more moderate the acceleration, and the more disrupted the supply chain, the more significant the downsizing. As a result, tasks that are not feasible in the current supply chain will not occupy the current cycle's capacity, significantly reducing idle machines and temporary rescheduling; when the supply chain recovers, it can rebound smoothly, improving the stability of the plan.

[0121] In step S440, resource conservation processing is performed on the adjusted capacity to generate capacity allocation results for production scheduling tasks in each region.

[0122] After the elastic constraints are applied, the sum of the capacity of each task may deviate from the available capacity of the production line. It is necessary to normalize it once to satisfy the conservation of resources and keep the relative tilt relationship intact.

[0123] Equation (11)

[0124] In the formula, Indicates the first The final capacity allocation results for production scheduling tasks in each region. This represents the sum of the flexible constraint capacity of all regional production scheduling tasks.

[0125] As shown in equation (11), if If the capacity is exceeded, it will be directly adopted; if it exceeds the capacity, it will be scaled up to the total capacity, and non-negativity and upper limit checks will be performed (the minimum / maximum percentage can be set at the task level). The total amount is strictly constrained to not exceed the available capacity of the production line to ensure the feasibility of capacity allocation.

[0126] In this embodiment, the standard deviation-driven capacity adjustment uses the mean, standard deviation, and skewness of the current task set as statistical baselines. First, a fair baseline is given according to priority share. Then, a directional skew is applied based on the standardized deviation from the relative mean, where the skewness term is measured against total capacity and is constrained by skewness and a threshold to avoid extreme amplification. Subsequently, supply chain smoothness is introduced to flexibly compress or freeze the allocation results, ensuring that infeasible tasks do not occupy current capacity. Finally, normalization conservation is used to strictly converge the adjusted total capacity of each task to the available capacity.

[0127] Therefore, under the premise of total capacity constraints, capacity allocation will steadily tilt towards tasks that are "significantly higher than the current average and have a feasible supply chain". When the priority difference between tasks is small, the tilt will automatically weaken and the allocation will be smoother. For tasks with supply chain disruptions, the allocation will be compressed or postponed in a timely manner, reducing waiting and rescheduling after entering the scheduling process.

[0128] Figure 5 A flowchart illustrating an example of generating a production schedule based on capacity allocation results according to an embodiment of this application is shown.

[0129] like Figure 5 As shown, in step S510, a process chain constraint matrix is ​​defined based on process resource constraints. This process chain constraint matrix includes a lower bound matrix for process chain duration and a process resource dependency matrix.

[0130] Specifically, the lower bound matrix of process chain duration Process resource dependency matrix Available capacity of each production line resource pool within the planning cycle ;in, This indicates the total number of processes on a standard production line. Indicates the first The task in the first The minimum allowable processing time for each process. Indicates the first The first task Resource feasibility indication for each process step This indicates the number of production line resource pools that can be run in parallel. Indicates production line resource pool Total schedulable duration in this period.

[0131] It should be noted that the scheduling must simultaneously satisfy two types of hard constraints: "process sequence" and "resource feasibility." These two constraints are first fixed using a matrix approach and then used as boundary conditions for loading. The lower bound matrix for process chain duration is shown below. For the task process Provide the shortest processing time / safe cycle time to ensure necessary time isolation between upstream and downstream processes. Process-Resource Dependency Matrix Marking process Which type of resource pool can process it? Resource pool capabilities. Provide the parallel resource pools for each planning period. The total schedulable working hours are determined. Therefore, the lower bound of time, available equipment, and upper limit of capacity are fixed before loading begins, avoiding infeasible solution space searches and subsequent rollbacks during the scheduling process.

[0132] In step S520, the capacity allocation results are converted into a time base.

[0133] The allocation algorithm outputs capacity / output units, which need to be converted into process durations to be used in scheduling. The conversion is performed for each task and each process using the following formula:

[0134] Equation (12)

[0135] In the formula, Indicates the first The regional production scheduling task is in the first The baseline processing time for each process. Indicates the first Standard production efficiency (units / hour) for each process. Indicates the first Inspection time for each process step.

[0136] when When there is a temporary shortage or fluctuation, the system will revert to the historical percentile or engineering quota, and all parameters will be accompanied by a version number for auditing purposes.

[0137] In step S530, under the condition of satisfying the constraints, the tasks are loaded in descending order based on priority, and the scheduling variables of task-process-resource-time are solved. The constraints include pre-process constraints, resource capacity constraints, and delivery date constraints.

[0138] Specifically, under the condition of satisfying constraints, based on comprehensive priority Load tasks in descending order and solve for the scheduling variables of task-process-resource-time. These constraints include pre- and post-process constraints, resource capacity constraints, and delivery time constraints. Indicates the first The regional production scheduling task is in the first The start time of each process, Indicates the first The regional production scheduling task is in the first The end time of each process, Represents a resource assignment variable if and only if the process Available in resource pool Its value is 1 during the processing.

[0139] Here, under the premise of meeting the hard constraints of technology and resources, the overall priority is determined. The task loads in descending order, allowing high-priority tasks to have priority access to limited available capacity and minimizing conflicts and delays. Tasks are arranged from largest to smallest priority, and for tasks of the same priority, a parallel rule is adopted that prioritizes tasks with tighter delivery dates and tasks with easier resource fulfillment.

[0140] For example, for the task process Select the resource pool from its feasible resource set that has "sufficient remaining capacity and minimal penalty for delays and load imbalances". , place ,renew Reserves, and advance If no feasible solution is found that satisfies the capacity or delivery time, the task is marked as "rolled to the next cycle" or a local rollback and reallocation is triggered.

[0141] Thus generated Simultaneously satisfying process sequence, resource capacity, and delivery time; high-priority tasks obtain feasible windows earlier, and due to unique assignment and capacity constraints, there will be no resource overlap or overload at the execution level.

[0142] In this embodiment, for the capacity allocation results at the task granularity, the "feasible boundary" is first fixed by matrix-based process chain constraints (duration lower bound matrix and process-resource dependency matrix) and the available time of each resource pool. Then, the task-level capacity is converted into the process-level baseline processing time. Subsequently, loading is performed from high to low according to the comprehensive priority. Through constraints of process before and after, capacity, delivery date and unique assignment, the assignment of task-process-resource-time is solved, forming the output from capacity results to executable schedule.

[0143] In some examples of embodiments of this application, pre- and post-process constraints are defined as:

[0144] Equation (13)

[0145] In the formula, For the first The buffer coefficient of each process step.

[0146] Here, forward propagation is used to calculate the earliest start time, when the process... Once the actual start / end is determined, As a downstream process A hard lower bound is established and updated immediately. Precedence and subsequent constraints are continuously maintained during loading, preventing downstream processes from being scheduled in impossible time windows, reducing order insertions and rework due to insufficient preparation on-site; simultaneously, through... The buffer makes the plan less sensitive to small fluctuations, resulting in a more stable execution cycle. This ensures the process sequence and safety intervals within the same task, avoiding the defects of insufficient inspection preparation caused by adjacent processes "touching together".

[0147] Resource capacity constraints are defined as:

[0148] For any resource pool Execute capacity constraints and unique assignment constraint In the formula, And only when the process Available in resource pool Its value is 1 during the upper processing, thus the solution is obtained. , Indicates task process The assigned production line resource pool.

[0149] Here, a remaining capacity bucket is maintained for each resource pool, and each task loaded is deducted from the corresponding bucket. If the deduction is not negative, the resource / time period is not feasible. Resource feasibility is determined by a matrix. Control: Only Candidates are enumerated from the allowed pool; forced inclusion is prohibited. Unique assignment is performed using a binary variable. Implementation: When a pool is selected, this variable is set to 1, and other pools are set to 0. The strategy for selecting a pool is to satisfy the capacity requirement and achieve the best score (the score takes into account both late penalty and load balancing). In addition, if all candidate pools have no capacity, a partial rollback / pool switching is triggered or the task is marked as rolling to the next cycle to avoid forcing infeasible plans into the current cycle.

[0150] Therefore, the available time of the resource pool within the planning cycle is limited, and each process can only be implemented on one production line (or one resource pool) to avoid conflicts caused by multiple lines occupying the same process at the same time.

[0151] Delivery time constraints are defined as follows:

[0152] Equation (14)

[0153] Equation (15)

[0154] In the formula, For the task The predicted delivery time Indicates task The completion time of the final process.

[0155] Here, a schedule slack is introduced into the loading score: if a candidate resource leads to... Exceed If the delivery date is not met, additional penalties will be imposed or the task will be deemed infeasible. The delivery date will be converted into a hard cutoff for the final process to ensure that tasks that can be completed on time are loaded first; for tasks nearing their delivery date, their time window will be prioritized through loading order and resource selection.

[0156] In step S540, the production scheduling plan is output. A production schedule is a production execution plan recorded in matrix form, which includes the identifier of each region's production task, the planned start and end times of each process, the corresponding delivery deadlines, and the production line resource number assigned to each process.

[0157] .

[0158] In this way, the output production schedule naturally meets the three conditions and advantages of no overlap, no overload, and on-time completion. In addition, high-priority tasks obtain earlier time windows and clear resource assignments while meeting constraints; the load of the parallel resource pool is more balanced due to capacity constraints, reducing congestion and cross-line interference; the plan is presented in a matrix form of start / end time and resource number, which can be directly issued to MES or APS for execution, reducing the probability of secondary adjustments and temporary rescheduling, and keeping the production cycle in line with the grid demand.

[0159] To verify the effectiveness of the proposed PWCR (Priority-Weighted Capacity Regulation) method, a comparative experiment was designed to evaluate the performance of the proposed PWCR method compared with traditional scheduling methods in meeting grid demand and risk management. The experiment used a synthetic dataset to simulate the demand, risks, and capacity of ring main units in a regional power grid over a 12-month period, and performed scheduling simulations.

[0160] 1) Experimental Data and Setup

[0161] A rolling simulation was conducted based on 12 production cycles per year (calculated monthly). The predicted demand (units) for months 1-12 are as follows: 100, 120, 80, 150, 130, 110, 140, 160, 90, 100, 120, 130; the corresponding equipment risk indices (0-1) are: 0.9, 0.8, 0.4, 0.7, 0.6, 0.5, 0.8, 0.9, 0.3, 0.4, 0.7, 0.6; and the monthly available capacity (units) are: 120, 110, 100, 130, 120, 100, 110, 140, 100, 90, 110, 120. Supply chain smoothness is fixed at 0.8 (good) to isolate variables. Production costs and delivery urgency are weighted equally in this experiment and are not used as differentiating factors. Cross-month pre-production is not allowed: each month's capacity can only be used to process unfinished demand; unfinished demand is automatically carried over to the next month. If the total remaining demand of the current month's activity set is less than the available capacity for the current month, the remaining capacity is idle and cannot be used for future months (no cross-month pre-production).

[0162] 2) Comparison Method

[0163] Baseline (FIFO-By-Month): Loading is based solely on the order of arrival by month. A monthly "activity queue" is formed to represent unfulfilled demand from the previous month. This queue consumes the current month's capacity sequentially from morning to night until capacity is exhausted or demand is zero. Any remaining capacity is carried over to the next month. If there is still remaining capacity after clearing the current month's queue, that capacity remains idle.

[0164] PWCR (This paper): Employs risk-guided capacity allocation. At the beginning of each month, an "activity set" is constructed to represent the unmet demand for the month. For each task in the activity set, the risk-weighted demand weight is calculated as: (unmet demand for the month) × max{0,1 + 0.5 × (risk index for the month - mean risk of the activity set)}. If the sum of all weights is 0, the allocation reverts to a proportional (or equal) allocation. The monthly capacity is allocated to each task according to a weighted normalized ratio, with the minimum value between the weighted and unmet demand. The sum of the allocated amounts equals the monthly capacity (or is truncated by the upper limit of total remaining demand), and any unmet demand is carried over to the next month. This process does not change the total monthly capacity; it only involves a skewed allocation within the current period.

[0165] It should be noted that the common constraints of the two methods mentioned above include: maintaining monthly capacity, not pre-producing across months, and rolling over all unfulfilled work. The only difference lies in whether or not risk information is used to adjust the proportion of current capacity.

[0166] 3) Simulation process

[0167] Starting from month 1, the process is rolled over monthly: Monthly requirements are read and added to the activity set; monthly production allocation is generated based on the selected strategy, and the remaining unfinished quantities for each month are updated; the on-time completion quantity (i.e., the quantity completed for the current month's requirements) and carry-over quantity are recorded for the current month. After 12 months, on-time performance and other indicators for the entire year are calculated.

[0168] 4) Experimental Results

[0169] Figure 6 A simulation comparison diagram showing the on-time completion rate for each month is provided. Figure 6 As shown, the two curves are close in months with smaller backlogs in the early stages; after entering the middle of the year (around June to October), the baseline significantly squeezes the monthly capacity due to prioritizing the clearing of historical carryovers, resulting in a precipitous drop in the on-time completion volume for that month. In contrast, PWCR reallocates capacity within the current period according to risk-weighted proportions, maintaining higher and smoother monthly deliveries, especially in months with concentrated high risks (such as July and August), where its advantages are more obvious; towards the end of the year, affected by the end-of-period boundary and remaining demand, the two curves converge again, which overall reflects PWCR's ability to provide advance guarantees for high-priority tasks in the current month without changing the total monthly capacity.

[0170] Figure 7 A simulation comparison diagram of cumulative risk-weighted hourly output is shown. For example... Figure 7 As shown, the two methods were close in the first 3-4 months. From the 6th month onwards, the PWCR method was generally higher than the baseline FIFO method. In the high-risk window of July-August, the gap between the PWCR method and the reference curve narrowed significantly. This indicates that without changing the total monthly capacity, the PWCR method can achieve faster on-time delivery in high-risk months through risk-weighted proportional allocation. By the end of the year, both methods were lower than the reference curve (due to the annual capacity being less than the boundary of risk-weighted total demand). However, the PWCR method still maintained its lead, reflecting its priority guarantee of high-risk demand and higher risk-weighted on-time fulfillment rate.

[0171] Experimental results show that, under the constraints of the same monthly capacity and no cross-month pre-production, the PWCR method, compared to the baseline scheme of first-come-first-served by month, can continuously increase the delivery of high-risk months and keep the cumulative risk-weighted on-time production curve above the baseline for a long period, thus resulting in a higher risk-weighted schedule reliability (WSR) and a lower average cross-month delay. Therefore, the PWCR method demonstrates advantages in prioritizing high-risk demands and robust delivery in rolling simulations of synthetic power grid scenarios, making it suitable for production scheduling decisions where power grid load and equipment health fluctuate simultaneously.

[0172] Based on a survey of existing ring main unit (RNB) operation and maintenance monitoring, topology modeling, traditional production scheduling optimization, and energy data space construction, this paper proposes a regional power grid RNB production scheduling decision-making method based on the data space, addressing the current industry's neglect of grid demand and equipment reliability. Through multi-source data fusion, combined with comprehensive priority indicators and priority-weighted capacity adjustment algorithms, a deep coupling between production plans and grid operation demands is achieved. Experiments show that the proposed algorithm, while considering risks, can better ensure the timely delivery of high-risk equipment and is expected to play a significant role in promoting the intelligent upgrading of power equipment manufacturing through the energy data space.

[0173] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0174] Figure 8 A structural block diagram of an example of a scheduling decision system for a regional power grid ring main unit in a data space, according to an embodiment of this application, is shown.

[0175] like Figure 8 As shown, the production scheduling decision system 800 for regional power grid ring network cabinets in the data space includes a data acquisition unit 810, a demand forecasting unit 820, a priority calculation unit 830, a capacity allocation unit 840, and a production scheduling plan generation unit 850.

[0176] The data acquisition unit 810 is used to acquire the grid-side associated parameters, supply chain status parameters and enterprise-side production parameters corresponding to the production scheduling tasks of each region. The grid-side associated parameters include the equipment reliability risk index of the ring network cabinet corresponding to the task in the current regional power grid, regional load forecast data and regional power grid planning data. The enterprise-side production parameters include the available capacity of the production line and process resource constraints.

[0177] The demand forecasting unit 820 is used to generate the predicted demand and delivery time of the ring main unit for the corresponding regional production task based on the grid-side associated data corresponding to each regional production task.

[0178] The priority calculation unit 830 is used to calculate a comprehensive priority index for production scheduling tasks in each region based on the predicted demand, equipment reliability risk index, supply chain status, and delivery deadline corresponding to the production scheduling tasks in each region, through a nonlinear coupling function. The nonlinear coupling function includes a risk enhancement function and a supply chain resilience function. The risk enhancement function is used to nonlinearly correct the priority in response to deviations from the mean of the equipment reliability risk index, so as to increase the priority of risky tasks. The supply chain resilience function is used to correct the priority based on an exponential function of supply chain smoothness, so as to reduce the priority of tasks when the supply chain is not smooth.

[0179] The capacity allocation unit 840 is used to process the comprehensive priority index of the production scheduling tasks in each region using a standard deviation-driven capacity adjustment algorithm to generate capacity allocation results for the production scheduling tasks in each region.

[0180] The production scheduling unit 850 is used to generate a production scheduling plan based on the capacity allocation results and the process resource constraints. The production scheduling plan includes the scheduling time and corresponding scheduling capacity of each region's scheduling tasks.

[0181] In some embodiments, this application provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions. The execution instructions can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of the scheduling decision method for regional power grid ring network cabinets in any of the data spaces described above in this application.

[0182] In some embodiments, this application also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of the scheduling decision method for regional power grid ring network cabinets in any of the above data spaces.

[0183] In some embodiments, this application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of a scheduling decision method for regional power grid ring network cabinets in a data space.

[0184] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0185] The electronic devices in this application can exist in various forms, including but not limited to: mobile communication devices, ultra-mobile personal computer devices, portable entertainment devices, or other airborne electronic devices with data interaction functions.

[0186] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. 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 embodiment according to actual needs.

[0187] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A production scheduling decision-making method for regional power grid ring network cabinets in a data space, characterized in that, The method includes: Obtain the grid-side associated parameters, supply chain status parameters, and enterprise-side production parameters corresponding to the production scheduling tasks in each region. The grid-side associated parameters include the equipment reliability risk index of the ring network cabinet corresponding to the task in the current regional power grid, regional load forecast data, and regional power grid planning data. The enterprise-side production parameters include the available capacity of the production line and process resource constraints. Based on the grid-side associated data corresponding to each regional production scheduling task, the predicted demand and delivery time of the corresponding regional production scheduling task for ring main units are generated. For production scheduling tasks in each region, a comprehensive priority index is calculated using a nonlinear coupling function based on the predicted demand, equipment reliability risk index, supply chain status, and delivery deadline corresponding to the regional production scheduling tasks. The nonlinear coupling function includes a risk enhancement function and a supply chain resilience function. The risk enhancement function is used to nonlinearly correct the priority in response to deviations from the mean of the equipment reliability risk index, thereby increasing the priority of risky tasks. The supply chain resilience function is used to correct the priority based on an exponential function of supply chain smoothness, thereby reducing the priority of tasks when the supply chain is not smooth. A standard deviation-driven capacity adjustment algorithm is used to process the comprehensive priority index of production scheduling tasks in each region in order to generate capacity allocation results for production scheduling tasks in each region. Based on the capacity allocation results and the process resource constraints, a production schedule is generated, which includes the production time and corresponding production capacity for each region's production tasks. The nonlinear coupling function is expressed by the following equation: ， In the formula, Indicates the first The comprehensive priority index for production scheduling tasks in each region. Indicates the first The equipment reliability risk index corresponding to the production scheduling task in each region ranges from (0,1). The larger the value, the older the equipment or the heavier the load, and the higher the risk. Indicates the first The supply chain status index for production scheduling tasks in each region ranges from (0,1) and reflects the raw material arrival rate and logistics support level. The larger the value, the smoother the supply chain. Indicates the first The predicted demand intensity corresponding to the production scheduling tasks in each region Indicates the first The predicted demand corresponding to the production scheduling tasks in each region. Indicates the baseline demand; Indicates the first The urgency of delivery deadlines for production tasks in a given region is defined as the ratio of remaining delivery time to the standard production cycle. For the first The risk enhancement function value corresponding to the production scheduling task in each region. For the first The supply chain elasticity function value corresponding to the production scheduling tasks in each region. To adjust the parameters; The risk enhancement function is expressed by the following formula: ， In the formula, This represents the average equipment reliability risk index for all regional production tasks awaiting scheduling. For steepness parameter, As the risk amplification coefficient, It is the hyperbolic tangent function; The supply chain resilience function is expressed by the following formula: ， In the formula, For supply chain sensitivity adjustment parameters; The acquisition of the equipment reliability risk index includes: The task instruction location information of the regional production scheduling task is analyzed, and the set of ring main units corresponding to the task instruction location information is determined. The ring main unit monitoring data of each ring main unit is obtained. The ring main unit monitoring data includes the cable joint hot spot temperature and ambient temperature at different sampling times. Calculate the thermal risk of each ring main unit in the ring main unit set, and obtain the equipment reliability risk index corresponding to the regional production scheduling task by arithmetic averaging: ， ， In the formula, Indicates the first The set of ring network cabinets with available monitoring data within the geographical area indicated by the regional production scheduling task. For set The number of Zhonghuan network cabinets, Indicates the first The thermal risk of a single unit in the Taiwan Ring power distribution cabinet. These represent the weighting coefficients for the temperature difference term and the temperature rise rate term, respectively. ; and These represent the number of times the data was monitored at the sampling time. The hot spot temperature of the cable joints in the Taihuan network cabinet and the ambient temperature. Indicates the temperature difference reference threshold. Indicates the number of monitored The rate of temperature rise of the cable joint hotspots in the Taiwan Ring main unit, and Indicates the reference threshold for the rate of temperature rise; The standard deviation-driven capacity adjustment algorithm processes the comprehensive priority index of production scheduling tasks in each region to generate capacity allocation results for each region's production scheduling tasks, including: Calculate the statistical characteristics of the comprehensive priority index for production scheduling tasks across all regions: ， ， ， In the formula, This indicates the total number of regional production tasks allocated within the current scheduling period. , and These represent the mean, standard deviation, and skewness of the comprehensive priority index for production scheduling tasks across all regions, respectively. Establish a capacity allocation benchmark based on priority distribution characteristics: ， In the formula, Indicates the first The baseline value for capacity allocation of production tasks in each region. Indicates the first The comprehensive priority index for production scheduling tasks in each region. This represents the sum of the priorities of all tasks; For the available capacity of the production line, The tilt coefficient; For a sign function, when the independent variable is positive, zero, or negative, the corresponding sign function value takes the values ​​of +1, 0, or -1, respectively. Introducing a supply chain status index to constrain capacity elasticity: ， In the formula, Indicates the first The production capacity of each region is subject to flexible constraints after supply chain limitations. The threshold for the supply chain status index. This is the supply chain state elasticity coefficient. This indicates taking the smaller of the two values; The adjusted production capacity is then processed to ensure resource conservation, resulting in a capacity allocation for production tasks in each region. ， In the formula, Indicates the first The final capacity allocation results for production scheduling tasks in each region. This represents the sum of the flexible constraint capacity of all regional production scheduling tasks.

2. The method according to claim 1, characterized in that, The process of generating the predicted demand and delivery deadline for ring main units for each regional production task based on the grid-side associated data corresponding to each regional production task includes: Extract the multimodal features corresponding to the grid-side associated data. The multimodal features include equipment risk features, load time series features, and planning text features. The equipment risk features include the equipment reliability risk index of the ring main unit corresponding to the task and its regional distribution standard deviation. The load time series features include the time series features of regional load forecast data in multi-scale time windows. The planning text features include the demand event semantic features corresponding to the regional power grid planning data. The multimodal features are input into the demand prediction neural network model to determine the predicted demand and delivery time of the corresponding ring main unit. The demand prediction neural network model includes: A cross-modal interactive attention layer is used to perform cross-attention among the device risk features, the load time series features, and the planning text features to complete time alignment and weight allocation, and obtain a fused feature representation; The multi-task prediction output layer is used to output the predicted demand vector and predicted delivery time of the ring main unit based on the fused feature representation.

3. The method according to claim 1, characterized in that, The step of generating a production schedule based on the capacity allocation results and the process resource constraints includes: Based on the process resource constraints, a process chain constraint matrix is ​​defined, which includes a lower bound matrix for process chain duration. Process resource dependency matrix Available capacity of each production line resource pool within the planning cycle ;in, This indicates the total number of processes on a standard production line. Indicates the first The task in the first The minimum allowable processing time for each process. Indicates the first The first task Resource feasibility indication for each process step This indicates the number of production line resource pools that can be run in parallel. Indicates production line resource pool The total schedulable duration in this period; Convert capacity allocation results to a time base: ， In the formula, Indicates the first The regional production scheduling task is in the first The baseline processing time for each process. Indicates the first Standard production efficiency for each process Indicates the first Inspection time for each process step; Under the condition of satisfying the constraints, based on Load tasks in descending order and solve for the scheduling variables of task-process-resource-time. The constraints include pre- and post-process constraints, resource capacity constraints, and delivery date constraints. Indicates the first The regional production scheduling task is in the first The start time of each process, Indicates the first The regional production scheduling task is in the first The end time of each process, Represents a resource assignment variable if and only if the process Available in resource pool Its value is 1 during the processing.

4. The method according to claim 3, characterized in that, The pre- and post-process constraints are defined as follows: ， In the formula, For the first The buffer coefficient of each process step; The resource capacity constraint is defined as: For any resource pool Execute capacity constraints and unique assignment constraint In the formula, And only when the process Available in resource pool Its value is 1 during the upper processing, thus the solution is obtained. , Indicates task process The assigned production line resource pool; The delivery date constraint is defined as follows: ， ， In the formula, For the task The predicted delivery time Indicates task The completion time of the final process; Output production scheduling plan; the production scheduling plan refers to the production execution plan recorded in matrix form, which includes the identifier of each region's production scheduling task, the planned start and end times of each process, the corresponding delivery deadline, and the production line resource number assigned to each process.

5. A production scheduling decision-making system for regional power grid ring network cabinets in a data space, used to implement the method as described in any one of claims 1-4; characterized in that, The system includes: The data acquisition unit is used to acquire grid-side associated parameters, supply chain status parameters, and enterprise-side production parameters corresponding to production scheduling tasks in each region. The grid-side associated parameters include the equipment reliability risk index of the ring network cabinet corresponding to the task in the current regional power grid, regional load forecast data, and regional power grid planning data. The enterprise-side production parameters include the available capacity of the production line and process resource constraints. The demand forecasting unit is used to generate the predicted demand and delivery time of the ring main unit for the corresponding regional production tasks based on the grid-side associated data corresponding to the production tasks in each region. The priority calculation unit is used to calculate a comprehensive priority index for production scheduling tasks in each region, based on the predicted demand, equipment reliability risk index, supply chain status, and delivery deadline corresponding to the regional production scheduling tasks, through a nonlinear coupling function. The nonlinear coupling function includes a risk enhancement function and a supply chain resilience function. The risk enhancement function is used to nonlinearly correct the priority in response to deviations from the mean of the equipment reliability risk index, thereby increasing the priority of risky tasks. The supply chain resilience function is used to correct the priority based on an exponential function of supply chain smoothness, thereby reducing the priority of tasks when the supply chain is not smooth. The capacity allocation unit is used to process the comprehensive priority index of production scheduling tasks in each region using a standard deviation-driven capacity adjustment algorithm to generate capacity allocation results for production scheduling tasks in each region. The production scheduling unit is used to generate a production scheduling plan based on the capacity allocation results and the process resource constraints. The production scheduling plan includes the production scheduling time and corresponding production capacity of each region's production scheduling tasks.

Images