Meteorological service knowledge graph generation method and system based on resource collaboration of production and innovation platform

Abstract

The present invention provides a method and system for generating a meteorological service knowledge graph based on resource collaboration of an industrial innovation platform. First, the industrial demand data set and the meteorological monitoring data set in the industrial innovation platform are obtained. Then, the semantics of the industrial demand features are aligned to generate industrial semantic association features. The meteorological element observation features are spatially and temporally integrated to generate a meteorological element feature set. Then, based on a preset resource collaboration network, the cross-domain features of the two are matched to generate a meteorological service sub-graph set containing disaster prevention and resource utilization related sub-graphs. Afterwards, according to the real-time interaction data of industrial entities, the sub-graph node connection weights are dynamically adjusted, and the priority ranking results are obtained in combination with the timeliness evaluation rules. Finally, according to the priority ranking results, the adapted meteorological service strategy is pushed to the service interface of the industrial innovation platform, covering disaster warning guidance parameters and climate resource allocation plans, so as to achieve accurate adaptation and dynamic response of meteorological services to industrial needs.

Description

Technical Field

[0001] The present invention relates to the field of meteorological service application technology, and in particular to a method and system for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform. Background Art

[0002] In the current context of the integration of industrial innovation and meteorological services, industrial activities are becoming increasingly dependent on meteorological services. However, the existing meteorological service technology system has exposed significant limitations when responding to complex and changing industrial needs.

[0003] For example, existing methods for constructing meteorological service knowledge graphs mainly focus on knowledge representation and reasoning within the meteorological field, and lack the ability to integrate and process cross-domain data features. In meteorological services, meteorological element observation characteristics have significant spatiotemporal characteristics, while industry demand characteristics have unique semantic connotations. Existing technologies have difficulty effectively integrating and matching these two different dimensional features, and are unable to generate meteorological service knowledge graphs with practical application value for industrial entities. Therefore, existing meteorological service knowledge graphs often cannot accurately reflect the potential correlation between meteorological elements and industrial activities, and are difficult to meet the diverse needs of the industry in disaster prevention and resource utilization. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform, the method comprising:

[0005] Obtaining an industrial demand data set and a meteorological monitoring data set from the industrial innovation platform, wherein the industrial demand data set includes resource description features and service demand features of multiple industrial entities, and the meteorological monitoring data set includes meteorological element observation features in the spatiotemporal dimension;

[0006] Performing semantic alignment processing on the resource description features and the service demand features to generate industry semantic association features, and performing spatiotemporal fusion processing on the meteorological element observation features to generate a meteorological element feature set;

[0007] Based on a preset resource collaboration network, cross-domain feature matching is performed on the industry semantic association features and the meteorological element feature set to generate a meteorological service sub-graph set for industry entities. The meteorological service sub-graph set includes a disaster prevention-related sub-graph and a resource utilization-related sub-graph.

[0008] Dynamically adjust the node connection weights in the meteorological service sub-graph set according to the real-time interactive data of the industrial entities, and generate a priority ranking result in combination with the preset timeliness evaluation rules;

[0009] Based on the priority ranking result, an adapted meteorological service strategy is pushed to the service interface of the production and innovation platform, and the meteorological service strategy includes disaster warning guidance parameters and climate resource allocation plan.

[0010] On the other hand, an embodiment of the present invention also provides a meteorological service knowledge graph generation system based on resource collaboration of a production and innovation platform, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present invention realizes the deep association and fusion of cross-domain data features by semantically aligning the resource description features and service demand features in the industrial demand data set, and spatiotemporal fusion of the meteorological element observation features in the meteorological monitoring data set, effectively solving the heterogeneity problem between industrial demand and meteorological data in semantic expression and spatiotemporal dimensions. On this basis, based on the preset resource collaboration network, the industry semantic association features are accurately matched with the meteorological element feature set, and a meteorological service sub-graph set containing a disaster prevention association sub-graph and a resource utilization association sub-graph is generated. This not only reveals the potential association between meteorological elements and industrial activities, but also realizes the transformation of meteorological services from a single disaster warning to a comprehensive service model that gives equal importance to disaster prevention and resource utilization. Furthermore, by introducing real-time interactive data of industrial entities to dynamically adjust the node connection weights in the meteorological service sub-graph set, and combining the timeliness evaluation rules to generate priority sorting results, the meteorological service strategy can respond to changes in industrial demand in real time, significantly improving the timeliness and accuracy of meteorological services. Finally, based on the priority sorting results, the adapted meteorological service strategy, including disaster warning guidance parameters and climate resource allocation plans, was pushed to the service interface of the production and innovation platform, realizing the transition of meteorological services from passive response to active adaptation, and effectively promoting the deep coordination and optimized allocation of meteorological resources and industrial needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform provided in an embodiment of the present invention.

[0013] Figure 2 It is a schematic diagram of exemplary hardware and software components of a meteorological service knowledge graph generation system based on resource collaboration of a production and innovation platform provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1This is a flow chart of a method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform provided by an embodiment of the present invention. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform is introduced in detail below.

[0015] Step S110: Obtain the industry demand data set and meteorological monitoring data set in the industry innovation platform, wherein the industry demand data set includes resource description characteristics and service demand characteristics of multiple industry entities, and the meteorological monitoring data set includes meteorological element observation characteristics in the time and space dimensions.

[0016] In this embodiment, a comprehensive production and innovation platform typically integrates various industries, including chemical, mechanical processing, and food processing. Specifically, taking the chemical industry entity as an example, in the industry demand data set, in terms of resource description features, production equipment includes reactors, whose production equipment category code is set to "CH-Reactor-001," and transfer pumps, whose category code is "CH-Pump-002." Raw materials include specific chemical A, whose category code is set to "CH-ChemicalA-010," and chemical solvent B, whose category code is "CH-SolventB-011." Resource allocation constraints include, for example, that the reactor's operating time cannot exceed 8 hours per run, and that the monthly purchase volume of chemical A must be between 1,000 and 1,500 liters. In terms of service demand features, from the demand data submitted by chemical companies, it is possible to identify disaster prevention demand keywords such as "fire prevention" and "leakage prevention," energy efficiency optimization keywords such as "optimize heating system energy consumption," and capacity cycle adjustment keywords such as "peak season production increase strategy."

[0017] In terms of meteorological monitoring data collection, the monitoring scope is the area covered by the production and innovation platform and its surrounding areas. In the temporal dimension, data such as temperature, wind speed, and air quality index are recorded every 30 minutes to form a time series. In the spatial dimension, the area covered by the production and innovation platform is divided into grids with a side length of 500 meters. Each grid records a corresponding meteorological element, such as ultraviolet intensity in grid E and atmospheric humidity in grid F. These data together constitute the meteorological monitoring data collection.

[0018] Step S120: semantically aligning the resource description features and the service demand features to generate industry semantic association features, and performing spatiotemporal fusion processing on the meteorological element observation features to generate a meteorological element feature set.

[0019] Step S121: Parse the entity attribute description text in the resource description feature to extract the industry resource type identifier and resource configuration constraint conditions, wherein the industry resource type identifier includes a production equipment category code and a raw material category code.

[0020] Taking the chemical industry as an example, from the text describing production equipment, "Reactors used to mix multiple chemicals require regular safety inspections," we extract the production equipment category code "CH-Reactor-001" and the resource allocation constraint "Requires regular safety inspections." From the text about raw materials, "Highly corrosive chemical A requires special material containers for storage," we extract the raw material category code "CH-ChemicalA-010" and the resource allocation constraint "Requires special material containers for storage."

[0021] Step S122: Identify a demand keyword set in the service demand feature, where the demand keyword set includes disaster prevention demand keywords, energy efficiency optimization keywords, and production capacity cycle adjustment keywords.

[0022] Continuing with the chemical industry as an example, the keyword "fire prevention" for disaster prevention needs was identified from the company's statement, "Focus on preventing fire accidents during high temperature seasons." The keyword "optimize heating system energy consumption" for energy efficiency optimization was identified from the statement, "reduce heating system energy consumption through process improvements." The keyword "peak season production increase strategy" for capacity cycle adjustment was identified from the statement, "develop a plan to expand production based on peak market demand."

[0023] Step S123: constructing an industry semantic mapping table, performing a many-to-many association mapping between the industry resource type identifier and the demand keyword set, and generating an initial semantic association matrix.

[0024] For example, step S1231: convert the production equipment category code and the raw material category code into binary code vectors respectively, and generate an equipment code vector set and a material code vector set.

[0025] For the chemical industry's production equipment category code "CH-Reactor-001," assuming a 12-bit binary encoding, according to specific encoding rules, it is converted into a binary code vector "000000000101," "CH-Pump-002" is converted to "000000000110," and so on, forming a set of equipment code vectors. For the raw material category code "CH-ChemicalA-010," the same 12-bit binary encoding is used, converting it to "000000010100," and "CH-SolventB-011" is converted to "000000010101," and so on, forming a set of material code vectors.

[0026] Step S1232: Input each keyword in the required keyword set into a pre-trained word embedding model to generate a keyword semantic vector set.

[0027] Assume that the word embedding model is trained based on a large amount of chemical industry text. For the keyword "fire prevention", after model processing, a 15-dimensional keyword semantic vector is generated, for example, "[0.1, 0.3, 0.2, 0.4, 0.1, 0.5, 0.3, 0.2, 0.4, 0.3, 0.2, 0.4, 0.1, 0.3, 0.2]", and "leakage prevention" generates "[0.2, 0.4, 0.1, 0.3, 0.2, 0.4, 0.1, 0.3, 0.2, 0.4, 0.1, 0.3, 0.2, 0.4, 0.1]", etc., forming a keyword semantic vector set.

[0028] Step S1233: Perform dimension unification processing on the device coding vector set and the material coding vector set, map the device coding vector and the material coding vector to the same feature dimension as the keyword semantic vector set through a fully connected neural network, and generate a resource coding vector set of unified dimension.

[0029] Assuming a 15-dimensional keyword semantic vector set, a fully connected neural network takes a device code vector set and a material code vector set as input. For example, a device code vector, such as "000000000101," is converted into a 15-dimensional vector, "[0.12, 0.25, 0.18, 0.31, 0.11, 0.29, 0.17, 0.22, 0.33, 0.19, 0.21, 0.27, 0.16, 0.30, 0.13]," through weight calculation and activation function processing within the network. The same processing is performed on the material coding vector. For example, “000000010100” is converted to “[0.22, 0.35, 0.28, 0.41, 0.21, 0.39, 0.27, 0.32, 0.43, 0.29, 0.31, 0.37, 0.26, 0.40, 0.23]” to form a resource coding vector set of unified dimension.

[0030] Step S1234: Calculate the cosine similarity between each resource encoding vector in the resource encoding vector set of the unified dimension and each keyword semantic vector in the keyword semantic vector set to generate a many-to-many similarity score matrix.

[0031] For a vector in the set of uniformly dimensional resource encoding vectors, such as "[0.12, 0.25, 0.18, 0.31, 0.11, 0.29, 0.17, 0.22, 0.33, 0.19, 0.21, 0.27, 0.16, 0.30, 0.13]," the similarity score between it and the keyword semantic vector "[0.1, 0.3, 0.2, 0.4, 0.1, 0.5, 0.3, 0.2, 0.4, 0.3, 0.2, 0.4, 0.1, 0.3, 0.2]" is calculated using the cosine similarity method. This calculation is performed for all resource encoding vectors and keyword semantic vectors to obtain a many-to-many similarity score matrix. Each element in the many-to-many similarity score matrix represents the similarity score between a resource encoding vector and a keyword semantic vector.

[0032] Step S1235: performing minimum-maximum normalization processing on the many-to-many similarity score matrix in the row direction, compressing the similarity score corresponding to each resource coding vector to the interval [0, 1], and generating a normalized similarity matrix.

[0033] For each row of the many-to-many similarity score matrix, find the minimum and maximum values ​​in that row. For each element in that row, convert it to the interval [0, 1] using the min-max normalization formula. For example, if the minimum value in a row is 0.1 and the maximum value is 0.8, and an element in that row is 0.4, the normalized value is (0.4 - 0.1) / (0.8 - 0.1) = 0.4286 (approximately). Perform this process on all rows of the many-to-many similarity score matrix to generate a normalized similarity matrix.

[0034] Step S1236: Marking the elements in the normalized similarity matrix that exceed a preset similarity threshold as valid association pairs, and generating the initial semantic association matrix based on the distribution of the valid association pairs.

[0035] Assuming a preset similarity threshold of 0.6, elements in the normalized similarity matrix with a score greater than 0.6 are marked as valid association pairs. Based on the position and distribution of these valid association pairs in the normalized similarity matrix, an initial semantic association matrix is ​​generated. The elements in the initial semantic association matrix represent the association between the industry resource type identifier and the demand keyword. For example, if the similarity score between the encoding vector corresponding to "CH-Reactor-001" and the semantic vector of the keyword "fire prevention" is greater than 0.6, then the corresponding position in the initial semantic association matrix is ​​marked as associated.

[0036] Step S124: Perform contextual semantic disambiguation processing on the initial semantic association matrix, eliminate redundant association pairs by setting a similarity threshold and a conflict detection algorithm, and generate optimized industry semantic association features. The optimized industry semantic association features include mapping weight parameters between industry entities and meteorological service needs.

[0037] Taking the initial semantic association matrix generated for the chemical industry as an example, a similarity threshold of 0.7 was set, and a conflict detection algorithm was employed. For example, considering the association between "CH-Reactor-001" and "leakage prevention," based on actual chemical production scenarios, if the reactor has virtually no leakage risk during normal operation and relevant protective measures are in place, then the conflict detection algorithm, combined with contextual information such as production processes and equipment characteristics, determines that this association pair is likely redundant. This analysis is repeated for all association pairs in the initial semantic association matrix, eliminating those with a similarity below 0.7 and deemed redundant by conflict detection. After processing, optimized industry semantic association features are obtained. These features not only clarify the association between industrial entities and meteorological service requirements but also include mapping weight parameters. For example, for the association between "CH-Reactor-001" and "fire prevention," a mapping weight parameter of 0.8 is assigned based on factors such as the equipment's material and operating environment, indicating a strong association between the equipment and the "fire prevention" requirement. This provides a more accurate semantic association basis for the subsequent generation of the meteorological service knowledge graph.

[0038] Step S125: Divide the time granularity of the meteorological element observation characteristics to generate hourly observation sequences and daily observation trends, wherein the hourly observation sequences include temperature change gradient parameters and precipitation intensity fluctuation parameters.

[0039] Taking meteorological monitoring data from the area covered by the production and innovation platform as an example, temperature data is recorded every 15 minutes. To generate hourly observation sequences, the temperature gradient parameters are calculated by calculating the temperature difference between adjacent recorded time points within each hour. For example, within a given hour, if the temperature is 20°C at 15 minutes, 22°C at 30 minutes, 23°C at 45 minutes, and 24°C at 60 minutes, the hourly temperature gradients are (22-20) / 0.25 = 8°C / hour, (23-22) / 0.25 = 4°C / hour, and (24-23) / 0.25 = 4°C / hour (where 0.25 indicates a 0.25-hour interval). Precipitation intensity fluctuation parameters are also recorded, such as the change in precipitation intensity within different 15-minute intervals. For daily observation trends, statistical analysis of all 15-minute records within a day is performed to obtain data such as daily average temperature and total daily precipitation, forming a daily observation trend.

[0040] Step S126: establishing a spatial grid index structure, spatially aggregating meteorological element observation features within the same geographical area, and generating a regional meteorological feature cluster, wherein the regional meteorological feature cluster includes an extreme weather event distribution map and a climate resource abundance distribution map.

[0041] The area covered by the production and innovation platform and its surrounding areas are divided into spatial grids with a side length of 1,000 meters. For each spatial grid, meteorological element data within the spatial grid, such as temperature, humidity, wind speed, precipitation, etc., are collected, and these data are aggregated to generate regional meteorological feature clusters. For the extreme weather event distribution map, whether extreme weather events such as heavy rain, strong winds, and lightning have occurred in the spatial grid, as well as information such as the time and intensity of occurrence. For the climate resource abundance distribution map, the abundance of climate resources such as solar energy, wind energy, and hydropower in the spatial grid is displayed. For example, the abundance of solar energy resources is measured by calculating the duration of sunshine within a certain period of time, and wind energy resources are evaluated by measuring the average wind speed.

[0042] Step S127: After downsampling the hourly observation sequence to the daily granularity through the sliding window average, perform time dimension convolution processing with the daily observation trend to generate time fused meteorological features. At the same time, perform spatial correlation analysis on the regional meteorological feature cluster and the regional meteorological feature clusters of the adjacent areas to generate spatial fused meteorological features. The spatial correlation analysis adopts the inverse distance weighted method to calculate the spatial fusion weight according to the geographical distance between the spatial grids.

[0043] Assume the sliding window size is four 15-minute intervals, or one hour. For the hourly temperature observation series, the hourly data is averaged to obtain downsampled daily granularity data. For example, from 9:00 AM to 12:00 AM on a given day, the temperatures recorded every 15 minutes are 20°C, 22°C, 23°C, 24°C, 25°C, 26°C, 27°C, 28°C, 29°C, 30°C, 31°C, 32°C, 33°C, 34°C, 35°C, and 36°C, respectively. For example, the downsampled average temperature for this period, taking 9:00 AM to 10:00 AM, is (20 + 22 + 23 + 24) / 4 = 22.25°C. The downsampled daily granularity data is convolved with the daily observation trend in the temporal dimension. The convolution kernel is then applied to the data to generate a temporally fused meteorological feature, which integrates meteorological information at different time scales.

[0044] For spatial correlation analysis, we take a grid as an example and calculate its geographic distance from adjacent grids. Assuming the distance between a grid and adjacent grid G ​​is 800 meters, and the distance between a grid and adjacent grid H is 1200 meters, using the inverse distance weighting method, the weight of grid G ​​is (1 / 800) / (1 / 800 + 1 / 1200) ≈ 0.6, and the weight of grid H is (1 / 1200) / (1 / 800 + 1 / 1200) ≈ 0.4. These weights are used to weight the meteorological characteristics of adjacent grids to generate spatially fused meteorological features.

[0045] Step S128: jointly encode the temporal fusion meteorological features and the spatial fusion meteorological features to generate a meteorological element feature set including spatiotemporal correlation weights.

[0046] Temporally fused meteorological features and spatially fused meteorological features are encoded according to predefined rules, such as binary encoding, to integrate their important information. Based on parameters used in the temporal and spatial fusion process, such as the convolution kernel parameters in the temporal convolution process and the distance weights in the spatial correlation analysis, spatiotemporal correlation weights are generated to form a set of meteorological element features.

[0047] Step S130: Based on the preset resource collaboration network, cross-domain feature matching is performed on the industry semantic association features and the meteorological element feature set to generate a meteorological service sub-graph set for industrial entities, wherein the meteorological service sub-graph set includes a disaster prevention-related sub-graph and a resource utilization-related sub-graph.

[0048] Step S131: creating an industry entity node and a meteorological element node in the resource collaboration network, wherein the industry entity node carries the industry semantic association feature, and the meteorological element node carries the meteorological element feature set.

[0049] In the pre-set resource collaboration network, taking the chemical industry as an example, an industry entity node is created. This industry entity node carries industry semantic association features, such as the mapping weight parameter 0.8 associated with "CH-Reactor-001" and "fire prevention." A meteorological element node is also created. This meteorological element node carries a meteorological element feature set that includes temporal fusion meteorological features, spatial fusion meteorological features, and spatiotemporal association weights. For example, this node contains spatiotemporal fusion feature information such as temperature and humidity in a specific area over a specific period of time.

[0050] Step S132: Calculate the demand matching degree between the industrial entity node and the meteorological element node, wherein the calculation of the demand matching degree includes: performing min-max normalization on the weight parameters in the industrial semantic association features, and performing Z-score normalization on the continuous values ​​in the meteorological element feature set, and then calculating the demand matching degree based on the weighted combination of the disaster impact factor and the resource utilization factor and through cosine similarity.

[0051] For the weight parameters in the industry semantic association features carried by the chemical industry entity node, we assume that their value range is 0-1. Using the min-max normalization formula, we further adjust them to a range of 0.2-0.8. For continuous values ​​in the meteorological element feature set, such as temperature and humidity, we use the Z-score normalization formula to eliminate dimensional differences and normalize them to a distribution with a mean of 0 and a standard deviation of 1. Assuming that the weight of the disaster impact factor is set to 0.6 and the weight of the resource utilization factor is set to 0.4, the relevant parameters are weighted and calculated. Finally, the cosine similarity method is used to calculate the demand matching between the industry entity node and the meteorological element node, resulting in a numerical value representing the degree of matching between the two.

[0052] Step S133: establishing a bidirectional connection channel between the industrial entity node and the meteorological element node according to the demand matching degree, wherein the bidirectional connection channel includes a disaster prevention association channel and a resource utilization optimization channel.

[0053] For example, if "CH-Reactor-001" (reactor) has a high degree of match with the demand of a certain meteorological element node and is related to the "fire prevention" demand, a disaster prevention association channel is established between the two; if it is related to resource utilization, such as the relationship between the reactor's operating energy consumption and meteorological conditions, a resource utilization optimization channel is established.

[0054] Step S134: Optimize the topology of the bidirectional connection channels, remove redundant channels with a matching degree lower than a preset threshold, and retain connection channels with a confidence level greater than a set confidence level to form the meteorological service sub-map set.

[0055] Assuming a preset threshold of 0.5 and a confidence level of 0.7, any connection channel with a matching degree below 0.5 is considered redundant and removed. For example, in the connection between the chemical industry and meteorological element nodes, if the matching degree between a connection channel corresponding to "CH-Reactor-001" (reactor) and a temperature-humidity combination node in the meteorological element is calculated to be 0.4, which is lower than the preset threshold of 0.5, then the connection channel is considered redundant and removed from the entire connection channel set.

[0056] For the retained channels, their confidence levels must be further assessed. This confidence level can be based on various factors, such as statistical data from similar past matches or an assessment of the importance of the industrial entities and meteorological elements involved in the current match. For example, for the connection channel established between "CH-Reactor-001" and the meteorological element node for the "fire prevention" requirement, a statistical analysis of historical data on the relationship between past fire accidents and meteorological conditions, as well as an assessment of the criticality of reactors in chemical production, determined that the confidence level for this connection was 0.75, exceeding the set confidence level of 0.7. Therefore, the connection channel was retained.

[0057] After processing all connected channels, the remaining channels form a meteorological service sub-graph set. The disaster prevention association sub-graph within this meteorological service sub-graph set illustrates the relationships between industry entities and meteorological elements in terms of disaster prevention. For example, in the chemical industry, "CH-Reactor-001" (reactor) is connected to specific meteorological conditions (such as meteorological element nodes corresponding to high temperature and dry weather) through a disaster prevention association channel, indicating that high temperature and dry weather may affect reactor fire prevention. The disaster prevention association sub-graph may also include information such as the strength of the association to quantify this impact.

[0058] The Resource Utilization Association submap shows the connection between industrial entities and meteorological factors in terms of resource utilization. For example, in a chemical production process, the energy consumption of a reactor may be related to factors such as weather temperature and wind speed. In the Resource Utilization Association submap, "CH-Reactor-001" is connected to the corresponding meteorological element node through a resource utilization optimization channel, demonstrating how to optimize the reactor's energy utilization based on meteorological conditions. Related parameter descriptions may also be provided, such as recommended energy adjustment ranges under different meteorological conditions.

[0059] Step S140: Dynamically adjust the node connection weights in the meteorological service sub-graph set according to the real-time interaction data of the industrial entities, and generate a priority ranking result in combination with the preset timeliness evaluation rules.

[0060] Step S141: monitor the real-time service request data submitted by the industrial entity in the production and innovation platform, and extract the timeliness identifier and priority parameter in the request content.

[0061] During the operation of the production and innovation platform, the real-time service request data submitted by chemical industry entities is continuously monitored. For example, a chemical company may submit a service request like this: "In the next three days, due to the trial production of new products, it is necessary to focus on ensuring the stable operation of the reactor to prevent failures due to weather factors." From this real-time service request, a timeliness identifier is extracted, which can be set to "within the next three days" to indicate the time limit for the service request. At the same time, a priority parameter is extracted. It is assumed that based on the company's emphasis on the trial production of new products and the assessment of possible losses, the priority parameter is set to "high."

[0062] Timeliness identifiers and priority parameters may have various representations and value ranges, depending on the specific design and business logic of the production and innovation platform. Timeliness identifiers can be specific time intervals, such as "next X hours" or "next X days," or relative time descriptions, such as "within the next production cycle." Priority parameters can be expressed in words, such as "high," "medium," or "low," or as numerical levels, such as 1-5, with 1 being the highest priority.

[0063] Step S142: parsing the meteorological service response time window corresponding to the timeliness identifier, and adjusting the connection weight coefficient of the disaster prevention associated channel according to the length of the meteorological service response time window.

[0064] For the extracted timeliness identifier "within the next three days," the corresponding weather service response window is parsed. Assuming the production and innovation platform sets the weather service response window to match the timeliness identifier, that is, within the next three days, the connection weight coefficient of the disaster prevention-related channel is adjusted based on the length of this time window.

[0065] Generally speaking, the shorter the time window, the more urgent the service demand, and the higher the connection weight coefficient of the disaster prevention-related channel should be adjusted. This adjustment can be achieved through a pre-defined functional relationship. For example, let the response time window length be T (unitless) and the connection weight coefficient be W. The functional relationship W = a + b / T (where a and b are predetermined constants). Using "the next three days" as an example, substitute T = 3 into the function to calculate the new connection weight coefficient W.

[0066] In this way, for the connection channels related to disaster prevention in the chemical industry, such as the disaster prevention-related channel "CH-Reactor-001" corresponding to high temperature meteorological conditions, its connection weight coefficient will be adjusted accordingly according to the length of the meteorological service response time window, so that in an emergency, the connection related to disaster prevention will be closer to ensure that industrial entities can respond to possible disaster threats.

[0067] Step S143: Identify the resource allocation level corresponding to the priority parameter, and dynamically weight the connection weight of the resource utilization optimization channel based on the resource allocation level to obtain a dynamic weighted result. The dynamic weighted processing includes: mapping the priority parameter to a normalized weight in the interval [0, 1], and then performing a weighted summation with the connection weight coefficient.

[0068] In this embodiment, after identifying the priority parameter "high", it is mapped to a resource allocation level according to the rules set by the production and innovation platform. Assume that the platform sets the resource allocation level corresponding to the priority "high" to level 3 (a total of 5 levels).

[0069] Next, the connection weights of the resource utilization optimization channel are dynamically reweighted. First, the priority parameter "High" (corresponding to resource allocation level 3) is mapped to a normalized weight in the range [0, 1]. Assuming a linear mapping function is used to map levels 1-5 to the range 0-1, the mapping formula is: Normalized Weight = (Level - 1) / (5 - 1). For level 3, substituting this into the formula yields a normalized weight = (3 - 1) / (5 - 1) = 0.5.

[0070] Next, obtain the original connection weight coefficient of the resource utilization optimization channel and set it as W0. Perform a weighted summation calculation, and set the weighted dynamic result as W1. The weighting formula is: W1 = α × normalized weight + (1-α) × W0 (where α is a pre-set weight factor, indicating the influence of priority on connection weight). For example, if α = 0.6, the normalized weight is 0.5, and W0 = 0.4, then W1 = 0.6 × 0.5 + (1-0.6) × 0.4 = 0.46.

[0071] Taking the resource utilization optimization channel corresponding to "CH-Reactor-001" and wind conditions in the chemical industry as an example, the above dynamic weighted processing is used to obtain new connection weights to reflect the adjustment of resource utilization optimization associations under different priorities, so that high-priority service requests can be more effectively guaranteed in terms of resource utilization.

[0072] Step S144: normalize the adjusted connection weight coefficient and the dynamic weighting result, and update the distribution state of the node connection weights in the meteorological service sub-graph set.

[0073] Normalize the adjusted connection weight coefficients of the disaster prevention-related channels (set as W2) and the dynamic weighted results of the resource utilization optimization channels (set as W1). Assuming that the normalization process uses the min-max normalization method, first determine the minimum value (min) and maximum value (max) of all connection weight coefficients and dynamic weighted results.

[0074] For W1 and W2, the normalized weight values ​​are calculated using the formula: Normalized weight = (current weight - min) / (max - min). For example, if min = 0.3, max = 0.7, W1 = 0.46, and W2 = 0.5, then the normalized W1' = (0.46 - 0.3) / (0.7 - 0.3) = 0.4, and the normalized W2' = (0.5 - 0.3) / (0.7 - 0.3) = 0.5.

[0075] After such normalization, these normalized weight values ​​are updated to the weights of the corresponding node connections in the meteorological service sub-graph set, thereby updating the distribution status of the node connection weights in the entire meteorological service sub-graph set. This ensures that all connection weights are on a unified scale, which is convenient for subsequent analysis and application. This enables the meteorological service sub-graph set to more accurately reflect the relationship between the real-time needs of industrial entities and meteorological elements. As the real-time interactive data of industrial entities changes, it is continuously and dynamically adjusted to provide a more suitable meteorological service knowledge graph.

[0076] Step S145: Obtain the service response history records of each node in the meteorological service sub-graph set, and extract the historical response delay time and service effect evaluation indicators.

[0077] During the operation of the production and innovation platform, the service response history of each node in the meteorological service sub-graph, whether it is an industrial entity node or a meteorological element node, is recorded. For example, the service response history of the "CH-Reactor-001" node in the chemical industry may include relevant information for each request for meteorological services.

[0078] From these service response history records, historical response delays are extracted. For example, a request for meteorological services to address the impact of high temperatures on a reactor may take two hours from the time the request is sent to the time the relevant meteorological service recommendations are actually received. This is a historical response delay record. Simultaneously, service effectiveness evaluation indicators are extracted, which can be determined in a variety of ways. For example, based on an assessment of the reactor's operational stability after receiving the meteorological service recommendations, if operational stability improves by 80%, this 80% is recorded as a value for the service effectiveness evaluation indicator. Service effectiveness evaluation indicators may also involve aspects such as the percentage of product quality improvement and the percentage of energy savings, depending on the business needs and evaluation criteria of the industrial entity.

[0079] The historical response delays and service effectiveness evaluation metrics for different nodes vary depending on the specific situation. For nodes related to different meteorological factors, such as temperature and humidity, their service response delays may vary depending on the complexity of meteorological data acquisition and analysis. Service effectiveness evaluation metrics will also vary depending on the impact on different aspects of the industrial entity.

[0080] Step S146: Construct a timeliness evaluation function, perform a nonlinear combination calculation on the historical response delay time and the service effect evaluation index to generate a node timeliness score, wherein the nonlinear combination calculation includes: converting the historical response delay time into a delay score, and normalizing the service effect evaluation index to the interval [0, 1] and performing a weighted summation with the delay score to generate a node timeliness score, wherein the delay score = 1-(delay time / maximum tolerable delay).

[0081] A timeliness evaluation function is constructed to comprehensively assess the timeliness of nodes. Taking the "CH-Reactor-001" node as an example, the historical response delay is first converted into a delay score. Assuming the maximum tolerable delay is set to 4 hours and a historical response delay is 2 hours, the formula for delay score = 1 - (delay time / maximum tolerable delay) yields a delay score of 1 - (2 / 4) = 0.5.

[0082] Then, normalize the service effectiveness evaluation index to the interval [0, 1]. If the service effectiveness evaluation index is an 80% improvement in operational stability, using the linear normalization formula: Normalized Index = Index Value / 100% (assuming the maximum index value is 100%), the normalized service effectiveness evaluation index is 0.8.

[0083] Assume the latency score weight is β1, and the normalized service effectiveness evaluation index weight is β2 (β1 + β2 = 1). Calculate the node timeliness score using the weighted summation formula: Node Timeliness Score = β1 × Latency Score + β2 × Normalized Service Effectiveness Evaluation Index. For example, if β1 = 0.4 and β2 = 0.6, then the node timeliness score = 0.4 × 0.5 + 0.6 × 0.8 = 0.68.

[0084] Different nodes calculate the corresponding node timeliness score based on their respective historical response delay time and service effect evaluation indicators through this timeliness evaluation function, providing basic data for subsequent priority sorting.

[0085] Step S147: Sort the nodes in the weather service sub-graph set according to the node timeliness score to generate a primary sorting list.

[0086] The node timeliness scores of all nodes in the meteorological service subgraph are aggregated. For example, consider chemical industry-related nodes and their associated meteorological element nodes. For example, assume there are industry entity nodes such as "CH-Reactor-001" and "CH-Pump-002," as well as meteorological element nodes such as temperature and humidity. Each node receives a node timeliness score calculated in step S212.

[0087] All nodes are sorted from high to low according to their timeliness scores. For example, the "CH-Reactor-001" node has a timeliness score of 0.68, the "CH-Pump-002" node has a timeliness score of 0.62, and the temperature and meteorological element node has a timeliness score of 0.75. After sorting, a preliminary ranking list is generated, which preliminarily reflects the relative importance of each node in terms of timeliness. Nodes with higher timeliness scores are positioned higher in the list, indicating that the corresponding service requests perform better in terms of timeliness.

[0088] Step S148: integrating the service request frequency data of the industrial entity and the load status parameters of the resource collaboration network, dynamically correcting the primary ranking list, and generating a final priority ranking result.

[0089] For example, step S1481: extract the number of requests per unit time and the proportion of request duration in the service request frequency data to generate a request frequency vector, and simultaneously extract the proportion of concurrent connections and resource occupancy in the load status parameters to generate a load status vector.

[0090] Continuing with the chemical industry as an example, we count the number of requests per unit time from the service request frequency data recorded on the production innovation platform. Assuming that the number of service requests corresponding to "CH-Reactor-001" was 20 in the past week, and a week is calculated as 168 hours, the number of requests per unit time = 20 / 168 (times / hour). We also calculate the request duration ratio. For example, if the total request duration of these 20 requests was 40 hours, and the total duration of a week is 168 hours, the request duration ratio = 40 / 168. The number of requests per unit time and the request duration ratio form a request frequency vector, such as [20 / 168, 40 / 168].

[0091] From the resource collaboration network's load status parameters, obtain the concurrent connection percentage and resource utilization. Assuming the maximum number of concurrent connections supported by the resource collaboration network is 1000 and the current number of concurrent connections for the "CH-Reactor-001" service request is 200, the concurrent connection percentage = 200 / 1000. Assuming the total resource capacity of the resource collaboration network is 100% and the "CH-Reactor-001" service request occupies 30% of the resources, the resource utilization is 30%. Combine the concurrent connection percentage and resource utilization to form a load status vector, such as [200 / 1000, 30%].

[0092] The request frequency vector and load state vector of different industry entities will vary depending on their business activities. For example, the request frequency and resource usage of the clothing manufacturing industry may be very different from those of the chemical industry, depending on factors such as the production scale, production process, and dependence on meteorological services.

[0093] Step S1482: Perform a logarithmic transformation on the number of requests per unit time in the request frequency vector and then perform min-max normalization processing to obtain a first normalized request frequency. Linearly scale the request duration ratio to the interval [0, 1] to obtain a second normalized request frequency.

[0094] For the number of requests per unit time in the request frequency vector, set it to X1 (e.g., 20 / 168), perform a logarithmic transformation, and let the result after the transformation be Y1 = log(X1 + 1). (The addition of 1 prevents the logarithmic function from being undefined when X1 is 0.) Then, perform min-max normalization on Y1. Assuming that the minimum value of the logarithmic transformation of the number of requests per unit time for all industrial entities is min1, and the maximum value is max1, then the first normalized request frequency = (Y1 - min1) / (max1 - min1).

[0095] For the request duration ratio, set it to X2 (e.g., 40 / 168) and linearly scale it to the interval [0, 1]. Assuming that the minimum value of the request duration ratio for all industry entities is min2 and the maximum value is max2, then the second normalized request frequency = (X2 - min2) / (max2 - min2).

[0096] Through such processing, the two elements in the request frequency vector are converted to a suitable unified scale, which is convenient for subsequent comprehensive calculation with other parameters.

[0097] Step S1483: performing Z-score normalization on the concurrent connection ratio and resource occupancy rate in the load state vector to obtain a first normalized load index and a second normalized load index.

[0098] For the concurrent connection ratio in the load state vector, set it to Z1 (e.g., 200 / 1000) and perform Z-score normalization. Assuming the mean of the concurrent connection ratio for all industry entities is μ1 and the standard deviation is σ1, the first normalized load metric is (Z1 - μ1) / σ1.

[0099] For resource utilization, set it to Z2 (e.g., 30%) and perform Z-score normalization. Assuming the mean of resource utilization for all industrial entities is μ2 and the standard deviation is σ2, the second normalized load index = (Z2 - μ2) / σ2.

[0100] Through Z-score normalization, the two parameters in the load state vector are converted into a standard distribution with a mean of 0 and a standard deviation of 1, so that they can be comprehensively analyzed on the same scale as other processed parameters.

[0101] Step S1484: linearly superimpose the first normalized request frequency, the second normalized request frequency, the first normalized load index, and the second normalized load index according to a preset weight ratio to generate a comprehensive load impact factor.

[0102] Assume the weight of the first normalized request frequency is γ1, the weight of the second normalized request frequency is γ2, the weight of the first normalized load index is γ3, and the weight of the second normalized load index is γ4 (γ1 + γ2 + γ3 + γ4 = 1). Calculate the comprehensive load impact factor using the formula: Comprehensive load impact factor = γ1 × first normalized request frequency + γ2 × second normalized request frequency + γ3 × first normalized load index + γ4 × second normalized load index.

[0103] For example, if γ1=0.2, γ2=0.2, γ3=0.3, γ4=0.3, the first normalized request frequency=0.4, the second normalized request frequency=0.5, the first normalized load index=-0.5, and the second normalized load index=0.5, then the comprehensive load impact factor=0.2×0.4+0.2×0.5+0.3×(-0.5)+0.3×0.5=0.28.

[0104] This comprehensive load impact factor comprehensively reflects the impact of the service request frequency of industrial entities and the load status of the resource collaboration network on node priority. Different weighting settings (γ1, γ2, γ3, and γ4) vary depending on the business focus and resource management strategy of the industrial innovation platform. For example, if the platform prioritizes service request frequency, the weights of γ1 and γ2 may be appropriately increased; if it focuses more on the rational allocation of resources and network load, the weights of γ3 and γ4 may be increased.

[0105] Step S1485: Dynamically attenuate and compensate the timeliness scores of the nodes in the primary sorting list according to the comprehensive load impact factor to generate a corrected timeliness score sequence.

[0106] For each node in the primary ranking list, its original node timeliness score is set to S (for example, the previously calculated timeliness score for the "CH-Reactor-001" node is 0.68), and its integrated load impact factor is set to ILF (for example, the previously calculated 0.28). The node timeliness score is adjusted using a pre-defined dynamic decay compensation function. Assume the dynamic decay compensation function is: Adjusted timeliness score = S × (1 + α × ILF), where α is an adjustment coefficient used to control the influence of the integrated load impact factor on the node timeliness score. For example, if α = 0.5, for the "CH-Reactor-001" node, the adjusted timeliness score = 0.68 × (1 + 0.5 × 0.28) = 0.68 × 1.14 = 0.7752.

[0107] This process is repeated for all nodes in the primary ranking list, resulting in a corrected timeliness score sequence. This process takes into account the service request frequency of industrial entities and the load status of the resource collaboration network. It optimizes the primary node timeliness scores, which were derived solely based on historical response delays and service effectiveness evaluation metrics, to make the scores more reflective of the actual node priority.

[0108] Step S1486: Arrange the nodes in descending order according to the corrected timeliness score sequence to generate the final priority sorting result.

[0109] The corrected scores of each node in the corrected timeliness score sequence are summed up, and then the nodes are rearranged in descending order. Taking the chemical industry-related nodes as an example, assuming that after dynamic attenuation compensation, the corrected timeliness score of the "CH-Reactor-001" node is 0.7752, the corrected timeliness score of the "CH-Pump-002" node is 0.72, and the corrected timeliness score of the temperature and meteorological element node is 0.75, etc.

[0110] After sorting in descending order, the final priority ranking results are obtained. In this final result, nodes ranked higher indicate higher priority, taking into account multiple factors such as historical response, service performance, service request frequency, and resource coordination network load. This final priority ranking result will be used to subsequently push adaptive weather service strategies to the production and innovation platform, ensuring that more important and urgent service requests receive priority.

[0111] Step S150: Pushing an adapted meteorological service strategy to the service interface of the production and innovation platform based on the priority sorting result, wherein the meteorological service strategy includes disaster warning guidance parameters and climate resource allocation plan.

[0112] Step S151: selecting a top N target service node set according to the final priority sorting result, wherein the target service node set includes disaster warning nodes and resource allocation nodes.

[0113] Assume that, based on the business needs and resource availability of the production and innovation platform, the top 10 nodes (N=10) are selected as the target service node set in the final priority sorting results. In the context of the chemical industry, these nodes might include disaster warning nodes related to reactors, meteorological element nodes related to hazards such as high temperatures and flammable gas leaks, and resource allocation nodes, such as those related to weather-related impacts on energy consumption and raw material supply.

[0114] For example, the high-temperature disaster warning node corresponding to "CH-Reactor-001" is associated with temperature-related elements in meteorological data, as well as associated resource allocation nodes, which may include configuration information for optimizing reactor energy consumption under different temperature conditions. These disaster warning nodes and resource allocation nodes together constitute the target service node set, providing key information for generating specific meteorological service strategies.

[0115] Step S152: extracting the real-time monitoring data and historical pattern data in the disaster warning node to generate a disaster occurrence probability curve and an impact range prediction map.

[0116] Taking the high temperature disaster warning node corresponding to "CH-Reactor-001" as an example, real-time monitoring data is obtained from this node, such as current temperature, humidity, air pressure and other meteorological data, as well as historical pattern data, that is, relevant data records under the same or similar meteorological conditions in the past period of time.

[0117] Suppose that by analyzing historical pattern data and using data mining techniques (the specific algorithms can be considered a black box, such as some machine learning algorithms specifically designed to analyze weather-disaster correlation patterns), we can identify the patterns characteristic of high-temperature disasters. Combined with real-time monitoring data, we can generate a disaster probability curve. For example, as temperature rises, the probability of a disaster may increase. By marking the disaster probability points corresponding to different temperature values ​​on a coordinate graph and then connecting these points, we can create a disaster probability curve.

[0118] For the impact range prediction map, based on the chemical company's plant layout, equipment distribution, and the spatial distribution of meteorological data (for example, temperature differences between different areas), relevant technologies (such as geographic information system-based analysis methods combined with machine learning prediction models) are used to predict the potential impact range of high temperature disasters. The potentially affected areas are marked on the map to generate the impact range prediction map. This impact range may include the workshop where the reactor is located and surrounding storage areas, providing intuitive guidance for companies to take preventative measures in advance.

[0119] Step S153: Analyze the climate resource distribution data and industry demand matching data in the resource allocation node to generate a resource scheduling path planning diagram and a utilization optimization plan.

[0120] Taking the resource allocation node related to "CH-Reactor-001" in the chemical industry as an example, climate resource distribution data, such as the local distribution of solar energy and wind energy, as well as industry demand matching data, are obtained from this node, that is, the relationship between the demand for energy, raw materials, etc. during the operation of the reactor and climate conditions.

[0121] The resource scheduling path planning map combines climate resource distribution data with industry demand, taking into account factors such as transportation costs and transportation time (transportation costs may be related to transportation distance and transportation method, while transportation time may be affected by road conditions and weather conditions). For example, if solar energy resources are relatively abundant in a certain area and the area is close to a chemical plant, and the reactor can be assisted by solar energy heating under specific weather conditions, then by analyzing these factors, a resource scheduling path from the solar energy collection area to the chemical plant can be planned, and the path, stations, and other information can be marked on the map to generate a resource scheduling path planning map.

[0122] For utilization optimization, we match data with industry needs and analyze the reactor's efficiency in utilizing resources like energy and raw materials under different climate conditions. For example, at lower temperatures, reactor heating consumes more energy. In this case, energy utilization can be improved by adjusting production processes and optimizing equipment parameters. By comprehensively considering various factors, we develop specific plans for optimizing resource utilization under different climate conditions, such as energy usage adjustment strategies for different temperature ranges and adjustments to raw material input ratios.

[0123] Step S154: Encapsulate the disaster probability curve, impact range prediction map, resource scheduling path planning map and utilization optimization plan into a standardized service data package, and push it in a targeted manner through the service interface of the production and innovation platform.

[0124] The generated disaster probability curve data (such as a series of temperature values ​​and their corresponding disaster probability values), impact range prediction map (stored in a certain image format, containing information such as geographical area tags), resource scheduling path planning map (also stored in an image format, with path and other information marked), and utilization optimization plan (documented in detail describing the resource utilization adjustment strategy under different climatic conditions) are integrated.

[0125] The data and documents are packaged according to the standardized service data package format pre-defined by the platform. For example, the data and documents may be organized according to a specific directory structure and necessary metadata information may be added, such as the data package generation time, applicable industry entity, and data source.

[0126] Once packaged, the standardized service data packages are pushed to relevant industry entities, such as chemical companies, through the platform's service interface. Chemical companies can receive these data packages through their clients on the platform and, based on the disaster probability curve and impact range forecast, make early disaster preparedness preparations. They can also rationally allocate resources based on resource scheduling path planning and utilization optimization plans, improving production efficiency and safety.

[0127] For example, the method further includes:

[0128] Step S210: Monitor the feedback data of industrial entities on the production and innovation platform on the push of meteorological service strategies, and extract the strategy effectiveness indicators and user satisfaction scores in the feedback data.

[0129] During the operation of the production and innovation platform, feedback data on the promoted weather service strategies from chemical industry entities will be continuously collected. For example, chemical companies may submit feedback information through the feedback interface provided by the platform, including changes in production conditions after the implementation of the weather service strategy and their satisfaction with the service.

[0130] From this feedback data, we extract indicators of strategy effectiveness. For example, if a company reports that applying disaster warning guidance parameters and climate resource allocation plans has reduced the incidence of production accidents by a certain percentage and increased energy utilization by several percentage points, these data can serve as indicators of strategy effectiveness. We also extract user satisfaction scores. Assuming the platform uses a 1-5 rating system, chemical companies can assign a score based on their overall experience with the meteorological service strategy, such as a 4.

[0131] Feedback from different industry entities will vary based on actual application results and subjective perceptions. For example, a larger chemical company may have more stringent evaluation criteria for strategy effectiveness, and its feedback on strategy effectiveness indicators and user satisfaction scores may differ from those of a smaller company, depending on factors such as the company's production complexity and its reliance on meteorological services.

[0132] Step S220: adjusting the parameter configuration of the resource collaboration network according to the policy effectiveness index, and revising the weight distribution ratio in the timeliness evaluation rule based on the user satisfaction score.

[0133] Step S221: extract the service response time reduction rate and resource utilization improvement rate from the policy effectiveness indicators to generate a policy effectiveness vector, and convert the user satisfaction score into a standardized satisfaction value in the interval [0, 1].

[0134] For the extracted strategy effectiveness indicators, using data from a chemical company as an example, assuming that after implementing the weather service strategy, service response time was reduced from an average of 3 hours to 2 hours, the service response time reduction rate = (3-2) / 3. Assuming that the resource utilization improvement rate increased from 70% to 75%, the resource utilization improvement rate = (75%-70%) / 70%. The service response time reduction rate and resource utilization improvement rate are combined to form a strategy effectiveness vector, such as [(3-2) / 3, (75%-70%) / 70%].

[0135] For the user satisfaction score, assuming the score is 4 points (out of 5 points), the formula: standardized satisfaction value = score / 5 is used to convert it into a standardized satisfaction value in the interval [0, 1], that is, standardized satisfaction value = 4 / 5 = 0.8.

[0136] Step S222: performing a logarithmic transformation on the service response time reduction rate and then performing min-max normalization processing to obtain a first normalized effectiveness index, and performing Z-score normalization processing on the resource utilization improvement rate to obtain a second normalized effectiveness index.

[0137] For the service response time reduction rate, let Xs (e.g., (3-2) / 3) be logarithmically transformed, and the result after the transformation be Ys = log(Xs+1). Then, perform min-max normalization on Ys. Assuming that the minimum value of the service response time reduction rate for all industrial entities after the logarithmic transformation is mins and the maximum value is maxs, the first normalized effectiveness index = (Ys-mins) / (maxs-mins).

[0138] For the resource utilization improvement rate, set it to Xu (e.g., (75% - 70%) / 70%) and perform Z-score normalization. Assuming the mean of the resource utilization improvement rate for all industrial entities is μu and the standard deviation is σu, the second normalized effectiveness indicator = (Xu - μu) / σu.

[0139] Through these processes, the two elements in the strategy effectiveness vector are converted to a suitable unified scale, which is convenient for subsequent comprehensive calculation with the standardized satisfaction value.

[0140] Step S223: The first normalized effectiveness index, the second normalized effectiveness index and the standardized satisfaction value are weighted and superimposed according to a preset ratio to generate a comprehensive adjustment factor.

[0141] Assume that the weight of the first normalized effectiveness indicator is ω1, the weight of the second normalized effectiveness indicator is ω2, and the weight of the standardized satisfaction value is ω3 (ω1 + ω2 + ω3 = 1). Calculate the comprehensive adjustment factor using the formula: Comprehensive adjustment factor = ω1 × first normalized effectiveness indicator + ω2 × second normalized effectiveness indicator + ω3 × standardized satisfaction value.

[0142] For example, if ω1=0.3, ω2=0.3, ω3=0.4, the first normalized effectiveness index=0.4, the second normalized effectiveness index=0.5, and the standardized satisfaction value=0.8, then the comprehensive adjustment factor=0.3×0.4+0.3×0.5+0.4×0.8=0.59.

[0143] Step S224: linearly scaling the node connection weight parameters in the resource collaboration network according to the comprehensive adjustment factor, and dynamically compensating and adjusting the matching threshold in the topology optimization process.

[0144] For the node connection weight parameter in the resource collaboration network, let the original weight parameter be Wn and the comprehensive adjustment factor be CAF (e.g., 0.59). The node connection weight parameter is linearly scaled using the formula: adjusted weight parameter = Wn × CAF. For example, a node connection weight parameter originally had a value of 0.6, but after adjustment, it becomes 0.6 × 0.59 = 0.354.

[0145] For the matching threshold during topology optimization, let the original matching threshold be T0. Dynamic compensation adjustment is performed using the formula: Adjusted matching threshold = T0 + β × (1-CAF), where β is a pre-set adjustment coefficient. For example, if T0 = 0.5, β = 0.1, and CAF = 0.59, then the adjusted matching threshold = 0.5 + 0.1 × (1-0.59) = 0.541.

[0146] Such adjustments enable the resource collaboration network to adaptively adjust node connection weights and topology structures based on feedback from industrial entities on meteorological service strategies, thereby improving the network's adaptability to actual needs and service quality.

[0147] Step S225: performing a correlation analysis on the standardized satisfaction value, the delay score weight parameter and the service effect evaluation weight parameter in the timeliness evaluation rule, and generating a weight correction coefficient.

[0148] Assume that the delay score weight parameter in the timeliness evaluation rule is β1, the service effectiveness evaluation weight parameter is β2 (β1 + β2 = 1), and the standardized satisfaction value is S (e.g., 0.8). Using correlation analysis methods (e.g., using the Pearson correlation coefficient calculation method, but treating the specific calculation process as a black box and focusing only on the results), calculate the correlation coefficients between the standardized satisfaction value and β1 and β2, respectively. Let the correlation coefficient with β1 be r1, and the correlation coefficient with β2 be r2.

[0149] A pre-defined functional relationship is then used to generate weight correction coefficients. For example, weight correction coefficient 1 = 1 + γ1 × r1, and weight correction coefficient 2 = 1 + γ2 × r2, where γ1 and γ2 are pre-determined adjustment coefficients. These weight correction coefficients will be used to adjust the delay scoring weight parameters and service effectiveness evaluation weight parameters to better reflect the actual needs of industry entities for service timeliness.

[0150] Step S226: Back-propagation update is performed on the delay score weight parameter and the service effect evaluation weight parameter according to the weight correction coefficient to generate a corrected weight distribution ratio.

[0151] The delay scoring weight parameter β1 is updated using the formula: Revised β1 = β1 × Weight Correction Factor 1. The service effect evaluation weight parameter β2 is updated using the formula: Revised β2 = β2 × Weight Correction Factor 2. For example, if β1 = 0.4, Weight Correction Factor 1 = 1.2, β2 = 0.6, and Weight Correction Factor 2 = 0.9, then Revised β1 = 0.4 × 1.2 = 0.48, and Revised β2 = 0.6 × 0.9 = 0.54.

[0152] After this backpropagation update, a revised weight distribution ratio is obtained. This revised weight distribution ratio will be applied to subsequent timeliness evaluation rules, allowing timeliness evaluation to more accurately reflect the actual feelings and needs of industry entities regarding the timeliness of meteorological services, further optimizing the generation and delivery of meteorological service knowledge graphs, and improving the service quality of the industry innovation platform and the satisfaction of industry entities.

[0153] Step S230: updating the optimized parameter configuration and the revised evaluation rules to the meteorological service processing system.

[0154] The optimized parameter configurations such as the adjusted node connection weight parameters in the resource collaboration network, the matching threshold of the topology optimization, and the revised evaluation rules such as the revised weight distribution ratio in the timeliness evaluation rules are transmitted and updated to the meteorological service processing system.

[0155] After receiving these updates, the meteorological service processing system will operate according to the new parameter configuration and evaluation rules. For example, in the subsequent generation of the meteorological service knowledge graph, the resource collaboration network will perform cross-domain feature matching based on the new node connection weight parameters, and the timeliness evaluation will calculate the node timeliness score based on the revised weight distribution ratio. By continuously updating these parameters and rules, the meteorological service processing system can better adapt to the changing needs of industrial entities, continuously provide more accurate and effective meteorological service strategies, and achieve a benign interaction and optimized development between the resource collaboration of the production and innovation platform and meteorological services.

[0156] Figure 2 A schematic diagram illustrates exemplary hardware and software components of a system 100 for generating a meteorological service knowledge graph based on collaborative production and innovation platform resources, which can implement the concepts of the present invention, according to some embodiments of the present invention. For example, the processor 120 can be used in the system 100 for generating a meteorological service knowledge graph based on collaborative production and innovation platform resources and perform the functions of the present invention.

[0157] The meteorological service knowledge graph generation system 100 based on resource collaboration of an industrial and creative platform can be a general-purpose server or a special-purpose server, both of which can be used to implement the meteorological service knowledge graph generation method based on resource collaboration of an industrial and creative platform of the present invention. Although the present invention only shows a single server, for convenience, the functions described in the present invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0158] For example, the meteorological service knowledge graph generation system 100 based on the resource collaboration of the production and innovation platform may include a network port 110 connected to the network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the meteorological service knowledge graph generation system 100 based on the resource collaboration of the production and innovation platform may also include program instructions stored in ROM, RAM, or other types of non-temporary storage media, or any combination thereof. The method of the present invention can be implemented according to these program instructions. The meteorological service knowledge graph generation system 100 based on the resource collaboration of the production and innovation platform also includes an input / output (I / O) interface 150 between the computer and other input and output devices.

[0159] For ease of explanation, only one processor is described in the meteorological service knowledge graph generation system 100 based on the collaboration of production and innovation platform resources. However, it should be noted that the meteorological service knowledge graph generation system 100 based on the collaboration of production and innovation platform resources in the present invention may also include multiple processors, so the steps performed by one processor described in the present invention may also be jointly performed by multiple processors or performed individually. For example, if the processor of the meteorological service knowledge graph generation system 100 based on the collaboration of production and innovation platform resources executes step A and step B, it should be understood that step A and step B may also be executed jointly by two different processors or executed individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0160] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned method for generating a meteorological service knowledge graph based on resource collaboration of the production and innovation platform is implemented.

[0161] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform, characterized in that: The method comprises: Obtaining an industrial demand data set and a meteorological monitoring data set from the industrial innovation platform, wherein the industrial demand data set includes resource description features and service demand features of multiple industrial entities, and the meteorological monitoring data set includes meteorological element observation features in the spatiotemporal dimension; Performing semantic alignment processing on the resource description features and the service demand features to generate industry semantic association features, and performing spatiotemporal fusion processing on the meteorological element observation features to generate a meteorological element feature set; Based on a preset resource collaboration network, cross-domain feature matching is performed on the industry semantic association features and the meteorological element feature set to generate a meteorological service sub-graph set for industry entities. The meteorological service sub-graph set includes a disaster prevention-related sub-graph and a resource utilization-related sub-graph. Dynamically adjust the node connection weights in the meteorological service sub-graph set according to the real-time interactive data of the industrial entities, and generate a priority ranking result in combination with the preset timeliness evaluation rules; Pushing an adapted meteorological service strategy to the service interface of the production and innovation platform based on the priority ranking result, wherein the meteorological service strategy includes disaster warning guidance parameters and climate resource allocation plan; The preset resource collaborative network is based on which cross-domain feature matching is performed between the industry semantic association features and the meteorological element feature set to generate a meteorological service sub-graph set for industry entities, including: Creating an industry entity node and a meteorological element node in the resource collaboration network, wherein the industry entity node carries the industry semantic association feature, and the meteorological element node carries the meteorological element feature set; Calculating a demand matching degree between the industry entity node and the meteorological element node, wherein the calculation of the demand matching degree includes: performing min-max normalization on the weight parameters in the industry semantic association features, performing Z-score normalization on the continuous values ​​in the meteorological element feature set, and calculating the demand matching degree based on a weighted combination of a disaster impact factor and a resource utilization factor by using cosine similarity; Establishing a bidirectional connection channel between the industrial entity node and the meteorological element node according to the demand matching degree, wherein the bidirectional connection channel includes a disaster prevention association channel and a resource utilization optimization channel; The topology structure of the bidirectional connection channels is optimized, redundant channels with a matching degree lower than a preset threshold are removed, and connection channels with a confidence degree greater than a set confidence degree are retained to form the meteorological service sub-map set.

2. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 1 is characterized in that: The performing semantic alignment processing on the resource description feature and the service requirement feature to generate industry semantic association features includes: Parsing the entity attribute description text in the resource description feature to extract the industry resource type identifier and resource configuration constraint conditions, wherein the industry resource type identifier includes a production equipment category code and a raw material category code; Identifying a demand keyword set in the service demand feature, the demand keyword set including disaster prevention demand keywords, energy efficiency optimization keywords, and capacity cycle adjustment keywords; Constructing an industry semantic mapping table, performing a many-to-many association mapping between the industry resource type identifier and the demand keyword set, and generating an initial semantic association matrix; The initial semantic association matrix is ​​subjected to contextual semantic disambiguation processing, and redundant association pairs are eliminated by setting a similarity threshold and a conflict detection algorithm to generate optimized industry semantic association features. The optimized industry semantic association features include mapping weight parameters between industry entities and meteorological service requirements.

3. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 1 is characterized in that: The performing spatiotemporal fusion processing on the meteorological element observation features to generate a meteorological element feature set includes: Dividing the time granularity of the meteorological element observation characteristics to generate hourly observation sequences and daily observation trends, wherein the hourly observation sequences include temperature change gradient parameters and precipitation intensity fluctuation parameters; Establishing a spatial grid index structure to spatially aggregate meteorological element observation features within the same geographical area to generate regional meteorological feature clusters, which include extreme weather event distribution maps and climate resource abundance distribution maps; After downsampling the hourly observation sequence to the daily granularity through sliding window averaging, the time dimension convolution processing is performed with the daily observation trend to generate a temporally fused meteorological feature. At the same time, the regional meteorological feature cluster is subjected to spatial correlation analysis with the regional meteorological feature clusters of adjacent regions to generate a spatially fused meteorological feature. The spatial correlation analysis adopts the inverse distance weighted method to calculate the spatial fusion weight according to the geographical distance between spatial grids. The temporal fusion meteorological features and the spatial fusion meteorological features are jointly encoded to generate a meteorological element feature set including spatiotemporal correlation weights.

4. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 1 is characterized in that: The dynamically adjusting the node connection weights in the meteorological service sub-graph set according to the real-time interactive data of the industrial entities includes: Monitoring real-time service request data submitted by industrial entities in the industrial innovation platform, and extracting timeliness identifiers and priority parameters in the request content; parsing the meteorological service response time window corresponding to the timeliness identifier, and adjusting the connection weight coefficient of the disaster prevention associated channel according to the length of the meteorological service response time window; Identifying a resource allocation level corresponding to the priority parameter, and dynamically weighting a connection weight of a resource utilization optimization channel based on the resource allocation level to obtain a dynamic weighted result, wherein the dynamic weighting includes: mapping the priority parameter to a normalized weight within the interval [0, 1], and then performing a weighted summation with a connection weight coefficient; The adjusted connection weight coefficient and the dynamic weighting result are normalized to update the distribution state of the node connection weights in the meteorological service sub-graph set.

5. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 1 is characterized in that: The prioritization result generated by combining the preset timeliness evaluation rules includes: Obtain the service response history of each node in the meteorological service sub-graph set, and extract the historical response delay time and service effect evaluation indicators; Constructing a timeliness evaluation function, performing a nonlinear combination calculation on the historical response delay time and the service effect evaluation index to generate a node timeliness score, wherein the nonlinear combination calculation includes: converting the historical response delay time into a delay score, and normalizing the service effect evaluation index to the interval [0, 1] and performing a weighted summation with the delay score to generate a node timeliness score, wherein the delay score = 1-(delay time / maximum tolerable delay); Sort the nodes in the meteorological service sub-graph set according to the node timeliness score to generate a primary sorted list; The service request frequency data of the industrial entity and the load status parameters of the resource collaboration network are integrated to dynamically correct the primary ranking list and generate a final priority ranking result.

6. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 5 is characterized in that: The method of pushing an adapted meteorological service strategy to a service interface of the production and innovation platform based on the priority sorting result includes: Selecting a top N target service node set according to the final priority sorting result, wherein the target service node set includes a disaster warning node and a resource allocation node; Extracting real-time monitoring data and historical pattern data from the disaster warning node to generate a disaster occurrence probability curve and an impact range prediction map; Analyze the climate resource distribution data and industry demand matching data in the resource allocation node to generate a resource scheduling path planning diagram and utilization optimization plan; The disaster probability curve, impact range prediction map, resource scheduling path planning map and utilization optimization plan are encapsulated into a standardized service data package and pushed in a targeted manner through the service interface of the production and innovation platform.

7. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 6 is characterized in that: The disaster warning guidance parameters include at least one of the following: Warning level classification parameters generated based on the disaster occurrence probability curve, wherein the warning level classification parameters set different response levels according to probability threshold intervals; A geographic area marking parameter generated based on the impact range prediction map, wherein the geographic area marking parameter is used to identify the location of industrial facilities affected by the disaster; The emergency resource configuration suggestion parameters are generated based on historical pattern data, and the emergency resource configuration suggestion parameters include material reserve and personnel scheduling plan.

8. The method for generating a meteorological service knowledge graph based on resource collaboration of a production and innovation platform according to claim 6 is characterized in that: The climate resource allocation plan includes at least one of the following: generating climate resource transportation route optimization parameters based on the resource scheduling path planning diagram, wherein the climate resource transportation route optimization parameters consider a balance between transportation cost and time efficiency; Renewable energy allocation ratio parameters generated based on the utilization optimization scheme, wherein the renewable energy allocation ratio parameters are dynamically adjusted according to the production capacity cycle of the industrial entity; The microclimate adjustment recommended parameters are generated in combination with real-time meteorological data, and the microclimate adjustment recommended parameters include temperature and humidity control indicators and energy consumption thresholds.

9. A meteorological service knowledge graph generation system based on resource collaboration of production and innovation platform, characterized by: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the meteorological service knowledge graph generation method based on production and innovation platform resource collaboration as described in any one of claims 1 to 8.

Images