CN-122019093-A - Storage management and intelligent scheduling method oriented to field intuition mode

CN122019093ACN 122019093 ACN122019093 ACN 122019093ACN-122019093-A

Abstract

The invention discloses a storage management and intelligent scheduling method for an intuitive mode in the field, which comprises the steps of establishing a storage unit comprising a mode definition area and a mode life cycle area which are physically isolated and logically associated for each mode, extracting situation characteristics in real time and matching the situation characteristics with a mode triggering condition through a matching calculation engine, comprehensively evaluating and scheduling a plurality of high-matching-degree modes based on a multi-objective optimization model, updating life cycle records and priorities of the modes according to efficiency feedback of mode application, and dynamically allocating resources in a hierarchical storage architecture according to call prediction. The method realizes the invariable storage of the mode core logic and the traceability of the whole life cycle, improves the decision quality through the refined matching and the multi-objective conflict resolution, and realizes the self-adaptive optimization and the efficient resource utilization of the system by means of the feedback closed loop and the dynamic resource scheduling.

Inventors

WANG XIAO
LUO HANMEI
LIU DAICHUN
SUN CHAO

Assignees

浙大城市学院

Dates

Publication Date: 20260512
Application Date: 20260128

Claims (9)

1. The storage management and intelligent scheduling method for the domain-oriented intuition mode is characterized by being suitable for managing a plurality of structured intuition mode units generated by a construction system and comprising the following steps of: S100, establishing a storage structure, and creating an independent storage unit for each structured intuitionistic mode unit, wherein the storage unit comprises a mode definition area and a mode life cycle area which are physically isolated and logically associated, the mode definition area is used for invariably storing mode core logic from the building system, and the mode life cycle area is used for recording running and evolution data of a mode in an event tracing mode; S200, performing situation matching, monitoring state changes of an application environment in real time, extracting situation features, matching the situation features with trigger conditions of pattern definition areas in all storage units through a matching calculation engine, and outputting real-time matching degree of all patterns; S300, performing optimized scheduling, wherein when the matching degree of a plurality of modes exceeds an activation threshold, the modes are comprehensively evaluated based on a multi-objective optimization model, the multi-objective optimization model at least comprehensively considers the historical efficiency data, the novelty coefficient and the correlation with the current situation of each mode, and scheduling sequencing is generated according to the evaluation result; S400, implementing feedback maintenance, receiving efficiency feedback after the mode is scheduled and applied, and synchronously updating the record of the corresponding mode in the mode life cycle area and the priority state of the record in a storage system based on the efficiency feedback; S500, dynamically allocating resources, and dynamically adjusting cache resources occupied by different priority modes in a hierarchical storage architecture according to historical call data and prediction information of the modes.
2. The method of claim 1, wherein the pattern core logic of the pattern definition area is stored invariably and comprises at least a trigger condition, a behavior execution sequence and an expected impact, wherein the operation and evolution data of the pattern lifecycle area record comprises at least a pattern activation history, a user feedback record and a pattern state change history, and wherein the pattern definition area is logically associated with the pattern lifecycle area by a unique identifier associated with the structured intuitive pattern unit.
3. The method for storage management and intelligent scheduling of a domain oriented intuition mode according to claim 2, wherein the mode life cycle area is implemented by an event tracing architecture, and the event of each time the mode is called, fed back or internal state changed is used as an immutable event record to be additionally stored, so as to support tracing and reconstruction of the mode full life cycle.
4. The storage management and intelligent scheduling method for the domain-oriented intuition mode according to claim 1 is characterized in that the matching calculation engine works by extracting a multidimensional situation feature vector from a current application environment state, performing mode matching on the situation feature vector and a mode triggering condition, calculating the real-time matching degree of each mode based on the matching degree of a predefined feature weight and the condition, and dynamically dividing the modes into three priority levels according to the calling frequency, the matching success rate and the data freshness index of the modes, wherein the high priority mode is stored in a high-speed memory, the medium priority mode is stored in a high-speed external memory, and the low priority mode is stored in a large-capacity archiving memory. .
5. The method for managing and intelligently scheduling storage oriented to the domain intuitionistic modes according to claim 4, wherein the mode matching adopts a two-stage screening strategy, the mode which does not meet the hard constraint condition is directly excluded by carrying out necessary matching screening according to the hard constraint condition in the triggering condition, then the similarity between the flexible constraint in the triggering condition and the context feature vector is calculated for the mode which passes the screening, and the dynamic allocation resource further comprises a predictive preloading mechanism, wherein the mode set which is possibly invoked in the subsequent period is predicted based on the analysis of the time law of the mode invocation and the semantic relevance among tasks, and the mode in the mode set is migrated from low-priority storage to high-priority storage in advance. .
6. The method for storage management and intelligent scheduling of a domain-oriented intuition model according to claim 1, wherein the evaluation dimensions of the multi-objective optimization model comprise historical successful execution rate of the model, recent call frequency, personalized fitness with a current operation user, and contribution degree of the model to diversity of recommendation results, and the model aggregates scores of the dimensions through a fusion function to generate comprehensive evaluation scores.
7. The method for storage management and intelligent scheduling of domain oriented intuition patterns according to claim 6, wherein the multi-objective optimization model integrates a novelty promoting factor, and for patterns with creation time later than a preset time point or cumulative call number lower than a preset threshold, an incentive score is added to the comprehensive evaluation score, and the incentive score decays with the increase of the call number of the patterns.
8. The method for managing and intelligently scheduling storage oriented to the domain intuitive mode according to claim 1, wherein the implementing feedback maintenance includes receiving structured feedback data and updating performance indexes of corresponding modes according to the feedback data, triggering degradation processing if one mode continuously receives feedback representing negative performance, reducing priority in scheduling and sequencing and adding an abnormal mark for the mode, and triggering upgrading processing if one mode continuously receives feedback representing positive performance, and improving priority of cache resources of the mode.
9. The method for domain-oriented intuitive schema storage management and intelligent scheduling as defined in claim 8, further comprising an exception handling mechanism that automatically initiates a review process when a schema is added to the exception flag, the review process comprising restricting the schema from being actively scheduled for recommendation, sending review notifications to associated responsible persons, and initiating enhanced monitoring of the schema application effect to collect decision data.

Description

Storage management and intelligent scheduling method oriented to field intuition mode Technical Field The invention relates to the technical field of building engineering informatization, in particular to a storage management and intelligent scheduling method oriented to an intuitive mode of the field. Background The artificial intelligence and expert system fields, especially in the intelligent auxiliary system oriented to specific business fields (such as financial wind control, medical diagnosis, industrial operation and maintenance, etc.), decision support based on 'field intuition mode' has become an important technical means. The domain intuition pattern refers to a standardized behavior logic or decision rule unit for coping with a specific situation, which is formed by a domain expert's experience or by machine learning refinement. In the prior art, a rule engine, a knowledge graph or a model library and other modes are generally used for storing, matching and calling, but when a large-scale and high-dynamic intelligent system is actually constructed, the following core bottlenecks to be solved are exposed by the existing architecture: Pattern storage stiffness and evolution tracking loss-the prior art mostly stores patterns as static logic units, and core definitions of patterns and operation life cycle data (such as call history and effect feedback) thereof are usually stored in a mixed mode or are mutually split. This results in a pattern that is difficult to implement with immutable core logic management, while lacking accurate, traceable records of state evolution (e.g., optimization, failure, version iterations) throughout its lifecycle. When the mode needs auditing, optimizing or diagnosing the problem, the historical state and decision basis cannot be quickly rebuilt, and maintainability and credibility of the system are reduced. Context matching is rough and conflicted with scheduling, in a complex dynamic application environment, multiple modes may match the current context at the same time. The existing matching mechanism is mostly based on simple rule priority or single similarity threshold, lacks of fine weight consideration on multidimensional situation characteristics, and cannot effectively process scheduling conflicts among a plurality of high-matching-degree modes. This often results in actions recommended or performed by the system that are not globally optimal, possibly resulting in decision conflicts, wasted resources, or missing more optimal but more novel modes, limiting the adaptive capacity and decision quality of the system. Feedback loop weakness and resource inefficiency-performance feedback after mode application often fails to be fed back to the management system in a structured and timely manner for dynamic adjustment of the priority of the mode or triggering of maintenance actions. Moreover, all modes typically occupy storage and computing resources in an equal manner, lacking dynamic resource allocation mechanisms based on their practical value, call popularity, and predicted demand. In a massive mode library scene, the high-frequency efficient mode response is delayed, and the cold door invalid mode continuously occupies valuable cache, so that the overall resource utilization efficiency of the system is low. Lack of systematic exception handling and novel facilitation mechanisms existing systems lack automated degradation, isolation and review procedures for patterns that continue to perform poorly, potentially continuously outputting low quality decisions. At the same time, the system tends to recommend mature, high frequency patterns, resulting in new introduced, potential novel patterns being "submerged" due to lack of exposure opportunities, which is detrimental to the continued innovation and evolution of pattern libraries. Therefore, there is an urgent need in the art for an innovative storage management and intelligent scheduling method, which can implement decoupling storage of definition and life cycle, accurate matching and conflict resolution based on multidimensional context, closed-loop self-optimization based on efficiency feedback, and dynamic resource scheduling based on value, so as to construct a high-maintainability, high-self-adaptive and high-efficiency mode operation system. Disclosure of Invention The invention aims to overcome the defects in the prior art and provide a storage management and intelligent scheduling method oriented to an intuitive mode in the field. In a first aspect, a storage management and intelligent scheduling method for a domain-oriented intuitive mode is provided, which is suitable for managing a plurality of structured intuitive mode units generated by a building system, and includes: And S100, establishing a storage structure, namely establishing an independent storage unit for each structured intuitionistic mode unit, wherein the storage unit comprises a mode definition area and a mode life cycle area which are physi