Intelligent control method for cost deviations in power plant retrofit projects oriented towards cloud platforms

Abstract

This invention relates to the field of power engineering cost control technology, specifically to an intelligent cost deviation control method for power plant renovation projects oriented towards a cloud platform. The method includes: establishing a cost monitoring process on an industrial cloud platform encompassing deviation detection, impact tracing, and control decision-making; identifying the initiation point and propagation path of cost deviations during the deviation detection stage; integrating the transmission relationship of project sub-items deviations with external market data flow disturbance signals during the impact tracing stage to map the multi-level impact chain of cost deviations; and analyzing the contribution weight and control sensitivity of deviations at each level during the control decision-making stage to generate a differentiated pre-control instruction set, collecting the cost flow response trajectory after instruction execution, iteratively correcting instructions based on their matching degree with the expected control trajectory, and updating the control strategy online. This method can fully present the deviation attribution logic, adapt control instructions to the characteristics of sub-items, achieve closed-loop intelligent control of cost deviations, and enhance the dynamic adaptive optimization capability of the control process.

Description

Technical Field

[0001] This invention relates to the field of power engineering cost control technology, and in particular to an intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms. Background Technology

[0002] Traditional power plant renovation projects often rely on local management systems for cost control, employing post-project deviation accounting and manual traceability. Cost personnel manually compare actual project costs with benchmark cost plans, and the impact of external market data on costs is only analyzed manually, without deep integration with the project's own cost control processes. This type of control model can only identify established cost deviations, failing to capture the emergence and spread of deviations in real time. The analysis of cost deviations is limited to the internal dimensions of a single project item.

[0003] Existing cost control schemes fail to integrate the transmission relationships of deviations between project sub-items with disturbance signals from external market data flows, making it difficult to form a complete multi-level impact chain of cost deviations. The depth and breadth of deviation tracing are limited. Cost control instructions often adopt a standardized formulation model, failing to differentiate settings based on the characteristics of deviation impacts at each level. After instruction execution, only static result verification is conducted, unable to collect dynamic response data of actual cost flows and match it with expected trajectories. Control strategies cannot achieve online iterative optimization, and cost deviation control remains in an open-loop operation.

[0004] It is necessary to rely on the industrial cloud platform to complete the multi-level impact chain mapping of cost deviation transmission relationship and external market disturbance signals. At the same time, it is necessary to complete the iterative correction of pre-control instructions and the online update of cost control strategies by matching the actual cost flow response trajectory with the expected trajectory, and build a closed-loop intelligent cost deviation control system. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms, comprising: Build a project cost monitoring process in the industrial cloud platform, including the deviation detection stage, the impact tracing stage, and the control and decision-making stage; During the deviation detection phase, the actual project cost flow continuously aggregated on the industrial cloud platform is compared with the benchmark cost plan to identify the initiation point and diffusion path of cost deviation. During the impact tracing phase, based on the diffusion path of the cost deviation, the transmission relationship of the cost deviation between project sub-items is tracked, and the disturbance signal generated by the external market data flow on the benchmark cost of project sub-items is monitored. Combining the transmission relationship of the cost deviation between project sub-items and the disturbance signal generated by the external market data flow, the multi-level impact chain formed by the cost deviation is mapped in the industrial cloud platform. During the regulatory decision-making stage, based on the multi-level impact chain of the cost deviation, the deviation contribution weight and regulatory sensitivity of each level in the chain are analyzed. Based on the aforementioned deviation contribution weight and control sensitivity, differentiated pre-control instruction sets are generated for different levels of projects in the industrial cloud platform; The pre-control instruction set is executed in the industrial cloud platform, and the response trajectory of the actual project cost flow is collected simultaneously. The matching degree of the response trajectory and the expected control trajectory is calculated. The pre-control instruction set is iteratively corrected according to the matching degree result to generate the final state control instruction. The cost control strategy running in the industrial cloud platform is updated online to complete the closed-loop control of cost deviation.

[0007] As a further aspect of the present invention, the identification of the germination point and diffusion path of cost deviation includes: Set a rolling time window, within which the actual project cost flow is aligned with the benchmark cost plan at the same time granularity; Calculate the relative difference between the actual project cost flow data and the benchmark cost plan data at each time granularity alignment point. When the relative difference exceeds a preset threshold continuously, the moment of the first exceedance is recorded as the budding point of the cost deviation. Starting from the point of emergence, the relative difference is tracked within a rolling time window in terms of the order of occurrence and the direction of transmission among the various sub-systems of the project. The order of occurrence and the direction of transmission together constitute the diffusion path of the cost deviation.

[0008] As a further aspect of the present invention, the tracking of cost deviations and their transmission relationships among project sub-items includes: With the total target cost of the project as the root node and each independent sub-project as a child node, construct a tree-like cost decomposition structure for the project. In the tree-like cost decomposition structure, starting from the child node where the cost deviation is detected, the process traces back upwards along the parent-child node link and laterally explores the related nodes at the same level along the sibling node link, recording the node sequence in which the deviation occurs. Based on the temporal and logical dependencies of cost data between adjacent nodes in the node sequence, the transmission direction and intensity of cost deviations between project sub-structures are determined, thus forming the transmission relationship.

[0009] As a further aspect of the present invention, the monitoring of disturbance signals generated by external market data streams on the benchmark cost of project sub-items includes: Access to external market data sources is achieved through the interface of the industrial cloud platform. These external market data sources include at least the raw material price index, the labor cost index, and the equipment rental price index. Associate one or more relevant external market data indices with the benchmark cost of each project item; The rate of change of each relevant external market data index relative to its benchmark value is calculated in real time. When the rate of change exceeds its corresponding threshold, a disturbance signal is generated for the benchmark cost of the project sub-item. The disturbance signal includes the disturbance source, disturbance direction and disturbance amplitude.

[0010] As a further aspect of the present invention, mapping the multi-level impact chain formed by cost deviations in the industrial cloud platform includes: Starting from the project sub-item corresponding to the germination time point, the upstream and downstream sub-items that have direct transmission in the transmission relationship are linked as secondary nodes; Within the same time period, any item that receives the disturbance signal and is identical to any node item in the current chain is marked as a coupled node affected by external factors and incorporated into the chain; The links continue to expand outward along the secondary nodes and coupling nodes until the weight of the deviation of the newly expanded node on the starting item is lower than the set threshold. The directed network formed at this time from the starting item to the end node is the multi-level influence chain.

[0011] As a further aspect of the present invention, the deviation contribution weight and control sensitivity of each level in the analytical chain include: In the multi-level influence chain, calculate all paths from the starting point, through each intermediate node, to the ending node; The historical cost deviation values ​​of each node item on each path are statistically analyzed, and the influence ratio of each node item on the total deviation of all end nodes is calculated according to the transmission strength in the transmission relationship, which is used as the deviation contribution weight of the node item. In the historical control records of the industrial cloud platform, the control operations applied to each node item are retrieved, and the average rate of change of the cost deviation value of the node item after each control is calculated. The average rate of change is used as the control sensitivity of the node item.

[0012] As a further aspect of the present invention, the generation of the differentiated pre-regulation instruction set includes: For each node in the multi-level influence chain, a comprehensive control priority is calculated based on its deviation contribution weight and control sensitivity. To comprehensively regulate high-priority sub-items, control instructions characterized by proactive and strong intervention are configured. These control instructions include specific cost adjustment ranges, adjustment steps, and execution priorities. To comprehensively regulate low-priority sub-items, regulation commands characterized by observation and fine-tuning are configured. These regulation commands mainly set monitoring thresholds and triggering conditions. The control instructions corresponding to all nodes are summarized to form the pre-control instruction set.

[0013] As a further aspect of the present invention, calculating the matching degree between the response trajectory and the expected control trajectory includes: After executing the pre-regulation instruction set for one complete cycle, extract the actual cost time series data of each node item within the complete cycle; Based on the adjustment targets set in the pre-regulation instruction set, the expected cost change trajectory of each node item within the same period is generated as the expected regulation trajectory. For each node item, calculate the cosine similarity between its actual cost time series and the expected control trajectory at each sampling point, and take the average of the similarities of all sampling points as the matching degree of the node item.

[0014] As a further aspect of the present invention, the iterative correction of the pre-regulation instruction set based on the matching degree result includes: The matching degree of each node item is multiplied by its overall control priority to obtain the urgency of instruction correction for the node item. Based on the urgency of the instruction correction, the control instruction parameters of the corresponding node items in the pre-control instruction set are adjusted in a targeted manner. The adjustment includes: for nodes with high instruction correction urgency, increasing their cost adjustment range or shortening the adjustment step; for nodes with low instruction correction urgency, relaxing their monitoring threshold. After a round of parameter adjustments, an updated pre-control instruction set is formed. The steps of executing, collecting response trajectories and calculating matching degree are repeated until the average matching degree of all node sub-items meets the convergence condition. The pre-control instruction set obtained at this time is determined as the final state control instruction.

[0015] As a further aspect of the present invention, the online updating of the cost control strategy running in the industrial cloud platform includes: The final state control instructions are compiled into rule descriptions that can be recognized by the industrial cloud platform management strategy engine; Without interrupting the existing monitoring process of the industrial cloud platform, the rule descriptions are loaded into the policy engine in an incremental hot deployment manner, replacing or overwriting the original policy entries related to the multi-level influence chain. Based on the new rules, the strategy engine makes real-time judgments and interventions on the actual cost flow of subsequent projects and the benchmark cost plan, thereby achieving closed-loop management of cost deviations.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on the diffusion path of cost deviations, the transmission relationship of deviations among project sub-items is traced. Simultaneously, the disturbance signal of external market data flow on the benchmark cost of project sub-items is monitored. After combining the two types of data, the multi-level impact chain of cost deviations is mapped in the industrial cloud platform. The transmission path of deviations among different sub-items can be fully presented. The disturbance effect of external market factors can be directly integrated into the deviation tracing process. The attribution logic of deviation formation covers both internal transmission and external disturbance dimensions. The level of deviation tracing is extended to multi-level transmission nodes, and the completeness of the tracing results is expanded.

[0017] This study analyzes the deviation contribution weights and control sensitivities at each level of a multi-level influence chain, generating differentiated pre-control instruction sets for different project sub-items. After executing the instructions, the response trajectory of the actual project cost flow is collected, and the matching degree between the response trajectory and the expected control trajectory is calculated. Based on the matching degree results, the pre-control instruction set is iteratively corrected and the final state control instruction is generated. The cost control strategy in the industrial cloud platform is updated online, allowing the control instructions to adapt to the deviation impact degree and control response characteristics of the corresponding sub-items. The dynamic response data of the actual cost directly drives the instruction optimization, and the adjustment of the control strategy can be completed online without manual intervention. The cost deviation control process is transformed from a fixed execution mode to a dynamic adaptive adjustment mode, and the autonomous optimization capability of the control process is enhanced. Attached Figure Description

[0018] Figure 1 The flowchart shows the intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms as described in this invention. Figure 2 A flowchart for tracking the transmission relationship of cost deviations among project sub-items; Figure 3 A trend chart showing the change in the matching degree of the control command execution; Figure 4 This indicates the trend of changes in the external market disturbance index. Figure 5 This is a trend chart illustrating the iterative correction effect during the decision-making stage. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 In the industrial cloud platform, the actual cost flow of a project is continuously aggregated and compared with a pre-set benchmark cost plan to identify the specific timing of cost deviations and their internal propagation paths. Based on this propagation path, the transmission relationship of cost deviations among various project components is tracked, and the disturbance signals generated by external market data flow on the benchmark cost of each project component are monitored simultaneously. Combining the internal transmission relationship with the external disturbance signals, a multi-level impact chain leading to cost deviations is mapped in the cloud platform. Based on this multi-level impact chain, the contribution weight of each node at each level to the total deviation and its sensitivity to control operations are analyzed. Based on the deviation contribution weight and control sensitivity of each node, differentiated pre-control instruction sets are generated for different levels of project components in the industrial cloud platform. These pre-control instruction sets are executed, and the response trajectory of the actual project cost flow is simultaneously collected. The matching degree between the response trajectory and the expected control trajectory is calculated. Based on the matching degree calculation results, the pre-control instruction set is iteratively corrected to generate final control instructions. This final state control command is used to update the cost control strategy running in the industrial cloud platform online, thereby completing the closed-loop control of the entire process of cost deviation from identification, analysis, decision-making to correction.

[0022] In one embodiment of the present invention, for cost monitoring of power plant flue gas desulfurization retrofit projects, the industrial cloud platform sets a rolling time window with a weekly time unit. The length of the rolling time window is set to six weeks. Within the rolling time window, the daily aggregated actual project cost flow is aligned with the benchmark cost plan at a daily time granularity. The relative difference between the actual project cost flow data and the benchmark cost plan data on each aligned day is calculated using the following formula: , in: Indicates the first The relative difference between days, Indicates the first Daily project actual cost flow data, Indicates the first The daily baseline cost plan data. The preset threshold is set to... When the relative difference Three consecutive days of alignment exceeding At that time, it will exceed for the first time The corresponding daily record indicates the point at which cost deviations begin to occur. For example, monitoring data shows that in the [number]th day of project execution... Day, No. day and day Daily relative difference , , In order , , Then determine the first The date is considered the precipitating point for cost deviations. From this precipitating point, the order and direction of the relative differences between the various sub-systems of the project are tracked within a rolling time window. For example, on the [date missing]... On the [date], the relative difference of the absorption tower body modification sub-item first exceeded the threshold, and on the [date]... On [date], the relative difference of the circulating slurry pump component exceeded the threshold. On that day, the relative difference of the limestone slurry sub-item exceeded the threshold. The order of occurrence and the direction of transmission of the deviations between this series of sub-items together constitute the diffusion path of the cost deviation. It can be understood that the diffusion path describes the process of the deviation being transmitted from one sub-item to another.

[0023] In some embodiments, the time granularity is not limited to "days," and the length of the rolling time window can be adjusted according to the total project cycle. For example, for a power plant main steam pipeline renovation project with a total cycle exceeding one year, the length of the rolling time window can be set to three months, and the time granularity can be set to "weeks." The relative difference between the weekly actual project cost flow data and the baseline cost plan data is calculated, and the preset threshold can be adjusted to [missing information]. When the relative difference exceeds for four consecutive weeks The germination time point is recorded. In some embodiments, the preset threshold is not a fixed value, but is set according to the cost sensitivity of the sub-system. For example, for core equipment sub-items such as "renovation of the steam turbine flow path", the preset threshold is set to [value missing]. For auxiliary engineering items such as "thermal insulation and corrosion prevention", the preset threshold is set to... After calculating the relative difference, it is compared with the corresponding threshold. Optionally, the starting point of the rolling time window is not fixed, but continuously scrolls forward as cloud platform data continues to accumulate. When each new time granularity data is added to the window, the earliest time granularity data in the window is removed to maintain the window length. Optionally, when tracking the transmission direction, in addition to the time sequence, the dependence between subsystems in terms of process, materials, or logic is also considered. For example, when the absorption tower body modification subsystem deviates, the next subsystem to deviate is usually the circulating slurry pump subsystem that is immediately adjacent in process, rather than the logically unrelated electrical instrumentation subsystem. This dependency helps determine the transmission direction.

[0024] In one embodiment of the present invention, for a power plant steam turbine flow path modification project, see [reference needed]. Figure 2 Using the total project target cost as the root node, and individual sub-projects such as boiler modification, turbine body modification, condenser modification, piping system modification, electrical system modification, thermal control system modification, insulation and corrosion protection, and scaffolding engineering as child nodes, a tree-like cost decomposition structure is constructed for the project. In this tree-like cost decomposition structure, starting from the turbine body modification sub-node where cost deviations are detected, the process traces upwards along the parent-child node link back to the root node of the total target cost, and then horizontally explores related nodes at the same level along the sibling node link. For example, the sibling nodes of the turbine body modification sub-project include boiler modification, condenser modification, and piping system modification. The sequence of nodes where deviations occur is recorded; a possible sequence is "turbine body modification sub-project" -> "boiler modification sub-project" -> "piping system modification sub-project". Based on the temporal and logical dependencies of cost data between adjacent nodes in the node sequence, the transmission direction and intensity of cost deviations between project sub-projects are determined. The transmission intensity can be understood as being calculated using the correlation coefficient and time interval of cost deviation values ​​between adjacent nodes in the sequence, thus forming a transmission relationship. In practical implementation, external market data sources are accessed through the industrial cloud platform's interface. These external market data sources include the national steel price composite index, the average daily wage index for senior welders in East China, and the rental price index for large lifting equipment. Relevant external market data indices are associated with the benchmark cost of each project item. For example, the benchmark cost for boiler modification is associated with the national steel price composite index and the average daily wage index for senior welders in East China; the benchmark cost for pipeline system modification is associated with the national steel price composite index; and the benchmark cost for turbine body modification is associated with the rental price index for large lifting equipment. The rate of change of each relevant external market data index relative to its benchmark value is calculated in real time. When the rate of change of the average daily wage index for senior welders in East China exceeds a threshold for two consecutive weeks... At that time, a disturbance signal is generated for the benchmark cost of the boiler renovation item. The disturbance signal includes the disturbance source "increased labor costs", the disturbance direction "positive" and the disturbance magnitude. The calculation formula is:

[0025] in: Indicates the first The rate of change of each related index during the statistical period Indicates the first The values ​​of the relevant indices in this statistical period. Indicates the first The preset benchmark values ​​for each relevant index.

[0026] In some embodiments, the hierarchical structure of the tree-like cost decomposition structure can be extended. For example, the turbine body modification item can be further decomposed into finer sub-nodes such as rotor replacement, cylinder modification, and blade replacement. Lateral exploration of cost deviations can be conducted not only between sibling nodes at the same level but also across different levels of nodes with process interface relationships. For example, a deviation in the rotor replacement item may propagate upwards to the turbine body modification item and laterally to the bearing lubrication item with which it interfaces. In some embodiments, access to external market data sources is dynamic. The industrial cloud platform automatically subscribes to relevant price indices for items in the procurement or construction window based on the project bill of materials and schedule. For example, the steel price index is associated only when the project enters the pipeline installation phase. Optionally, the threshold for disturbance signals can be set in relation to the cost proportion of the item. For core items with a high cost proportion, such as turbine body modification, the threshold for the rate of change of its associated index is set lower. For auxiliary items with a low cost proportion, the threshold for the rate of change of their correlation index can be set as follows: Optionally, the determination of the transmission relationship depends not only on the timing but also on a weighting factor for process logic dependencies. For example, there is a strong coupling relationship in the steam flow between the boiler modification sub-project and the turbine body modification sub-project. The weighting factor for the transmission strength between these two nodes is set as follows: The boiler renovation project and the electrical system renovation project have a weak power supply relationship, and the conduction strength weighting factor is set to... It can be understood that the weighting factor is used to correct the transmission strength value calculated solely based on the correlation of time-series deviations.

[0027] In one embodiment of the present invention, after identifying the "steam turbine rotor replacement" item as the project item corresponding to the initiation point, the upstream and downstream items with direct transmission relationships in the transmission relationship are linked as secondary nodes, starting from the steam turbine rotor replacement item. For example, according to the transmission relationship, the direct downstream item of the steam turbine rotor replacement item is the steam turbine body assembly and commissioning item, and the direct upstream item is the special alloy steel forging procurement item. Therefore, the steam turbine body assembly and commissioning item and the special alloy steel forging procurement item are linked to the starting point as secondary nodes. Items that receive disturbance signals within the same time period and are the same as any node item in the current chain are marked as coupling nodes affected by external factors and incorporated into the chain. For example, the special alloy steel forging procurement item receives a disturbance signal originating from the "global nickel price index" exceeding a threshold during the same period. Therefore, the special alloy steel forging procurement item is marked as a coupling node. The links continue to expand outward along secondary and coupling nodes. For example, from the turbine assembly and commissioning sub-item to the downstream unit linkage test sub-item, from the special alloy steel forging procurement sub-item to the upstream raw material international logistics sub-item, and so on until a new node is added, such as the "factory lighting renovation sub-item". The deviation impact weight on the starting turbine rotor replacement sub-item is calculated to be below the set threshold. When the expansion stops, the directed network formed from the starting item to the ending node is called a multi-level influence chain.

[0028] In practical implementation, within a multi-level influence chain, the calculation involves all paths from the starting point (turbine rotor replacement) through each intermediate node to the final node. Possible paths include "turbine rotor replacement -> special alloy steel forging procurement -> international raw material logistics" and "turbine rotor replacement -> turbine assembly and commissioning -> unit commissioning." The historical cost deviation values ​​for each node on each path are statistically analyzed. For example, in three similar past projects, the historical cost deviation values ​​for the special alloy steel forging procurement item were as follows: , , The historical cost deviations for the turbine assembly and commissioning sub-item are as follows: , , The influence ratio of each node's component on the total deviation of all terminal nodes is calculated based on the weighted average of the conduction strength in the conduction relationship. The conduction strength values ​​are derived from correlation analysis of historical data. The calculation formula is as follows:

[0029] in: Indicates the first The deviation contribution weight of each node item Indicates the first The node sub-item in the Historical cost deviation values ​​along the path, Indicates the first The node sub-item in the Conductivity coefficient along the path (range of values) arrive ), Indicates containing the first The total number of paths to each node. This represents the total number of nodes in a multi-level influence chain. The weighted calculation result serves as the deviation contribution weight for each node's component.

[0030] In some embodiments, the expansion of multi-level influence chains is based not only on direct transmission relationships but also on indirect logical connections. For example, when a deviation occurs in the procurement of special alloy steel forgings and becomes a coupling node, the deviation may indirectly affect the "turbine sealing parts procurement" item, which belongs to the same procurement package but is not directly adjacent, through the logical event of "procurement contract change." This indirect influence relationship is also incorporated into the rules for chain expansion. In some embodiments, when calculating all paths, an upper limit is set for the path length, for example, only considering paths starting from the starting point and passing through a path that does not exceed a certain length. The path to each intermediate node is controlled to manage computational complexity. Optionally, the search scope of historical control records on the industrial cloud platform is not limited to the current project, but can be searched within the cloud platform database for all completed records of similar power plant renovation projects.

[0031] See Figure 3 In the analysis of the matching degree of control command execution during the control decision-making stage, cosine similarity was used as a metric to achieve quantitative evaluation and trend visualization of the matching between the actual response trajectory and the expected control trajectory. Specifically, four complete control cycles were used as observation windows. Actual cost time-series data for each node item after command execution were collected synchronously. Based on the adjustment targets of the pre-control command set, the expected control trajectory for the same period was generated. The cosine similarity was calculated for each sampling point and the average was taken to obtain the overall actual trajectory matching degree for each cycle. From the trend changes, the actual trajectory matching degree gradually increased from 0.62 in the first cycle to 0.94 in the fourth cycle, showing a continuous convergence trend. Meanwhile, the expected trajectory matching degree remained stable at the target level of 0.90, reflecting that during the iterative correction process of the pre-control command set, the actual execution effect gradually approached the expected control target. In the iterative correction logic, the matching degree result will be multiplied by the comprehensive control priority of each node item to obtain the urgency of instruction correction, and then the control instruction parameters will be adjusted in a targeted manner: for nodes with both high matching degree and high priority, the intervention intensity will be strengthened to maintain the convergence trend; for nodes with low matching degree, the cost adjustment range will be expanded or the adjustment step will be shortened until the overall matching degree meets the convergence condition, and finally the final state control instruction will be generated and the online update of the cloud platform control strategy will be completed.

[0032] In one embodiment of the present invention, for each node component in the multi-level influence chain, a comprehensive control priority is calculated based on the deviation contribution weight of the node component and the control sensitivity of the node component. The formula for calculating the comprehensive control priority is as follows:

[0033] in: Indicates the first The overall control priority of each node and sub-item Indicates the first The deviation contribution weight of each node item Indicates the first The control sensitivity of each node item (absolute value to represent the degree of sensitivity). and These are preset weighting coefficients used to balance the contribution of deviations and the effectiveness of control. For high-priority nodes in comprehensive control, control instructions characterized by proactive and strong intervention are configured. For example, the control instructions for high-priority nodes such as the procurement of special alloy steel forgings include specific cost adjustment ranges such as "reducing the target cost." to The adjustment of the step size will be implemented in two phases, with each phase lowering the step size by no less than [amount missing]. "Highest priority" is assigned to the lowest priority nodes. For lower priority nodes, control commands characterized by observation and fine-tuning are configured. For example, the control commands for lower priority nodes, such as factory lighting renovation, primarily set monitoring thresholds such as "weekly cost deviation exceeding..." The system includes a "timely alarm" and a trigger condition: "cost review triggered by alarms for two consecutive weeks." All control instructions corresponding to each node item are summarized to form a pre-control instruction set, which contains specific operational guidelines for all items in the chain. After executing a complete cycle of the pre-control instruction set, the actual cost time series data for each node item within the complete cycle is extracted. The actual cost time series data is recorded with daily or weekly sampling points. Based on the adjustment targets set for each node item in the pre-control instruction set, the expected cost change trajectory for each node item within the same period is generated. The basis for generating the expected control trajectory is the adjustment targets and the preset execution rate model. For each node item, the cosine similarity between the actual cost time series of the node item and the expected control trajectory at each sampling point is calculated. The average similarity of all sampling points is taken as the matching degree of the node item. The matching degree calculation formula is:

[0034] in: Indicates the first The matching degree of each node item. This represents the total number of sampling points within a complete period. This represents a multidimensional feature (such as daily cost, cumulative cost, and trend). The node sub-item in the The actual cost feature vector of each sampling point This represents the feature vector of expected construction cost for the corresponding sampling point. Representing vectors The model, Representing vectors The model.

[0035] In a specific example, for a multi-level influence chain containing key nodes, refer to Table 1, which shows the comprehensive control priority, control instructions and matching degree after execution of some node items.

[0036] Table 1: Node-Specific Control Commands and Matching Degrees

[0037] In some embodiments, the configuration of control instructions not only considers the overall control priority but also the current project stage of each node item. For example, for a node item in the procurement finalization stage, even if the overall control priority is high, strong intervention instructions such as "change supplier" will no longer be configured; instead, limited instructions such as "negotiate changes to payment terms" will be configured. In some embodiments, the generation of the expected control trajectory adopts a pattern matching method based on historical successful cases, rather than a simple linear model. For example, successful control cases with similar deviation patterns and supplier types to the current special alloy steel forging procurement node item are found in historical data, and their cost recovery trajectories are used as the expected control trajectory for the current node item. Optionally, the feature vector used to calculate the matching degree not only includes cost data but can also include derivative indicators related to progress and quality, such as unit progress cost and key material delivery rate, to form a multi-dimensional feature vector for calculation. Optionally, a complete cycle is defined as the expected duration of the main adjustment actions set by each instruction in the pre-control instruction set. For example, if the instruction requires "two phases, two weeks per phase," then a complete cycle is four weeks. It can be understood that the instruction cycles for different node items may be different, and the matching degree calculation needs to be performed separately based on their respective complete cycles.

[0038] See Figure 4In monitoring external market disturbance indices during the source tracing phase, three core market indices—raw material prices, labor costs, and equipment leasing—are accessed through an industrial cloud platform. Using a benchmark value of 100 as a reference, the system quantifies and tracks disturbance signals related to the benchmark cost of each project component. From a time-series perspective, the three indices exhibit differentiated fluctuation characteristics: The raw material price index peaked at 105 in week 3, then steadily declined to 97 in week 8, showing an overall trend of initial rise followed by a gradual decline. This reflects the disturbance characteristics of a gradual return to normalcy after a short-term supply-demand imbalance in the raw material market. Its fluctuation range is the largest, and its potential transmission effect on project cost deviations is the strongest. The labor cost index steadily increased from week 1, gradually rising from a benchmark value of 100 to 104 in weeks 7-8, showing a sustained and moderate upward trend. This reflects a long-term disturbance signal of tight labor market supply and demand. Its fluctuation stability is higher, representing a gradual source of cost pressure. The equipment leasing index initially plummeted to a low of 97 in week 4, then gradually rebounded to 101 in week 8, exhibiting an overall "V-shaped" reversal trend. This reflects the gradual recovery of the equipment leasing market after a period of supply-demand mismatch. Its fluctuation range falls between the previous two indices, indicating a medium-term reversible disturbance. The time-series changes of these three indices collectively constitute a set of disturbances in the external market data flow, providing a quantitative basis for subsequently mapping the multi-level impact chain of cost deviations, analyzing the contribution weight of each sub-item deviation, and assessing regulatory sensitivity.

[0039] In one embodiment of the present invention, the matching degree of each node item is multiplied by the comprehensive control priority of the node item to obtain the instruction correction urgency of the node item. The formula for calculating the instruction correction urgency is as follows:

[0040] in: Indicates the first The urgency of instruction correction for each node item. Indicates the first The overall control priority of each node and sub-item Indicates the first The matching degree of each node item. Based on the urgency of instruction correction, the control instruction parameters of the corresponding node items in the pre-control instruction set are adjusted in a targeted manner. For node items with high instruction correction urgency, the cost adjustment range set in the corresponding control instruction is increased or the adjustment step size set in the control instruction is shortened. For example, for the special alloy steel forging procurement item, the initial instruction correction urgency is... The calculation result is If the situation is deemed highly urgent, then the scope of cost adjustment in the regulatory directive will be expanded from "lowering the target cost" to include other aspects. to Expanded to "lowering target costs" to The adjustment will be implemented in two phases, with each phase lowering the step size by no less than [amount missing]. The reduction has been shortened to "implementation in three phases, with each phase lowering the price by no less than..." "For node items with low urgency of instruction correction, relax the monitoring thresholds set in the corresponding control instructions. For example, for the factory lighting renovation item, the initial instruction correction urgency..." The calculation result is If the urgency level is low, then the monitoring threshold in its control instructions will be changed from "weekly cost deviation exceeds..." The "timely warning" has been relaxed to "weekly cost deviation exceeding..." "Timely alarm". After a round of parameter adjustments, an updated pre-control instruction set is formed. The updated pre-control instruction set replaces the original instruction set. The steps of repeatedly executing the updated pre-control instruction set, collecting the actual cost flow response trajectory after execution, and calculating the matching degree of the node sub-items are repeated until the average matching degree of all node sub-items meets the convergence condition. The convergence condition is set to the average matching degree of all node sub-items being greater than 1. And the mean change between two consecutive iterations is less than The pre-control instruction set obtained at this time is determined as the final state control instruction.

[0041] In practical implementation, the final-state control instructions are compiled into rule descriptions that can be recognized by the industrial cloud platform's management and control strategy engine. These rule descriptions use the "IF-THEN" format. For example, the final-state control instruction for the special alloy steel forging procurement item is compiled as: "IF Item Name == 'Special Alloy Steel Forging Procurement' AND Weekly Cumulative Deviation Rate > Then, this triggers a second round of price negotiations and implements a three-phase cost reduction plan, with the first phase targeting a price reduction. Without interrupting the existing monitoring processes of the industrial cloud platform, rule descriptions are loaded into the strategy engine using incremental hot deployment. Incremental hot deployment means that only the new rule description file is uploaded to the designated directory of the strategy engine. The engine automatically detects and loads the new rules, replacing or overwriting the original strategy entries related to the multi-level impact chain. Based on the newly loaded rule descriptions, the strategy engine performs real-time judgment and intervention on the actual cost flow of subsequent projects and the benchmark cost plan, realizing closed-loop control of cost deviations. For example, if the real-time data of the current week shows that the weekly cumulative deviation rate of the special alloy steel forging procurement item reaches a certain level... At that time, the strategy engine automatically triggers the action of "starting a second price negotiation" and notifies the procurement manager, while pushing the first phase target of the "three-stage cost reduction plan" to the cost control module.

[0042] In some embodiments, the calculation of instruction correction urgency can incorporate a correction factor for the rate of decrease in matching degree. For node components whose matching degree decreases rapidly in consecutive iterations, even if the current absolute value of the matching degree is still acceptable, their instruction correction urgency will be increased to address potential risks of runaway. In some embodiments, the convergence condition not only considers the average matching degree of all node components but also sets a minimum matching degree requirement for individual node components, such as requiring that the matching degree of any node component must not be lower than a certain threshold. Otherwise, even if the mean meets the target, iterative correction will still be necessary. Optionally, the iterative correction of the pre-regulation instruction set can be performed in a hierarchical and batch-based manner, with each iteration only correcting the instructions ranked highest in urgency. Adjusting the node-specific instruction parameters can be understood as accelerating the overall convergence process and reducing system oscillations. Optionally, during rule description compilation, specific numerical parameters such as monitoring thresholds can be included. Adjusting goals These parameters are separated from the rule logic and stored in an independent configuration table. When these parameters need to be adjusted in subsequent iterations, only the configuration table needs to be updated without recompiling the entire rule description file, which enhances the flexibility of policy maintenance.

[0043] See Figure 5 In the control and decision-making stage of intelligent cost deviation management for power plant renovation projects oriented towards cloud platforms, the iterative correction effect is quantitatively evaluated through two core indicators: average matching degree and the number of high-urgency nodes. Specifically: Average matching degree (line graph): Represents the overall degree of fit between the actual cost response trajectory after the execution of pre-control instructions and the expected control trajectory. The initial value was 0.65, which continuously climbed to 0.92 after 4 rounds of iterative correction, gradually approaching the set convergence threshold (0.9, dashed line), reflecting that the accuracy and effectiveness of control instructions are continuously enhanced with iteration. Number of high-urgency nodes (bar chart): Represents the number of project sub-nodes with high urgency of instruction correction and requiring key intervention. Initially, there were 4 nodes, which gradually decreased to 1 in the 3rd round as iteration progressed, reflecting that high-priority deviation nodes were continuously eliminated and the overall system risk was effectively converged. From an iterative logic perspective, this stage adjusts the pre-control command parameters in a targeted manner by modifying the urgency of the command: expanding the cost adjustment range and shortening the adjustment step for high-urgency nodes; relaxing the monitoring threshold for low-urgency nodes, thereby driving the average matching degree to rise and the number of high-urgency nodes to decrease, ultimately achieving the convergence condition for the average matching degree of all nodes, generating the final state control command and completing the online update of the cloud platform control strategy.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A cloud platform-oriented power plant retrofit project cost deviation intelligent management and control method, characterized in that, include: Build a project cost monitoring process in the industrial cloud platform, including the deviation detection stage, the impact tracing stage, and the control and decision-making stage; During the deviation detection phase, the actual project cost flow continuously aggregated on the industrial cloud platform is compared with the benchmark cost plan to identify the initiation point and diffusion path of cost deviation. During the impact tracing phase, based on the diffusion path of the cost deviation, the transmission relationship of the cost deviation between project sub-items is tracked, and the disturbance signal generated by the external market data flow on the benchmark cost of project sub-items is monitored. Combining the transmission relationship of the cost deviation between project sub-items and the disturbance signal generated by the external market data flow, the multi-level impact chain formed by the cost deviation is mapped in the industrial cloud platform. During the regulatory decision-making stage, based on the multi-level impact chain of the cost deviation, the deviation contribution weight and regulatory sensitivity of each level in the chain are analyzed. Based on the aforementioned deviation contribution weight and control sensitivity, differentiated pre-control instruction sets are generated for different levels of projects in the industrial cloud platform; The pre-control instruction set is executed in the industrial cloud platform, and the response trajectory of the actual project cost flow is collected simultaneously. The matching degree of the response trajectory and the expected control trajectory is calculated. The pre-control instruction set is iteratively corrected according to the matching degree result to generate the final state control instruction. The cost control strategy running in the industrial cloud platform is updated online to complete the closed-loop control of cost deviation.

2. The cloud platform-oriented power plant retrofit project cost deviation intelligent management and control method according to claim 1, characterized in that, The identification of the initiation point and propagation path of cost deviations includes: Set a rolling time window, within which the actual project cost flow is aligned with the benchmark cost plan at the same time granularity; Calculate the relative difference between the actual project cost flow data and the benchmark cost plan data at each time granularity alignment point. When the relative difference exceeds a preset threshold continuously, the moment of the first exceedance is recorded as the budding point of the cost deviation. Starting from the point of emergence, the relative difference is tracked within a rolling time window in terms of the order of occurrence and the direction of transmission among the various sub-systems of the project. The order of occurrence and the direction of transmission together constitute the diffusion path of the cost deviation.

3. The intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms according to claim 2, characterized in that, The transmission relationship of cost deviation tracking among project sub-items includes: With the total target cost of the project as the root node and each independent sub-project as a child node, construct a tree-like cost decomposition structure for the project. In the tree-like cost decomposition structure, starting from the child node where the cost deviation is detected, the process traces back upwards along the parent-child node link and laterally explores the related nodes at the same level along the sibling node link, recording the node sequence in which the deviation occurs. Based on the temporal and logical dependencies of cost data between adjacent nodes in the node sequence, the transmission direction and intensity of cost deviations between project sub-structures are determined, thus forming the transmission relationship.

4. The intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms as described in claim 3, characterized in that, The disturbance signals generated by the monitoring of external market data streams on the project's sub-item benchmark cost include: Access to external market data sources is achieved through the interface of the industrial cloud platform. These external market data sources include at least the raw material price index, the labor cost index, and the equipment rental price index. Associate one or more relevant external market data indices with the benchmark cost of each project item; The rate of change of each relevant external market data index relative to its benchmark value is calculated in real time. When the rate of change exceeds its corresponding threshold, a disturbance signal is generated for the benchmark cost of the project sub-item. The disturbance signal includes the disturbance source, disturbance direction and disturbance amplitude.

5. The intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms according to claim 4, characterized in that, The multi-level impact chain formed by cost deviations mapped in the industrial cloud platform includes: Starting from the project sub-item corresponding to the germination time point, the upstream and downstream sub-items that have direct transmission in the transmission relationship are linked as secondary nodes; Within the same time period, any item that receives the disturbance signal and is identical to any node item in the current chain is marked as a coupled node affected by external factors and incorporated into the chain; The links continue to expand outward along the secondary nodes and coupling nodes until the weight of the deviation of the newly expanded node on the starting item is lower than the set threshold. The directed network formed at this time from the starting item to the end node is the multi-level influence chain.

6. The intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms according to claim 5, characterized in that, The deviation contribution weights and control sensitivity of each level in the analytical chain include: In the multi-level influence chain, calculate all paths from the starting point, through each intermediate node, to the ending node; The historical cost deviation values ​​of each node item on each path are statistically analyzed, and the influence ratio of each node item on the total deviation of all end nodes is calculated according to the transmission strength in the transmission relationship, which is used as the deviation contribution weight of the node item. In the historical control records of the industrial cloud platform, the control operations applied to each node item are retrieved, and the average rate of change of the cost deviation value of the node item after each control is calculated. The average rate of change is used as the control sensitivity of the node item.

7. The intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms according to claim 6, characterized in that, The generated differentiated pre-regulation instruction set includes: For each node in the multi-level influence chain, a comprehensive control priority is calculated based on its deviation contribution weight and control sensitivity. To comprehensively regulate high-priority sub-items, control instructions characterized by proactive and strong intervention are configured. These control instructions include specific cost adjustment ranges, adjustment steps, and execution priorities. To comprehensively regulate low-priority sub-items, regulation commands characterized by observation and fine-tuning are configured. These regulation commands mainly set monitoring thresholds and triggering conditions. The control instructions corresponding to all nodes are summarized to form the pre-control instruction set.

8. The intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms according to claim 7, characterized in that, Calculating the matching degree between the response trajectory and the expected control trajectory includes: After executing the pre-regulation instruction set for one complete cycle, extract the actual cost time series data of each node item within the complete cycle; Based on the adjustment targets set in the pre-regulation instruction set, the expected cost change trajectory of each node item within the same period is generated as the expected regulation trajectory. For each node item, calculate the cosine similarity between its actual cost time series and the expected control trajectory at each sampling point, and take the average of the similarities of all sampling points as the matching degree of the node item.

9. The intelligent cost deviation control method for power plant renovation projects oriented towards cloud platforms according to claim 8, characterized in that, The iterative correction of the pre-adjustment instruction set based on the matching degree result includes: The matching degree of each node item is multiplied by its overall control priority to obtain the urgency of instruction correction for the node item. Based on the urgency of the instruction correction, the control instruction parameters of the corresponding node items in the pre-control instruction set are adjusted in a targeted manner. The adjustment includes: for nodes with high instruction correction urgency, increasing their cost adjustment range or shortening the adjustment step; for nodes with low instruction correction urgency, relaxing their monitoring threshold. After a round of parameter adjustments, an updated pre-control instruction set is formed. The steps of executing, collecting response trajectories and calculating matching degree are repeated until the average matching degree of all node sub-items meets the convergence condition. The pre-control instruction set obtained at this time is determined as the final state control instruction.

10. The intelligent control method for cost deviations in power plant renovation projects oriented towards cloud platforms according to claim 1, characterized in that, The online updating of the cost control strategy running in the industrial cloud platform includes: The final state control instructions are compiled into rule descriptions that can be recognized by the industrial cloud platform management strategy engine; Without interrupting the existing monitoring process of the industrial cloud platform, the rule descriptions are loaded into the policy engine in an incremental hot deployment manner, replacing or overwriting the original policy entries related to the multi-level influence chain. Based on the new rules, the strategy engine makes real-time judgments and interventions on the actual cost flow of subsequent projects and the benchmark cost plan, thereby achieving closed-loop management of cost deviations.

Images