A Power Green Supply Chain Scheduling Optimization Method Based on Multimodal Data Fusion

Abstract

This application relates to the field of multimodal data-driven power dispatching technology, specifically to a power green supply chain dispatching optimization method based on multimodal data fusion. This method includes: acquiring repair work orders for each piece of equipment, all types of power data, and all types of meteorological environmental data during the power green supply chain dispatching process; using a multimodal difference coefficient to filter matching cycles as historical references, and weighted evaluation of equipment normality to filter faulty equipment; and using an ant colony algorithm, considering the total distance traveled by each power emergency repair vehicle during the power green supply chain dispatching process, total pollution emissions, and time urgency costs, to plan the optimal travel path for the power emergency repair vehicles. This application solves the problem of dispatching optimization difficulties caused by the lack of multimodal data, and improves the dispatching quality and overall efficiency of the green supply chain through fault identification and path planning.

Description

Technical Field

[0001] This application relates to the field of multimodal data power dispatching technology, specifically to a power green supply chain dispatching optimization method based on multimodal data fusion. Background Technology

[0002] Against the backdrop of a global push for energy transition strategies, the power system, as the core carrier of this transition, is accelerating the construction of a new power system dominated by new energy sources. Green supply chain management has become a key path to achieving low-carbon and sustainable development in the power industry, covering the entire lifecycle from raw material extraction, equipment manufacturing, logistics and transportation to engineering construction, operation and maintenance, and decommissioning and recycling. In this process, massive, multi-source, and heterogeneous big data is continuously generated, providing a data foundation for accurate carbon footprint tracking, green supplier assessment, equipment health monitoring, and intelligent dispatch optimization. However, this data is collected by different entities at different times, locations, and business contexts, exhibiting highly fragmented and asymmetrical characteristics, urgently requiring the release of its collaborative value through advanced multimodal fusion technologies.

[0003] Currently, big data analysis in the green power supply chain faces a core bottleneck: the semantic alignment of multimodal data and the lack of multimodal data hinder the improvement of scheduling optimization quality. Specifically, various types of data differ significantly in temporal granularity and spatial reference systems. Furthermore, different modalities of data also exhibit certain deficiencies or semantic discrepancies. These factors lead existing fusion methods to often be limited to single-objective or shallow splicing, making it difficult to accurately identify fault modes and quantify node anomalies in scenarios with missing or distorted work orders. Consequently, scheduling optimization tends to favor local efficiency or single objectives, failing to support collaborative decision-making across the entire supply chain and multiple objectives. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a power green supply chain scheduling optimization method based on multimodal data fusion to solve existing problems.

[0005] The power green supply chain scheduling optimization method based on multimodal data fusion proposed in this application adopts the following technical solution:

[0006] Acquire repair work orders for each piece of equipment, all types of power data, and all types of meteorological and environmental data during the green power supply chain dispatch process;

[0007] The differences in power data, meteorological data, and text of repair orders for each device in the current period compared to any historical period are analyzed separately to determine the difference coefficient between the current period and any historical period for each device. This allows for the selection of matching periods for the current period of each device from historical periods, which can then be used as historical reference standards to determine whether each device is in a fault state in the current period.

[0008] The difference coefficient between each device in each cycle and its matching cycle is used as the weight of the fault severity of each device in any matching cycle of each cycle, so as to evaluate the normality of each device in the current cycle and to screen out faulty devices from all devices.

[0009] Using the ant colony algorithm, the optimal travel path of each power repair vehicle is planned, taking into account the total distance traveled, total pollution emissions, and time urgency cost during the power green supply chain scheduling process.

[0010] Preferably, the historical period includes all periods prior to the current period of each device, as well as periods of devices belonging to the same category as each device prior to the current period.

[0011] Preferably, the method for determining the difference coefficient of each device between the current period and any historical period is as follows:

[0012] Based on the differences in power data and meteorological environment data of each device in the current period compared to any historical period, the power fluctuation of each device in the current period is determined.

[0013] The difference coefficient between each device in the current period and any historical period is positively correlated with the difference in power fluctuation of each device in the current period compared to the power fluctuation in any historical period, as well as the text difference of the repair work order.

[0014] Preferably, the power fluctuation of each device in the current period is positively correlated with the differences in power data and meteorological environment data.

[0015] The preferred method for obtaining differences in power data and meteorological environmental data is as follows:

[0016] Calculate the sum of differences between each device in the current period and any historical period across all types of power data, and the sum of differences between each device in all types of meteorological and environmental data, respectively, and record them as the power difference index and environmental difference index of each device between the current period and any historical period.

[0017] The sum of the power difference index and the sum of the environmental difference index for each device between the current cycle and all historical cycles are denoted as power data difference and meteorological environment data difference, respectively.

[0018] Preferably, the step of filtering the matching period for the current period of each device from historical periods includes:

[0019] The difference coefficient between each device in the current period and all historical periods is used as the input of the threshold segmentation algorithm, and the output is the segmentation threshold. The historical periods with a difference coefficient less than or equal to the segmentation threshold are used as the matching period for the current period of each device.

[0020] Preferably, the method for evaluating the normality of each device in the current period is as follows:

[0021] The result of weighted fusion of the difference coefficient between each device and all its matching cycles in each cycle and the degree of fault severity is denoted as the abnormal matching factor of each device in each cycle.

[0022] The difference between the abnormal matching factor corresponding to each device in the current period and the average abnormal matching factor in all matching periods of the current period is taken as the normality of each device in the current period.

[0023] Preferably, the step of screening out faulty devices from all devices includes:

[0024] The normality of all devices in the current period is used as the input of the threshold segmentation algorithm. The output segmentation threshold is recorded as the normal threshold. Devices with a normality less than the normal threshold are recorded as faulty devices.

[0025] Preferably, the optimal driving route for the planned power emergency repair vehicle includes:

[0026] Based on the total distance traveled, total pollution emissions, and time urgency cost of each power repair vehicle during the power green supply chain dispatch process, the objective function of the ant colony algorithm is determined.

[0027] The ant colony algorithm is used to determine the optimal travel path for each power repair vehicle when the objective function is minimized.

[0028] Preferably, the expression for the objective function F is: In the formula, These represent the total distance traveled, total pollution emissions, and time urgency cost of the e-th power repair vehicle in the current cycle during the power green supply chain dispatch process; E represents the total number of power repair vehicles participating in the dispatch process.

[0029] One embodiment of this application provides a power green supply chain scheduling optimization method based on multimodal data fusion, the method comprising the following steps:

[0030] This application has at least the following beneficial effects:

[0031] This application calculates the difference coefficient by analyzing power, meteorological, and textual data, and accurately selects historical matching cycles as reference benchmarks. This effectively solves the bottleneck of difficulty in accurately identifying fault modes and quantifying anomalies in scenarios with missing or distorted work orders, significantly improving the quality of multimodal data alignment and fusion, and providing reliable support for subsequent global collaborative scheduling. Furthermore, this application uses the difference coefficient as a weight to evaluate equipment normality, accurately selecting faulty equipment whose state characteristics highly match those of historical typical faults. This effectively overcomes the impact of missing or distorted work orders, realizing the transformation of fault identification from manual serial decision-making to data-driven parallel collaborative decision-making, which helps improve the accuracy of global scheduling optimization. Finally, this application uses the ant colony algorithm to construct an objective function that integrates total distance, total pollution emissions, and weighted time urgency costs, optimizing the planning of the optimal travel path for power repair vehicles. This achieves a balance between low-carbon environmental protection, transportation timeliness, and global collaboration in green supply chain scheduling, improving the scheduling quality and comprehensive benefits of the green power supply chain. Attached Figure Description

[0032] To more clearly illustrate the technical solutions and advantages 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 only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A flowchart illustrating the steps of a power green supply chain scheduling optimization method based on multimodal data fusion, as provided in one embodiment of this application;

[0034] Figure 2 A flowchart of a matching period filtering process provided in one embodiment of this application. Detailed Implementation

[0035] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the power green supply chain scheduling optimization method based on multimodal data fusion proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0037] The following description, in conjunction with the accompanying drawings, details the specific scheme of the power green supply chain scheduling optimization method based on multimodal data fusion provided in this application.

[0038] This application provides an embodiment of a power green supply chain scheduling optimization method based on multimodal data fusion. Specifically, the method is described below. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0039] Step S1: Obtain the repair work order for each piece of equipment, all types of power data, and all types of meteorological and environmental data during the green power supply chain scheduling process.

[0040] In the process of optimizing the scheduling of the green power supply chain, the main approach is to analyze the changes in three types of data to optimize the scheduling process. These three types of data are power system data, meteorological environmental data, and maintenance work order data. Among them, power operation data and maintenance work order data are used to measure abnormal conditions of the power system to determine whether the green supply chain needs to be scheduled. Meteorological environmental data affects the diagnosis of power system anomalies and the scheduling and transportation process.

[0041] First, the system acquires repair work orders for each piece of equipment, all types of power data, and all types of meteorological environmental data during the power green supply chain dispatch process. The power data includes current and voltage, while the meteorological environmental data includes wind speed, temperature, and humidity. The acquisition frequency for all of these data is set to f, and the length of each cycle is T seconds. In this embodiment, the acquisition frequency and cycle length are 50Hz and 1h, respectively. In practical applications, as other implementation methods, implementers can set their own values ​​according to specific circumstances; this embodiment does not impose any special restrictions. For text data such as repair work orders, the system uses a keyword dictionary and jieba word segmentation to perform sentence segmentation, word segmentation, part-of-speech tagging, and stop word deletion on the obtained repair work order text. The system also classifies fault types and equipment types according to the power keyword dictionary.

[0042] Furthermore, the collected data, excluding repair work orders, are normalized. In this embodiment, the maximum-minimum value normalization method is used to normalize each collected data item. In practical applications, as other implementation methods, implementers may also use other normalization methods depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of normalization methods.

[0043] The processes of using keyword dictionaries and jieba tools to segment repair work orders into sentences, words, parts-of-speech tags, and stop word deletion for fault types and equipment types, as well as the process of normalizing data using the maximum-minimum value normalization method, are all well-known technologies and will not be elaborated further.

[0044] Step S2: Calculate the multimodal difference coefficient to screen the matching period as a historical reference, and weight the evaluation of equipment normality to screen faulty equipment.

[0045] The green power supply chain spans the entire lifecycle of power assets, participating in both the construction and operation phases, such as fault response and equipment replacement. While multimodal data exists within this supply chain, the differences in trends and semantic descriptions make data alignment difficult. Furthermore, accurately identifying fault modes and anomalies at different locations within the power system is challenging in scenarios with missing or distorted data. This identification error leads intelligent systems to focus on local efficiency or single objectives when response times for anomaly handling are critical, hindering collaborative decision-making across the entire supply chain and multiple objectives. Therefore, optimizing the fusion quality of the total multimodal data in the green supply chain scheduling process is necessary. It should be noted that this method is primarily applied to scenarios where power repair work orders are missing or vaguely described, such as in remote mountainous areas. In these cases, work orders alone are insufficient for accurate scheduling optimization. It is necessary to calculate multimodal difference coefficients, filter matching periods as historical references, and weightedly assess equipment normality to accurately identify faulty equipment and adjust the scheduling process. The specific process is as follows:

[0046] S2.1 Analyze the differences in power data, meteorological environment data, and repair work order text for each device in the current period compared to any historical period, and determine the difference coefficient between each device in the current period and any historical period, so as to select the matching period for each device in the current period from the historical periods.

[0047] In the operation and maintenance of power systems, various modalities of data are generated, such as work order data and power data. Work order data serves as a bridge between the power system and the green supply chain. Based on the fault diagnosis and description of power equipment damage in the work order data, the green supply chain facilitates the transportation and allocation of corresponding equipment. However, in remote areas of the power system, this work order data is not readily available; power data is often the first thing obtained. Furthermore, the content of the work order data is related to the description provided by the reporting personnel, which can lead to different descriptions for the same type of fault. These factors all affect the alignment quality of the data. Therefore, to improve the alignment quality between different modalities of data, it is necessary to first analyze the characteristics and correlations between each modality.

[0048] To address the issues of missing repair work order data and text discrepancies, since there are certain similarities between the power data fluctuations corresponding to each type of fault, different types of faults can be distinguished based on the similarity of power data changes, and the specific fault type can be determined based on the distinction results. However, because the environmental conditions at each fault location are inconsistent, these differences can cause disturbances to the detection results, thus affecting the quality of fault distinction based on the similarity of power data changes. Therefore, it is also necessary to consider the influence of environmental factors and mitigate the corresponding disturbances.

[0049] Based on the above analysis, this embodiment analyzes the differences in power data, meteorological environmental data, and textual differences in repair work orders for each device in the current cycle compared to any historical cycle. This determines the difference coefficient between the current cycle and any historical cycle for each device, allowing for the selection of matching cycles for the current cycle from historical cycles. This enhances the matching quality based on multimodal data and improves the accuracy of correlation measurement between data from different cycles. The specific process is as follows:

[0050] First, this embodiment determines the power fluctuation of each device in the current period based on the differences in power data and meteorological environment data of each device in the current period compared to any historical period;

[0051] Furthermore, based on the differences between the power fluctuation and the text of the repair work order, a difference coefficient is determined. Specifically, the difference coefficient of each device between the current period and any historical period is positively correlated with the difference between the power fluctuation of each device in the current period and the power fluctuation in any historical period, as well as the text difference of the repair work order.

[0052] Preferably, as one implementation method, in this embodiment, the difference coefficient between the p-th device in the current period and the historical period i is... The expression is: In the formula, This represents the textual difference in the repair work order for the p-th device between the current cycle and the historical cycle i; This represents the difference in power fluctuation of the p-th device in the current period compared to the power fluctuation in the historical period i; k represents the preset smoothing factor, and exp() represents the exponential function with the natural constant as the base.

[0053] Specifically, if the current repair work order is missing, then... This option is set to 0 to disregard the impact of differences in repair work orders.

[0054] It should be noted that the preset smoothing factor is set manually. In this embodiment, the preset smoothing factor is set to 0.2, which aims to optimize the decay rate of the exponential function in the calculation of the difference coefficient. This ensures that when a repair work order exists, the text difference can effectively regulate the difference in power fluctuation rather than completely dominate it. This prevents the difference coefficient from drastically fluctuating due to minor semantic fluctuations in the work order text, thereby maintaining the stability and distinguishability of the matching degree assessment during different historical periods, and making the selected matching period more representative and robust.

[0055] The specific process for obtaining power fluctuation data and the specific process for obtaining textual differences in repair work orders are as follows:

[0056] (1) Based on the differences in power data and meteorological environment data of each device in the current period compared to any historical period, determine the power fluctuation of each device in the current period, specifically:

[0057] The power fluctuation of each device in the current period is positively correlated with the differences in power data and meteorological environment data.

[0058] Preferably, as one implementation method, in this embodiment, the power fluctuation of the p-th device in the current period is... The expression is: In the formula, These represent the differences in power data and meteorological environment data of the p-th device in the current period compared to the historical period i, respectively. This represents the sum of the differences in power data for the p-th device in the current period compared to all historical periods; These represent all types of power data and corresponding standard power data for the p-th device in the current period, respectively. The standard power data are values ​​preset based on experimental data. This represents a preset weight, with a value ranging from 1.5 to 2.5. In this embodiment, it is set to 2. The purpose is to apply a penalty weight to cases that deviate from standard power data, i.e., cases with low power fluctuations, especially for... Below In severe fault scenarios such as power outages or voltage drops, the difference value is amplified. This ensures that such faults can still generate high power fluctuations, thereby avoiding misjudging serious faults as normal states, ensuring the accuracy of fault identification and the safety of the system. In practical applications, implementers can also set their own settings according to specific circumstances, and this embodiment does not impose any special restrictions.

[0059] It should be noted that the methods for obtaining the power data difference and environmental difference indices are as follows: calculate the sum of the differences of each device in all types of power data and the sum of the differences in all types of meteorological and environmental data between the current period and any historical period, respectively, and record them as the power difference index and environmental difference index of each device between the current period and any historical period.

[0060] The sum of the power difference index and the sum of the environmental difference index for each device between the current period and all historical periods are denoted as power data difference and meteorological environmental data difference, respectively. It should be noted that in this embodiment, the DTW distance of each device in the current period compared to any historical period in various types of power data is used as the difference between each device in the current period and any historical period in various types of power data. Similarly, the meteorological environmental data difference is also calculated using the DTW distance method. In practical applications, as other implementation methods, implementers may also use other methods such as Euclidean distance to measure the difference between data groups, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods for measuring the difference between data groups.

[0061] The calculation process for the DTW distance is a well-known technique and will not be elaborated further.

[0062] (2) The calculation process of the text difference of the repair work order is as follows: the text difference coefficient of the p-th equipment for the c-th fault type in the current cycle and the historical cycle i. The expression is: In the formula, This represents the intersection and union ratio of the repair work order of the p-th device under the c-th fault type in the current cycle with respect to the standard repair work order of the p-th device under the c-th fault type, with respect to the n-th keyword. This represents the intersection and union ratio of the repair work order of the p-th device under the c-th fault type in historical period j with respect to the n-th keyword; N represents the total number of keyword types, where keywords are divided into three categories: fault keywords, equipment keywords, and residual keywords. Residual keywords are keywords other than fault keywords and equipment keywords.

[0063] The average of the text difference coefficients for all fault types for each device in the current cycle and any historical cycle is used as the text difference of the repair work order for each device between the current cycle and any historical cycle.

[0064] It should be noted that there are many methods to measure the difference between data. In this embodiment, the absolute value of the difference between the power fluctuation of the p-th device in the current period and the power fluctuation in the historical period i is taken as the difference between the power fluctuation of the p-th device in the current period and the power fluctuation in the historical period i. In practical applications, as other implementation methods, implementers may also use other methods such as the square of the difference to measure the difference between data, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods for measuring the difference between data.

[0065] Based on the difference coefficient, it can be understood that the difference coefficient is used to characterize the matching quality of multimodal data of equipment in different periods. It reflects the overall deviation of the current period from the historical period, and its value directly determines whether the historical period can be selected as the reference standard for judging the current state. The calculation logic of the difference coefficient is to first obtain the power fluctuation difference based on the DTW distance between the power data and the meteorological environment, and then combine the semantic difference of the repair work order text, and use an exponential function to dynamically fuse the text difference and the power fluctuation difference. The calculation of the difference coefficient is mainly affected by two factors: the difference in the repair work order text and the difference in power fluctuation. The larger these factors are, the more inconsistent the current period is with the historical period in terms of text description or power fluctuation pattern, resulting in a larger difference coefficient. This reflects that the current state is not similar to the historical state, and the historical period cannot be used as a matching period. Conversely, the smaller these factors are, the more consistent the current period is with the historical period in terms of fault characteristics and environmental background, resulting in a smaller difference coefficient. This reflects that the current state is very close to the historical state, and the historical period will be selected as a matching period to provide a historical reference for fault judgment.

[0066] Furthermore, in this embodiment, the difference coefficient between each device in the current period and all historical periods is used as the input of the threshold segmentation algorithm, and the segmentation threshold is output. The historical periods with difference coefficients less than or equal to the segmentation threshold are used as the matching period of each device in the current period. The matching period is used to characterize the set of historical fault or abnormal records that are highly similar to the device in the current period in terms of multimodal characteristics (including power fluctuations, environmental background and text description) selected from massive historical data, and serves as the benchmark anchor point for judging the device status in the current period.

[0067] Preferably, the flowchart of the matching period filtering process provided in this embodiment is as follows: Figure 2 As shown.

[0068] It should be noted that there are many commonly used threshold segmentation algorithms. In this embodiment, the Otsu threshold segmentation algorithm is used to divide the historical period. In actual applications, as other implementation methods, implementers may also use other threshold segmentation algorithms according to specific circumstances. This embodiment does not impose any special restrictions.

[0069] Among them, the Otsu threshold segmentation algorithm is a well-known technology, and its specific principles and processes will not be elaborated here.

[0070] Thus, this embodiment, by analyzing power, meteorological, and text data to calculate the difference coefficient, accurately selects historical matching cycles as reference benchmarks, effectively solving the bottlenecks of difficulty in accurately identifying fault modes and quantifying anomalies in scenarios with missing or distorted work orders. It significantly improves the quality of multimodal data alignment and fusion, providing reliable support for subsequent global collaborative scheduling.

[0071] S2.2: The difference coefficient between each device in each cycle and its matching cycle is used as the weight of the fault severity of each device in any matching cycle of each cycle, so as to evaluate the normality of each device in the current cycle and to screen out faulty devices from all devices.

[0072] When optimizing the dispatch of power equipment using a green supply chain, the faults and abnormalities of different nodes within the same cycle are not entirely the same. After a fault occurs, the dispatch center needs to retrieve the corresponding spare parts from the green spare parts warehouse as soon as possible. During the transportation of spare parts, it is necessary to ensure not only that the energy consumption and pollution of the green supply chain are minimized, but also that the timeliness of spare parts transportation is guaranteed. In the traditional dispatch optimization process, spare parts for faulty equipment are often allocated based on the repair work orders filled out by maintenance personnel. However, in remote areas of the power system, if the entire process of abnormal alarm, dispatching maintenance personnel, on-site inspection, work order filling, and spare parts allocation and transportation is carried out, it will seriously affect the efficiency of handling abnormalities. At the same time, this method is strictly dependent on the accuracy of maintenance personnel in anomaly investigation and the accuracy of work order filling. In this case, the traditional process is mainly a serial and local decision driven by events. In order to improve the efficiency of anomaly repair, it can only perform single-objective or local optimization on the faulty node, and it is difficult to guarantee the global multi-objective optimization of the green supply chain in the dispatch process.

[0073] In the above analysis, based on the power data fluctuations and text data characteristics of repair work orders in the same type of nodes, the similarity of faults or anomalies between different cycles was measured. Therefore, if the fault status of each node and the equipment to be dispatched can be adjusted by combining the cycle difference coefficient during the scheduling process, it will help to accurately optimize the scheduling of standby equipment even in the case of missing or incorrect work order data. In this embodiment, the difference coefficient between each device in each cycle and its matching cycle is used as the weight of the fault severity of each device in any matching cycle of each cycle to evaluate the normality of each device in the current cycle, which is used to screen out faulty devices from all devices. The specific process is as follows:

[0074] First, the difference coefficient between each device in each cycle and its matching cycle is used as the weight of the fault severity of each device in any matching cycle of each cycle, in order to assess the normality of each device in the current cycle. Specifically:

[0075] The result of weighted fusion of the difference coefficient between each device and all its matching cycles in each cycle and the degree of fault severity is denoted as the abnormal matching factor of each device in each cycle.

[0076] The difference between the abnormal matching factor corresponding to each device in the current period and the average abnormal matching factor in all matching periods of the current period is taken as the normality of each device in the current period.

[0077] To facilitate understanding of the calculation principle of the abnormal matching factor, a specific expression is given below for illustration:

[0078] Abnormal matching factor of the p-th device under period m The expression is: In the formula, This represents the fault severity of the p-th device under the matching period j of period m. The fault severity is an empirical value obtained from a large number of experiments, and the value ranges from 0 to 1. The larger the value, the more serious the fault. This represents the difference coefficient between the period m and the period j of the p-th device; This represents the number of all matching cycles of cycle m of the p-th device.

[0079] It should be noted that there are many methods to measure the difference between data. In this embodiment, the absolute value of the difference between the abnormal matching factor corresponding to the current period of each device and the mean of the abnormal matching factor under all matching periods of the current period is used as the difference between the abnormal matching factor corresponding to the current period of each device and the mean of the abnormal matching factor under all matching periods of the current period. In practical applications, as other implementation methods, implementers may also use other methods such as the square of the difference or the ratio, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods for measuring the difference between data.

[0080] Based on the concept of normality, it can be understood that normality characterizes whether the equipment is in a typical fault or abnormal state in the current cycle. It reflects the degree to which the equipment state deviates from the historical fault baseline in the current cycle, and its value directly determines whether the equipment is identified as a faulty object that needs to be scheduled. The calculation logic of normality is to first use the difference coefficient as a weight to calculate the fault severity of the matching cycle to obtain the abnormal matching factor, and then calculate the absolute difference between the abnormal matching factor of the current cycle and the mean of the factors of all matching cycles. The calculation of normality is mainly affected by the difference between the abnormal matching factor of the current cycle and the mean of the matching cycle group. The larger the difference, the more outlier and atypical the current cycle state appears compared to the historical fault group of the matching cycle, resulting in a larger normality. This reflects that the equipment may not have experienced a serious fault or the abnormal state is not obvious, and it is identified as a non-faulty equipment. Conversely, the smaller the difference, the more the current cycle state closely matches the fault characteristics of the matching cycle and is in a typical fault concentration area, resulting in a smaller normality. This reflects that the equipment is in a high-confidence fault state and needs to be immediately screened as a faulty equipment for scheduling.

[0081] Furthermore, in this embodiment, the normality of all devices in the current cycle is used as the input of the Otsu threshold segmentation algorithm. The output segmentation threshold is recorded as the normal threshold. Devices with a normality less than the normal threshold are recorded as faulty devices. Faulty devices are used to characterize abnormal entities that are accurately identified in the current cycle after multimodal data fusion and historical matching analysis. Their device status characteristics are highly consistent with typical historical fault cycles and their normality is lower than the threshold. These devices are determined to be faulty and have a high risk of harm, and are thus identified as the final target objects in the green supply chain that require spare parts transportation and emergency repair services. This realizes the transformation from fuzzy abnormal signals to clear scheduling instructions.

[0082] Thus, this embodiment uses the difference coefficient as a weight to evaluate the normality of equipment, accurately screens out faulty equipment whose status characteristics highly match those of historical typical faults, effectively overcomes the impact of missing or distorted work orders, and realizes the transformation of fault identification from manual serial decision-making to data-driven parallel collaborative decision-making, which helps to improve the accuracy of global scheduling optimization.

[0083] S3: Using the ant colony algorithm, the optimal travel path of the power repair vehicle is planned, taking into account the total distance, total pollution emissions, and time urgency cost of each power repair vehicle in the power green supply chain scheduling process.

[0084] Step S3: Using the ant colony algorithm, the optimal travel path of the power repair vehicle is planned, taking into account the total distance traveled, total pollution emissions, and time urgency cost of each power repair vehicle during the power green supply chain scheduling process.

[0085] The coordinates of all faulty equipment and all green spare parts warehouses are used as inputs to the ant colony algorithm. The number of ants is set to 100, the pheromone importance is set to 1, the pheromone evaporation rate is set to 0.3, and the maximum number of iterations is set to 300. During the optimization of the ant colony algorithm, it is necessary to control the energy consumption pollution and transportation path during the scheduling and transportation process, so as to construct the following objective function. The optimization goal is to minimize the objective function. The path corresponding to each power repair vehicle when the objective function reaches its minimum value is taken as the optimal driving path.

[0086] The expression for the objective function F is: In the formula, These represent the total distance traveled, total pollution emissions, and time urgency cost of the e-th power repair vehicle in the current cycle during the power green supply chain dispatch process; E represents the total number of all power repair vehicles participating in the dispatch process. These are the preset first weighting coefficient, the preset second weighting coefficient, and the preset third weighting coefficient, respectively. In this embodiment The values ​​are 0.3, 0.3, and 0.4 respectively. In practical applications, implementers can also set the values ​​according to the importance of each item. This embodiment does not impose any special restrictions.

[0087] in, In the formula, These represent the severity of the fault of the u-th and r-th faulty devices in the current cycle, respectively. This represents the total distance traveled by the e-th power repair vehicle during the power green supply chain dispatch process to repair faulty equipment in the current cycle. Let r represent the average speed of the r-th power repair vehicle; U represents the number of faulty devices in the current period.

[0088] Thus, this embodiment utilizes the ant colony algorithm to construct an objective function that integrates total distance, total pollution emissions, and weighted time urgency costs, optimizing the planning of the optimal travel path for power repair vehicles. This achieves a balance between low-carbon environmental protection, transportation timeliness, and global collaboration in green supply chain scheduling, significantly improving the overall efficiency and decision-making quality of scheduling.

[0089] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0090] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0091] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A power green supply chain scheduling optimization method based on multimodal data fusion, characterized in that, The method includes the following steps: Acquire repair work orders for each piece of equipment, all types of power data, and all types of meteorological and environmental data during the green power supply chain dispatch process; The differences in power data, meteorological data, and text of repair orders for each device in the current period compared to any historical period are analyzed separately to determine the difference coefficient between the current period and any historical period for each device. This allows for the selection of matching periods for the current period of each device from historical periods, which can then be used as historical reference standards to determine whether each device is in a fault state in the current period. The difference coefficient between each device and its matching cycle in each cycle is used as the weight of the fault severity of each device in any matching cycle of each cycle to evaluate the normality of each device in the current cycle, which is used to screen out faulty devices from all devices; the evaluation method of the normality of each device in the current cycle is as follows: the result of weighted fusion of the difference coefficient between each device and all matching cycles of each cycle and the fault severity is recorded as the abnormal matching factor of each device in each cycle; the difference between the abnormal matching factor corresponding to each device in the current cycle and the mean of the abnormal matching factor in all matching cycles of the current cycle is used as the normality of each device in the current cycle. Using the ant colony algorithm, the optimal travel path of each power repair vehicle is planned, taking into account the total distance traveled, total pollution emissions, and time urgency cost during the power green supply chain scheduling process.

2. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The historical period includes all periods prior to the current period for each device, as well as periods for devices belonging to the same category as each device prior to the current period.

3. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The method for determining the difference coefficient of each device between the current period and any historical period is as follows: Based on the differences in power data and meteorological environment data of each device in the current period compared to any historical period, the power fluctuation of each device in the current period is determined. The difference coefficient between each device in the current period and any historical period is positively correlated with the difference in power fluctuation of each device in the current period compared to the power fluctuation in any historical period, as well as the text difference of the repair work order.

4. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 3, characterized in that, The power fluctuation of each device in the current period is positively correlated with the differences in power data and meteorological environment data.

5. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The methods for obtaining differences in power data and meteorological environmental data are as follows: Calculate the sum of differences between each device in the current period and any historical period across all types of power data, and the sum of differences between each device in all types of meteorological and environmental data, respectively, and record them as the power difference index and environmental difference index of each device between the current period and any historical period. The sum of the power difference index and the sum of the environmental difference index for each device between the current cycle and all historical cycles are denoted as power data difference and meteorological environment data difference, respectively.

6. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The process of selecting the matching period for the current period for each device from historical periods includes: The difference coefficient between each device in the current period and all historical periods is used as the input of the threshold segmentation algorithm, and the output is the segmentation threshold. The historical periods with a difference coefficient less than or equal to the segmentation threshold are used as the matching period for the current period of each device.

7. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The process of filtering out faulty devices from all devices includes: The normality of all devices in the current period is used as the input of the threshold segmentation algorithm. The output segmentation threshold is recorded as the normal threshold. Devices with a normality less than the normal threshold are recorded as faulty devices.

8. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 1, characterized in that, The optimal driving route for the planned power emergency repair vehicle includes: Based on the total distance traveled, total pollution emissions, and time urgency cost of each power repair vehicle during the power green supply chain dispatch process, the objective function of the ant colony algorithm is determined. The ant colony algorithm is used to determine the optimal travel path for each power repair vehicle when the objective function is minimized.

9. The power green supply chain scheduling optimization method based on multimodal data fusion as described in claim 8, characterized in that, The expression for the objective function F is: In the formula, , , These represent the total distance traveled, total pollution emissions, and time urgency cost of the e-th power emergency repair vehicle for repairing faulty equipment in the current cycle during the power green supply chain dispatch process. E represents the total number of all power emergency repair vehicles participating in the power green supply chain dispatch process.

Images