A power asset blockchain financial evaluation method

Abstract

The present application relates to the technical field of electric power financial information processing, and particularly relates to a kind of electric power asset blockchain financial evaluation method.The method comprises: forming minimum verifiable input set from asset operation monitoring data, operation and maintenance record mirror and metering settlement account certificate mirror;Based on the input set, complete evaluation model input mapping, abnormal segment labeling, model version mirror locking and parameter snapshot registration, perform step commitment calculation and concatenation, generate evaluation hash chain and recalculation slice list;Accordingly, perform slice extraction recalculation, source signature verification and consistency comparison, aggregate threshold signature to generate audit pass ticket;Under the ticket trigger, read parameter snapshot to carry out asset value calculation and trusted interval generation, output report fingerprint and complete evidence chain index binding.The scheme realizes three-layer evidence closed loop of input, process and result, strengthens evaluation transparency and verifiability, reduces valuation manipulation risk, and facilitates traceable archive.

Description

Technical Field

[0001] This invention relates to the field of power financial information processing technology, and in particular to a blockchain-based financial evaluation method for power assets. Background Technology

[0002] In the field of power financial information processing technology, existing solutions for blockchain-based financial valuation of power assets typically revolve around processing asset operation monitoring data, mirrored operation and maintenance records, and mirrored metering-settlement ledgers. These solutions first standardize field definitions, bind source signatures, and register time anchors, then generate valuation results or reports based on the input mapping of the valuation model. However, these solutions suffer from limitations such as incomplete minimum exposure screening and index construction on the input side, non-standardized model version mirror locking and parameter snapshot registration on the process side, and a lack of step commitment calculations and process concatenation on the audit and contract sides, failing to form a hash chain and recalculation slice list structure for the valuation process. Existing methods often implement irreversible hash calculations at the level of some fields or only record source voucher pointers, frequently generating single-result archives through one-time process summaries. They lack end-to-end organization for random slice extraction and recalculation task assignment, source signature verification and process consistency comparison, threshold signature aggregation, and result adjudication. In scenarios where contract-triggered verification is not pre-emptively implemented, asset value calculation and trusted interval generation are often disconnected from upstream audit status, making it difficult to support the stable implementation of report fingerprint generation and evidence chain index binding. For the joint processing of asset value calculation and trusted interval generation based on asset operation monitoring data, operation and maintenance record mirrors and metering-settlement account mirrors, through the minimum verifiable input set structure, evaluation process hash chain and audit through bill structure under contract trigger verification constraints, existing technologies generally have shortcomings in the coordination of unified field caliber and time anchor registration, continuity of model version mirror locking and parameter snapshot registration, and closed-loop connection of step commitment calculation and process to audit through bill structure. It is difficult to form a consistent process of collection-alignment-commitment-extraction-recalculation-adjudication-triggering-generation-archiving in the application scenario of blockchain financial evaluation of power assets, resulting in insufficient coupling between valuation and evidence expression and instability of traceable archiving structure. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a blockchain-based financial valuation method for power assets, comprising:

[0004] Acquire asset operation monitoring data, operation and maintenance record mirrors and metering-settlement account mirrors, perform field standardization, source signature binding, time anchor registration, key field extraction, irreversible hash calculation, source voucher pointer generation, minimum exposure filtering and index construction processing, and generate a minimum verifiable input set structure;

[0005] Obtain the minimum verifiable input set structure, perform evaluation model input mapping, abnormal fragment annotation, model version requirement extraction, model version mirror locking, parameter snapshot registration, step commitment calculation and process concatenation processing, and generate an evaluation process hash chain and recalculation slice list structure including stage commitment records, sub-commitment records, concatenation node descriptions, commitment log pointers and recalculation slice list structures.

[0006] Obtain the hash chain and recalculation slice list structure of the evaluation process, perform random slice extraction, recalculation task assignment, recalculation result extraction, source signature verification, process consistency comparison, threshold signature aggregation and result adjudication processing, and generate an audit pass invoice structure.

[0007] The audit process involves obtaining the audited invoice structure, executing contract trigger verification, reading parameter snapshots, extracting parameter snapshots and process commitments, calculating asset value and generating a trusted range in the contract execution environment, generating report fingerprints and binding evidence chain indexes, and generating report fingerprints and traceable archive structures.

[0008] Furthermore, the asset operation monitoring data, operation and maintenance record mirroring, and metering-settlement voucher mirroring include:

[0009] The asset operation monitoring data refers to a multi-dimensional record set from status monitoring, alarm records, fault conditions, environmental sampling, and inside and outside the maintenance window, with data granularity covering all layers of measurement points, loops, equipment, and asset identification; the operation and maintenance record mirror refers to a read-only snapshot of work orders, replacement parts registrations, inspection logs, and defect closure loops formed on the maintenance platform at a specified time point; the metering-settlement account mirror refers to a read-only snapshot of electricity consumption records, billing rule mappings, reconciliation voucher images, and settlement lists formed from the metering acquisition platform and financial settlement module at a specified time point.

[0010] Furthermore, the minimum verifiable input set structure includes:

[0011] The minimum verifiable input set structure includes a summary column, a source credential pointer column, a time anchor element column, an anomaly flag column, a verification status column, and a multi-level index.

[0012] Furthermore, the evaluation process hash chain and recalculation slice list structure includes:

[0013] The evaluation process hash chain includes stage commitment records, sub-commitment records, concatenated node descriptions, and commitment log pointers; the recalculation slice list structure includes slice number, slice type, time anchor interval, feature slot segment reference, and abnormal segment number set.

[0014] Furthermore, the audit examined the bill structure, which included:

[0015] The audit process uses a document structure that includes an aggregate signature, a threshold decision policy number, a session number, a pass / fail slice list, a fail / no slice list, a buffer queue list, an evaluation process hash chain reference, a model parameter snapshot table reference, and a time anchor interval.

[0016] Furthermore, the process of calculating asset value and generating a trusted interval, generating report fingerprints, and binding evidence chain indexes within the contract execution environment includes:

[0017] The contract execution environment is the contract virtual machine. After receiving the valuation calculation trigger package, the VM executes the parameter parsing process. The parameter parsing process reads the parameter snapshot reference and verifies it on the chain. After the verification is successful, a parameter shadow area is built in the contract. The parameter shadow area stores window parameters, abnormal participation strategies, feature participation strategies and aggregation rule strategies according to the parameter template number, and establishes deduplication and merging rules for the iteration area marker.

[0018] Furthermore, the contract reads the commitment records and concatenation node descriptions within the truncation range referenced by the hash chain of the evaluation process, and constructs a process commitment view within the contract. The process commitment view is organized by stage and time anchor interval. The contract locates the summary sequence from the back-pointing key of the model input mapping packet according to the feature participation strategy and abnormal participation strategy of the parameter shadow area, and organizes the located summary sequence into a set of valuation input fragments according to the window parameters.

[0019] Furthermore, the process of calculating asset value and generating a trusted interval, generating report fingerprints, and binding evidence chain indexes within the contract execution environment also includes:

[0020] The asset valuation calculation process involves calling the evaluation logic identified by the image digest within the VM, evaluating each piece of the valuation input fragment set, and generating fragment valuation results. When the abnormal participation strategy sets a participation scope restriction for a certain fragment, the fragment is evaluated only within the participation scope, and a fragment valuation result with an abnormal participation tag is generated. When the iteration zone marker exists, the deduplication and merging rules are executed on the iteration zone entries, and a merged fragment valuation result is generated for duplicate fragments in the same time anchor interval, and the deduplication and merging description pointer is recorded.

[0021] Furthermore, the trusted interval generation step calls the trusted interval generation logic within the VM. Based on the aggregation rule strategy and window parameters configured in the parameter shadow area, the fragment estimation results are resampled and aggregated in multiple rounds. The resampling and aggregation process uses the structured reference of commitment records and summary sequences to carry out repeated sampling. In each round of resampling, the sampling round mark and commitment node reference are recorded. After the aggregation is completed, a structured description of the boundary of the estimated trusted interval is obtained.

[0022] Furthermore, the contract compiles valuation results and evidence elements, which include parameter snapshot references, mirror summary references, process commitment view capture range, fragment valuation result summary sequence, abnormal participation tags, and deduplication and merging description pointers. The contract organizes the above evidence elements into an evidence list. The evidence list has a session number and trigger time anchor written in the structure header, an evidence list fingerprint summary written in the structure tail, and a two-way reference to the commitment log pointer and the model parameter snapshot table established.

[0023] The key innovations of this invention include:

[0024] (1) After obtaining the audit pass bill structure, construct a post-triggered valuation link of contract trigger verification and parameter snapshot reading → asset value calculation and confidence interval generation → report fingerprint generation and evidence chain index binding, and use the audit pass bill structure as the contract trigger condition to connect the objects and reference relationships of the entire process of valuation and archiving.

[0025] (2) Based on the step commitment calculation and process connection to generate the evaluation process hash chain and recalculation slice list structure, further implement the random sampling of slices and recalculation task assignment, source signature verification and process consistency comparison, threshold signature aggregation and result adjudication based on the evaluation process hash chain and recalculation slice list structure to form a verifiable generation path of audit-passed invoice structure.

[0026] (3) Based on the asset operation monitoring data, operation and maintenance record mirror and metering-settlement account mirror, complete the field standardization, source signature binding and time anchor registration, irreversible hash calculation and source voucher pointer generation, minimum exposure screening and index construction, generate the minimum verifiable input set structure, and use it as the only input reference surface for evaluation model input mapping and abnormal fragment annotation, model version mirror locking and parameter snapshot registration, and step commitment calculation and process connection.

[0027] The following are its main beneficial effects:

[0028] (1) The operation link that is applied to the contract trigger verification and parameter snapshot reading to report fingerprint generation and evidence chain index binding, the asset value calculation and credibility interval generation are carried out only when the audit passes the bill structure. This ensures that the valuation process and source signature verification and process consistency comparison are referenced from the same source, which facilitates the connection between the valuation result and the traceable archive structure within the same process. It is suitable for the blockchain financial evaluation scenario of power assets with the evaluation process hash chain and the minimum verifiable input set structure as upstream inputs.

[0029] (2) It operates on the audit link from the evaluation process hash chain and recalculation slice list structure to the audit pass bill structure. Through random sampling of slices and recalculation task assignment, source signature verification and process consistency comparison, threshold signature aggregation and result adjudication, it provides an end-to-end verifiable expression. It can form clear records in the difference positioning of inputs, steps and parameters, reduce the dependence on one-time process summary, and is suitable for audit review scenarios based on time anchor registration and index construction.

[0030] (3) It acts on the continuous link from the input side to the process side. Through the minimum verifiable input set structure, plaintext data is transformed into an input surface that can be referenced by unified field caliber, source signature binding and time anchor registration, irreversible hash calculation and source voucher pointer generation. This provides a consistent data baseline for subsequent evaluation model input mapping and abnormal fragment labeling, model version mirror locking and parameter snapshot registration, as well as step commitment calculation and process connection. It reduces the interference of plaintext exposure and inconsistent caliber on the valuation and audit link. It is suitable for the joint processing scenario of asset operation monitoring data, operation and maintenance record mirror and measurement-settlement account voucher mirror. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating a blockchain-based financial valuation method for power assets, provided as an embodiment of this application. Detailed Implementation

[0032] Example 1: Refer to Figure 1 This is a flowchart illustrating a blockchain-based financial evaluation method for power assets provided in an embodiment of the present invention. The process may include at least steps S100-S400:

[0033] S100: Obtain asset operation monitoring data, operation and maintenance record mirror and metering-settlement account mirror, perform field standardization, source signature binding, time anchor registration, key field extraction, irreversible hash calculation, source voucher pointer generation, minimum exposure filtering and index construction processing, and generate minimum verifiable input set structure;

[0034] S200. Obtain the minimum verifiable input set structure, perform evaluation model input mapping, abnormal fragment annotation, model version requirement extraction, model version mirror locking, parameter snapshot registration, step commitment calculation and process concatenation processing, and generate an evaluation process hash chain and recalculation slice list structure including stage commitment records, sub-commitment records, concatenation node descriptions, commitment log pointers and recalculation slice list structures.

[0035] S300: Obtain the hash chain and recalculation slice list structure of the evaluation process, perform random slice extraction, recalculation task assignment, recalculation result extraction, source signature verification, process consistency comparison, threshold signature aggregation and result adjudication processing, and generate audit pass ticket structure;

[0036] S400: Obtain the audit-approved invoice structure, execute contract trigger verification, parameter snapshot reading, parameter snapshot and process commitment extraction, perform asset value calculation and trusted interval generation, report fingerprint generation and evidence chain index binding processing in the contract execution environment, and generate report fingerprint and traceable archive structure.

[0037] Step S100 includes at least steps S110-S130:

[0038] S110. Obtain asset operation monitoring data, operation and maintenance record mirror and metering-settlement account mirror, perform field standardization, source signature binding and time anchor registration to obtain preprocessed input package;

[0039] The input sources for this section are asset operation monitoring data stored on the asset-side nodes, operation and maintenance record images exported from the maintenance system, and metering-settlement voucher images exported from the settlement system. The asset operation monitoring data refers to a multi-dimensional record set from status monitoring, alarm records, fault conditions, environmental sampling, and records inside and outside the maintenance window, with data granularity covering measurement points, loops, equipment, and asset identification layers. The operation and maintenance record images refer to read-only snapshots of work orders, replacement part registrations, inspection logs, defect closures, and other items generated on the maintenance platform at a specified time point. The metering-settlement voucher images refer to read-only snapshots of electricity consumption records, billing rule mappings, reconciliation voucher images, and settlement lists generated from the metering acquisition platform and financial settlement module at a specified time point. Specifically, the above three types of inputs are written to a temporary buffer through the access gateway, triggering a field probing task to read the field name, field type, unit information, dimensional constraints, allowed value range, and missing measurement flag of each data source, and reading the source identifier, acquisition node identifier, and acquisition time associated with the data source from the access side. Furthermore, the read results undergo unified field specification processing. In this embodiment, unified field specification refers to the standardized mapping of fields with the same name but inconsistent meanings or synonymous but different names. This includes three aspects: field renaming, unit conversion, dimension integration, value range pruning, enumeration value mapping, fixed null value strategy, and fixed missing test labeling rules. Field renaming is achieved through matching using a maintained naming specification table. Unit conversion is achieved through numerical conversion using dimension mapping relationships. Value range pruning writes anomaly fragment markers for abnormal out-of-bounds records. Enumeration value mapping generates a unified code for cases where the same business meaning is encoded differently in multiple systems. Null value strategy fixing writes missing test labels for missing data without estimation. Fixed missing test labeling rules require all missing test records to be explicitly marked and included in the subsequent abnormal fragment labeling stage. Then, source signature binding is performed on the data after standardization. In this embodiment, source signature binding refers to attaching a digital signature to each record, including a source identifier, a collection node identifier, and the digital signature of the person or role responsible for collection. The signature material includes four elements: signature algorithm marker, signature digest, signature time, and certificate serial number. Records with failed signatures or expired certificates are marked as source abnormal and included in the abnormal fragment record area. Subsequently, time anchor registration is performed on the signature-bound data. In this embodiment, time anchor registration refers to generating an independent time anchor structure for each record. The time anchor structure records elements such as the original collection time, access time, standard time zone marker, daylight saving time marker, time sequence number, and cross-domain synchronization sequence number. The time anchor is bound to the record key value, and cross-source data is corrected using the standard time zone marker. Synchronization conflicts are resolved through a sequence number adjudication strategy. Entries with unresolved conflicts are entered into a manual review queue and marked as pending review.To address anomalies during the access process, an anomaly classification module is constructed to record anomaly categories such as missing fields, inconsistent units, mismatched dimensions, failed signature verification, invalid certificates, and time conflicts. These anomalies are then written into the anomaly fragment recording area, which is referenced in subsequent anomaly fragment annotation stages. Finally, the integrated data, after field standardization, source signature binding, and time anchor registration, is written into a structured container. This container uses a unified table structure to carry field values, source identifiers, signature elements, time anchor elements, and anomaly flags, forming a preprocessing input package. This preprocessing input package is recorded as the output field name and is directly read by the preprocessing input package in the next subsection S120. Simultaneously, this output is referenced in the evaluation model input mapping stage of the subsequent main step S200, passively used for source signature verification in the challenge-response third-party audit stage of the main step S300, and passively used for evidence chain tracing pointer generation in the valuation smart contract execution stage of the main step S400.

[0040] S120. Extract a list of key fields from the preprocessed input package, perform irreversible hash calculation and generate source credential pointers, and generate a field hash alignment table.

[0041] The input source for this section is the preprocessing input package output from the previous section. In this embodiment, the key field list refers to the minimum set of fields involved in the evaluation, auditing, and evidence storage of asset operation monitoring data, maintenance record mirrors, and metering-settlement voucher mirrors. This set covers items such as asset identifier, equipment identifier, measurement point identifier, data acquisition time anchor, source identifier, signature digest, certificate serial number, numerical field, unit marker, dimensional marker, value range marker, missing measurement flag, anomaly flag, maintenance work order number, replacement part number, billing rule number, settlement voucher number, and reconciliation number. Specifically, the preprocessing input package is sent to the field extraction module. The field extraction module reads the corresponding fields according to the key field whitelist and generates a key field list in a fixed order. In this embodiment, the fixed order refers to a unified field sequence given by the naming convention table. The sequence order does not change due to different data sources. If a field is missing, a missing measurement flag is written into the list; if an anomaly flag exists in a field, an anomaly flag is retained in the list. Subsequently, irreversible hash calculations are performed on the list of key fields. In this embodiment, irreversible hash calculations refer to generating digest values ​​at two granularities: single field and field group. Single field digests bind the field name, field value, unit marker, dimensional marker, time anchor element, and source identifier. Field group digests bind the field group name, group order, group member digests, and group time sequence information. Digest calculations use fixed parameter configurations and are written to the calculation strategy number. Records that fail to complete the calculation are entered into the abnormal fragment record area, and a digest missing flag is written to the list. Then, source credential pointer generation is performed. In this embodiment, the source credential pointer points to reference information that can be verified on-chain, including the source identifier, certificate serial number, signature digest, signature time, evidence storage location, and read method flag. The read method flag records the access protocol, authentication requirements, and rate limit constraints. The source credential pointer is bound to the digest value through a combined key value of the field and the time anchor. Entry that fails to bind is written to the abnormal fragment record area. During the irreversible hash calculation and source credential pointer generation process, the sampling verification module extracts sample entries from the key field list and performs verifications for duplicate summary values, source credential mismatches, and time anchor misalignments. Verification results are written to the verification log, and a recalculation flag is set for abnormal entries. Next, a field hash alignment table is constructed. In this embodiment, the field hash alignment table points to a structured index that aligns all summary values, source credential pointers, and time anchor elements within the key field list. The table header is fixed with six columns: field name, single-field summary, field group summary, source credential pointer, time anchor element, anomaly flag, and verification status. The table body expands in record number order, with record numbers arranged according to time anchor order and retaining cross-source sequence numbers. After the field hash alignment table is written to storage, a unique version number is generated. The version number is generated from a combined sequence of time anchors and source identifiers. When the version number is not consecutive, a version skip flag is written, and the reason is recorded in the log.Finally, the field hash alignment table is recorded as the output field name and can be directly read by the field hash alignment table in the next subsection S130. At the same time, this output is passively referenced in the step commitment calculation stage of the main step S200, used as a comparison basis in the process consistency comparison stage of the main step S300, and used as the source index input of the report fingerprint in the evidence chain index binding stage of the main step S400.

[0042] S130. Perform minimum exposure filtering and index construction on the field hash alignment table to generate a minimum verifiable input set structure;

[0043] The input source for this section is the field hash alignment table output from the previous section. In this embodiment, the minimum exposure filter refers to forming a minimum set of fields that can be used by the three stages of evaluation, auditing, and evidence preservation without copying the original plaintext. This set rejects specific plaintext field values, retaining only the summary, source credential pointer, time anchor element, anomaly flag, and necessary structural indexes. Specifically, the field hash alignment table is loaded into the filtering module. The filtering module reads the columns allowed to be exposed according to the policy whitelist. The policy whitelist is provided by the data governance configuration and covers the summary column, source credential pointer column, time anchor element column, verification status column, and anomaly flag column. Columns not on the whitelist are directly removed. Redundancy pruning is performed on the retained columns. Redundancy pruning generates folded mappings and records folded relationships for situations such as duplicate source, duplicate summary, and duplicate time anchor. The folded relationships are written to the mapping log and maintained in the structure for traceability. Hierarchical labeling is performed on the anomaly flag column. Hierarchical labeling distinguishes between categories such as abnormal source, missing summary, time conflict, invalid certificate, and field mismatch and records hierarchical codes. There is a one-to-one mapping relationship between the hierarchical codes and subsequent anomaly fragment labeling stages. Subsequently, index construction is performed. In this embodiment, index construction refers to generating multi-level search keys for the minimum exposure set. The outermost index uses a composite key composed of asset identifier, device identifier, and time anchor sequence number. The middle layer index uses a composite key composed of source identifier, certificate sequence number, and signature time. The inner layer index generates a fast location key for the summary column. The three layers of indexes reference each other and record the level number and parent-child relationship. The level number is incremented and does not coincide with the time anchor sequence number. At the same time, access control tags are generated during index construction. In this embodiment, access control tags refer to the annotation of three types of constraints at the record layer: visitor role, access window, and access action. The visitor role is passed in by the subsequent audit node or evaluation service when it is called. The access window comes from the time anchor element. The access action is limited to three types: read-only, recalculation read, and evidence query.Next, after the minimum exposure filtering and index construction are completed, a minimum verifiable input set structure is formed. In this embodiment, the minimum verifiable input set structure refers to a structured set composed of a summary column, a source credential pointer column, a time anchor element column, an anomaly flag column, a verification status column, and a multi-level index. The set has four meta-attributes: version number, generation time, policy whitelist number, and mapping log pointer. After the set is written to storage, a structure summary and a structure signature are generated. The structure summary is bound to the version number and the time anchor range, and the structure signature is signed by the data governance service. The signature element is written to the structure header. After successful writing, the minimum verifiable input set structure is output. The minimally verifiable input set structure serves as the output field name and is directly read by the minimally verifiable input set structure in S210 of the subsequent main step S200. This structure is used to evaluate model input mapping and anomaly segment annotation processing. It also serves as an input reference in the model version mirror locking and parameter snapshot registration stage in S220, as a commitment input set in the step commitment calculation and process concatenation stage in S230, as a slice location basis in the random sampling and recalculation task assignment stage in the main step S300, and as one of the source indexes for the report fingerprint in the contract trigger verification and evidence chain index binding stage in the main step S400. In summary, this step's technical effect is: through minimal exposure filtering and index construction, an input set without plaintext but possessing verifiable and referable characteristics is formed, supporting the collaborative invocation of subsequent mapping, commitment, recalculation, and evidence storage stages.

[0044] Step S200 includes at least steps S210-S230:

[0045] S210. Obtain the minimum verifiable input set structure, perform evaluation model input mapping and abnormal segment labeling processing, and obtain the model input mapping package;

[0046] The input source for this section is the minimum verifiable input set structure output from the preceding steps. This minimum verifiable input set structure includes a summary column, a source credential pointer column, a time anchor element column, an anomaly flag column, a verification status column, and a multi-level index, all of which carry verifiable information in a plaintext-free manner. Specifically, the minimum verifiable input set structure is loaded into the mapping assembly module. The mapping assembly module first reads the naming convention table and the field mapping dictionary, and performs a one-to-one mapping between the standard field names pointed to by the summary column and the evaluation feature names. In this embodiment, the mapping dictionary points to a fixed correspondence maintained by the data governance service, and includes four types of meta-elements: field aliases, dimension markers, unit markers, and value range markers, and establishes a sequential reference relationship with the time anchor elements. Subsequently, the mapping and assembly module calls the feature construction sub-process. This sub-process aligns and segments adjacent records according to window parameters for time-anchored sequences under the same asset and device identifiers. These window parameters are derived from the window parameter table in the configuration library, which records the window length, sliding step size, sampling threshold, and out-of-bounds handling strategy. The summary columns within the window are not reverse-expanded; instead, secondary aggregation and order rearrangement are performed according to the configuration. The rearrangement rules are written to the mapping log, and a feature slot index for the evaluation model is generated. To ensure the relevance of anomaly information in subsequent stages, the anomaly flag column is not removed during mapping but is copied to the feature slot index management area. This serves as guidance for controlling the scope and weight of anomaly samples during model calculation. The management area also records the source document pointer and the backreference key of the time-anchor element, forming a unified entry point that can be read and written by audit nodes and contracts.

[0047] Furthermore, the mapping assembly module introduces an abnormal fragment annotation sub-process. In this embodiment, abnormal fragment annotation refers to the joint judgment action based on the abnormal flag column, the verification status column, and the source credential pointer. When multiple records simultaneously have any of the flags of source abnormality, summary missing, or time conflict within a continuous time anchor range, the continuous range is defined as an abnormal fragment. The abnormal fragment is assigned a fragment number, a fragment start and end time anchor, an asset identifier and equipment identifier to which the fragment belongs, a fragment source set, and a fragment cause set, and the fragment number is written into the management area of ​​the feature slot index. When the abnormal flag only involves a single record and does not form a continuous range, the record is defined as an isolated abnormal entry. Isolated abnormal entries only participate in the temporal structure placeholder in subsequent model input mapping and do not participate in statistical aggregation. When the abnormal flag involves cross-source conflict and cannot be implemented by the time anchor order adjudication strategy, the conflict is recorded as pending review and a manual review pointer is attached. The manual review pointer establishes a reference with the interface of the external review system through the source credential pointer. In this embodiment, the interface is exposed in the form of an Application Programming Interface (API). Access control and rate limiting are registered when the API is called for the first time. After the abnormal fragment annotation sub-process is completed, the mapping assembly module writes the feature slot index, abnormal fragment annotation record and naming convention reference relationship into the mapping cache, generating a structured carrier that can be read downstream.

[0048] Understandably, after completing the above actions, the mapping assembly module calls the model input packaging sub-process to merge and package the structures that can be directly input into the evaluable model. The packaged content includes feature slot indexes, window parameter labels, abnormal fragment annotation records, source voucher pointer back keys, and time anchor element interval labels. Plaintext is not introduced during packaging; only summaries and pointers are used to express data entities. If the window parameters are inconsistent with the model's expected parameters, a compatibility conversion is performed on the window parameters before packaging. Windows that fail to convert are marked as unusable windows and recorded in the packaging log. Finally, the merged and packaged structure is recorded as the output field name model input mapping package, and the system registers that the output field name is directly referenced by the model input mapping package in the subsequent step S220. At the same time, in the subsequent main steps, when the challenge-response third-party audit reads the mapping evidence, the model input mapping package is passively referenced as auxiliary evidence for process consistency comparison. When the valuation smart contract needs to establish the binding of the evidence chain and the feature slot index, the model input mapping package is passively referenced as a retrieval entry point to form a cross-main step call path.

[0049] S220. Extract the model version requirements from the model input mapping package, perform model version mirror locking and parameter snapshot registration, and generate a model parameter snapshot table.

[0050] The input source for this section is the model input mapping package output from the preceding steps. This package carries the structural definitions of window parameter labels and feature slot indices, and the mapping log contains placeholder information for the model's expected parameters. Specifically, the version requirement parsing sub-process first reads the structure of the feature slot indices in the model input mapping package. The read content includes slot names, slot order, window parameter labels, and participation markers for abnormal fragment annotation records. Then, it calls the version requirement parsing rules. In this embodiment, the version requirement parsing rules refer to a rule set maintained by the evaluation service. The rule set provides the model version identifier and parameter template number corresponding to different asset types, equipment types, window parameters, and abnormal participation strategy combinations. Successful parsing yields the model version identifier and parameter template number; failed parsing scenarios are written to the version parsing log and fall back to the default model version identifier. Next, the version locking subprocess retrieves the corresponding image from the model version image library based on the model version identifier. The model version image library stores model components that have been reviewed and entered the whitelist. In this embodiment, the image is packaged using a container image that conforms to the Open Container Specification. The image metadata includes the image summary, build time, build chain, and source credential pointer. To prevent non-whitelisted images from entering the evaluation path, the version locking subprocess verifies the image whitelist. The verification action reads the signature record and validity label of the image whitelist. Images that fail the verification are returned to the version parsing process for re-matching.

[0051] Furthermore, the parameter snapshot registration sub-process generates an actual parameter snapshot based on the parameter template number. In this embodiment, the parameter template records three types of parameter fields: feature participation method, abnormal fragment participation strategy, window parameters and resampling strategy, and aggregation rule strategy and threshold strategy. During the registration of the actual parameter snapshot, the template content is compared item by item with the window parameter label. If the window length, sliding step size, or abnormal participation strategy is inconsistent with the template, a difference label is recorded in the parameter snapshot, and the reason for the difference is written in the parameter snapshot header. During the parameter snapshot registration process, the Key Management Service (KMS) is called to complete the writing of the snapshot signature and the signature material storage pointer. The signature material includes the signature digest, signature time, and certificate serial number, and is bound to the version-locked mirror digest to form a mirror-snapshot joint binding relationship. After the parameter snapshot registration is completed, a model parameter snapshot table is generated. In this embodiment, the model parameter snapshot table points to a structured table containing the model version identifier, mirror digest, parameter template number, window parameter label, abnormal participation strategy label, snapshot signature material pointer, and difference label. After registration, this table is granted read-only permission and an auditable mark so that third-party audit nodes can independently read and verify the signature in subsequent stages.

[0052] Understandably, when a report fingerprint and traceable archive structure exist for a previous cycle, the version requirement parsing subprocess allows reading the version reference segment of this structure to compare the mirror summary and parameter template number used in the previous assessment. If there are significant version changes, a version migration tag is added to the model parameter snapshot table, and a migration reason pointer is recorded. The migration reason pointer establishes a reference relationship with the change record in the evidence base for comparison of the change trajectory during the audit phase. After the above registration action is completed, the model parameter snapshot table is written to storage as an output field name, indicating that this output field name is directly referenced by the model parameter snapshot table in subsequent step S230. At the same time, in the context of cross-main steps, the model parameter snapshot table is passively referenced by the challenge-response third-party audit as one of the sources of the recalculation input parameter set, and when the valuation smart contract reads parameters, the model parameter snapshot table is passively referenced as the parameter reading entry point.

[0053] S230. Perform step commitment calculation and process concatenation on the model parameter snapshot table to generate an evaluation process hash chain and a recalculation slice list structure.

[0054] The input source for this section is the model parameter snapshot table output from the previous steps. This model parameter snapshot table has locked the model version mirror and parameter template number, and completed the mirror-snapshot joint binding. Specifically, the step commitment calculation sub-process first reads the window parameter label, abnormal participation strategy label, and mirror summary information from the model parameter snapshot table, and requests the recalcible summary sequence and source credential pointer back key from the model input mapping package in the order of feature slot index. Based on this, the step commitment calculation sub-process generates a stage commitment record for each fixed stage according to preprocessing, feature assembly, model evaluation, and result orchestration. The stage commitment record includes a stage identifier, input reference key, output reference key, mirror summary reference, parameter snapshot reference, and stage result summary. When a stage performs differentiated execution of the abnormal fragment participation strategy, the stage commitment record includes an abnormal participation label and writes a difference description pointer at the record header. When a stage execution encounters an unavailable window, the stage commitment record writes a window unavailable label in the result summary and adds the window to a separate slice during recalculation slice orchestration. To ensure that the granularity of commitment calculation balances audit recalculation costs and operability, the commitment calculation sub-process generates sub-commitment records within each stage based on time anchor intervals. These sub-commitment records use a combination of time anchor intervals, feature slot segments, and anomaly fragment numbers as keys to form fine-grained recalculation units that can be randomly selected. All commitment records are written to the commitment log after generation. The commitment log records record numbers and parent-child relationships in a sequential manner and generates commitment log pointers for subsequent processes to reference.

[0055] Subsequently, the process concatenation sub-process concatenates all stage commitment records and sub-commitment records. The concatenation order follows the stage order and the arrangement of time anchor intervals. At each concatenation point, a concatenation node description is written. The concatenation node description records the commitment record numbers before and after the concatenation, the mirror digest reference, and the parameter snapshot reference, allowing for breakpoint recalculation at any concatenation node. After concatenation is completed, the process concatenation sub-process generates an evaluation process hash chain. In this embodiment, the evaluation process hash chain points to a sequential structure set of commitment records and concatenation node descriptions. The set is assigned a unique version number and a writing time anchor interval upon writing, and records the reference relationship with the model parameter snapshot table. The evaluation process hash chain writes a commitment log pointer at the beginning of the structure and a key pointing back to the model input mapping package at the end of the structure, so as to quickly locate the mapping evidence when calling across main steps. The recalculation slice list structure, generated synchronously with the evaluation process hash chain, supports the random sampling and task assignment of challenge-response third-party audits. In this implementation, the recalculation slice list structure is a list-based collection. Each item in the collection includes a slice number, slice type, time anchor interval, feature slot segment reference, set of abnormal fragment numbers, required mirror digest reference, required parameter snapshot reference, and expected output reference key. Slice types are distinguished into two categories: stage slices and sub-commitment slices. Stage slices are used for rapid recalculation at a coarser granularity, while sub-commitment slices are used to locate the source of differences at a finer granularity. For the time anchor interval corresponding to the window unavailable label, the recalculation slice list structure lists abnormal slices separately and writes an unavailable description pointer on the entry so that the audit node can provide different processing branches when reading.

[0056] Understandably, after the recalculation slice list structure is generated, the process concatenation sub-processes also write extraction strategy tags in the header. These tags record the random extraction ratio, priority extraction rules, and duplicate extraction control. Priority extraction rules tend to cover slice entries with increased cross-source references, dense abnormal fragments, or version migration tags. Duplicate extraction control is used to prevent resource congestion caused by multiple audit nodes extracting the same entry in a short period. After these actions are completed, the evaluation process hash chain and recalculation slice list structure are recorded as output field names, indicating that these output field names are directly referenced by the evaluation process hash chain and recalculation slice list structure in subsequent step S310 for random slice extraction and recalculation task assignment. Simultaneously, in the context of cross-main steps, this structure is passively referenced by the valuation smart contract when reading process commitments and evidence references, serving as one of the input sources for subsequent evidence chain index binding and report fingerprint generation.

[0057] In one embodiment, the input source for S230 is the model parameter snapshot table output from the preceding step. This model parameter snapshot table has locked the model version mirror and parameter template number, and completed the mirror-snapshot joint binding. The step commitment calculation sub-process first reads the window parameter label, abnormal participation strategy label, and mirror summary information from the model parameter snapshot table, and requests a recalcible summary sequence and source credential pointer back key from the model input mapping package in the order of feature slot indexes. Based on this, the step commitment calculation sub-process generates a stage commitment record for each fixed stage, following preprocessing, feature assembly, model evaluation, and result orchestration. The stage commitment record includes a stage identifier, input reference key, output reference key, mirror summary reference, parameter snapshot reference, and stage result summary. To quantify the completeness of the stage commitment record, a topological sorting constraint is introduced to ensure the verifiability of the stage order. Formula ① is used to calculate the order consistency score of the stage commitment record. Let the set of stage identifiers be... These correspond to four stages: preprocessing, feature assembly, model evaluation, and result orchestration. The time anchor interval sequence is ,in Let be the number of intervals. The summary sequence of feature slot indices extracted from the model input mapping package is denoted as . And extract window parameter labels from the model parameter snapshot table. and abnormal participation strategy tags Consistency score of the phase commitment records. Defined as:

[0058]

[0059] in, It is the first Each stage is marked. This originates from the definition of a fixed stage; It is the first Phase 1 Each time anchor interval takes a positive integer value and is derived from the time anchor elements of the model input mapping package. It is the first Phase 1 The summary value of each interval is a hash summary sequence derived from the feature slot index of the model input mapping package; It is the first The window parameter labels for each stage are real numbers and are derived from the window parameter labels in the model parameter snapshot table. It is the first The abnormal participation strategy label for each stage has an enumerated value, which is derived from the abnormal participation strategy label in the model parameter snapshot table. It is an indicator function that maps policy labels to binary weights; It is the topological sorting weight coefficient, which takes a positive real number and is derived from the system configuration. It is a topological sorting function that calculates the linear score of the stage order; It is the score for the sequential consistency of the phase commitment records; It is a time-anchored range index; It represents the number of time anchor intervals.

[0060] Extract the summary sequence from the model input mapping package and denot it as Extract window parameter labels from the model parameter snapshot table and record them as follows: Together, they form the formula ① Formula ① As an output item of the stage commitment record, it is used for subsequent process concatenation. Subsequently, the step commitment calculation sub-process generates sub-commitment records within the stage according to time anchor intervals. Each sub-commitment record uses a combination of time anchor interval, feature slot segment, and anomaly fragment number as a key to form a randomly extractable recalculation unit. Formula ② is used to calculate the fine-grained aggregate value of the sub-commitment record. Assume the sub-commitment unit is indexed by the time anchor interval. Feature slot segment index and abnormal fragment number Unique identifier, whose summary subset is The weight of the abnormal policy is The fine-grained aggregation value recorded by the sub-commitment. Enhance timing consistency through cross-correlation operations:

[0061]

[0062] in, It is a stage identifier index; It is a time anchor interval index, derived from the time anchor sequence of the model input mapping package; It is a feature slot segment index, which comes from the feature slot index of the model input mapping package; It is the abnormal segment number, which comes from the abnormal flag column of the model input mapping package; It is the first Phase, First Interval, number The first feature segment Each digest value is a hash digest derived from the digest column of the model input mapping package; It is the first Phase, First The policy weights for abnormal segments are real numbers, derived from the abnormal participation policy labels in the model parameter snapshot table; It is a cross-correlation function that calculates the similarity between two summary sequences; It is a fine-grained aggregate value of the sub-commitment record.

[0063] Formula ② directly uses the summary sequence of Formula ① and strategy weights calculate Formula ② This is the output of the sub-commitment record. After all commitment records are generated, they are written to the commitment log. The commitment log records record numbers and parent-child relationships in a sequential manner and generates a commitment log pointer. The output of this section consists of stage commitment records and sub-commitment records, which are then used as inputs to the sub-processes in the workflow.

[0064] Following the aforementioned stage commitment records and sub-commitment records, the process concatenates all commitment records in a sequential order based on the stage sequence and the arrangement of time anchor intervals. A concatenation node description is written at each concatenation point. The concatenation node description records the preceding and following commitment record numbers, mirror digest references, and parameter snapshot references. Formula ③ is used to calculate the network flow weight of the concatenation node description. Let the concatenation node... Corresponding to stage and time interval Its input is the stage commitment record score of formula ①. The aggregate value of the sub-commitment records in Formula ② Network flow weights described by concatenated nodes. Optimization through flow balancing constraints:

[0065]

[0066] in It is a node The set of predecessor nodes, It is the concatenation node index, which takes a positive integer value and is derived from the concatenation order. It is a node The phase markers are derived from the phase commitment records; It is a node The time anchor interval is derived from the time anchor element; The stage commitment record score is obtained from formula ①; It is the aggregated value of the sub-commitment records obtained from formula ②. Index for the predecessor node; It is the network flow weight coefficient, which takes a positive real number and is derived from the system configuration. It is a network flow function that calculates the flow values ​​between nodes; These are the network flow weights described by the concatenated nodes; Take the minimum value.

[0067] Formula ③ directly uses formula ① And formula ② calculate After concatenation, the process concatenation sub-processes generate an evaluation process hash chain, which is a sequential set of commitment records and concatenation node descriptions. Formula ④ is used to calculate the overall integrity hash of the evaluation process hash chain. Let the chain version number be... The time anchor interval is The set of weights for the concatenated nodes is The overall integrity of the hash chain is evaluated during the hash assessment process. Fusing temporal features through convolution operations:

[0068]

[0069] in, This represents the convolution operation. These are convolutional kernels that are time-anchored; It's a version number, generated by the system. It is a time anchor interval, taking values ​​from a time series, derived from the time anchor element; The weights of the concatenated nodes are obtained from formula ③; It is a cryptographic hash function that outputs a fixed-length digest; It is a concatenation operator; It is the overall integrity hash of the evaluation process hash chain; It is a concatenated node index.

[0070] Formula ④ uses formula ③ calculate Simultaneously, a recalculation slice list structure is generated. This structure is a list-based collection containing slice number, slice type, time anchor interval, feature slot segment reference, abnormal segment number set, required mirror summary reference, required parameter snapshot reference, and expected output reference key. The output of this part, the evaluation process hash chain and the recalculation slice list structure, are input to the evaluation process hash chain and the recalculation slice list structure in subsequent step S310.

[0071] Following the aforementioned evaluation process hash chain and recalculation slice list structure, the process connects sub-processes to further optimize the extraction strategy of the recalculation slice list. The recalculation slice list structure includes an extraction strategy label at the header, recording the random extraction ratio, priority extraction rules, and duplicate extraction control. Formula ⑤ is used to calculate the priority extraction weight of the slice. Let the slice... Has slice type (Stage slice is 1, sub-commitment slice is 2), time anchor interval length Abnormal fragment density (Based on the size of the set of abnormal fragment IDs). Weights are extracted preferentially. Enhance anomaly sensitivity through weighted cross-spectral density:

[0072]

[0073] in, It is the power spectral density function, used to quantify the frequency domain characteristics of anomaly density. It is a signal-to-noise ratio function used to evaluate the quality of the abstract sequence; It is a slice index, derived from the recalculated slice list; It is the slice type code, taking the value 1 or 2, which comes from the slice type; It is the length of the time anchor interval, taking a positive real number, and derived from the time anchor interval; It is the density of abnormal fragments, which is derived from the size of the set of abnormal fragment IDs; It is a slice The summary sequence is derived from the model input mapping package; It is the signal-to-noise ratio weighting coefficient, which takes a positive real number and is derived from the strategy configuration; It is the priority extraction weight of the slice.

[0074] Formula ⑤ Used for extraction strategy optimization. Formula 6 is used to calculate the actual extraction probability to avoid resource conflicts. Let the number of audit nodes be... Repeated sampling control factor is Actual sampling probability Constraints through logistic mapping:

[0075]

[0076] in: This is the number of audit nodes, a positive integer value, derived from a third-party audit node register. It is a repeated extraction control factor, with a positive real number value, derived from the extraction strategy label; It is an exponential function; It is the actual sampling probability.

[0077] Formula ⑥ directly uses formula ⑤ calculate Finally, the output evaluation process hash chain and recalculation slice list structure are directly referenced by the S310's evaluation process hash chain and recalculation slice list structure for slice random extraction and recalculation tasks.

[0078] In summary, the technical effects of this step are as follows: By calculating the commitments of each step and connecting the processes, the evaluation activities are solidified into a structured process expression that can be referenced and recalculated. At the same time, recalculation slices that meet the needs of challenge-response auditing are compiled, so that downstream audit recalculation and contract reading and writing have a unified input entry and a consistent reference path.

[0079] Step S300 includes at least steps S310-S330:

[0080] S310. Obtain the hash chain and recalculation slice list structure of the evaluation process, perform random slice extraction and recalculation task assignment processing, and obtain the audit recalculation task package.

[0081] The input sources for this section are the evaluation process hash chain and the recalculation slice list structure output from the previous steps. The evaluation process hash chain records the stage commitment record, sub-commitment record, concatenated node description, commitment log pointer, and key pointing back to the model input mapping package. The recalculation slice list structure records the slice number, slice type, time anchor interval, feature slot segment reference, abnormal segment number set, required mirror summary reference, required parameter snapshot reference, and expected output reference key, and writes the extraction strategy label in the header. Specifically, the evaluation process hash chain and recalculation slice list structure are loaded into the extraction scheduling module. The extraction scheduling module constructs a candidate set based on the random sampling ratio, priority sampling rules, and duplicate sampling control of the extraction strategy labels. During the candidate set construction process, batches are divided according to time anchor intervals, a priority sequence is established according to the density of abnormal segments and version migration labels, and a suppression factor is established for duplicate slices of the same feature slot segment. When recalculation resources are tight, the extraction scheduling module reads the resource threshold configuration, which comes from the scheduling configuration library, and records the concurrency limit, queue depth, and number of retries. Based on this, the extraction scheduling module dynamically shrinks the size of the candidate set and registers the pruning reason in the log. Subsequently, the scheduling module invokes the node orchestration sub-process. This sub-process reads the third-party audit node roster and node health status view. The third-party audit node roster records node identifiers, node public keys, node roles, and node geographic tags. The node health status view records latency, throughput, failure rate, and the number of active sessions. The node orchestration sub-process assigns candidate slices according to geographic tag partitioning and health status scores. The assignment results are written to the task orchestration table, which includes task number, slice number, assigned node identifier, retry window, and acknowledgment deadline. After the task orchestration table is generated, the task distribution sub-process distributes tasks through a dual-channel approach: Message Queue (MQ) and Application Programming Interface (API). The MQ channel handles batch distribution, while the API channel handles acknowledgment notifications and exception feedback. During distribution, a distribution time anchor and routing identifier are written. When a node acknowledgment times out or is rejected, the task distribution sub-process performs a secondary assignment according to the retry window. This secondary assignment is recorded in the task orchestration table, indicating the reason for the source switch. After extraction and assignment are completed, the system constructs an audit recalculation task package. In this embodiment, the audit recalculation task package refers to a structured carrier consisting of a task list, slice details, mirror summary reference, parameter snapshot reference, expected output reference key, session number, and return address. The session number is bound to the extraction batch, and the return address is used to receive the recalculation result summary and process log.Ultimately, the audit recalculation task package is registered as an output field name and is called by the audit recalculation task package input location in the next subsection S320; at the same time, in the cross-main step context, the audit recalculation task package is passively referenced by the evidence chain index binding link of the main step S400 to record the pointer relationship between the recalculation session and the subsequent report fingerprint.

[0082] S320. Extract the recalculation results from the audit recalculation task package, perform source signature verification and process consistency comparison, and generate an audit comparison record table.

[0083] The input source for this section is the audit recalculation task package output from the previous section. This package carries slice details, image digest references, parameter snapshot references, and expected output reference keys. Specifically, after receiving the audit recalculation task package, the third-party audit node loads the model version image pointed to by the image digest reference through its node execution container and reads the parameter set pointed to by the parameter snapshot reference. In this embodiment, the node execution container is an isolated runtime environment that records the container instance number, image digest, snapshot signature material pointer, and startup time anchor. Subsequently, the node locates the digest sequence and source credential pointer from the model input mapping package backreference key according to the time anchor interval and feature slot segment reference in the slice details. This location process is accessed through a read-only channel and does not request any plaintext fields. After successful localization, the node triggers a recalculation sub-process. This sub-process executes either stage-level recalculation or sub-commitment-level recalculation based on the slice type. Stage-level recalculation executes each of the four stages—preprocessing, feature assembly, model evaluation, and result orchestration—segment by segment, generating a stage result summary. Sub-commitment-level recalculation performs local evaluation on specific time anchor intervals and feature slot segments, generating a sub-result summary. All result summaries are written to the node's local recalculation log. The recalculation log records the result summary, container instance number, image summary, parameter snapshot reference, and time anchor interval. After generation, the recalculation log is sent back to the comparison endpoint via API. The comparison endpoint loads the source signature verification sub-process. This sub-process verifies the source credential pointer in the returned record. Signature verification requires reading the certificate serial number, signature summary, and signature time, and cross-matching them with the source credential pointer record in the minimum verifiable input set structure. Entries that fail signature verification are marked as source anomalies, and an anomaly reason pointer is recorded. Subsequently, the consistency comparison sub-process runs within the Consistency Checking Engine (CCE). CCE locates the corresponding commitment record from the evaluation process hash chain and compares the recalculated stage result summary or sub-result summary with the stage result summary or sub-result summary in the commitment record. The comparison method employs a dual-channel approach: summary matching and structure matching. Summary matching compares summary values, while structure matching compares field order, time anchor intervals, and anomaly participation tags. When a mismatch occurs, CCE generates a difference description pointer and a difference classification tag. The difference classification tag distinguishes between situations such as source anomalies, inconsistent container images, parameter snapshot differences, time anchor misalignment, and slice boundary offsets. To form a cross-node readable evidence trail, CCE writes the input reference key, expected output reference key, recalculated result summary, and comparison status of each comparison unit into the comparison cache and generates a comparison unit record number at the end of the unit. Within the same audit session, the comparison unit record numbers are arranged in time anchor order and bound to the session number for easy subsequent aggregation.After completing the source signature verification and process consistency comparison, the system constructs an audit comparison record table. In this embodiment, the audit comparison record table is a structured table that records the task number, slice number, comparison unit record number, comparison status, difference classification label, difference description pointer, node signature summary, and time anchor sequence. The table generates a version number and marks the receipt deadline upon writing. Finally, the audit comparison record table is registered as an output field name and is used for the input position of the audit comparison record table in the next subsection S330. Simultaneously, in the cross-main step context, the audit comparison record table is passively referenced by the contract trigger verification step in the main step S400, serving as a preliminary verification basis for the audit-passed document structure.

[0084] S330. Perform threshold signature aggregation and result adjudication on the audit comparison record table to generate an audit pass document structure;

[0085] The input source for this section is the audit comparison record table output from the previous section. The audit comparison record table records the task number, slice number, comparison unit record number, comparison status, difference classification label, difference description pointer, node signature summary and time anchor sequence. Specifically, the aggregation orchestration sub-process first reads all node signature digests and corresponding comparison statuses from the audit comparison record table, and establishes an aggregation view according to session number, slice number, and time anchor sequence. Subsequently, the aggregation orchestration sub-process calls the Signature Aggregation Service (SAS) to aggregate signature digests from different audit nodes within the same session. SAS reads the threshold adjudication policy, which records the minimum number of signatures, geographical dispersion requirements, and node role matching requirements, and records the upper limit of abnormal concentration. When the same slice has a highly consistent failure status within a short period of time and the source is concentrated in a single region or single role, SAS writes a concentration alarm label in the aggregation view and triggers a review path. The review path reads the task orchestration table to add a backup node for the slice. When the aggregation reaches the minimum number of signatures and dispersion requirements of the threshold adjudication policy, SAS outputs the aggregated signature and preliminary adjudication judgment. After the initial ruling is generated, the result ruling subprocess summarizes the comparison status of each slice. The summarization logic is executed in order of the comparison unit record number. If all comparison unit records are in a pass status, the slice is marked as pass. If there is a difference classification label, but the difference description pointer points to an inconsistent container image and the image digest references a change number already registered in the corresponding whitelist, the slice enters the buffer queue, which waits for the image whitelist to be synchronized before making a ruling. If there is a source anomaly or parameter snapshot difference mark, the slice is marked as fail and a reason summary is written. After completing the result ruling at the slice level, the result ruling subprocess continues to summarize at the session level, generating a session-level pass rate, a session-level difference distribution summary, and a session-level anomaly group overview. The summary results are written to the ticket generator along with the aggregated signature. The invoice generator constructs an audit pass invoice structure, which in this implementation is a structured credential. It records the aggregate signature, threshold adjudication strategy number, session number, pass / fail slice list, fail / reject slice list, buffer queue list, evaluation process hash chain reference, model parameter snapshot table reference, and time anchor interval. The invoice generator writes the issuance time anchor and issuer identifier at the structure header and a publicly verifiable fingerprint digest at the structure tail, establishing references to the commitment log pointer and comparison cache for subsequent traceability. Finally, the audit pass invoice structure is registered as an output field name and is called at the S410 audit pass invoice structure input position of main step S400 to generate the valuation calculation trigger package. Simultaneously, in cross-main step contexts, the audit pass invoice structure is recorded as a prerequisite credential for the report fingerprint by the evidence chain archiving stage and, when needed, refers back to the evaluation process hash chain and model parameter snapshot table for verification.In summary, the technical effects of this step are as follows: by aggregating threshold signatures and adjudicating results, a verifiable credential expression with adjudication basis is formed, and a strict reference relationship is established between the audit results of the session layer and the slice layer and the existing commitment structure, thereby providing a unified credential entry point for subsequent contract triggering and evidence archiving.

[0086] Step S400 includes at least steps S410-S430:

[0087] S410. Obtain the audited invoice structure, perform contract trigger verification and parameter snapshot reading processing to obtain the valuation calculation trigger package;

[0088] The input source for this section is the audit pass ticket structure output from the previous steps. The audit pass ticket structure records the aggregate signature, threshold adjudication strategy number, session number, pass slice list, fail slice list, buffer queue list, evaluation process hash chain reference, model parameter snapshot table reference, and time anchor interval. Its structure header contains the issuance time anchor and issuer identifier, and the structure tail carries a publicly verifiable fingerprint digest. Specifically, the audit is loaded into the contract-triggered verification unit through the bill structure. The contract-triggered verification unit first verifies the aggregated signature. The verification action reads the number of signatures corresponding to the threshold adjudication strategy number, the geographical distribution requirements, and the node role matching requirements, and verifies and deduplicates the node signature digests in the aggregated signature item by item. When any node signature digest does not match the registered node public key or is on the revocation list, the node signature digest is marked as invalid and written into the invalid signature segment of the verification log. At the same time, the invalidity reason pointer is retained in the verification buffer for subsequent review. When the distribution requirements of the aggregated signature are not met, the contract-triggered verification unit adds a supplementary verification instruction without interrupting the current process. The geographical label to be verified and the node role matching are written into the supplementary verification suspension segment and the time anchor is recorded synchronously. The supplementary verification result is incorporated into the incremental record of the same session number in subsequent sessions. Furthermore, the contract trigger verification unit performs a consistent cross-read of the evaluation process hash chain reference through the slice list. In this embodiment, the consistent cross-read refers to sampling and verifying the stage commitment record, sub-commitment record, and concatenated node description of each slice entry, and writing the sampling results to the cross-read log. When the sampling results deviate from the slice status specified in the audit-passed document structure, the deviating entry is transferred to the mitigation queue list and triggers a secondary verification. The secondary verification results are incrementally appended to the same session number. Subsequently, the parameter snapshot reading unit reads the model version identifier, mirror digest, parameter template number, window parameter label, abnormal participation strategy label, and snapshot signature material pointer according to the model parameter snapshot table reference, and verifies the snapshot signature material pointer. When the snapshot signature is in an expired state or the certificate serial number appears in the revocation list, the parameter snapshot reading unit registers the snapshot as pending update and marks it with a rollback flag in this contract trigger verification. The rollback flag is bound to the mirror digest reference and parameter template number of the previous valid snapshot, allowing the previous valid snapshot to be called to perform business calculations without changing the session number. Understandably, the contract trigger verification unit also performs window overlap checks on the session number and time anchor interval. The window overlap check records the overlap relationship between the current valuation round and the previous valuation round on the time anchor. Overlapping entries are marked as iteration areas. During the valuation calculation stage, the iteration area entries are subject to deduplication and merging rules and the source mapping segment is marked in the evidence list.After the above verification actions are completed, the system enters the trigger package orchestration stage. This stage involves structurally merging the input elements required for contract execution, specifically including the session number, contract call identifier, image digest reference, parameter snapshot reference, the truncation range of the evaluation process hash chain reference, the digest set from the slice list, the rollback flag and iteration zone flag, resource scheduling parameters, and the valuation result return address. After structural merging, a valuation calculation trigger package is formed. The trigger time anchor and trigger source identifier are written to the trigger package header, and the trigger package fingerprint digest is written to the trigger package tail to support subsequent peer verification. Finally, the output of this section is recorded as the output field name valuation calculation trigger package, indicating that this output field name is directly referenced by the valuation calculation trigger package in subsequent step S420. Simultaneously, in the cross-main step context, the valuation calculation trigger package is indexed and mapped by the evidence chain archiving subsystem to locate the association path between this contract trigger and the preceding audit session during the report fingerprint generation stage.

[0089] S420. Extract parameter snapshots and process commitments from the valuation calculation trigger package, perform asset value calculation and confidence interval generation, and generate valuation results and evidence list.

[0090] The input source for this section is the valuation calculation trigger package output from the previous section. This package contains the session number, image digest reference, parameter snapshot reference, the truncation range of the evaluation process hash chain reference, the digest set from the slice list, the rollback flag and iteration zone flag, resource scheduling parameters, and the valuation result return address. Specifically, the contract execution environment is a Virtual Machine (VM). After receiving the valuation calculation trigger package, the VM first executes the parameter parsing process. The parameter parsing process reads the parameter snapshot reference and verifies it on the chain. After successful verification, a parameter shadow zone is constructed within the contract. The parameter shadow zone stores window parameters, exception participation strategies, feature participation strategies, and aggregation rule strategies according to parameter template numbers, and establishes deduplication and merging rules for the iteration zone flag. When the trigger package carries a rollback flag, the parameter parsing process establishes a rollback mapping in the shadow zone based on the previous valid snapshot information of the rollback flag. The rollback mapping is only invoked when the current snapshot verification fails or the image digest reference is missing. Subsequently, the contract reads the commitment records and concatenation node descriptions within the truncation range referenced by the hash chain of the evaluation process, and constructs a process commitment view within the contract. The process commitment view is organized by stage and time anchor interval, and the commitment records and concatenation node descriptions are marked with callable positions in the view. After the process commitment view is constructed, the contract locates the summary sequence from the back-pointing key of the model input mapping packet according to the feature participation strategy and abnormal participation strategy of the parameter shadow area. The location action does not read any plaintext fields, but only reads the summary and source credential pointer, and organizes the located summary sequence into a set of valuation input fragments according to the window parameters. After the set of valuation input fragments is established, the contract enters the asset value calculation stage. In this embodiment, the asset value calculation stage refers to calling the evaluation logic identified by the mirror digest reference within the VM, evaluating each fragment of the valuation input fragment set and generating fragment valuation results. When the abnormal participation strategy sets a participation scope restriction for a certain fragment, the fragment is evaluated only within the participation scope and a fragment valuation result with an abnormal participation tag is generated. The tag is retained in the evidence list. When the iteration zone marker exists, the asset value calculation stage performs deduplication and merging rules on the iteration zone entries, generates merged fragment valuation results for duplicate fragments in the same time anchor interval, and records the deduplication and merging description pointer. Furthermore, the trusted interval generation step calls the trusted interval generation logic within the VM. The trusted interval generation logic performs multiple rounds of resampling and aggregation on the fragment estimation results based on the aggregation rule strategy and window parameters configured in the parameter shadow area. The resampling and aggregation process does not introduce plaintext fields, and uses structured references of commitment records and summary sequences to carry out repeated sampling. In each round of resampling, the sampling round mark and commitment node reference are recorded. After aggregation, a structured description of the boundary of the estimated trusted interval is obtained. For slice entries with references in the buffer queue list, the trusted interval generation step skips unresolved entries during the aggregation process and writes a pending explanation segment in the evidence list. The pending explanation segment carries the secondary verification reference and the truncation range.After the asset value calculation and confidence interval generation are completed, the contract compiles the valuation results and evidence elements. In this embodiment, the valuation results point to the structural expression of the asset valuation point value and the confidence interval boundary. The evidence elements include parameter snapshot references, mirror summary references, process commitment view capture range, fragment valuation result summary sequence, abnormal participation tags, and deduplication and merging description pointers. The contract organizes the above evidence elements into an evidence list. The evidence list writes the session number and trigger time anchor at the beginning of the structure and the evidence list fingerprint summary at the end of the structure, and establishes a bidirectional reference to the commitment log pointer and the model parameter snapshot table. Understandably, the contract performs resource reclamation and state registration before output. Resource reclamation releases the shadow area and process commitment view cache, and state registration records the status code, time fragment, and abnormal overview of this contract execution, which are used to reference the execution status when generating the report fingerprint later. Finally, the valuation results and evidence list output by the contract are recorded as output field names, and it is indicated that the output field names are directly referenced by the valuation results and evidence list in the subsequent step S430. At the same time, in the cross-main step context, this output is passively read by the evidence chain archiving module and the evidence storage contract to complete the report fingerprint generation and evidence chain index binding.

[0091] In one embodiment, the input source of S420 is the valuation calculation trigger package output in the previous section. This package contains the session number, image digest reference, parameter snapshot reference, the truncation range of the evaluation process hash chain reference, the digest set of the slice list, the rollback flag and iteration zone flag, resource scheduling parameters, and the valuation result return address. Specifically, the contract execution environment is a contract virtual machine (VM). After receiving the valuation calculation trigger package, the VM first executes the parameter parsing process. The parameter parsing process reads the parameter snapshot reference and verifies it on the chain. After successful verification, a parameter shadow area is constructed within the contract. The parameter shadow area stores window parameters, anomaly participation strategies, feature participation strategies, and aggregation rule strategies according to parameter template numbers, and establishes deduplication and merging rules for the iteration zone flag. To quantify the parameter loading quality of the parameter shadow area, a parameter loading score is introduced. Formula ⑦ is used to calculate the parameter loading score. Let the parameter set contained in the parameter shadow area be... , where each parameter The value of this entry corresponds to one of the window parameters, exception participation strategy, feature participation strategy, or aggregation rule strategy, and is derived from the actual value loaded after the parameter snapshot references the signature. Parameter loading score The deviation between the parameter values ​​and the template standard values ​​is calculated using a weighted sum of squares:

[0092]

[0093] in, The number of parameters in the parameter shadow area is determined by the parameter template number; It is the parameter index, with values ​​from 1 to... ; It is the first The actual value of each parameter comes from the value loaded after the parameter snapshot references the signature verification. It is the first The template standard value of each parameter, which takes the value of a real number, is derived from the parameter template; It is the first The weighting factor of each parameter takes the value of a positive real number and comes from the priority setting of the parameter template; This is the parameter loading score, and its value is a non-negative real number. The estimation calculation triggers the extraction of a parameter snapshot reference, which is then loaded and recorded after signature verification. The standard value is extracted from the parameter template and recorded as... Together, they form the formula in ⑦. .

[0094] Formula ⑦ As an indicator of the quality of the parameter shadow area, it is used in the subsequent construction of the process commitment view. Subsequently, the contract reads the commitment records and concatenation node descriptions within the truncation range referenced by the evaluation process hash chain, and constructs the process commitment view within the contract. The process commitment view is organized by stage and time anchor interval, and the commitment records and concatenation node descriptions are marked with their callable positions in the view. Formula ⑧ is used to calculate the integrity metric of the process commitment view. Let the set of commitment records be... The set of concatenated node descriptions is ,in and It's an index. Integrity metric. Topology sorting score and network flow balance constraints based on commitment records and node descriptions:

[0095]

[0096] in, It is a commitment record index, derived from the commitment record sequence of the hash chain in the evaluation process; It is the concatenated node description index, which takes a positive integer value and comes from the concatenated node description sequence of the hash chain in the evaluation process; It is the first Each commitment record includes elements such as stage identifiers and input reference keys; It is the first Each serial node is described, recording information such as the previous and subsequent commitment record numbers; It is a topological sorting function that calculates the order score of records and takes a real number value. It is a network flow function that calculates the flow values ​​between nodes and takes real numbers. These are network flow weight coefficients, derived from system configuration; It is a measure of the integrity of the process commitment view, and takes a real number value.

[0097] Formula ⑧ directly uses the parameters of Formula ⑦ to load the score. As a weighting adjustment factor, i.e. ,in These are the basic weights. Formula ⑧ As an output of view construction quality, this section outputs the shadow area and process commitment view, which are inputs to the asset valuation stage.

[0098] Following the aforementioned parameter shadow area and process commitment view, the contract locates the summary sequence from the model input mapping packet's back-pointing key based on the characteristic participation strategy and abnormal participation strategy of the parameter shadow area, and organizes the located summary sequence into a valuation input fragment set according to window parameters. After the valuation input fragment set is established, the contract enters the asset value calculation stage. This stage calls the evaluation logic identified by the mirrored summary reference within the VM to evaluate each fragment of the valuation input fragment set. Formula ⑨ is used to calculate the fragment valuation result. Assume the valuation input fragment set contains... Each segment For a given time anchor interval, its summary sequence is as follows: ,in This is the length of the abstract sequence. Fragment estimation results. The weights are calculated through a weighted linear combination, and the weights are derived from the feature participation strategy of the parameter shadow region.

[0099]

[0100] in, It is the index of the estimated input fragment, with a value from 1 to... It originates from the set of valuation input fragments; This is the number of segments, a positive integer value determined by the window parameters; It is a summary sequence index, with values ​​from 1 to... ; It is the length of the abstract sequence, a positive integer, determined by the feature slot segment; It is a fragment The Each digest value is a hash digest derived from the digest column of the model input mapping package; These are weights, which take real numbers and originate from the feature participation strategy of the parameter shadow region; It is an exception participation tag, with a binary value derived from the exception participation strategy; It is an indicator function that returns 0 or 1; It is the result of fragment estimation, and the value is a real number.

[0101] Formula 9 directly uses the integrity measure of Formula 8. As a weight scaling factor, i.e. ,in These are the basic weights. Formula 9 This serves as the fragment estimation output. Further, the confidence interval generation stage calls the confidence interval generation logic within the VM, performing multiple rounds of resampling and aggregation on the fragment estimation results based on the aggregation rule strategy and window parameters configured in the parameter shadow region. Formula 10 is used to calculate the confidence interval boundary. Let the fragment estimation result set be... The number of resampling rounds is Each resampling from Randomly select samples from the dataset, and use the aggregation function as follows: Confidential interval boundary Defined as the quantile of the resampled aggregate value:

[0102]

[0103] in, It represents the resampling round, with a positive integer value, derived from the aggregation rule strategy; It is a resampling function that randomly samples from the fragment estimation results; It is a quantile function that calculates a specified quantile of aggregate values ​​and takes real numbers. It represents the boundary of the reliable interval and takes the value of a real number.

[0104] Formula 10 directly uses the fragment estimation result of Formula 9. As input. Formula 10 This section outputs the estimation results and confidence interval boundaries of the output fragments, which are then input into the evidence list compilation stage.

[0105] Following the aforementioned fragment valuation results and confidence interval boundaries, the contract compiles the valuation results and evidentiary elements. The valuation results point to the structural expression of the asset valuation point value and the confidence interval boundary. Evidence elements include parameter snapshot references, mirror summary references, process commitment view capture ranges, fragment valuation result summary sequences, abnormal participation tags, and deduplication / merging instructions. The contract organizes these evidentiary elements into an evidence list and writes an evidence list fingerprint summary at the end of the structure. Formula 11 is used to calculate the evidence list fingerprint summary. Let the asset valuation point value be... It is the aggregated value (such as the average) of the fragment estimation results, and the confidence interval boundary is... The set of evidence elements is List of Evidence, Fingerprint Summary Calculated using a cryptographic hash function:

[0106]

[0107] in, Indicates the concatenation operator; It is the asset valuation point value, which is derived from the aggregation of fragment valuation results from formula ⑨; It is the boundary of the confidence interval, derived from formula 10; These are evidentiary elements, including parameter snapshot references, etc. It is a cryptographic hash function that outputs a fixed-length digest; It is a fingerprint summary of the evidence list.

[0108] Formula 11 can be used directly from Formula 9. And formula 10 Formula 11 This serves as a completeness indicator for the evidence list. Ultimately, the valuation result and evidence list are output and directly referenced by the valuation result and evidence list in subsequent step S430.

[0109] This section summarizes the technical effects: By implementing parameter shadow areas and process commitment views, the verifiability of the valuation process and minimal data exposure are ensured; asset value calculation and confidence interval generation are based on commitment-based resampling, outputting verifiable results and a list of evidence, forming a closed loop of post-audit triggering.

[0110] S430. Generate a report fingerprint and bind the evidence chain index to the valuation result and the evidence list to generate a report fingerprint and a traceable archive structure.

[0111] The input source for this section is the valuation results and evidence list output from the preceding steps. These results and evidence list carry a session number, trigger time anchor, parameter snapshot reference, mirror summary reference, process commitment view cutoff range, fragment valuation result summary sequence, abnormal participation tag, deduplication and merging explanation pointer, and evidence list fingerprint summary. Specifically, the report generation unit first constructs a report assembly view. In this embodiment, the report assembly view places the asset valuation point value and the confidence interval boundary of the valuation results in a unified display area, places the parameter snapshot reference, mirror summary reference, and process commitment view cutoff range in the evidence area, and places the fragment valuation result summary sequence, abnormal participation tag, and deduplication and merging explanation pointer in the notes area. During the assembly process, the report generation unit performs a consistency check on the reference path of the evidence list. The consistency check reads the commitment log pointer and the model parameter snapshot table, checking whether the commitment record and snapshot signature material pointer are consistent with the references recorded in the evidence list. Items that fail the check are transferred to the temporary revision area with a correction reason pointer. The temporary revision area does not change the valuation results; it only presents the revision explanation at the report level. Subsequently, the fingerprint generation unit generates a report summary sequence based on the structured content of the report assembly view. The report summary sequence is processed sequentially according to the display area, evidence area, and notes area. The summary value generated by the sequential processing is coupled with the fingerprint summary of the evidence list. A time anchor, session number, and contract call identifier are written in the header, and the fingerprint summary and public verification entry are written in the tail. In this embodiment, the public verification entry points to a read-only interface provided to the outside world. It is exposed in the form of an Application Programming Interface (API). The first call requires the registration of the visitor role and access window. The interface only returns the summary value, reference key, and time anchor, and does not return any plaintext fields. After fingerprint generation is complete, the evidence chain index binding unit begins to execute index building and binding actions. The index building action establishes a multi-level index relationship between the report fingerprint and the minimum verifiable input set structure, the model input mapping package, the evaluation process hash chain and the recalculation slice list structure, the audit pass document structure, and the model parameter snapshot table. The multi-level index uses the session number and time anchor interval as a composite key at the outermost layer, the reference key and the structure version number as a composite key in the middle layer, and the record number of the fragment valuation result summary sequence as the positioning key in the inner layer. The binding action places the above multi-level indexes item by item into the evidence chain archiving module. The evidence chain archiving module interacts with the evidence storage contract, writes the report fingerprint, index relationship and necessary reference keys on the chain, and registers the back pointer position of the off-chain evidence list and commitment log pointer in the archive record. The back pointer position adopts a read-only storage strategy and carries an access control flag, which restricts the access role and access window.Furthermore, to support subsequent version traceability, the report generation unit writes a version migration tag in the report header. When the current parameter snapshot or mirror summary is out of sync with the previous report, the version migration tag records the migration reason pointer and a difference overview. When there are pending items in the mitigation queue list, the report assembly view places the pending description section in the notes area and reserves an incremental slot for the pending description section in the evidence chain index binding stage. The incremental slot can be incrementally written after subsequent secondary verification. After completing the above operations, the system outputs the report fingerprint and traceable archive structure, and indicates in the output statement that the output field name can be called by any subsequent replay verification session, and is also passively referenced as an external verification entry point in the upper-level asset securitization scenario. In the cross-main step context, the report fingerprint and traceable archive structure are read by the review calls of main step S200 and main step S300, and are used to provide historical references for version requirement analysis, extraction strategy labeling, and threshold strategy review in future rounds. In summary, the technical effects of this step are as follows: By generating report fingerprints and binding them to the evidence chain index, the valuation results and evidence throughout the process are linked together to form a unified reference system on and off the chain. This completes the closed-loop expression from audit vouchers to contract execution and report solidification, and provides a stable entry point for subsequent replay verification and historical comparison.

Claims

1. A blockchain-based financial valuation method for power assets, characterized in that, include: Acquire asset operation monitoring data, operation and maintenance record mirrors and metering-settlement account mirrors, perform field standardization, source signature binding, time anchor registration, key field extraction, irreversible hash calculation, source voucher pointer generation, minimum exposure filtering and index construction processing, and generate a minimum verifiable input set structure; Obtain the minimum verifiable input set structure, perform evaluation model input mapping, abnormal fragment annotation, model version requirement extraction, model version mirror locking, parameter snapshot registration, step commitment calculation and process concatenation processing, and generate an evaluation process hash chain and recalculation slice list structure including stage commitment records, sub-commitment records, concatenation node descriptions, commitment log pointers and recalculation slice list structures. Obtain the hash chain and recalculation slice list structure of the evaluation process, perform random slice extraction, recalculation task assignment, recalculation result extraction, source signature verification, process consistency comparison, threshold signature aggregation and result adjudication processing, and generate an audit pass invoice structure. The audit process involves obtaining the audited invoice structure, executing contract trigger verification, reading parameter snapshots, extracting parameter snapshots and process commitments, calculating asset value and generating a trusted range in the contract execution environment, generating report fingerprints and binding evidence chain indexes, and generating report fingerprints and traceable archive structures.

2. The method according to claim 1, characterized in that, Asset operation monitoring data, operation and maintenance record mirroring, and metering-settlement voucher mirroring include: The asset operation monitoring data refers to a multi-dimensional record set from status monitoring, alarm records, fault conditions, environmental sampling, and inside and outside the maintenance window, with data granularity covering all layers of measurement points, loops, equipment, and asset identification; the operation and maintenance record mirror refers to a read-only snapshot of work orders, replacement parts registrations, inspection logs, and defect closure loops formed on the maintenance platform at a specified time point; the metering-settlement account mirror refers to a read-only snapshot of electricity consumption records, billing rule mappings, reconciliation voucher images, and settlement lists formed from the metering acquisition platform and financial settlement module at a specified time point.

3. The method according to claim 1, characterized in that, The minimum verifiable input set structure includes: The minimum verifiable input set structure includes a summary column, a source credential pointer column, a time anchor element column, an anomaly flag column, a verification status column, and a multi-level index.

4. The method according to claim 1, characterized in that, The evaluation process hash chain and recalculation slice list structure includes: The evaluation process hash chain includes stage commitment records, sub-commitment records, concatenated node descriptions, and commitment log pointers; the recalculation slice list structure includes slice number, slice type, time anchor interval, feature slot segment reference, and abnormal segment number set.

5. The method according to claim 1, characterized in that, The audit approved the following document structure: The audit process uses a document structure that includes an aggregate signature, a threshold decision policy number, a session number, a pass / fail slice list, a fail / no slice list, a buffer queue list, an evaluation process hash chain reference, a model parameter snapshot table reference, and a time anchor interval.

6. The method according to claim 1, characterized in that, The process of calculating asset value and generating trusted intervals, generating report fingerprints, and binding evidence chain indexes in a contract execution environment includes: The contract execution environment is the contract virtual machine. After receiving the valuation calculation trigger package, the VM executes the parameter parsing process. The parameter parsing process reads the parameter snapshot reference and verifies it on the chain. After the verification is successful, a parameter shadow area is built in the contract. The parameter shadow area stores window parameters, abnormal participation strategies, feature participation strategies and aggregation rule strategies according to the parameter template number, and establishes deduplication and merging rules for the iteration area marker.

7. The method according to claim 6, characterized in that, The contract reads the commitment records and concatenated node descriptions within the truncation range referenced by the hash chain of the evaluation process, and constructs a process commitment view within the contract. The process commitment view is organized by stage and time anchor interval. The contract locates the summary sequence from the back-pointing key of the model input mapping packet according to the feature participation strategy and abnormal participation strategy of the parameter shadow area, and organizes the located summary sequence into a set of valuation input fragments according to the window parameters.

8. The method according to claim 1, characterized in that, The process of calculating asset value and generating trusted intervals, generating report fingerprints, and binding evidence chain indexes in the contract execution environment also includes: The asset valuation calculation process involves calling the evaluation logic identified by the image digest within the VM, evaluating each piece of the valuation input fragment set, and generating fragment valuation results. When the abnormal participation strategy sets a participation scope restriction for a certain fragment, the fragment is evaluated only within the participation scope, and a fragment valuation result with an abnormal participation tag is generated. When the iteration zone marker exists, the deduplication and merging rules are executed on the iteration zone entries, and a merged fragment valuation result is generated for duplicate fragments in the same time anchor interval, and the deduplication and merging description pointer is recorded.

9. The method according to claim 8, characterized in that, The confidence interval generation process calls the confidence interval generation logic within the VM. Based on the aggregation rule strategy and window parameters configured in the parameter shadow area, it performs multiple rounds of resampling and aggregation on the fragment estimation results. The resampling and aggregation process uses the structured reference of commitment records and summary sequences to carry out repeated sampling. In each round of resampling, the sampling round mark and commitment node reference are recorded. After the aggregation is completed, a structured description of the boundary of the confidence interval is obtained.

10. The method according to claim 9, characterized in that, The contract compiles valuation results and evidence elements, which include parameter snapshot references, mirror summary references, process commitment view capture range, fragment valuation result summary sequence, abnormal participation tags, and deduplication and merging description pointers. The contract organizes the above evidence elements into an evidence list. The evidence list has a session number and trigger time anchor written in the structure header, an evidence list fingerprint summary written in the structure tail, and a two-way reference to the commitment log pointer and the model parameter snapshot table is established.

Images