An intelligent service scheduling method and system for park multi-dimensional data integration
By using an intelligent service scheduling method that integrates multi-dimensional data in the park, multi-source data is collected and analyzed to construct dynamic scheduling instructions. Risk assessment is then conducted in a digital twin sandbox, which solves the problem of insufficient adaptability and robustness in park scheduling and improves both economy and security.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG ZHIYI BIG DATA TECHNOLOGY CO LTD
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies cannot keenly perceive the dynamic game relationship between business events and energy indicators in multi-dimensional data integration in the park, resulting in the lack of adaptability and robustness of the generated scheduling instructions under specific working conditions, making it difficult to balance economy and security.
Collect multi-source heterogeneous operation data from the park, construct a multi-dimensional data tensor, generate initial scheduling instructions through low-rank completion operation and semantic-aware adaptive weighted decomposition model, and perform future time period state trajectory extrapolation and deviation risk assessment in a digital twin sandbox. Finally, execute the instructions through an edge control terminal and perform adaptive iterative optimization.
It enables adaptive adjustment of scheduling strategies, improves the economy and security of park scheduling, shifts the safety paradigm from post-event remediation to pre-event prevention, and overcomes the scheduling lag and lack of flexibility of traditional methods under complex working conditions.
Smart Images
Figure CN122390301A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data science technology, and in particular to an intelligent service scheduling method and system for multi-dimensional data integration in industrial parks. Background Technology
[0002] With the deepening of the "dual carbon" strategy and the acceleration of the digital transformation of industrial parks, intelligent service scheduling technology for parks has evolved from traditional single energy monitoring to comprehensive management of multi-energy complementarity and source-grid-load-storage coordination. Currently, academia and industry widely adopt data acquisition architecture based on the Internet of Things, combined with big data analysis, cloud computing and digital twin technology, in an attempt to build a real-time mapping between physical parks and virtual spaces to achieve refined control of multi-dimensional energy flows such as electricity, heat, cooling and gas.
[0003] When dealing with the complex and dynamic environment of a park, the sensor network of the park often exhibits highly sparse and unstructured data due to communication interference, equipment failure, or asynchronous sampling. Traditional methods often use static multi-objective optimization models with fixed weights, which cannot keenly perceive the dynamic game relationship between business events and energy indicators. As a result, the generated scheduling instructions lack adaptability and robustness under specific working conditions, making it difficult to balance economy and security. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, this invention provides an intelligent service scheduling method for multi-dimensional data integration in industrial parks to address the problem of the inability to keenly perceive the dynamic game relationship between business events and energy indicators, which leads to the lack of adaptability and robustness of the generated scheduling instructions under specific operating conditions, making it difficult to balance economic efficiency and security.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an intelligent service scheduling method for multi-dimensional data integration in a park, which includes collecting multi-source heterogeneous operation data of the park and mapping it into an initial sparse tensor. A low-rank completion operation is performed on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct the generated state matrix. Analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve it to generate the initial scheduling instructions; The initial scheduling instructions are input into the digital twin sandbox to extrapolate the state trajectory for future time periods in order to calculate the deviation risk value. Based on the deviation risk value, the initial scheduling instructions are corrected to generate a safety verification instruction. Security verification commands are sent to the park edge control terminal to drive physical devices to execute them, and adaptive iterative optimization is performed using the actual operation feedback data generated during execution.
[0007] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the initial sparse tensor includes: Collect multi-source operational data generated by the park's power system, environmental monitoring system, and equipment control system, and perform unified time alignment and standardization processing on the data; Based on the preset spatial mapping rules, the processed data is constructed into a multidimensional data tensor structure, and the positions of valid observation data and missing data in the data tensor are identified, thereby generating an initial sparse tensor.
[0008] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the generated state matrix includes: A dynamic semantic evolution monitoring mechanism is constructed to perform online clustering analysis on the historical spatiotemporal distribution patterns of each spatial node in the initial sparse tensor, calculate the Euclidean distance between the current data distribution feature vector and the historical semantic cluster center, and determine the semantic drift metric. An adaptive semantic recalibration operation is performed. When the semantic drift metric exceeds a preset dynamic threshold, it is determined that a semantic evolution event has occurred in the corresponding spatial node. The original static label of the spatial node is stripped and marked as an unknown dynamic region. At the same time, the instantaneous semantic embedding vector of the spatial node is reconstructed using unsupervised embedding learning method based on real-time observation data within the sliding time window. When the semantic drift metric does not exceed the dynamic threshold, the center of the historical semantic cluster of the spatial node is used as the steady-state semantic embedding vector. A semantically aware adaptive weighted low-rank decomposition model is constructed, which maps the instantaneous semantic embedding vector or steady-state semantic embedding vector to a mode-specific semantic constraint matrix, and weights the latent factor correlation strength in the low-rank decomposition model according to the semantic constraint matrix. A regularized optimization objective function is constructed, and a dynamic coupling coefficient based on the semantic constraint matrix is introduced into the nuclear norm constraint term of the low-rank component tensor expression to be solved. The coupling weight and temporal mode constraint weight are adjusted for the spatial nodes marked as unknown dynamic regions. Based on the low-rank constraint, a noise suppression mechanism is introduced. By constructing a sparse noise penalty term based on the residual standard deviation, the abnormal observation data is modeled to form a joint optimization objective function. The alternating direction multiplier method solver is invoked to iteratively solve the optimization objective function under the constraint of the observation mask tensor. The information aggregation range of the spatial nodes is adjusted according to the dynamic coupling coefficient until the latent factor matrix converges, and the low-rank component tensor is obtained. Perform a matrix expansion operation on the low-rank component tensor along the time modes to generate a state matrix.
[0009] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the generation of the initial scheduling instruction includes: Extract business event information generated by the park's business system and construct a business event feature vector; and perform statistical analysis on the operational data collected by the energy system to obtain an energy load index vector. Perform an adaptive similarity calculation step based on semantic drift awareness to construct a spatiotemporally adaptive semantic association matrix; The semantic embedding vectors of each spatial node are analyzed and separated into steady-state semantic sub-vectors and instantaneous semantic sub-vectors; Calculate the steady-state fundamental similarity and transient mutation similarity between any two spatial nodes, respectively; Obtain the semantic drift intensity coefficient and map it to a dynamic adjustment factor. Then, perform nonlinear weighted fusion of the steady-state basic similarity and transient mutation similarity based on the dynamic adjustment factor to generate the target similarity value. Based on the target similarity values of spatial nodes, a spatiotemporally adaptive semantic association matrix is generated. Based on the spatiotemporal adaptive semantic association matrix, spectral clustering is performed to divide the spatial nodes into several dynamic semantic clusters and determine the cluster center semantic vector of each dynamic semantic cluster. Based on the distribution characteristics of the dynamic semantic clusters and the corresponding cluster center semantic vectors, combined with the energy load index vectors, an initial scheduling instruction is generated.
[0010] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the step of inputting the initial scheduling instruction into a digital twin environment for deduction and generating a security verification instruction includes: In a digital twin simulation environment, the initial scheduling instructions are used to predict future operating states and assess equipment operating risks. The initial scheduling instructions are adjusted based on the risk assessment results, and a security verification instruction is generated.
[0011] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the adaptive iterative optimization includes: The security verification command is sent to the park edge control terminal driver device for execution. Collect actual operational feedback data after the device is executed, and calculate the deviation characteristics between the data and the prediction results of the digital twin sandbox; Based on the aforementioned deviation characteristics, the boundary constraints or objective function weights in subsequent scheduling tasks are dynamically adjusted to achieve continuous evolution of the scheduling strategy.
[0012] As a preferred embodiment of the intelligent service scheduling method for multi-dimensional data integration in a park as described in this invention, the calculation and dynamic correction of deviation characteristics includes: Construct a multidimensional deviation feature vector that reflects the difference between the actual operating state and the predicted state; Attribution decoupling analysis is performed on the multidimensional deviation feature vector to identify the dominant anomaly patterns that cause scheduling deviations; Based on the type of dominant abnormal mode, the corresponding target adjustment strategy is matched from the preset parameter mapping relationship, and the key coupling coefficient in the multi-objective optimization function or the limiting threshold in the boundary constraint condition is dynamically adjusted accordingly.
[0013] Secondly, the present invention provides an intelligent service scheduling system for multi-dimensional data integration in a park, including a tensor mapping module, a state reconstruction module, an optimization solution module, a deduction and correction module, and an execution iteration module; The tensor mapping module is used to collect multi-source heterogeneous operation data of the park and map it into an initial sparse tensor; The state reconstruction module is used to perform low-rank completion operation on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct and generate a state matrix. The optimization solution module is used to analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve to generate initial scheduling instructions. The simulation correction module is used to input the initial scheduling command into the digital twin sandbox to simulate the state trajectory of future time periods in order to calculate the deviation risk value, and to perform correction processing on the initial scheduling command based on the deviation risk value to generate a safety verification command. The execution iteration module is used to send security verification commands to the park edge control terminal to drive physical devices to execute, and to perform adaptive iterative optimization using the actual operation feedback data generated during execution.
[0014] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the intelligent service scheduling method for multi-dimensional data integration in a park as described in the first aspect of the present invention.
[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent service scheduling method for multi-dimensional data integration in a park as described in the first aspect of the present invention.
[0016] The beneficial effects of this invention are as follows: By performing low-rank completion operation on the initial sparse tensor, the hidden spatiotemporal correlation patterns are accurately mined and a highly reliable state matrix is reconstructed in the case of missing data. This eliminates the decision blind spot caused by data noise from the source. Furthermore, by analyzing business events and load characteristics, a multi-objective optimization function with dynamic coupling weights is constructed. This enables the scheduling strategy to adaptively adjust priorities according to the urgency of the working conditions, overcoming the shortcomings of traditional static weights that cannot balance multiple conflicting objectives. Initial instructions with both economic efficiency and business assurance capabilities are generated. Subsequently, a digital twin sandbox is used to extrapolate future trajectories and quantify deviation risks. Potential equipment overload or instability hazards are exposed in advance in the virtual space and the instructions are corrected in reverse. This realizes a safety paradigm shift from "post-event remediation" to "pre-event prevention". Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 A flowchart for an intelligent service scheduling method for multi-dimensional data integration in a park; Figure 2 This is a schematic diagram of an intelligent service scheduling system for multi-dimensional data integration in the park. Detailed Implementation
[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0022] Reference Figures 1-2As one embodiment of the present invention, this embodiment provides an intelligent service scheduling method for multi-dimensional data integration in a park, comprising the following steps: S1. Collect multi-source heterogeneous operation data of the park and map it into an initial sparse tensor.
[0023] Furthermore, it collects multi-source operational data generated by the park's power system, environmental monitoring system, and equipment control system, and performs unified time alignment and standardization processing on the data; Based on the preset spatial mapping rules, the processed data is constructed into a multidimensional data tensor structure, and the positions of valid observation data and missing data in the data tensor are identified, thereby generating an initial sparse tensor.
[0024] It should be noted that it is mainly used to complete the unified access, time alignment and tensor organization of multi-source heterogeneous operation data in the park, so as to provide a standardized data foundation for subsequent state recovery, semantic analysis and scheduling optimization.
[0025] In the actual operation of the park, the power system, environmental monitoring system and equipment control system have different data sources, sampling frequencies and data formats. If a unified time reference alignment and standardization process is not performed first, it will be difficult for the subsequent model to identify the correlation between multiple data sources in the same spatiotemporal coordinate system.
[0026] The processed data is constructed into a multidimensional data tensor structure, and the locations of valid observation data and missing data are identified.
[0027] Data gaps in park scenarios are not entirely random; they may originate from communication interruptions or the absence of sensors in certain areas. Therefore, it is necessary to explicitly retain the structural information of "observed" and "unobserved" data during the data organization phase so that subsequent low-rank completion and semantic evolution analysis processes can be constrained within the observation boundaries.
[0028] S2. Perform low-rank completion operation on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct the generated state matrix.
[0029] Furthermore, a dynamic semantic evolution monitoring mechanism is constructed to perform online clustering analysis on the historical spatiotemporal distribution patterns of each spatial node in the initial sparse tensor, calculate the Euclidean distance between the current data distribution feature vector and the historical semantic cluster center, and determine the semantic drift metric. An adaptive semantic recalibration operation is performed. When the semantic drift metric exceeds the preset dynamic threshold, it is determined that a semantic evolution event has occurred in the corresponding spatial node. The original static label of the spatial node is stripped and marked as an unknown dynamic region. At the same time, the instantaneous semantic embedding vector of the spatial node is reconstructed by using the real-time observation data within the sliding time window through an unsupervised embedding learning method. When the semantic drift metric does not exceed the dynamic threshold, the center of the historical semantic cluster of the spatial node is used as the steady-state semantic embedding vector. A semantically aware adaptive weighted low-rank decomposition model is constructed, which maps instantaneous semantic embedding vectors or steady-state semantic embedding vectors to modality-specific semantic constraint matrices, and weights the latent factor association strengths in the low-rank decomposition model according to the semantic constraint matrices. A regularized optimization objective function is constructed. Dynamic coupling coefficients based on semantic constraint matrices are introduced into the nuclear norm constraint terms of the low-rank component tensor expression to be solved. The coupling weights and temporal mode constraint weights are adjusted for the spatial nodes marked as unknown dynamic regions. A noise suppression mechanism is introduced based on low-rank constraints. Anomaly observation data is modeled by constructing a sparse noise penalty term based on residual standard deviation, forming a joint optimization objective function. The alternating direction multiplier method solver is invoked to iteratively solve the objective function under the constraint of the observation mask tensor. The information aggregation range of the spatial nodes is adjusted according to the dynamic coupling coefficient until the latent factor matrix converges, and the low-rank component tensor is obtained. Perform a matrix expansion operation on the low-rank component tensor along the time modes to generate the state matrix.
[0030] It should be noted that traditional park data recovery methods usually assume that the business attributes of spatial nodes are stable and unchanged, such as considering meeting rooms as high-traffic, high-energy-consumption areas in the long term.
[0031] However, in actual industrial parks, the functions of the same physical space may change over time, depending on tasks and temporary events. If static tags are still used for data completion, outdated semantics will be imposed on the current data, resulting in a recovery result that is numerically continuous but distorted in its business meaning.
[0032] Therefore, the core of introducing a dynamic semantic evolution monitoring mechanism before the low-rank completion process is to first quantify "whether the meaning of the data has changed" and then decide "how the data can complement each other to complete the data".
[0033] By comparing the current distribution characteristics with the historical semantic cluster centers, we can identify whether spatial nodes have undergone "semantic drift".
[0034] Semantic drift metric It can be represented as: ; in, Indicates that the i-th spatial node is in The data distribution feature vector at time point Indicates the historical semantic cluster to which the node belongs. The center vector.
[0035] The indicator accurately measures the degree of deviation of the current operating state from the historical normal pattern, and serves as the basis for subsequent determination of whether semantic embedding needs to be reconstructed.
[0036] An adaptive semantic recalibration strategy was designed based on the metric values: when When the threshold is exceeded, semantic evolution is determined, the old label is immediately removed and the instantaneous embedding vector is reconstructed using real-time window data; otherwise, the steady-state center is used to avoid error propagation caused by label failure.
[0037] In the low-rank decomposition stage, an "adaptive information aggregation mechanism oriented towards semantic evolution" was constructed, so that the intensity of information sharing between different spatial nodes is no longer fixed, but dynamically adjusted according to semantic relations.
[0038] Dynamic coupling coefficient between nodes It can be represented as: ; in, The aforementioned semantic drift metric, Representing spatial nodes and Basic semantic similarity, This is the sensitivity adjustment parameter.
[0039] When a node undergoes a drastic semantic shift (i.e.) When it is relatively large, The constraints decay exponentially, thereby automatically reducing the node's dependence on historical semantic constraints and preventing old patterns from spreading to new scenarios; conversely, they strengthen constraints to maintain stability.
[0040] This dynamic adjustment mechanism is embedded in the overall goal of low-rank completion.
[0041] The joint optimization objective function corresponding to the low-rank completion process can be expressed as: ; in, Let be the low-rank component tensor to be determined. From The extracted first The latent factor vectors of each node are obtained through dynamic coupling coefficients. Adaptively adjust the smooth constraint strength between nodes. To observe the masking operator, For sparse noise penalty term, , , These are the weight parameters.
[0042] In summary, by embedding a dynamic mechanism in the regularization term, the output state matrix not only retains the spatiotemporal low-rank regularity of the park's operation, but also accurately adapts to the dynamic changes of the current business scenario, fundamentally solving the technical problem of distortion in the traditional method under sudden operating conditions.
[0043] S3. Analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve to generate the initial scheduling instructions.
[0044] Furthermore, business event information generated by the park's business system is extracted and a business event feature vector is constructed, and the energy load index vector is obtained by statistical analysis of the operational data collected by the energy system. Perform an adaptive similarity calculation step based on semantic drift awareness to construct a spatiotemporally adaptive semantic association matrix; The semantic embedding vectors of each spatial node are analyzed and separated into steady-state semantic sub-vectors and instantaneous semantic sub-vectors; Calculate the steady-state fundamental similarity and transient mutation similarity between any two spatial nodes, respectively; Obtain the semantic drift intensity coefficient and map it to a dynamic adjustment factor. Then, based on the dynamic adjustment factor, perform nonlinear weighted fusion of steady-state basic similarity and transient mutation similarity to generate the target similarity value. Based on the target similarity values of spatial nodes, a spatiotemporally adaptive semantic association matrix is generated. Based on the spatiotemporal adaptive semantic association matrix, spectral clustering is performed to divide spatial nodes into several dynamic semantic clusters and determine the cluster center semantic vector of each dynamic semantic cluster. Based on the distribution characteristics of dynamic semantic clusters and the corresponding cluster center semantic vectors, combined with the energy load index vector, initial scheduling instructions are generated.
[0045] It should be noted that the core of further transforming the state matrix into executable scheduling decisions is not simply to perform conventional optimizations based on business events and energy consumption indicators, but to pass the semantic evolution results to the scheduling layer, enabling the scheduling model to have the ability to adaptively organize and dynamically cluster based on semantic drift.
[0046] If semantic changes have been identified, but optimization is still performed using fixed partitioning, then innovation cannot be translated into control benefits.
[0047] To this end, an "adaptive similarity calculation step based on semantic drift awareness" was set up.
[0048] First, the semantic embedding is split into steady-state sub-vectors and instantaneous sub-vectors, and the basic similarity and mutation similarity are calculated respectively, aiming to generate target similarity values that can reflect the dynamic association strength between nodes in real time.
[0049] Target similarity value It can be represented as: ; in, Indicates the steady-state fundamental similarity, reflecting the consistency of long-term functions. It represents the similarity of transient changes and captures the covariance of short-term states. It is a dynamic adjustment factor.
[0050] By utilizing semantic drift signals, we can adaptively determine whether to place more faith in long-term patterns or focus more on current changes.
[0051] Dynamic adjustment factor It can be represented as: ; in, This is the semantic drift intensity coefficient. The sensitivity gain coefficient, This is the offset.
[0052] When the park is operating smoothly ( Hour, →0, target similarity is mainly dominated by the steady-state term, maintaining topological stability; when sudden semantic drift occurs ( When (large), →1, the target similarity quickly switches to be dominated by transient terms, thereby instantly reconstructing the node connection relationship in the association matrix and capturing sudden collaborative patterns.
[0053] Based on this dynamic similarity matrix Spectral clustering is performed, and the resulting "dynamic semantic clusters" replace the traditional fixed functional partitions.
[0054] The multi-objective scheduling function is optimized for these dynamic clusters to ensure that the generated initial scheduling instructions have stronger scenario adaptability. In one embodiment, the multi-objective scheduling function can be expressed as: ; in, For scheduling decision variables, The number of dynamic semantic clusters, For the first Energy cost per cluster; The load response of the cluster; The ideal load pattern represented by the cluster center semantic vector; It is a dynamic weight based on semantic association, and its size depends on the degree of semantic consistency of nodes within the cluster.
[0055] By concretizing the abstract "semantic drift signal" into dynamic adjustment factors and dynamic coupling weights in the formula, an operable dynamic scheduling unit is constructed.
[0056] The seamless closed loop from "perceiving change" to "responding to change" ensures that changes in business scenarios perceived by the upper layer can directly drive the real-time reorganization of control strategies at the lower layer, effectively solving the technical problems of lagging scheduling strategies and insufficient flexibility in the face of complex and ever-changing working conditions in the park.
[0057] S4. Input the initial scheduling command into the digital twin sandbox to extrapolate the state trajectory for future time periods to calculate the deviation risk value, and perform correction processing on the initial scheduling command based on the deviation risk value to generate a safety verification command.
[0058] Furthermore, the initial scheduling instructions are used to predict future operating states and assess equipment operating risks within a digital twin simulation environment; The initial scheduling instructions are adjusted based on the risk assessment results, and a security verification instruction is generated.
[0059] It should be noted that before the dispatch instructions are actually issued, a digital twin sandbox is used to rehearse the future operating status and verify the risks, thereby forming a predictive safety screening mechanism before the execution layer.
[0060] The scheduling problem in the park involves not only energy consumption optimization, but also equipment safety, operational stability and the satisfaction of boundary constraints. Therefore, the initial scheduling instructions obtained solely from the optimization function are not necessarily suitable for direct execution. It is necessary to set up a virtual verification step before physical execution.
[0061] The initial scheduling instructions are input into the digital twin simulation environment and the operational risks of the equipment are assessed. The optimization model usually gives the "optimum" from the perspective of the objective function, while the digital twin sandbox is more suitable for verifying "whether it can operate safely" from the perspective of state evolution.
[0062] The former focuses on minimizing costs, while the latter focuses on whether the trajectory crosses the boundary and whether the state becomes unstable. The two are complementary in function.
[0063] Therefore, the initial scheduling instructions are input again, so that the scheduling results first undergo a simulation of the future state trajectory, and then the instructions are corrected according to the risk assessment results to form a safety verification instruction.
[0064] This enables the system to have a closed-loop structure of "optimize first, then verify, and then execute".
[0065] By using a digital twin sandbox as an intermediate buffer layer connecting the decision-making layer and the execution layer, the semantically aware scheduling results are verifiable and correctable before entering the real equipment, thereby improving the overall engineering reliability of the solution.
[0066] S5 sends security verification commands to the park edge control terminal to drive physical devices to execute them, and uses the actual operation feedback data generated by the execution to perform adaptive iterative optimization.
[0067] Furthermore, the security verification command is sent to the park edge control terminal driver device for execution; Collect actual operational feedback data after the equipment is executed, and calculate the deviation characteristics between the data and the prediction results of the digital twin sandbox; Based on the deviation characteristics, the boundary constraints or objective function weights in subsequent scheduling tasks are dynamically adjusted to achieve continuous evolution of the scheduling strategy.
[0068] Construct a multidimensional deviation feature vector that reflects the difference between the actual operating state and the predicted state; Attribution decoupling analysis is performed on the multidimensional deviation feature vector to identify the dominant anomaly patterns that cause scheduling deviations; Based on the type of dominant anomaly mode, the corresponding target adjustment strategy is matched from the preset parameter mapping relationship, and the key coupling coefficients or limit thresholds in the multi-objective optimization function or boundary constraints are dynamically adjusted accordingly.
[0069] It should be noted that by establishing a feedback loop from the physical execution end back to the digital decision-making end, this invention can not only perform one-time scheduling based on historical data, but also continuously revise subsequent scheduling strategies based on actual execution results.
[0070] The park's operating environment is subject to uncertainties such as aging equipment, changes in personnel activities, the insertion of temporary tasks, and external disturbances. Even after digital twin verification, the actual execution results may still deviate from the predicted results. Therefore, it is necessary to continuously correct the model's understanding through a feedback update mechanism.
[0071] Furthermore, instead of simply comparing whether the "execution result is correct", a multi-dimensional deviation feature vector reflecting the difference between the actual operating state and the predicted state is constructed, and attribution decoupling analysis is performed to identify the dominant abnormal patterns that cause scheduling deviations.
[0072] Deviation itself is only a symptom. Only by further identifying whether the source is environmental disturbance, equipment response lag, local semantic mismatch, or unreasonable constraint settings can subsequent updates be targeted.
[0073] Therefore, by mapping multidimensional deviation characteristics to a preset parameter adjustment strategy and adjusting the key coupling coefficients or boundary constraint thresholds in the multi-objective optimization function accordingly, the system can perform differentiated updates based on the source of anomalies.
[0074] This embodiment also provides an intelligent service scheduling system for multi-dimensional data integration in a park, including: a tensor mapping module, a state reconstruction module, an optimization solution module, a deduction and correction module, and an execution iteration module; The tensor mapping module is used to collect multi-source heterogeneous operational data from the park and map it into an initial sparse tensor. The state reconstruction module is used to perform low-rank completion operations on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct the generated state matrix. The optimization and solution module is used to analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve it to generate the initial scheduling instructions. The simulation and correction module is used to input the initial scheduling instructions into the digital twin sandbox to simulate the state trajectory of future time periods in order to calculate the deviation risk value, and to perform correction processing on the initial scheduling instructions based on the deviation risk value to generate a safety verification instruction. The execution iteration module is used to send security verification commands to the park edge control terminal to drive physical devices to execute them, and to perform adaptive iterative optimization using the actual operation feedback data generated during execution.
[0075] This embodiment also provides a computer device applicable to the intelligent service scheduling method for multi-dimensional data integration in a park, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent service scheduling method for multi-dimensional data integration in a park as proposed in the above embodiment.
[0076] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0077] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the intelligent service scheduling method for multi-dimensional data integration in a park, as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0078] In summary, this invention achieves the following: by performing low-rank completion operations on the initial sparse tensor, it accurately mines the hidden spatiotemporal correlation patterns and reconstructs a highly reliable state matrix in data-deficient scenarios, eliminating decision-making blind spots caused by data noise at the source. Furthermore, it analyzes business events and load characteristics to construct a multi-objective optimization function with dynamically coupled weights, enabling the scheduling strategy to adaptively adjust priorities according to the urgency of the work conditions. This overcomes the shortcomings of traditional static weights in balancing multiple conflicting objectives, generating initial instructions that combine economy and business assurance capabilities. Subsequently, it uses a digital twin sandbox to extrapolate future trajectories and quantify deviation risks, exposing potential equipment overload or instability hazards in advance in the virtual space and correcting instructions in reverse, thus realizing a safety paradigm shift from "post-event remediation" to "pre-event prevention."
[0079] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A smart service scheduling method for multi-dimensional data integration in a park, characterized in that: include: Collect multi-source heterogeneous operational data from the park and map it into an initial sparse tensor; A low-rank completion operation is performed on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct the generated state matrix. Analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve it to generate the initial scheduling instructions; The initial scheduling instructions are input into the digital twin sandbox to extrapolate the state trajectory for future time periods in order to calculate the deviation risk value. Based on the deviation risk value, the initial scheduling instructions are corrected to generate a safety verification instruction. Security verification commands are sent to the park edge control terminal to drive physical devices to execute them, and adaptive iterative optimization is performed using the actual operation feedback data generated during execution.
2. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 1, characterized in that: The initial sparse tensor includes: Collect multi-source operational data generated by the park's power system, environmental monitoring system, and equipment control system, and perform unified time alignment and standardization processing on the data; Based on the preset spatial mapping rules, the processed data is constructed into a multidimensional data tensor structure, and the positions of valid observation data and missing data in the data tensor are identified, thereby generating an initial sparse tensor.
3. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 2, characterized in that: The generated state matrix includes: A dynamic semantic evolution monitoring mechanism is constructed to perform online clustering analysis on the historical spatiotemporal distribution patterns of each spatial node in the initial sparse tensor, calculate the Euclidean distance between the current data distribution feature vector and the historical semantic cluster center, and determine the semantic drift metric. An adaptive semantic recalibration operation is performed. When the semantic drift metric exceeds a preset dynamic threshold, it is determined that a semantic evolution event has occurred in the corresponding spatial node. The original static label of the spatial node is stripped and marked as an unknown dynamic region. At the same time, the instantaneous semantic embedding vector of the spatial node is reconstructed using unsupervised embedding learning method based on real-time observation data within the sliding time window. When the semantic drift metric does not exceed the dynamic threshold, the center of the historical semantic cluster of the spatial node is used as the steady-state semantic embedding vector. A semantically aware adaptive weighted low-rank decomposition model is constructed, which maps the instantaneous semantic embedding vector or steady-state semantic embedding vector to a mode-specific semantic constraint matrix, and weights the latent factor correlation strength in the low-rank decomposition model according to the semantic constraint matrix. A regularized optimization objective function is constructed, and a dynamic coupling coefficient based on the semantic constraint matrix is introduced into the nuclear norm constraint term of the low-rank component tensor expression to be solved. The coupling weight and temporal mode constraint weight are adjusted for the spatial nodes marked as unknown dynamic regions. Based on the low-rank constraint, a noise suppression mechanism is introduced. By constructing a sparse noise penalty term based on the residual standard deviation, the abnormal observation data is modeled to form a joint optimization objective function. The alternating direction multiplier method solver is invoked to iteratively solve the optimization objective function under the constraint of the observation mask tensor. The information aggregation range of the spatial nodes is adjusted according to the dynamic coupling coefficient until the latent factor matrix converges, and the low-rank component tensor is obtained. Perform a matrix expansion operation on the low-rank component tensor along the time modes to generate a state matrix.
4. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 3, characterized in that: The generation of the initial scheduling instruction includes: Extract business event information generated by the park's business system and construct a business event feature vector; and perform statistical analysis on the operational data collected by the energy system to obtain an energy load index vector. Perform an adaptive similarity calculation step based on semantic drift awareness to construct a spatiotemporally adaptive semantic association matrix; The semantic embedding vectors of each spatial node are analyzed and separated into steady-state semantic sub-vectors and instantaneous semantic sub-vectors; Calculate the steady-state fundamental similarity and transient mutation similarity between any two spatial nodes, respectively; Obtain the semantic drift intensity coefficient and map it to a dynamic adjustment factor. Then, perform nonlinear weighted fusion of the steady-state basic similarity and transient mutation similarity based on the dynamic adjustment factor to generate the target similarity value. Based on the target similarity values of spatial nodes, a spatiotemporally adaptive semantic association matrix is generated. Based on the spatiotemporal adaptive semantic association matrix, spectral clustering is performed to divide the spatial nodes into several dynamic semantic clusters and determine the cluster center semantic vector of each dynamic semantic cluster. Based on the distribution characteristics of the dynamic semantic clusters and the corresponding cluster center semantic vectors, combined with the energy load index vectors, an initial scheduling instruction is generated.
5. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 4, characterized in that: The step of inputting the initial scheduling instructions into the digital twin environment for simulation and generating security verification instructions includes: In a digital twin simulation environment, the initial scheduling instructions are used to predict future operating states and assess equipment operating risks. The initial scheduling instructions are adjusted based on the risk assessment results to generate security verification instructions.
6. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 5, characterized in that: The adaptive iterative optimization includes: The security verification command is sent to the park edge control terminal driver device for execution. Collect actual operational feedback data after the device is executed, and calculate the deviation characteristics between the data and the prediction results of the digital twin sandbox; Based on the aforementioned deviation characteristics, the boundary constraints or objective function weights in subsequent scheduling tasks are dynamically adjusted to achieve continuous evolution of the scheduling strategy.
7. The intelligent service scheduling method for multi-dimensional data integration in a park as described in claim 6, characterized in that: The calculation of deviation characteristics and dynamic correction includes: Construct a multidimensional deviation feature vector that reflects the difference between the actual operating state and the predicted state; Attribution decoupling analysis is performed on the multidimensional deviation feature vector to identify the dominant anomaly patterns that cause scheduling deviations; Based on the type of dominant abnormal mode, the corresponding target adjustment strategy is matched from the preset parameter mapping relationship, and the key coupling coefficient in the multi-objective optimization function or the limiting threshold in the boundary constraint condition is dynamically adjusted accordingly.
8. An intelligent service scheduling system for multi-dimensional data integration in a park, based on the intelligent service scheduling method for multi-dimensional data integration in a park as described in any one of claims 1 to 7, characterized in that: It includes a tensor mapping module, a state reconstruction module, an optimization solution module, a deduction and correction module, and an execution iteration module; The tensor mapping module is used to collect multi-source heterogeneous operation data of the park and map it into an initial sparse tensor; The state reconstruction module is used to perform low-rank completion operation on the initial sparse tensor to uncover hidden spatiotemporal correlation patterns and reconstruct and generate a state matrix. The optimization solution module is used to analyze the business event characteristics and energy load indicators in the state matrix, construct a multi-objective optimization function with dynamic coupling weight coefficients, and solve to generate initial scheduling instructions. The simulation correction module is used to input the initial scheduling command into the digital twin sandbox to simulate the state trajectory of future time periods in order to calculate the deviation risk value, and to perform correction processing on the initial scheduling command based on the deviation risk value to generate a safety verification command. The execution iteration module is used to send security verification commands to the park edge control terminal to drive physical devices to execute, and to perform adaptive iterative optimization using the actual operation feedback data generated during execution.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent service scheduling method for multi-dimensional data integration in a park as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent service scheduling method for multi-dimensional data integration in a park as described in any one of claims 1 to 7.