Method for generating flexible resource influence map and related device

Abstract

The present disclosure provides a flexible resource influence map generation method and related device, the method comprising: determining the original data set of the flexible resource regulation of the target industrial user, constructing the core causal node based on the original data set, and obtaining the standardized causal analysis data matrix; performing causal skeleton recognition on the standardized causal analysis data matrix to obtain a time sequence expansion causal diagram; performing dynamic edge weight calculation on the time sequence expansion causal diagram to obtain a dynamic causal influence weight; generating an influence map based on the time sequence expansion causal diagram and the dynamic causal influence weight, so as to adjust the scheduling strategy of the flexible resource regulation based on the influence map. The present disclosure can accurately identify the actual effect of economic incentives and industrial physical constraints on load response changes, and provide an interpretable basis for adjusting the scheduling strategy.

Description

Technical Field

[0001] This disclosure relates to the field of power data analysis and processing technology, and in particular to a method and apparatus for generating flexible resource influence maps. Background Technology

[0002] This section is intended to provide background or context for the embodiments of this disclosure as set forth in the claims. The description herein is not intended to be a prior art simply because it is included in this section.

[0003] Against the backdrop of the construction of new power systems, the volatility of new energy output has increased, and the frequent occurrence of extreme weather has led to a widening of the peak-valley load difference. The power grid needs to guide key industrial loads with regulation capabilities to participate in peak shaving, peak shifting, or peak transfer through price incentives. As an important part of the power grid's flexible resources, the response behavior of industrial users is not only affected by electricity price signals, but also constrained by multiple industrial physical factors such as production processes, equipment operating boundaries, material warehouse inventory, gas storage tank capacity, equipment ramp-up capabilities, order delivery, and carbon emission constraints.

[0004] However, in the relevant technologies, when conducting flexible resource impact analysis, there is a problem that it is difficult to reveal the dynamic interaction mechanism between economic incentives, industrial physical state and actual response behavior, which leads to the inability to accurately identify the dominant factors of changes in response capability. Summary of the Invention

[0005] In view of this, the purpose of this disclosure is to propose a method and related apparatus for generating flexible resource influence maps, which at least to some extent solves one of the technical problems in the related technologies.

[0006] To achieve the above objectives, a first aspect of the exemplary embodiments of this disclosure provides a method for generating flexible resource influence maps, the method comprising:

[0007] The original dataset for flexible resource regulation of target industrial users is determined, and core causal nodes are constructed based on the original dataset to obtain a standardized causal analysis data matrix. Causal skeleton identification is performed on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph; Dynamic edge weights are calculated on the temporal unfolded causal graph to obtain dynamic causal influence weights; An influence graph is generated based on the time-series unfolded causal graph and the dynamic causal influence weights, so that the scheduling strategy for the flexible resource regulation can be adjusted based on the influence graph.

[0008] In some exemplary embodiments, the original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data.

[0009] In some exemplary embodiments, the construction of core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix includes: The original dataset is time-scale aligned to obtain an aligned multi-source feature matrix; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is subjected to industrial physical constraint verification to obtain the verified core causal node sequence. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

[0010] In some exemplary embodiments, the core causal node sequence includes: external disaster-causing stimuli and grid supply and demand tension nodes, economic incentive intensity nodes, industrial physical security margin nodes, compliance and low-carbon pressure nodes, and industrial user actual response action nodes.

[0011] In some exemplary embodiments, causal skeleton identification is performed on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph, including: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

[0012] In some exemplary embodiments, the step of dynamically calculating the edge weights of the temporal unfolded causal graph to obtain dynamic causal influence weights includes: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

[0013] In some exemplary embodiments, generating an influence map based on the time-series unfolded causal graph and the dynamic causal influence weights includes: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; The influence graph is obtained based on the aggregated influence path set and the aggregated dynamic edge weights.

[0014] Based on the same inventive concept, a second aspect of the exemplary embodiments of this disclosure provides a flexible resource influence map generation apparatus, comprising: The data matrix determination module is configured to determine the original dataset of flexible resource regulation for target industrial users, and construct core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix. The causal graph determination module is configured to perform causal skeleton identification on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph. The influence weight determination module is configured to perform dynamic edge weight calculation on the time-series unfolded causal graph to obtain dynamic causal influence weights. The influence graph determination module is configured to generate an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights, so as to adjust the scheduling strategy of the flexible resource regulation based on the influence graph.

[0015] In some exemplary embodiments, the data matrix determination module is specifically configured as follows: The original dataset for flexible resource regulation of target industrial users is determined, and the original dataset is aligned to a time scale to obtain an aligned multi-source feature matrix; wherein, the original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is verified by industrial physical constraints to obtain the verified core causal node sequence; wherein, the core causal node sequence includes: external disaster-causing stimuli and the degree of tension between power grid supply and demand, economic incentive intensity, industrial physical safety margin, compliance and low-carbon pressure, and actual response actions of industrial users. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

[0016] In some exemplary embodiments, the cause-effect graph determination module is specifically configured as follows: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

[0017] In some exemplary embodiments, the influence weight determination module is specifically configured to: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

[0018] In some exemplary embodiments, the influence map determination module is specifically configured as follows: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; Based on the aggregated influence path set and the aggregated dynamic edge weights, the influence graph is obtained, so that the flexible resource regulation and scheduling strategy can be adjusted based on the influence graph.

[0019] Based on the same inventive concept, a third aspect of the exemplary embodiments of this disclosure provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in the first aspect.

[0020] Based on the same inventive concept, a fourth aspect of the exemplary embodiments of this disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method as described in the first aspect.

[0021] Based on the same inventive concept, a fifth aspect of the exemplary embodiments of this disclosure provides a computer program product including computer program instructions that, when run on a computer, cause the computer to perform the method as described in the first aspect.

[0022] As can be seen from the above, the flexible resource influence map generation method and related apparatus provided in this disclosure include: This invention identifies the original dataset for flexible resource regulation of target industrial users, constructs core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix, identifies the causal skeleton of the standardized causal analysis data matrix to obtain a time-series unfolded causal graph, calculates dynamic edge weights in the time-series unfolded causal graph to obtain dynamic causal influence weights, and generates an influence map based on the time-series unfolded causal graph and the dynamic causal influence weights, so as to adjust the scheduling strategy for flexible resource regulation based on the influence map. This disclosure can accurately identify the actual effects of economic incentives and industrial physical constraints on load response changes and provide an interpretable basis for adjusting the scheduling strategy. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic diagram illustrating an application scenario of the flexible resource influence map generation method provided as an exemplary embodiment of this disclosure; Figure 2 A flowchart illustrating a flexible resource influence map generation method provided as an exemplary embodiment of this disclosure; Figure 3 A schematic diagram of a flexible resource influence map generation device provided as an exemplary embodiment of this disclosure; Figure 4 A schematic diagram of the structure of an electronic device hardware provided for an exemplary embodiment of this disclosure. Detailed Implementation

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0026] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this application's technical solution, based on the prompt message.

[0027] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0028] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this application. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this application.

[0029] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0030] To make the objectives, technical solutions, and advantages of this disclosure clearer, the principles and spirit of this disclosure will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement this disclosure, and are not intended to limit the scope of this disclosure in any way. Rather, these embodiments are provided to make this disclosure more thorough and complete, and to fully convey the scope of this disclosure to those skilled in the art.

[0031] In this article, it is important to understand that any number of elements in the accompanying figures is for illustrative purposes and not for limitation, and any naming is for distinction only and has no limiting meaning.

[0032] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar words used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly. The article "a" or "an" preceding an element does not exclude the existence of multiple such elements.

[0033] The principles and spirit of this disclosure will be explained in detail below with reference to several representative embodiments.

[0034] As described in the background section, related technologies, when conducting flexible resource impact analysis, suffer from the problem of failing to reveal the dynamic interaction mechanism between economic incentives, industrial physical conditions, and actual response behavior, thus making it impossible to accurately identify the dominant factors influencing changes in response capacity. Specifically, while traditional load forecasting models can output future load or response quantities, the models themselves only concern themselves with forecast accuracy and do not explain the underlying reasons for load changes. When dispatchers observe insufficient response from a particular industrial user, they cannot determine from the forecast results whether it is due to insufficient compensation prices, depleted internal inventory, limited equipment ramp-up capabilities, or safety margins approaching their limits. The black-box nature of forecasting models obscures the true mechanisms at play.

[0035] In related technologies, statistical correlation methods, such as Pearson correlation coefficients or Granger causality tests, can measure the dependency between variables, but this dependency is not equivalent to the actual mechanism of action. During long-term continuous regulation, temperature, grid load gap, compensation prices, industrial physical conditions, and actual response quantities often change synchronously, easily forming common trends or spurious correlations. For example, price increases and load decreases may occur simultaneously, but the statistical relationship between them may be driven by weather changes rather than prices directly causing the load decrease. Therefore, relying solely on statistical correlation is insufficient to distinguish the true impact of economic incentives and industrial physical conditions on response behavior.

[0036] In related technologies, interpretable AI methods such as SHAP or LIME, while able to demonstrate the contribution of features to model predictions, interpret the trained predictive model rather than real-world industrial production systems. The model may learn incorrect associations in the data, such as mistaking a daily temperature fluctuation for a primary driving force, causing the feature importance ranking to deviate from the actual physical process. These methods can only explain "why the model predicts this way," not "why the real system responds this way."

[0037] In related technologies, time-series causal discovery algorithms, such as PCMCI or dynamic Bayesian networks, mainly rely on statistical data testing for structure search. Given the prevalence of strong autocorrelation, non-stationary trends, and limited sample sizes in industrial data, these algorithms easily identify candidate paths that do not conform to industrial physical logic. For example, the algorithm might output clearly illogical edges such as "equipment ramp-up capability affects temperature" or "actual response actions affect price signals from the previous few hours." Lacking prior knowledge such as process topology, material flow, energy flow, and equipment safety boundaries, the physical consistency of causal analysis results is difficult to guarantee, thus failing to accurately determine the true cause of changes in response capability.

[0038] Industrial mechanistic models or digital twin models can effectively describe equipment boundaries, material balance, and production constraints, and can simulate whether industrial users can continue to reduce load under given conditions. However, these models typically lack the ability to automatically learn the strength of different factors' influence from historical data, and cannot dynamically track whether the effectiveness of price incentives diminishes during continuous regulation. Therefore, even knowing that physical margins are decreasing, it is difficult to quantify whether the current inadequate response is due to insufficient price incentives or if industrial physical constraints have become dominant.

[0039] In summary, due to the limitations of each of the aforementioned technologies, when faced with industrial demand response scenarios involving long cycles and multiple coupled factors, these technologies struggle to effectively reveal the dynamic mechanisms between economic incentives, industrial physical conditions, and actual response behavior, making it impossible to accurately determine why response capabilities change.

[0040] To address the aforementioned issues, this disclosure provides a method and related apparatus for generating flexible resource influence maps. The method specifically includes: This disclosure first determines the original dataset for flexible resource regulation of target industrial users, and constructs core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix. Then, it identifies the causal skeleton of the standardized causal analysis data matrix to obtain a time-series unfolded causal graph. Finally, it calculates the dynamic edge weights of the time-series unfolded causal graph to obtain dynamic causal influence weights. Based on the time-series unfolded causal graph and the dynamic causal influence weights, it generates an influence map to enable the adjustment and scheduling strategy of flexible resource regulation based on the influence map. This disclosure first determines the original dataset for flexible resource regulation of target industrial users and constructs core causal nodes based on this dataset, transforming the originally scattered multi-source data such as meteorological, power grid, electricity price, equipment status, inventory, and production performance into a standardized analysis matrix with clear physical meaning. This step is not for simply predicting load, but for explicitly extracting key factors such as economic incentives (e.g., compensation prices) and industrial physical constraints (e.g., material warehouse inventory, equipment ramp-up capability), so that subsequent analysis can focus on the real relationships between these variables, rather than simply observing changes in load values.

[0041] Based on this, this disclosure identifies the time-varying influence paths between various nodes from causal analysis data. For example, it examines whether economic incentives affect actual response behavior after several time steps, and whether industrial physical conditions change first, thereby constraining response capabilities. By constructing a causal graph with temporal relationships, this disclosure can distinguish between correlations and true driving relationships at the statistical level, avoiding misjudging synchronously changing spurious correlations as actual effects.

[0042] Subsequently, this disclosure performs dynamic edge weight calculations on the time-series unfolded causal graph to obtain dynamic causal influence weights. This step goes beyond simply determining whether there is an impact at the structural level; instead, it quantifies the strength of each influence path at different times. For example, the weight of economic incentives may be higher in the early stages of regulation, while the weight of industrial physical constraints gradually increases in the later stages of continuous regulation. The calculation results of these dynamic weights can intuitively reflect which factor has a stronger dominant role in response behavior at the current moment, thereby accurately identifying the true cause of changes in response capability.

[0043] Finally, this disclosure generates an influence graph based on time-series unfolded causal graphs and dynamic causal influence weights, and adjusts scheduling strategies accordingly. This graph visually displays the dynamic influence relationships between variables such as economic incentives and industrial physical constraints using nodes and weighted directed edges. This allows schedulers to not only see whether the load response is insufficient, but also clearly see whether it is due to a decline in the effectiveness of price incentives or a tightening of equipment safety boundaries. Based on this interpretable graph result, schedulers can adjust strategies more effectively, for example, instead of simply increasing compensation prices, they can shift towards allocating other resources or alleviating physical bottlenecks.

[0044] In summary, this disclosure can accurately identify the actual effects of economic incentives and industrial physical constraints on load response changes, and provides an interpretable basis for adjusting scheduling strategies.

[0045] After introducing the basic principles of this disclosure, various non-limiting embodiments of this disclosure will be described in detail below.

[0046] refer to Figure 1 This is a schematic diagram of an application scenario of the flexible resource influence map generation method provided by the exemplary embodiments of this disclosure.

[0047] This application scenario includes a terminal device 101 and a server 102. The terminal device 101 and the server 102 can be connected via a wired or wireless communication network to achieve data interaction.

[0048] Terminal device 101 may be an electronic device located close to the user side, possessing data transmission and multimedia input / output functions, including but not limited to desktop computers, mobile phones, portable computers, tablet computers, media players, smart wearable devices, personal digital assistants (PDAs), or other electronic devices capable of performing the aforementioned functions. This electronic device may include a processor and a display screen with touch input functionality. The display screen is used to present a graphical user interface (GUI), which can display an application interface. The processor is used to process application data, generate the GUI, and control the display of the GUI on the screen.

[0049] Server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0050] In some exemplary embodiments, the flexible resource influence map generation method can run on terminal device 101 or server 102.

[0051] When the flexible resource influence map generation method runs on server 102, server 102 is used to provide flexible resource influence map generation services to users of terminal device 101.

[0052] Server 102 determines the original dataset of flexible resource regulation transmitted by the target industrial user to terminal device 101. Server 102 constructs core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix. Server 102 performs causal skeleton identification on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph. Server 102 performs dynamic edge weight calculation on the time-series unfolded causal graph to obtain dynamic causal influence weights; After generating an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights, server 102 transmits the influence graph to terminal device 101 so that terminal device 101 can adjust the scheduling strategy of flexible resource regulation based on the influence graph.

[0053] It should be noted that the above application scenarios are shown only to facilitate understanding of the spirit and principles of this disclosure, and the implementation of this disclosure is not limited in any way. On the contrary, the implementation of this disclosure can be applied to any applicable scenario.

[0054] refer to Figure 2 A method for generating flexible resource influence maps, the method comprising the following steps: Step S210: Determine the original dataset of flexible resource regulation for the target industrial user, construct core causal nodes based on the original dataset, and obtain a standardized causal analysis data matrix.

[0055] In practice, flexible resource regulation refers to the process by which key industrial users (especially those in continuous production industries such as steel, chemical, cement, and non-ferrous metal smelting) respond to economic incentives such as grid compensation prices, spot market price signals, demand response invitations, or interruptible load agreements in response to long-cycle demand response and extreme weather supply guarantee scenarios. This is based on constraints of industrial physical conditions such as production processes, equipment operating boundaries, material and warehouse inventory, gas storage capacity, equipment ramp-up capabilities, order delivery, and safety production requirements. By adjusting some processes, utilizing intermediate buffer inventory, or releasing by-product energy, they can implement load regulation by responding to economic incentives such as grid compensation prices, spot market price signals, demand response invitations, or interruptible load agreements.

[0056] In some embodiments, the original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data.

[0057] In practice, the subset of external environmental data refers to meteorological and environmental data that covers the entire process of demand response events, mainly including time-series data such as ambient temperature and high / low temperature impact thresholds used to characterize the intensity of extreme weather disasters.

[0058] In practice, the power grid operation status data subset refers to power grid operation data that covers the entire process of demand response events. It mainly includes time-series data such as power grid forecast load, available power supply capacity, and load gap, which are used to characterize the degree of power grid supply and demand tension.

[0059] In practice, the market incentive and dispatch instruction data subset refers to the price incentive and control instruction data released by the power grid or market to industrial flexible resources, covering the entire process of demand response events. It mainly includes time-series data such as compensation electricity price, spot price, response invitation price, dispatch instruction status, target response quantity, and cumulative compensation cost.

[0060] In practical implementation, industrial digital twin and physical state data subset refers to the underlying physical state data of industrial production process covering the entire process of demand response events. It mainly includes time-series data such as intermediate material warehouse inventory, gas storage tank or energy storage unit capacity, key equipment operating power, equipment ramp-up capability, equipment output boundary and safe operation lower limit, which are used to characterize the physical feasible space for industrial users to continue to participate in demand response.

[0061] In practice, the production performance and low-carbon constraint data subset refers to the production delivery and low-carbon performance data of industrial users that cover the entire process of demand response events. It mainly includes time-series data such as cumulative orders or planned output, actual cumulative output, allowed cumulative delays or output deviations, cumulative carbon emissions or converted carbon emissions, and carbon quotas or allowed emissions as of the current moment, which are used to characterize the production delivery, contract performance and low-carbon constraint pressures experienced by industrial users in the continuous response process.

[0062] In practice, the apparent response data subset of industrial users refers to the actual power load response data of industrial users covering the entire process of demand response events. It mainly includes time-series data such as baseline load, actual power consumption of industrial users, actual response quantity (the difference between baseline load and actual power consumption) and response deviation, which are used to characterize the actual response degree of industrial users relative to the baseline load at time t, under conditions where they are not involved in the current demand response or are not affected by the current dispatch instructions.

[0063] In some embodiments, core causal nodes are constructed based on the original dataset to obtain a standardized causal analysis data matrix, including: The original dataset is time-scale aligned to obtain an aligned multi-source feature matrix; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is subjected to industrial physical constraint verification to obtain the verified core causal node sequence. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

[0064] In practice, the original dataset is aligned by time scale to obtain the aligned multi-source feature matrix as follows: Minimum scheduling cycle for demand response As a unified time step, construct discrete time series Each scheduling cycle corresponds to a time window Represented as:

[0065] For continuously operating variables such as load, equipment power, and ambient temperature, the average value or time-weighted average value is used for alignment. For safety constraint variables such as material warehouse inventory lower limit distance, gas holder margin, and equipment ramp-up margin, minimum margin or extreme value sampling methods are preferred to retain short-term risks. For variables such as compensation price, spot price, response invitation price, and dispatch instruction status, the effective price, last effective value, or time-weighted value within the corresponding dispatch period is used. For cumulative variables such as cumulative compensation cost, cumulative output, and cumulative carbon emissions, the end value of the window or the window increment value is used.

[0066] For continuously running variables, the time window average alignment representation is as follows:

[0067] For safety margin type variables, the minimum value within the window is used for alignment as follows:

[0068] For risk intensity variables, the representation using maximum value alignment within the window is as follows:

[0069] For cumulative variables, the window end value or window increment value is used to represent them, thereby unifying multi-source heterogeneous data to the same time scale and obtaining an aligned multi-source feature matrix.

[0070] in, This indicates the number of time samples after alignment. This indicates the number of original or intermediate features involved in constructing the core causal node.

[0071] In specific implementation, the aligned multi-source feature matrix is ​​cleaned to obtain the cleaned multi-source feature matrix in the following way: For variables with clearly defined physical upper and lower limits, anomaly detection is performed based on equipment parameters and process boundaries. When the variable value exceeds the physical lower limit... Or physical limit The time stamp is marked as an outlier and corrected using physical boundary truncation, where physical boundary truncation is expressed as:

[0072] Simultaneously, it combines interpolation of adjacent time windows, correction of data from similar devices, or processing based on estimates of device state equations; for short-term missing data, forward filling or linear interpolation is used to fill in the missing data; for long-term missing data segments or data segments with unobservable key physical states, they are marked as low-confidence samples or directly removed, thereby reducing the risk of false causal relationships introduced by data quality issues and obtaining a cleaned multi-source feature matrix.

[0073] In practice, the power grid dispatching business logic refers to guiding industrial loads with adjustment capabilities to participate in peak shaving, peak shifting, or peak transfer through methods such as compensation electricity prices, spot market price signals, demand response invitations, and interruptible load agreements. Based on the dominant factors of changes in the response capabilities of industrial users, it is decided whether to continue using price incentive strategies or switch to alternative control strategies such as cross-regional resource rotation, industry-specific peak shifting, reserve resource mobilization, and interruptible load contingency plans.

[0074] In practical implementation, the industrial physical mechanism refers to the thermal inertia, production inertia and self-regression characteristics of the industrial production process. The changes in equipment status, material inventory, gas holder pressure and load are affected by historical conditions. The industrial load response is constrained by physical factors such as production process, equipment operating boundary, material warehouse inventory, gas holder capacity, equipment ramping ability and safety production requirements. During continuous production, enterprises can reduce load by adjusting processes, utilizing intermediate buffer inventory or releasing by-product energy. However, under long-term regulation, after the gradual consumption of buffer resources such as material warehouses, gas holders and equipment ramping margin, the physical feasible space will shrink. Moreover, materials or energy must meet the conservation relationship. The operating power and ramping rate of equipment are constrained by the upper and lower limits of safety, and the flexible response capability is limited by the weakest physical link.

[0075] In specific implementation, the core causal node sequence is constructed from the cleaned multi-source feature matrix based on power grid dispatching business logic and industrial physical mechanisms, as follows: The multi-source feature matrix is ​​divided into five categories according to its role in the demand response process: external drivers, economic incentives, physical constraints, production pressures, and response results. For each category of features, physical meaning-clear normalization, weighted fusion, or bottleneck aggregation calculation are performed, and physical upper and lower limits are used for truncation. In this way, the original features are transformed into a standard time-series node sequence that can characterize the flexible resource response mechanism.

[0076] In specific implementation, the core causal node sequence is subjected to industrial physical constraint verification to obtain the verified core causal node sequence in the following way: For any buffer unit Its state should satisfy Furthermore, the state transition should satisfy the material or energy conservation relationship:

[0077] in, and These represent the material flow rate, gas flow rate, or energy flow rate entering and exiting the buffer unit, respectively. For any critical equipment... Its operating power should meet the requirements Its climbing constraint should satisfy:

[0078] The amount of positive reduction that industrial users can implement should also meet the lower limit constraint for safe operation:

[0079] in, Indicates the amount of positive reduction that can be implemented. This represents the minimum power consumption required to meet industrial safety production constraints. If the actual observation or calculation results violate the above physical constraints, the corresponding sample will be marked as an anomalous sample, a low-confidence sample, or undergo physical boundary correction, thereby obtaining the verified core causal node sequence.

[0080] In specific implementation, the core causal node sequence after verification is subjected to stationarity processing to obtain a stationarized node sequence in the following way: For each core node sequence in the verified core causal nodes Perform an ADF test, setting the significance level to be [value missing]. If the ADF test yields If the sequence has a unit root, it is considered non-stationary. In this case, the first difference of the sequence is performed as follows:

[0081] If the stationarity requirement is still not met after first-order differencing, further processing is carried out by adopting higher-order differencing, detrending, rolling window modeling, or scenario-specific modeling, depending on the business scenario. At the same time, the maximum differencing order is limited to avoid excessive differencing from destroying the physical meaning, thereby obtaining a stationary node sequence.

[0082] In specific implementation, the standardized causal analysis data matrix is ​​obtained by standardizing the stationary node sequence as follows: set up Represented as the th after differencing or detrending processing Given a sequence of nodes, calculate the sample mean of that sequence. and sample standard deviation Then, the value at each time step is standardized, and the formula is expressed as:

[0083] in, To prevent extremely small positive numbers with a denominator of zero, thus obtaining a standardized causal analysis data matrix. ,in This indicates the number of valid time samples after data cleaning and stabilization.

[0084] In some embodiments, the core causal node sequence includes: external disaster-causing stimuli and grid supply and demand tension nodes, economic incentive intensity nodes, industrial physical security margin nodes, compliance and low-carbon pressure nodes, and industrial user actual response action nodes.

[0085] In practice, the method for obtaining the nodes of external disaster-causing stimuli and the tension between power grid supply and demand is as follows: For high-temperature scenarios, the intensity of meteorological disasters can be expressed as:

[0086] For cold wave scenarios, the intensity of meteorological disasters can be expressed as:

[0087] in, Indicates ambient temperature. This indicates the threshold of influence of high temperature. This indicates the upper limit of the normalized reference for high temperature. This indicates the threshold of low temperature effect. This indicates the lower limit of the normalized reference for low temperature. To prevent extremely small positive numbers with a denominator of zero.

[0088] The intensity of the power grid load gap can be expressed as:

[0089] in, Indicates the predicted load on the power grid. Indicates available power supply capacity. This represents the normalized reference value for the load gap.

[0090] The node, which combines external disaster-causing stimuli with the degree of power grid supply and demand tension, can be represented as:

[0091] in, and These represent the weights of the intensity of meteorological disasters and the intensity of the power grid load gap, respectively. This node is used to characterize the driving effect of the external environment and the power grid supply and demand status on demand response events.

[0092] External disaster stimuli and the degree of tension between power grid supply and demand It is used to characterize the dispatch pressure caused by extreme weather and the power grid supply and demand gap.

[0093] In practice, the method for obtaining the economic incentive intensity node is as follows: The intensity of economic incentives is represented as follows:

[0094] in, Indicates time The corresponding compensation electricity price, response invitation price, or market incentive price, and These represent the lower and upper bounds of price normalization, respectively. This represents the target response quantity issued by the power grid. This represents the normalized reference value of the target response. ,and , To ensure numerical stability, it can be... Cut off to Interval.

[0095] Economic incentive intensity node This is used to characterize the intensity of price incentives and regulation in the release of flexible resources from the power grid or market to industry. If demand response compensation prices are the primary metric, then... Compensation price Normalization yields the result; if both price level and target response strength are considered simultaneously, the target response quantity is further reduced. Incorporate into the structure.

[0096] In practice, the industrial physical safety margin node is obtained as follows: For the For each silo, gas holder, or buffer unit, its inventory or capacity margin is expressed as follows:

[0097] in, Indicates the first Each buffer unit at time... The inventory, liquid level, gas volume, or capacity status. and These represent its lower safety limit and upper safety limit, respectively. For the... The ramp margin of each key piece of equipment is expressed as follows:

[0098] in, Indicates the first Key equipment at all times Operating power This represents the maximum allowable rate of power change for the equipment. For the equipment's output boundary, its output margin can be expressed as:

[0099] in, and These represent the lower and upper limits of the equipment's safe operating power, respectively. The above... , and It can be truncated to the [0,1] interval as needed to ensure that the node values ​​have a uniform scale.

[0100] Considering that the flexible response capability in continuous industrial production processes is often limited by the weakest physical link, this scheme preferably adopts a short-board constraint approach to construct a comprehensive industrial physical safety margin:

[0101] in, This represents the set of buffer units participating in demand response. This represents a collection of key energy-consuming equipment. When... When the value approaches 0, it indicates that at least one critical physical link is nearing the safety boundary, and the physical feasible space for industrial users to continue reducing load has significantly shrunk; when A value close to 1 indicates that the internal physical buffer of the industrial user is relatively sufficient. If the importance of different devices or buffer units varies significantly in a specific industrial scenario, a weighted fusion method using expert weighting or data-driven weighting can also be used to construct the buffer. As an alternative implementation method.

[0102] Industrial physical safety margin node This node is used to characterize the physical feasible space for industrial users to continue participating in demand response at the present moment. It is mainly determined by factors such as intermediate material warehouse inventory, gas storage tank or energy storage unit capacity, ramp-up capability of key equipment, and equipment output boundary. It is a key node that distinguishes this solution from analysis based solely on electricity price, weather, and historical load data.

[0103] In practice, the methods for obtaining compliance and low-carbon pressure points are as follows: Order fulfillment pressure can be expressed as:

[0104] in, Indicates the end time The cumulative orders or planned production volume that should be completed. Indicates the end time Actual cumulative output This indicates the allowable cumulative delay or production deviation.

[0105] The pressure to meet low-carbon commitments can be expressed as:

[0106] in, Indicates the end time The cumulative carbon emissions or equivalent carbon emissions, Indicates the end time The corresponding carbon quota or allowable emissions.

[0107] The overall compliance and low-carbon pressure points are as follows:

[0108] in, and These represent the normalized reference values ​​for order fulfillment pressure and low-carbon fulfillment pressure, respectively. ,and , To ensure the stability of the node value range, it can be... Cut off to the interval [0,1].

[0109] Compliance and Low-Carbon Pressure Nodes It is used to characterize the pressures of production delivery, contract fulfillment, and low-carbon constraints faced by industrial users during continuous response processes.

[0110] In practice, the method for obtaining the actual response action nodes of industrial users is as follows: set up This represents the baseline load under conditions where it is not involved in the current demand response or affected by the current dispatch instructions. If the actual power consumption of an industrial user is represented, then the actual response quantity is expressed as:

[0111] when When, it indicates that the user has achieved a load reduction response relative to the baseline load; when This indicates that the user is experiencing a load rebound or that the actual load is higher than the baseline. The actual response action node for industrial users is defined as follows:

[0112] in, This represents the baseline load normalized reference value. If this scheme is only used to analyze the positive peak-shaving response capability, it can also be used.

[0113] As a preferred implementation, to simultaneously identify the risks of insufficient response and load rebound, [the following is retained] The symbolic information is more helpful for subsequent causal analysis.

[0114] Industrial user actual response action nodes Used to characterize the actual response of industrial users relative to the baseline load at any given time.

[0115] Step S220: Perform causal skeleton identification on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph.

[0116] In some embodiments, causal skeleton identification is performed on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph, including: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

[0117] In specific implementation, the standardized causal analysis data matrix is ​​constructed with candidate parent nodes based on the maximum candidate causal time lag, and the complete set of candidate parent nodes is obtained as follows: Combining the maximum physical propagation time in industrial production and scheduling response Minimum scheduling cycle for demand response The maximum candidate time delay step number can be set as follows:

[0118] in, This indicates rounding up. If economic incentives, industrial physical conditions, and performance pressures have different transmission durations in a specific scenario, maximum time delays can be set for different categories of variables, and corresponding time delay upper limits can be applied to the corresponding candidate paths.

[0119] For each target node Construct the complete set of candidate parent nodes:

[0120] in, This indicates that the same node is prohibited from pointing to itself at the same time, i.e., it is prohibited. The self-loop relationship at the same moment. For Its own history items, such as If so, it is allowed to enter the testing process as a candidate, because variables such as industrial load, inventory, equipment status and performance pressure usually have certain autoregressive characteristics.

[0121] In specific implementation, the prior mask matrix is ​​constructed based on the industrial physical mechanism to obtain the time-delay causal prior mask matrix by constructing the entire set of candidate parent nodes. Construct a Boolean time-delay causal prior mask matrix , used to represent variables Is it allowed to be the target node? Candidate causal parent nodes.

[0122] The mask matrix is ​​defined as follows:

[0123] The first type of constraint is used to prohibit nodes from self-looping at the same time; the second type of constraint is used to protect nodes from external disaster stimuli and the degree of power grid supply and demand tension. The exogenous properties, namely the internal state of the industry and user response behavior, should not negatively influence external meteorological conditions or the initial supply and demand tension of the power grid; the third type of constraint is used to limit synchronous edges at the same time, and is only allowed when there is a clear physical transmission direction, process sequence relationship, or scheduling business sequence relationship. The fourth type of constraint is used to exclude impossible industrial transmission paths based on the industrial digital twin model, process topology, material flow and energy flow direction; the fifth type of constraint is used to exclude unreasonable paths based on scheduling business rules, such as actual response actions should not be the direct cause of scheduling price formation at the same time.

[0124] To avoid the existence of mutually pointing synchronization edges within the same time slice, for The candidate paths should also satisfy the constraint of uniqueness of the synchronization edge direction: .

[0125] In specific implementation, the initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix as follows: For each target node According to the mask matrix Filter the values ​​and keep only those that meet the criteria. candidate parent node That is, the initial set of candidate parent nodes is defined as:

[0126] This process eliminates candidate paths that violate time sequence, industrial physical mechanisms, and scheduling business logic, resulting in an initial set of candidate parent nodes.

[0127] In practice, the initial set of candidate parent nodes is conditionally independent to obtain the filtered set of candidate parent nodes as follows: For the target node With any candidate parent node Set a subset of condition variables Construct a conditional independence test to test the null hypothesis:

[0128] in, Indicates a given set of condition variables In the case of candidate causal variables With target variable Conditional independence. Start testing with a lower-dimensional set of conditions and gradually increase the size of the set to reduce the computational burden of testing conditional independence in higher dimensions.

[0129] Starting with a lower-dimensional condition set, gradually increase the size of the condition set, and compute... Let the significance level be . ,like Then the null hypothesis cannot be rejected, and the candidate parent node is removed from the current candidate set; if Then retain it.

[0130] Iterate until the candidate set no longer changes or the preset maximum condition set dimension is reached. Finally, the target node is obtained. The filtered set of candidate parent nodes: .

[0131] In practice, the transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain the set of candidate causal edges in the following way: For each candidate edge ,make , Construct a set of control variables that simultaneously includes the target node's other candidate parent nodes and the historical parent nodes of the causal variable itself. :

[0132] in For the target node Other candidate parent nodes are used to control other direct factors that affect the target node; Causal variable The candidate parent node in its corresponding time slice is used to control the historical inertia and historical driving factors of the causal variable itself; for The construction of the time index should be shifted accordingly to the time index. Or an earlier film.

[0133] Under the above control set, construct the null hypothesis for transient conditional independence testing:

[0134] If in the given back, and If a candidate edge still exhibits a significant conditional dependency, then it is considered a candidate edge. It has value for retention; if the two are no longer significantly correlated under this condition set, it indicates that the candidate relationship may be mainly caused by other parent nodes, autoregressive inertia or historical common driving factors, and should be removed from the causal skeleton.

[0135] Subsequently, conditional mutual information was used as a test for conditional independence. For variables... , and condition variable set Conditional mutual information is defined as:

[0136] In the formula, The closer to 0, the better the value of the given condition. back, and The closer to conditional independence; if If it is significantly greater than 0, it means and In control Conditional dependencies still exist.

[0137] As a specific embodiment, it can be based on Nearest neighbor conditional mutual information estimator pair Make an estimate. Let the number of valid samples be... The nearest neighbor parameter is , For the Digamma function, , and They represent the first The conditional mutual information estimate can be expressed as: (The count of neighboring samples in the corresponding subspace for each sample is given.)

[0138] To reduce the impact of estimation error on significance judgment under limited sample conditions, this scheme can use a random permutation test to calculate the significance level. Specifically, maintaining... and The sample correspondence remains unchanged, for Perform random permutations within the time block to obtain Group the permutation samples and calculate the mutual information of the permutation conditions for each group. The significance level of the permutation test is defined as:

[0139] in, This represents the conditional mutual information estimate calculated from the original sample. Indicates the first The conditional mutual information estimate corresponding to the permutation sample. For indicator functions, Indicates the number of permutations.

[0140] Let the significance level for the transient conditional independence test phase be . If candidate edge The test results satisfy If so, then retain the candidate edge; if If the candidate edge is not found, then the candidate edge is removed, thus obtaining the set of candidate causal edges.

[0141] In specific implementation, the temporal causal edge set is determined by establishing the temporal causal edge set as follows: For each candidate path The path is determined as a candidate causal edge in the temporal causal skeleton if it simultaneously meets the following conditions: First, it satisfies the physical prior mask constraint, i.e. Second, preliminary screening is conducted through candidate parent nodes, i.e. Third, through the transient conditional independence test, i.e. Fourth, for The synchronization edges satisfy the uniqueness of the synchronization edge direction and the constraint of the industrial physics transmission direction; thus, the set of temporal causal edges is obtained:

[0142] In specific implementation, the temporal unfolded causal graph is constructed based on the set of temporal causal edges, and the temporal unfolded causal graph is obtained in the following way: All core node instances after being expanded by time slice are treated as a node set. The established set of temporal causal edges As a set of directed edges, a temporal unfolded causal graph can be directly constructed:

[0143] Each node and Each edge has a clear time index, and each edge retains the causal time delay. Information is used to obtain a time-series unfolded causal graph that can characterize the cross-time influence relationship between variables.

[0144] Step S230: Perform dynamic edge weight calculation on the temporal unfolded causal graph to obtain dynamic causal influence weights.

[0145] In some embodiments, dynamic edge weight calculation is performed on the temporal unfolded causal graph to obtain dynamic causal influence weights, including: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

[0146] In specific implementation, the causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node in the following way: For time-series unfolded causal graph Each target node in Traverse the edge set Extract all The source node and its time delay corresponding to the directed edge ending at a given point are defined as follows:

[0147] This yields the set of causal parent nodes for each target node.

[0148] In specific implementation, a structure mask is constructed based on the set of causal parent nodes, and the structure mask is obtained in the following way: For each target node and its causal parent node set , To ensure that the model input satisfies the causal structure constraints of the output in the above embodiments, the structure mask can be defined as:

[0149] in, Representing variables Allowed as a target node The structure function input, This indicates that the variable is masked. Through this mask, the model front end only retains the parent node input confirmed by the causal skeleton in the above embodiment, and masks variables that have not passed the physical prior constraints or conditional independence tests.

[0150] In specific implementation, the nonlinear structure function model is constructed based on the structure mask to obtain an approximate structure function model as follows: For each target node Using structural masks By disabling input variables that fail the causality test and allowing only variables from the set of causal parent nodes as model input, an input vector is constructed:

[0151] in, Represents the target node The causal parent node input vector, This indicates that the time delay retained from the above embodiments parent node Subsequently, an approximate model of the nonlinear structure function is constructed:

[0152] This model can employ nonlinear architectures such as temporal fusion Transformer, temporal convolutional networks, and long short-term memory networks; it minimizes the objective function:

[0153] Supervised training is performed to obtain an approximate model of the structure function. Among these, This represents the optimal parameters obtained through training. Represents the target node The number of effective training samples, Represents the set of valid sample time indices. Represents the regularization coefficient. This represents the L2 norm regularization term for the parameter.

[0154] In practice, the marginal causal effect is obtained by calculating the local intervention effect based on the structure function approximation model as follows: For established causal edges At sample points The structure function model is calculated using automatic differentiation or finite difference. For causal variables The partial derivatives are expressed as:

[0155] in, Marginal causal effect, indicating that at time... Nearby, parent node Small changes to the target node The intensity is affected by the local edges.

[0156] Alternatively, a finite difference approximation can be used when the model is not differentiable.

[0157] This yields the local marginal causal effect of the causal variable on the target variable at each time step; where, This represents a small perturbation. Indicates only in variables A unit perturbation vector with 1 at the corresponding position and 0 at other positions.

[0158] In specific implementation, the dynamic causal influence weights are obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effects as follows: In time As the endpoint, take a length of Scrolling time window ,in, Indicates the window length. For causal edges... The average absolute causal influence weight of its scrolling window is defined as:

[0159] in, Indicates time Causal edges within the corresponding scrolling window The average influence strength. Absolute values ​​are used to measure the magnitude of the influence strength and are used to display the thickness of lines or the magnitude of weights in subsequent graphs.

[0160] Furthermore, the signed average, obtained through calculation, can be expressed as:

[0161] Where, in the formula, This is used to indicate the average positive or negative influence direction of the causal edge within the current window. If This indicates that an increase in the causal variable tends to increase the target variable; if This indicates that an increase in the causal variable tends to decrease the target variable. This can be used in subsequent graph visualizations. The thickness of the mapping lines is determined by... Mapping affects direction or color attributes.

[0162] To facilitate relative comparisons between different edges, the weights of incoming edges to the same target node can be normalized. For the target node... Its normalized edge weights can be expressed as:

[0163] in, This represents the normalized relative influence weight. To prevent extremely small positive numbers with a denominator of zero, this weight is used to characterize the relative proportion of the influence of a parent node on the target node among all incoming edge influences within the current time window.

[0164] If the temporal causal skeleton identified in the above embodiments contains If the direct edge is defined, then the dynamic direct influence of economic incentives on response actions is defined as:

[0165] If the temporal causal skeleton identified in the above embodiments contains If the direct edge is defined, then the dynamic direct influence of industrial physical margin on the response action is defined as: ) in, This indicates the dynamic strength of the impact of economic incentives on response actions within the current rolling window. This indicates the dynamic influence of the industrial physical safety margin factor on the response action within the current rolling window. If multiple time-delay edges exist, the weights of each time-delay edge are summed to represent the combined influence of the factor on the response action at different lag periods.

[0166] In some scenarios, the impact of economic incentives or physical margins on response actions may be indirectly transmitted through intermediate nodes such as performance pressure and production status. To enhance the interpretability of the graph, this invention can further calculate the impact from the cause node. To the response action node The path influence. Let... Indicates from arrive The set of all influence paths that conform to the time series direction, for any path A path influence can be represented as consisting of several edges:

[0167] In the formula, Represents a node Transmitted to the response action node via direct or indirect path The overall influence This represents the normalized dynamic edge weight of a single edge in the path. This path influence can serve as a supplement to the direct edge weights to explain situations where price incentives or physical margins indirectly affect response actions through intermediate mechanisms.

[0168] In a preferred embodiment, if or and If a stable direct edge exists, then it is preferred. and Implement early warning for the effectiveness of price incentives; if a direct edge does not exist but a stable indirect path exists, then adopt... and As a supplementary indicator.

[0169] Step S240: Generate an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights, so as to adjust the scheduling strategy of the flexible resource regulation based on the influence graph.

[0170] In some embodiments, generating an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights includes: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; The influence graph is obtained based on the aggregated influence path set and the aggregated dynamic edge weights.

[0171] In specific implementation, the method for extracting aggregated influence paths from the time-series unfolded causal graph to obtain the aggregated influence path set is as follows: Time-series unfolded causal graph Each causal edge with a time index Mapped as aggregate edges at the variable level That is, the set of aggregated influence paths is defined as:

[0172] This ignores specific time slice indices and retains the causal time lag information between variable pairs for subsequent visualization of dynamic influence maps.

[0173] In specific implementation, the dynamic causal influence weights of the aggregated influence path set are aggregated to obtain the aggregated dynamic edge weights in the following way: For variable pairs with multiple time delay edges By aggregating the influence weights under different lag periods, we obtain the aggregated dynamic edge weights at the graph display level:

[0174] in, Indicates time variable For variables The strength of the aggregation effect is used to map the thickness or weight of directed edges in the graph. Correspondingly, the direction of the aggregated signed effect can be represented as:

[0175] This allows the effects of the same variable on different lag periods to be combined into a single aggregate dynamic edge weight, where, Used to represent variables For variables The combined positive or negative influence direction. If ,express Increase tends to improve ;like ,express Increase tends to decrease .

[0176] In specific implementation, the influence graph is obtained based on the aggregated influence path set and the aggregated dynamic edge weights as follows: Five core causal nodes As a graph node, it will aggregate the set of influence paths. As directed edges, and pair each variable with Corresponding aggregate dynamic edge weight Mapping to the thickness or weight of directed edges will aggregate the signed influence on the direction. The mapping is represented by edge colors or markers (distinguishing between positive and negative impacts), while the current physical state (such as industrial physical safety margin values, actual response quantities, etc.) is overlaid on each node, thereby generating a dynamic impact map of flexible resources:

[0177] This tool is used to visually demonstrate the changing degree to which various factors influence the response actions of industrial users over different time periods.

[0178] In specific implementation, the method for adjusting the scheduling strategy of flexible resource regulation based on the influence map is as follows: The dynamic influence map of flexible resources The output is sent to a scheduling-aided decision-making system or a human-computer interaction interface. Five core causal nodes are included. Each node is rendered as a graph node, and the aggregated influence path set is identified through the above steps. Rendered as directed edges; the thickness of each edge and the aggregated dynamic edge weight. Positive correlation is used to characterize the influence strength of the influence path within the current scrolling time window; the color or label of each edge is determined by the aggregated signed influence direction. The decision is used to distinguish between positive and negative effects; for variable pairs that fail the physical prior constraints and conditional independence tests, causal connections are not displayed on the graph. Simultaneously, each node's current physical state is overlaid: Node Displays the current value of industrial physical safety margin, minimum margin for silos or gas holders, equipment ramp margin, and proximity to safety boundaries; nodes Displays actual response volume, response deviation, and load rebound status; Node Displays the current compensation price, target response amount, and cumulative compensation cost.

[0179] Subsequently, the economic incentive effect obtained through the above embodiments and industrial physical margin influence Calculate the dominance ratio of the physical margin effect:

[0180] in, Indicates time The degree to which the influence of industrial physical margin outweighs the influence of economic incentives; Indicates the industrial physical safety margin for response action nodes Dynamic influence; This indicates the effect of the intensity of economic incentives on the response action node. Dynamic influence; To prevent extremely small positive numbers from having unstable values ​​due to excessively small denominators And its growth trend indicators:

[0181] in, Indicates from arrive Within the rolling window, the growth trend of the physical margin impact relative to the economic incentive impact is... Indicates the length of the warning judgment window. When... When the value is large, it indicates that the dominant role of physical constraints on the response action is increasing, while the relative role of economic incentives on the response action is decreasing.

[0182] When the following conditions are met simultaneously, it can be determined that there is a risk of diminishing marginal effect of price incentives and dominant industrial physical constraints:

[0183]

[0184]

[0185] in, This represents the dominance threshold of physical margin influence relative to economic incentive influence. This indicates the threshold for the growth of the dominant trend. This indicates the warning threshold for industrial physical safety margin. The three conditions mentioned above correspond to "the impact of physical margin has become dominant", "the degree of dominance of physical margin is increasing", and "the industrial physical margin itself is approaching the safety boundary".

[0186] And set an early warning status. ;otherwise Preferably, the indicator of the decline in the influence of economic incentives is calculated. This serves as an auxiliary basis for judgment. Among them, when... And exceeding the preset threshold When this timeframe is reached, it indicates that the influence of economic incentives on response actions is showing a downward trend within the current window. This indicator can serve as supplementary evidence of declining effectiveness of price incentives, in conjunction with... , and Used together for early warning judgment when At the same time, further pinpoint the bottlenecks in industrial physics: for the set of buffer units Calculate the margins of each silo, gas holder, or energy storage unit. For key equipment sets Calculate the ramp margin of each piece of equipment and output margin The current physical bottleneck is represented as:

[0187] in, This indicates the physical bottleneck closest to the safety boundary, which could be a silo, gas holder, energy storage unit, equipment ramp-up constraint, or equipment output boundary. This physical bottleneck information is used to explain why industrial users continue to reduce their load capacity, thus avoiding attributing inadequate response solely to insufficient price incentives.

[0188] The generated attribution evidence includes: and relative size, and Trends in industrial physical security margin Current value, minimum physical margin stage This includes information on corresponding material warehouse inventory, gas holder capacity, equipment ramp-up margin, or equipment output boundary. The aforementioned attribution evidence is presented in the dynamic influence map as node annotations, bottleneck warning boxes, or diagnostic reports.

[0189] when At the same time, it can also output a set of strategy suggestions:

[0190] in, Indicates the first Dispatch assistance suggestions include, but are not limited to, reducing the priority of a single price increase strategy, switching to other flexible resources, initiating cross-regional rotation dispatch, calling standby units, triggering interruptible load contingency plans, or delaying the re-call of the user. This set of strategy suggestions serves only as auxiliary decision-making output and does not change the dispatcher's final control authority.

[0191] The proposed strategies include, but are not limited to: indicating that the marginal effect of a single price incentive is decreasing for current industrial users or resource groups, and that it is not advisable to rely solely on further increasing compensation prices to expand the response volume; indicating that the current decline in response capability is related to the industrial physical security margin approaching its limit, and providing key physical bottlenecks. It is recommended that dispatchers consider cross-regional resource rotation, cross-industry peak shifting, calling up backup adjustment resources, activating interruptible load protocols, or adjusting subsequent demand response invitation targets; it is also advised that when continuing to call upon this industrial user to participate in the response, attention should be paid to recovery time, buffer resource replenishment status, and production performance risks. And the economic incentive effect If the current price incentive strategy remains dominant, it indicates that the strategy is still effective. Demand response scheduling can continue based on the power grid supply-demand gap and compensation budget, while continuous monitoring is maintained. , and The changes in [the system] prevent the industrial physical margin from gradually approaching the safety boundary during continuous regulation. The set of strategy recommendations is only used as an auxiliary decision-making output and does not change the final control authority of the scheduler.

[0192] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0193] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0194] Based on the same inventive concept, corresponding to any of the above embodiments, this disclosure also provides a flexible resource influence map generation device.

[0195] refer to Figure 3 The flexible resource influence map generation device includes: The data matrix determination module 310 is configured to determine the original dataset of flexible resource regulation for target industrial users, and construct core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix. The causal graph determination module 320 is configured to perform causal skeleton identification on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph. The influence weight determination module 330 is configured to perform dynamic edge weight calculation on the time-series unfolded causal graph to obtain dynamic causal influence weights. The influence graph determination module 340 is configured to generate an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights, so as to adjust the scheduling strategy of the flexible resource regulation based on the influence graph.

[0196] In this exemplary embodiment, the data matrix determination module 310 is specifically configured as follows: The original dataset for flexible resource regulation of target industrial users is determined, and the original dataset is aligned to a time scale to obtain an aligned multi-source feature matrix; wherein, the original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is verified by industrial physical constraints to obtain the verified core causal node sequence; wherein, the core causal node sequence includes: external disaster-causing stimuli and the degree of tension between power grid supply and demand, economic incentive intensity, industrial physical safety margin, compliance and low-carbon pressure, and actual response actions of industrial users. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

[0197] In this exemplary embodiment, the cause-effect graph determination module 320 is specifically configured as follows: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

[0198] In this exemplary embodiment, the influence weight determination module 330 is specifically configured as follows: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

[0199] In this exemplary embodiment, the influence map determination module 340 is specifically configured as follows: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; Based on the aggregated influence path set and the aggregated dynamic edge weights, the influence graph is obtained, so that the flexible resource regulation and scheduling strategy can be adjusted based on the influence graph.

[0200] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0201] The apparatus described above is used to implement the corresponding flexible resource influence map generation method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0202] Based on the same inventive concept, corresponding to any of the above embodiments, this disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the flexible resource influence map generation method described in any of the above embodiments.

[0203] Figure 4 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0204] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0205] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0206] The input / output interface 1030 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0207] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0208] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0209] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0210] The electronic devices described above are used to implement the corresponding flexible resource influence map generation method in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0211] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the flexible resource influence map generation method as described in any of the above embodiments.

[0212] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0213] The aforementioned non-transitory computer-readable storage media can be any available medium or data storage device that a computer can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0214] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the flexible resource influence map generation method as described in any of the embodiments in the exemplary method section above, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0215] Based on the same inventive concept, corresponding to the flexible resource influence map generation method described in any of the above embodiments, this disclosure also provides a computer program product, which includes computer program instructions. In some embodiments, the computer program instructions can be executed by one or more processors of a computer to cause the computer and / or the processor to execute the flexible resource influence map generation method. Corresponding to the execution entity for each step in each embodiment of the flexible resource influence map generation method, the processor executing the corresponding step can belong to the corresponding execution entity.

[0216] The computer program product of the above embodiments is used to cause the computer and / or the processor to execute the flexible resource influence map generation method as described in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0217] Those skilled in the art will recognize that embodiments of this disclosure can be implemented as a system, method, or computer program product. Therefore, this disclosure can be implemented as entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this disclosure can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0218] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (not exhaustive) of a computer-readable storage medium may include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0219] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0220] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0221] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0222] It should be understood that each block of a flowchart and / or block diagram, as well as combinations of blocks in a flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine that, when executed by a computer or other programmable data processing device, creates means for implementing the functions / operations specified in the blocks of the flowchart and / or block diagram.

[0223] These computer program instructions may also be stored in a computer-readable medium that enables a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable medium produce a product comprising an instruction apparatus that implements the functions / operations specified in the boxes of a flowchart and / or block diagram.

[0224] Computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, such that the instructions that execute on the computer or other programmable apparatus can provide a process for implementing the functions / operations specified in the boxes of a flowchart and / or block diagram.

[0225] Furthermore, although the operations of the methods of this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be executed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0226] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0227] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0228] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0229] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0230] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0231] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

[0232] While the spirit and principles of this disclosure have been described with reference to several specific embodiments, it should be understood that this disclosure is not limited to the disclosed specific embodiments, and the division of aspects does not imply that features in these aspects cannot be combined for benefit; such division is merely for convenience of expression. This disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. The scope of the appended claims is to be interpreted in the broadest sense, thereby encompassing all such modifications and equivalent structures and functions.

Claims

1. A method for generating flexible resource influence maps, characterized in that, include: The original dataset for flexible resource regulation of target industrial users is determined, and core causal nodes are constructed based on the original dataset to obtain a standardized causal analysis data matrix. Causal skeleton identification is performed on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph; Dynamic edge weights are calculated on the temporal unfolded causal graph to obtain dynamic causal influence weights; An influence graph is generated based on the time-series unfolded causal graph and the dynamic causal influence weights, so that the scheduling strategy for the flexible resource regulation can be adjusted based on the influence graph.

2. The method according to claim 1, characterized in that, The original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data.

3. The method according to claim 1, characterized in that, The construction of core causal nodes based on the original dataset yields a standardized causal analysis data matrix, including: The original dataset is time-scale aligned to obtain an aligned multi-source feature matrix; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is subjected to industrial physical constraint verification to obtain the verified core causal node sequence. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

4. The method according to claim 3, characterized in that, The core causal node sequence includes: external disaster stimuli and the degree of tension in power grid supply and demand, economic incentive intensity, industrial physical security margin, compliance and low-carbon pressure, and actual response actions of industrial users.

5. The method according to claim 3, characterized in that, The step of identifying the causal skeleton of the standardized causal analysis data matrix to obtain a time-series unfolded causal graph includes: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

6. The method according to claim 1, characterized in that, The step of dynamically calculating the edge weights of the time-series unfolded causal graph to obtain dynamic causal influence weights includes: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

7. The method according to claim 1, characterized in that, The generation of an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights includes: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; The influence graph is obtained based on the aggregated influence path set and the aggregated dynamic edge weights.

8. A flexible resource influence map generation device, characterized in that, include: The data matrix determination module is configured to determine the original dataset of flexible resource regulation for target industrial users, and construct core causal nodes based on the original dataset to obtain a standardized causal analysis data matrix. The causal graph determination module is configured to perform causal skeleton identification on the standardized causal analysis data matrix to obtain a time-series unfolded causal graph. The influence weight determination module is configured to perform dynamic edge weight calculation on the time-series unfolded causal graph to obtain dynamic causal influence weights. The influence graph determination module is configured to generate an influence graph based on the time-series unfolded causal graph and the dynamic causal influence weights, so as to adjust the scheduling strategy of the flexible resource regulation based on the influence graph.

9. The apparatus according to claim 8, characterized in that, The data matrix determination module is specifically configured as follows: The original dataset for flexible resource regulation of target industrial users is determined, and the original dataset is aligned to a time scale to obtain an aligned multi-source feature matrix; wherein, the original dataset includes: a subset of external environment data, a subset of power grid operation status data, a subset of market incentive and dispatch instruction data, a subset of industrial digital twin and physical status data, a subset of production performance and low-carbon constraint data, and a subset of industrial user apparent response data; The aligned multi-source feature matrix is ​​cleaned to obtain a cleaned multi-source feature matrix. Based on the power grid dispatching business logic and industrial physical mechanism, the core causal node sequence is constructed on the cleaned multi-source feature matrix to obtain the core causal node sequence. The core causal node sequence is verified by industrial physical constraints to obtain the verified core causal node sequence; wherein, the core causal node sequence includes: external disaster-causing stimuli and the degree of tension between power grid supply and demand, economic incentive intensity, industrial physical safety margin, compliance and low-carbon pressure, and actual response actions of industrial users. The verified core causal node sequence is subjected to stationarity processing to obtain a stationarized node sequence; The stationary node sequence is standardized to obtain the standardized causal analysis data matrix.

10. The apparatus according to claim 8, characterized in that, The cause-effect graph determination module is specifically configured as follows: Based on the maximum candidate causal time delay, candidate parent nodes are constructed from the standardized causal analysis data matrix to obtain the complete set of candidate parent nodes; Based on the industrial physical mechanism, a priori mask matrix is ​​constructed for the entire set of candidate parent nodes to obtain a time-delay causal priori mask matrix. The initial set of candidate parent nodes is obtained by filtering the entire set of candidate parent nodes based on the time-delay causal prior mask matrix. The initial set of candidate parent nodes is subjected to conditional independence filtering to obtain the filtered set of candidate parent nodes; A transient conditional independence test is performed on the filtered set of candidate parent nodes to obtain a set of candidate causal edges; Temporal causal edge determination is performed on the candidate causal edge set to obtain the temporal causal edge set; The temporal unfolded causal graph is constructed based on the set of temporal causal edges to obtain the temporal unfolded causal graph.

11. The apparatus according to claim 8, characterized in that, The influence weight determination module is specifically configured as follows: The causal parent node set is extracted from the temporal unfolded causal graph to obtain the causal parent node set of each target node; The structure mask is constructed based on the set of causal parent nodes to obtain the structure mask; Based on the structure mask, a nonlinear structure function model is constructed to obtain an approximate structure function model. Local intervention effects are calculated based on the structure function approximation model to obtain marginal causal effects; The dynamic causal influence weight is obtained by performing a rolling window dynamic edge weight calculation on the marginal causal effect.

12. The apparatus according to claim 8, characterized in that, The influence map determination module is specifically configured as follows: The time-series unfolded causal graph is subjected to aggregated influence path extraction to obtain an aggregated influence path set; The dynamic causal influence weights are aggregated by weighting the aggregated influence path set to obtain aggregated dynamic edge weights; Based on the aggregated influence path set and the aggregated dynamic edge weights, the influence graph is obtained, so that the flexible resource regulation and scheduling strategy can be adjusted based on the influence graph.

13. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 7.

14. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

15. A computer program product, characterized in that, It includes computer program instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

Images