A distributed integrated energy cross-scale collaborative control method and system

Abstract

The application discloses a distributed integrated energy cross-scale collaborative control method and system, which is applicable to a park, a micro-grid, a transformer area and a district energy station with high proportion of renewable energy access. The method deploys an edge routing control terminal at an electrical coupling point, a key feeder or a tie line, uniformly collects node physical quantities, constructs a node net active power sequence, and generates a frequency domain risk characteristic identifier through time-frequency analysis. The risk characteristic identifier is uploaded to a cloud-edge collaborative control platform in a lightweight message, a partition weighted risk index is synthesized, a dynamic time resolution aggregation is driven, and then the risk characteristic is embedded into a multi-time scale planning and scheduling model to obtain a planned power trajectory, a fast resource reservation capacity and an allowed deviation band through rolling optimization. The edge terminal switches the control mode according to the local risk and power deviation, routes high-frequency compensation power to fast resources for execution, and returns the results to correct the subsequent optimization boundary, thereby improving the system stability and scheduling economy.

Description

Technical Field

[0001] This specification relates to the field of power system automation and integrated energy operation control technology. More specifically, this application relates to a distributed integrated energy cross-scale collaborative control method and system. Background Technology

[0002] With the large-scale integration of distributed photovoltaic, wind power, energy storage, electric vehicles, combined cooling, heating and power, heat pumps and interruptible loads, the coupling of multiple energy flows such as electricity, heat, cooling and gas in parks, microgrids, distribution areas and regional energy stations has become closer, and the system operation exhibits multi-timescale fluctuation characteristics.

[0003] Existing integrated energy systems typically employ a hierarchical control approach: planning layer – scheduling layer – real-time control layer. The upper layers are often optimized based on aggregated data at the 15-minute, 30-minute, or hourly level, while the lower layers perform local compensation based on local measurements. This approach is prone to causing high-frequency disturbance information from the field to be smoothed out or lost during the uploading and modeling process, making it difficult for the upper-level scheduling to accurately perceive risks at the second or minute level caused by factors such as photovoltaic shading, sudden load changes, and concentrated access of electric vehicles.

[0004] Meanwhile, existing fixed-time-resolution optimization models struggle to balance computational efficiency with high-frequency information retention. Furthermore, the response speeds of various energy storage systems, such as electric energy storage, gas turbines, heat pumps, and thermal and cold storage devices, vary considerably. Modeling with a uniform time scale can easily generate scheduling instructions that are physically difficult to execute.

[0005] Therefore, there is an urgent need for a distributed integrated energy cross-scale collaborative control method that can transform high-frequency disturbance characteristics at the edge into identifiable risk variables at the upper layer and achieve cloud-edge closed-loop correction. Summary of the Invention

[0006] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0007] Firstly, this application proposes a distributed integrated energy cross-scale collaborative control method, including: Multiple edge routing control terminals are deployed at electrical coupling points, key feeders, or tie lines in the distributed integrated energy system, and node physical quantities are collected under unified time synchronization. The collected multi-source heterogeneous data undergoes protocol conversion, point mapping, and object normalization to construct a node state sequence with a unified time scale, and generates a node net active power sequence based on power-side sampled values. Time-frequency analysis was performed on the net active power sequence of the above nodes to generate frequency domain risk feature identifiers characterizing the micro high-frequency disturbances of the nodes; The node identifier, timestamp, the above-mentioned node net active power sequence, the above-mentioned frequency domain risk characteristic identifier, and the device status quantity are organized into a lightweight operation message and uploaded to the cloud-edge collaborative control platform. The aforementioned cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node. Based on the above-mentioned partitioned weighted risk indicators, dynamic time-resolution aggregation is performed on the forecast of new energy output, load and resource status, and a common decision-making time granularity is generated by combining the minimum response time of equipment. Based on the aggregated multi-granularity time series dataset, a multi-time-scale planning and scheduling model is constructed. The above-mentioned frequency domain risk feature identifiers are embedded into the above multi-time-scale planning and scheduling model as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty coefficient. Rolling optimization is performed to obtain the planned power trajectory, the fast resource reservation capacity, and the allowable deviation band. The above-mentioned planned power trajectory, the above-mentioned rapid resource reserved capacity, and the above-mentioned allowable deviation band are sent to the corresponding edge routing control terminal. The aforementioned edge routing control terminal switches between planned tracking mode and filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, and routes high-frequency compensation power to fast resource execution in the aforementioned filtered smoothing mode; The node execution results are sent back to the aforementioned cloud-edge collaborative control platform, which then updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle.

[0008] In one feasible implementation, the above-mentioned acquisition of node physical quantities under unified time synchronization includes: The sampling times of each of the aforementioned edge routing control terminals are aligned using a unified timing unit; Collect three-phase voltage, three-phase current, active power, reactive power and frequency at a sampling frequency greater than or equal to 100Hz; Synchronously collect data on energy storage status of charge, thermal energy storage status, equipment start / stop status, and available capacity of interruptible loads; For nodes that do not have the conditions for three-phase waveform sampling, the synchronous active power data sent by smart meters or protection and control devices with unified time synchronization is used as the equivalent input.

[0009] In one feasible implementation, the above-mentioned generation of the node net active power sequence based on power-side sampled values ​​includes: The instantaneous active power of node i at sampling time k is constructed according to the following formula:

[0010] in, , , These are the sampled values ​​of the three-phase voltage at node i at sampling time k. , , These are the sampled values ​​of the three-phase current at node i at sampling time k; Applying a polarity convention to the above instantaneous active power, we obtain the node net active power Pnet,i(k), where generation injection is positive and consumption absorption is negative; Simultaneously, a sequence of thermal power, cooling power, gas consumption, energy storage state parameters, and adjustable capacity boundary parameters is generated.

[0011] In one feasible implementation, the above-mentioned time-frequency analysis of the net active power sequence of the nodes generates frequency domain risk feature identifiers characterizing the micro-high-frequency disturbances of the nodes, including: The net active power sequence Pnet,i(k) of the above nodes is subjected to anti-aliasing pre-filtering and sliding window processing; Perform a short-time Fourier transform with a preset window length L and a preset step size S to obtain the time-frequency coefficients Xi(m,f); For each time frame m, extract the normalized energy percentage Ei(m) of the risk band from 0.1Hz to 2Hz; The normalized energy percentage Ei(m) above is standardized by using the rolling mean μi and rolling standard deviation σi within a preset historical window to obtain the standardized risk value Zi(m); The frequency domain risk feature identifier of node i is generated through bounded mapping and first-order hysteresis filtering. Among them, the aforementioned frequency domain risk characteristic identifier Used to characterize the degree of high-frequency disturbance risk of node i within the corresponding time frame.

[0012] In one feasible implementation, the aforementioned cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node, including: The nodes are divided into zones according to feeders, transformer areas, microgrids, and park or regional energy stations; Node weights are set based on node rated capacity, network sensitivity coefficient, or criticality weight. ; The partition-weighted risk index for partition r is synthesized using the following formula. :

[0013] in, For partitioning internal node set The frequency domain risk characteristic identifier for node i. Let be the node weight of node i; The aforementioned lightweight operation message includes at least the node identifier, timestamp, node net active power, frequency domain risk characteristic identifier, energy storage charge status, thermal energy storage status, equipment available capacity, and alarm status.

[0014] In one feasible implementation, the above-mentioned dynamic time-resolution aggregation of new energy output forecast, load forecast, and resource status forecast based on the aforementioned partition-weighted risk indicators, and the generation of a common decision-making time granularity by combining the minimum equipment response time, includes: For rapidly changing variables on the power side, the time granularity on the power side is selected according to the relationship between the above-mentioned partitioned weighted risk indicators and the preset risk threshold. When the above partition-weighted risk index is greater than the first risk threshold, the first time granularity is selected. When the above partition-weighted risk index is less than or equal to the first risk threshold and greater than the second risk threshold, the second time granularity is selected. When the above partition-weighted risk index is less than or equal to the second risk threshold, the third time granularity is selected. Wherein, the first time granularity is smaller than the second time granularity, and the second time granularity is smaller than the third time granularity. For slow-changing variables and slow-response equipment on the heat, cold, and air sides, the equipment response time granularity is generated by combining the equipment's preset minimum adjustment cycle. By coordinating the aforementioned power-side time granularity with the aforementioned equipment response time granularity, the aforementioned common decision-making time granularity is generated.

[0015] In one feasible implementation, the above-mentioned multi-timescale planning and scheduling model includes electrical power balance constraints, thermal power balance constraints, cold power balance constraints, and gas volume balance constraints. In scenarios involving coupled resources of electricity, heat, cold, and gas, the above multi-timescale planning and scheduling model simultaneously satisfies the above-mentioned power balance constraints, heat power balance constraints, cold power balance constraints, and gas volume balance constraints. In scenarios involving only power-side resources, the aforementioned multi-timescale planning and scheduling model retains only the power balance constraints and power equipment constraints. The aforementioned constraints on power equipment include at least the energy storage state of charge constraints, energy storage charging and discharging power constraints, grid connection point switching power constraints, equipment ramping constraints, and interruptible load dispatch constraints.

[0016] In one feasible implementation, the aforementioned frequency domain risk characteristic identifier is embedded as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty coefficient into the aforementioned multi-timescale planning and scheduling model, including: Construct an objective function that includes electricity purchase and sale costs, fuel costs, operation and maintenance costs, planning smoothing costs, and risk relaxation penalty costs; Identify the frequency domain risk characteristics of node i at decision time t. As a weight for the planned smoothing term, the changes in planned net active power corresponding to high-risk nodes are subject to stronger constraints. Set risk slack variables And mark the above frequency domain risk characteristics. As the aforementioned risk slack variable Penalty coefficient; The boundary of node plan changes should be corrected according to the following formula:

[0017] in, Let be the planned net active power of node i at decision time k. Let be the planned net active power of node i at the previous decision time. Let i be the basic plan change boundary. To preset the risk tightening coefficient, For risk slack variables.

[0018] In one feasible implementation, the aforementioned edge routing control terminal switches between planned tracking mode and filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, including: Obtain local frequency domain risk feature identifier Real-time node net active power Planned power trajectory and allowable deviation band ; Based on the above real-time node net active power With respect to the above-mentioned planned power trajectory Calculate the power deviation ΔPi(t); When the above local frequency domain risk characteristics are identified The power deviation ΔPi(t) is less than or equal to the preset safety threshold, and the power deviation ΔPi(t) is located within the allowable deviation band. During this period, execute the above-mentioned plan tracking mode; When the above local frequency domain risk characteristics are identified If the power deviation is greater than the preset safety threshold, or if the power deviation ΔPi(t) exceeds the allowable deviation range... When the time comes, switch to the above-mentioned filtering and smoothing mode; Under the aforementioned filtering and smoothing mode, based on the aforementioned local frequency domain risk characteristic identifier... The cutoff frequency of the low-pass filter is dynamically adjusted based on the aforementioned power deviation. Calculate the high-frequency compensation power; The aforementioned high-frequency compensation power is preferentially routed to energy storage converters, electric vehicle charging facilities, and interruptible loads. It continuously transmits back the node's net active power, compensation power, energy storage state of charge change, number of equipment actions, thermal energy storage margin, plan tracking error, and fault or alarm information after execution. When communication is interrupted or the upper-layer plan is temporarily unavailable, it enters degraded operation based on the most recent valid planned power trajectory, the current frequency domain risk characteristic identifier, and local security constraints.

[0019] Secondly, this application proposes a distributed integrated energy cross-scale collaborative control system, comprising: The acquisition unit is used to deploy multiple edge routing control terminals at electrical coupling points, key feeders or tie lines of a distributed integrated energy system, and to acquire node physical quantities under unified time synchronization. The first generation unit is used to perform protocol conversion, point mapping and object normalization on the collected multi-source heterogeneous data, construct a node state sequence with a unified time scale, and generate a node net active power sequence based on the power side sampling value. The second generation unit is used to perform time-frequency analysis on the net active power sequence of the above nodes and generate frequency domain risk feature identifiers that characterize the micro high-frequency disturbances of the nodes. The upload unit is used to organize the node identifier, timestamp, the net active power sequence of the above-mentioned nodes, the frequency domain risk characteristic identifier and the device status quantity into a lightweight operation message and upload it to the cloud-edge collaborative control platform. The synthesis unit is used by the cloud-edge collaborative control platform to synthesize a partition-weighted risk index based on the frequency domain risk characteristics of each node. The third generation unit is used to perform dynamic time resolution aggregation on the new energy output forecast, load forecast and resource status forecast based on the above-mentioned partition weighted risk indicators, and generate common decision time granularity in combination with the minimum response time of the equipment. The optimization unit is used to construct a multi-timescale planning and scheduling model based on the aggregated multi-granularity time series dataset. The frequency domain risk feature identifier is embedded into the multi-timescale planning and scheduling model as the weight of the planning smoothing term, the constraint boundary correction amount and the risk relaxation penalty term coefficient. Rolling optimization is performed to obtain the planned power trajectory, the fast resource reservation capacity and the allowable deviation band. The sending unit is used to send the above-mentioned planned power trajectory, the above-mentioned rapid resource reserved capacity and the above-mentioned allowable deviation band to the corresponding edge routing control terminal. The switching unit is used for the edge routing control terminal to switch between the planned tracking mode and the filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, and to route the high-frequency compensation power to the fast resource execution in the filtered smoothing mode. The control unit is used to transmit the node operation results after execution back to the cloud-edge collaborative control platform, which then updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle based on these results.

[0020] In summary, this invention extracts high-frequency power disturbances into frequency domain risk feature identifiers through edge-side time-frequency analysis. This preserves on-site high-frequency risk information while reducing the communication and computational burden caused by continuous uploading of the original waveform. By using partitioned weighted risk indicators and dynamic temporal resolution aggregation, cloud scheduling can retain finer-grained data during high-risk periods and reduce model size during low-risk periods. By embedding the frequency domain risk feature identifiers into the smoothing terms, constraint boundaries, and relaxation penalty terms of the multi-timescale planning and scheduling model, high-risk nodes obtain smoother planned power trajectories and more sufficient dynamic margins. Through the edge routing control terminal switching between planned tracking mode and filtered smoothing mode, high-frequency compensation power is preferentially allocated to fast resources such as energy storage, electric vehicle charging facilities, and interruptible loads, achieving local isolation of high-frequency disturbances and cloud-based coordination of low-frequency energy. Through execution result feedback and optimized boundary updates, a cloud-edge closed-loop correction is formed, improving system operational stability, scheduling executability, and economy.

[0021] The distributed integrated energy cross-scale coordinated control method proposed in this application, along with other advantages, objectives, and features of this application, will be partly apparent from the following description and partly understood by those skilled in the art through research and practice of this application. Attached Figure Description

[0022] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic diagram of a distributed integrated energy cross-scale collaborative control method provided in this application embodiment; Figure 2 A schematic diagram of a distributed integrated energy cross-scale collaborative control system provided in this application embodiment; Figure 3 A schematic diagram of the edge-side frequency domain risk feature extraction process and mathematical mapping provided for embodiments of this application; Figure 4This is a schematic diagram illustrating the coupling of risk-driven dynamic temporal resolution aggregation and cross-scale rolling optimization in an embodiment of this application. Figure 5 A schematic diagram of the state machine and frequency band routing control of the edge routing control terminal switching between planning tracking mode and filtering smoothing mode provided in the embodiments of this application; Figure 6 A timing diagram illustrating cloud-edge closed-loop communication, command issuance, and feedback correction provided in an embodiment of this application; Figure 7 This is a structural schematic diagram of a distributed integrated energy cross-scale collaborative control system provided in an embodiment of this application. Detailed Implementation

[0023] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will now be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0024] Please see Figure 1 , Figure 1 This application provides a flowchart illustrating a distributed integrated energy cross-scale collaborative control method, which may specifically include: S101. Deploy multiple edge routing control terminals at the electrical coupling points, key feeders or tie lines of the distributed integrated energy system, and collect node physical quantities under unified time synchronization. S102. Perform protocol conversion, point mapping and object normalization on the collected multi-source heterogeneous data, construct a node state sequence with a unified time scale, and generate a node net active power sequence based on the power side sampling value. S103. Perform time-frequency analysis on the net active power sequence of the above nodes to generate frequency domain risk feature identifiers that characterize the micro high-frequency disturbances of the nodes. S104. Organize the node identifier, timestamp, the above-mentioned node net active power sequence, the above-mentioned frequency domain risk characteristic identifier and equipment status quantity into a lightweight operation message and upload it to the cloud-edge collaborative control platform. S105. The cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node. S106. Based on the above-mentioned partitioned weighted risk indicators, perform dynamic time resolution aggregation on new energy output forecast, load forecast and resource status forecast, and generate common decision time granularity by combining the minimum response time of equipment. S107. Based on the aggregated multi-granularity time series dataset, construct a multi-time-scale planning and scheduling model. Embed the above-mentioned frequency domain risk feature identifier as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty term coefficient into the above-mentioned multi-time-scale planning and scheduling model. Perform rolling optimization to obtain the planned power trajectory, fast resource reservation capacity, and allowable deviation band. S108. Send the above-mentioned planned power trajectory, the above-mentioned rapid resource reserved capacity and the above-mentioned allowable deviation band to the corresponding edge routing control terminal. S109. The aforementioned edge routing control terminal switches between the planned tracking mode and the filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, and routes the high-frequency compensation power to the fast resource execution in the aforementioned filtered smoothing mode. S110. The node execution results after execution are sent back to the cloud-edge collaborative control platform, which then updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle.

[0025] For example, this embodiment is applied to scenarios such as park microgrids, regional energy stations, integrated energy systems in distribution areas, or user-side energy units. This distributed integrated energy system may include distributed photovoltaic systems, energy storage converters, electric vehicle charging facilities, combined cooling, heating and power (CCHP) units, heat pumps, thermal and cold storage equipment, interruptible loads, and grid-connected tie lines. To achieve cross-scale collaborative control, multiple edge routing control terminals are deployed at the system's electrical coupling points, key feeders, or tie lines. Each edge routing control terminal can then collect node physical quantities such as three-phase voltage, three-phase current, active power, reactive power, frequency, energy storage state of charge, equipment start / stop status, and adjustable capacity from the nearest node.

[0026] When executing S101, each edge routing control terminal synchronizes its time via BeiDou, GPS, or IEEE 1588PTP time synchronization, ensuring a unified time reference for the sampled data from different nodes. For nodes with high-frequency sampling capabilities, power-side physical quantities can be acquired at a sampling frequency of no less than 100Hz using high-precision acquisition units and analog-to-digital converters. For nodes lacking three-phase waveform sampling capabilities, synchronized active power data uploaded by smart meters or protection and control devices with unified time synchronization can be used as equivalent input. This avoids phase deviations caused by inconsistent sampling times between different nodes, providing a unified time-scale basis for subsequent frequency domain analysis and zonal risk synthesis.

[0027] During S102 execution, the edge routing control terminal performs protocol conversion and semantic parsing on data from devices such as smart meters, protection and control devices, photovoltaic inverters, energy storage converters, charging facility controllers, and cold and heat source controllers. For example, for data uploaded using different communication protocols such as Modbus, CAN, and IEC61850, the original message is first parsed, and then, based on preset point table mapping rules or a unified object model, register addresses, device locations, and original fields are converted into node objects, device objects, and state variables with clear physical meanings. Subsequently, instantaneous active power is calculated based on the sampled values ​​of three-phase voltage and three-phase current, and a node net active power sequence is generated according to the rule of taking positive values ​​for generation injection and negative values ​​for consumption absorption. At the same time, node state sequences such as thermal power, cold power, gas consumption, energy storage state variables, and adjustable capacity boundaries are formed.

[0028] When executing S103, the edge routing control terminal does not directly upload all high-frequency raw waveforms. Instead, it performs anti-aliasing pre-filtering, sliding window slicing, and time-frequency analysis on the node's net active power sequence locally. For example, short-time Fourier transform can be used to extract the energy proportion within a preset risk frequency band, and the energy proportion can be standardized using the mean and standard deviation within the historical window. Then, a frequency domain risk characteristic identifier is obtained through bounded mapping and first-order lag filtering. This frequency domain risk characteristic identifier can take values ​​between 0 and 1. The larger the value, the more severe the high-frequency power fluctuation of the node within the current time window, and the more significant the impact on frequency support, voltage stability, rapid energy storage response, and grid connection point power smoothing.

[0029] When executing S104, the edge routing control terminal organizes information such as node identifier, timestamp, node net active power, frequency domain risk characteristic identifier, energy storage state of charge, thermal energy storage state, available device capacity, and alarm status into a lightweight operation message, and uploads it to the cloud-edge collaborative control platform via an encrypted communication link. Because the message retains the frequency domain risk characteristic identifier reflecting the intensity of high-frequency disturbances, instead of continuously uploading complete high-frequency waveforms, it can reduce communication bandwidth and cloud storage pressure, while allowing the cloud to retain the key semantic information needed to identify micro-level high-frequency risks.

[0030] When executing S105, the cloud-edge collaborative control platform receives lightweight operation messages uploaded by multiple edge routing control terminals and partitions the nodes according to feeder, distribution area, microgrid, park, or regional energy station. For multiple nodes within the same partition, the cloud-edge collaborative control platform can set node weights based on node rated capacity, network sensitivity coefficient, critical load level, or grid connection support importance, and then weight and synthesize the frequency domain risk characteristics of each node to obtain a partition-weighted risk index. This partition-weighted risk index can reflect the overall high-frequency disturbance risk level of a certain area, avoiding the judgment of system operating status based solely on a single node or low-frequency average power.

[0031] When executing S106, the cloud-edge collaborative control platform dynamically aggregates the forecasts of new energy output, load, and resource status based on the zone-weighted risk indicators, using a time resolution. When a zone has a high risk, it indicates that the zone has significant short-term fluctuations or rapid ramp-up risks. In this case, a finer time granularity is used for fast-changing variables on the power side to preserve the characteristics of power peaks, ramp-up changes, and short-term fluctuations. When the zone risk is low, a coarser time granularity is used to reduce the number of optimization variables and the solution burden. For slow-changing variables such as heat, cold, and gas, as well as slow-response equipment such as heat pumps, boilers, and thermal storage tanks, it is also necessary to generate a common decision-making time granularity based on the minimum response time of the equipment to avoid the optimization model outputting rapid adjustment commands that the equipment cannot physically execute.

[0032] When executing S107, the cloud-edge collaborative control platform constructs a multi-timescale planning and scheduling model based on the aggregated multi-granularity time-series dataset. This model can simultaneously consider factors such as electricity purchase and sale costs, fuel costs, equipment operation and maintenance costs, energy storage lifespan loss, combined cooling, heating and power (CCHP) output, heat pump power, thermal and cold storage plans, flexible load control, and the operating status of gas-electric coupling equipment. During model construction, frequency domain risk characteristics are embedded into the planning smoothing term, constraint boundary correction term, and risk relaxation penalty term. This makes the planned power trajectory corresponding to high-risk nodes smoother and tightens their planned change boundaries accordingly, thereby reserving more dynamic adjustment margins for underlying rapid resources. After rolling optimization, the cloud-edge collaborative control platform outputs the planned power trajectory, rapid resource reserved capacity, and allowable deviation band for future control cycles.

[0033] When executing S108, the cloud-edge collaborative control platform decomposes the aforementioned planned power trajectory, rapid resource reservation capacity, and allowable deviation band to the corresponding edge routing control terminals. For each edge routing control terminal, the received planned power trajectory is used to indicate the low-frequency power target that the node should track in the future; the rapid resource reservation capacity is used to constrain the available range of rapidly responding resources such as local energy storage, electric vehicle charging facilities, and flexible loads; and the allowable deviation band is used to determine whether the current actual operating status is still within an acceptable tracking range.

[0034] When executing S109, the edge routing control terminal switches modes based on the frequency domain risk characteristic identifier and actual power deviation calculated locally in real time. When the local frequency domain risk characteristic identifier is low and the actual power deviation is within the allowable deviation band, the edge routing control terminal executes the planned tracking mode, that is, performs regular tracking control according to the planned power trajectory issued by the cloud-edge collaborative control platform. When the local frequency domain risk characteristic identifier increases, or the actual power deviation exceeds the allowable deviation band, the edge routing control terminal switches to the filtering and smoothing mode, decomposes the power deviation into frequency bands, and prioritizes the routing of high-frequency compensation power to fast resources such as energy storage converters, electric vehicle charging facilities, supercapacitors, flywheel energy storage, or interruptible loads. This allows high-frequency disturbances to be quickly absorbed or compensated locally, while low-frequency energy deviations are reserved for the cloud-edge collaborative control platform to optimize in the next rolling cycle.

[0035] During the execution of S110, the edge routing control terminal transmits the following data back to the cloud-edge collaborative control platform: node net active power, compensation power, energy storage state of charge change, number of equipment actions, thermal energy storage margin, plan tracking error, and fault or alarm information. The cloud-edge collaborative control platform updates the risk baseline statistics based on the transmitted results to correct the standardized benchmark for subsequent frequency domain risk feature identification; updates the equipment availability boundary based on the number of equipment actions, energy storage status, and alarm status; updates the rapid resource reservation capacity requirement based on the actual compensation power and plan tracking error; and adjusts the optimization boundary for the next rolling cycle accordingly. Thus, this embodiment forms a closed-loop control process from on-site high-frequency sensing, edge feature extraction, cloud-based risk aggregation, cross-scale rolling optimization, edge frequency band routing execution, to execution result feedback correction, achieving synergy between microscopic high-frequency disturbance risks and macroscopic multi-timescale scheduling.

[0036] In one feasible implementation, the above-mentioned acquisition of node physical quantities under unified time synchronization includes: The sampling times of each of the aforementioned edge routing control terminals are aligned using a unified timing unit; Three-phase voltage, three-phase current, active power, reactive power, and frequency are collected at a sampling frequency greater than or equal to 100Hz. Synchronously collect data on energy storage status of charge, thermal energy storage status, equipment start / stop status, and available capacity of interruptible loads; For nodes that do not have the conditions for three-phase waveform sampling, the synchronous active power data sent by smart meters or protection and control devices with unified time synchronization is used as the equivalent input.

[0037] In one feasible implementation, the above-mentioned generation of the node net active power sequence based on power-side sampled values ​​includes: The instantaneous active power of node i at sampling time k is constructed according to the following formula:

[0038] in, , , These are the sampled values ​​of the three-phase voltage at node i at sampling time k. , , These are the sampled values ​​of the three-phase current at node i at sampling time k; Applying a polarity convention to the above instantaneous active power, we obtain the node net active power Pnet,i(k), where generation injection is positive and consumption absorption is negative; Simultaneously, a sequence of thermal power, cooling power, gas consumption, energy storage state parameters, and adjustable capacity boundary parameters is generated.

[0039] In one feasible implementation, the above-mentioned time-frequency analysis of the net active power sequence of the nodes generates frequency domain risk feature identifiers characterizing the micro-high-frequency disturbances of the nodes, including: The net active power sequence Pnet,i(k) of the above nodes is subjected to anti-aliasing pre-filtering and sliding window processing; Perform a short-time Fourier transform with a preset window length L and a preset step size S to obtain the time-frequency coefficients Xi(m,f); For each time frame m, extract the normalized energy percentage Ei(m) of the risk band from 0.1Hz to 2Hz; The normalized energy percentage Ei(m) above is standardized by using the rolling mean μi and rolling standard deviation σi within a preset historical window to obtain the standardized risk value Zi(m); The frequency domain risk feature identifier of node i is generated through bounded mapping and first-order hysteresis filtering. Among them, the aforementioned frequency domain risk characteristic identifier Used to characterize the degree of high-frequency disturbance risk of node i within the corresponding time frame.

[0040] In one feasible implementation, the aforementioned cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node, including: The nodes are divided into zones according to feeders, transformer areas, microgrids, and park or regional energy stations; Node weights are set based on node rated capacity, network sensitivity coefficient, or criticality weight. ; The partition-weighted risk index for partition r is synthesized using the following formula. :

[0041] in, For partitioning internal node set The frequency domain risk characteristic identifier for node i. Let be the node weight of node i; The aforementioned lightweight operation message includes at least the node identifier, timestamp, node net active power, frequency domain risk characteristic identifier, energy storage charge status, thermal energy storage status, equipment available capacity, and alarm status.

[0042] In one feasible implementation, the above-mentioned dynamic time-resolution aggregation of new energy output forecast, load forecast, and resource status forecast based on the aforementioned partition-weighted risk indicators, and the generation of a common decision-making time granularity by combining the minimum equipment response time, includes: For rapidly changing variables on the power side, the time granularity on the power side is selected according to the relationship between the above-mentioned partitioned weighted risk indicators and the preset risk threshold. When the above partition-weighted risk index is greater than the first risk threshold, the first time granularity is selected. When the above partition-weighted risk index is less than or equal to the first risk threshold and greater than the second risk threshold, the second time granularity is selected. When the above partition-weighted risk index is less than or equal to the second risk threshold, the third time granularity is selected. Wherein, the first time granularity is smaller than the second time granularity, and the second time granularity is smaller than the third time granularity. For slow-changing variables and slow-response equipment on the heat, cold, and air sides, the equipment response time granularity is generated by combining the equipment's preset minimum adjustment cycle. By coordinating the aforementioned power-side time granularity with the aforementioned equipment response time granularity, the aforementioned common decision-making time granularity is generated.

[0043] In one feasible implementation, the above-mentioned multi-timescale planning and scheduling model includes electrical power balance constraints, thermal power balance constraints, cold power balance constraints, and gas volume balance constraints. In scenarios involving coupled resources of electricity, heat, cold, and gas, the above multi-timescale planning and scheduling model simultaneously satisfies the above-mentioned power balance constraints, heat power balance constraints, cold power balance constraints, and gas volume balance constraints. In scenarios involving only power-side resources, the aforementioned multi-timescale planning and scheduling model retains only the power balance constraints and power equipment constraints. The aforementioned constraints on power equipment include at least the energy storage state of charge constraints, energy storage charging and discharging power constraints, grid connection point switching power constraints, equipment ramping constraints, and interruptible load dispatch constraints.

[0044] In one feasible implementation, the aforementioned frequency domain risk characteristic identifier is embedded as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty coefficient into the aforementioned multi-timescale planning and scheduling model, including: Construct an objective function that includes electricity purchase and sale costs, fuel costs, operation and maintenance costs, planning smoothing costs, and risk relaxation penalty costs; Identify the frequency domain risk characteristics of node i at decision time t. As a weight for the planned smoothing term, the changes in planned net active power corresponding to high-risk nodes are subject to stronger constraints. Set risk slack variables And mark the above frequency domain risk characteristics. As the aforementioned risk slack variable Penalty coefficient; The boundary of node plan changes should be corrected according to the following formula:

[0045] in, Let be the planned net active power of node i at decision time k. Let be the planned net active power of node i at the previous decision time. Let i be the basic plan change boundary. To preset the risk tightening coefficient, For risk slack variables.

[0046] In one feasible implementation, the aforementioned edge routing control terminal switches between planned tracking mode and filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, including: Obtain local frequency domain risk feature identifier Real-time node net active power Planned power trajectory and allowable deviation band ; Based on the above real-time node net active power With respect to the above-mentioned planned power trajectory Calculate the power deviation ΔPi(t); When the above local frequency domain risk characteristics are identified The power deviation ΔPi(t) is less than or equal to the preset safety threshold, and the power deviation ΔPi(t) is located within the allowable deviation band. During this period, execute the above-mentioned plan tracking mode; When the above local frequency domain risk characteristics are identified If the power deviation is greater than the preset safety threshold, or if the power deviation ΔPi(t) exceeds the allowable deviation range... When the time comes, switch to the above-mentioned filtering and smoothing mode; Under the aforementioned filtering and smoothing mode, based on the aforementioned local frequency domain risk characteristic identifier... The cutoff frequency of the low-pass filter is dynamically adjusted based on the aforementioned power deviation. Calculate the high-frequency compensation power; The aforementioned high-frequency compensation power is preferentially routed to energy storage converters, electric vehicle charging facilities, and interruptible loads. It continuously transmits back the node's net active power, compensation power, energy storage state of charge change, number of equipment actions, thermal energy storage margin, plan tracking error, and fault or alarm information after execution. When communication is interrupted or the upper-layer plan is temporarily unavailable, it enters degraded operation based on the most recent valid planned power trajectory, the current frequency domain risk characteristic identifier, and local security constraints.

[0047] In one feasible implementation, such as Figure 2 As shown, this embodiment adopts a cloud-edge-device collaborative architecture. At the field device layer, it connects to execution objects such as photovoltaic inverters, energy storage converters, electric vehicle charging facilities, interruptible load switches, combined cooling, heating and power unit controllers, heat pump controllers, cold and heat storage equipment controllers, and gas-electric coupling equipment. At the edge control layer, multiple edge routing control terminals 100 are set up; and at the cloud side, a cloud-edge collaborative control platform 200 is set up. The edge routing control terminals 100 are connected to field sensors, measurement and control devices, and execution device controllers. The cloud-edge collaborative control platform 200 is connected to multiple edge routing control terminals 100 through encrypted communication links, thereby forming a closed-loop control system encompassing high-frequency sensing at the field, edge feature extraction, cloud-based cross-scale optimization, edge frequency band routing execution, and feedback correction.

[0048] Specifically, such as Figure 2As shown, each edge routing control terminal 100 may include a multi-source synchronous acquisition unit 101, a unified time synchronization unit 102, a multi-protocol access and semantic parsing unit 103, an edge feature extraction unit 104, a local execution and frequency band routing control unit 105, a device control interface and security interlocking unit 106, and an encrypted communication unit 107. The multi-source synchronous acquisition unit 101 is used to acquire three-phase voltage, three-phase current, active power, reactive power, frequency, DC bus voltage, energy storage state of charge, thermal energy storage state, temperature, flow rate, and equipment start / stop status. The unified time synchronization unit 102 is used to synchronize the time of each sampling channel through BeiDou, GPS, network time synchronization, or IEEE 1588PTP precise time synchronization. The multi-protocol access and semantic parsing unit 103 is used to access the raw data from smart meters, protection and control devices, photovoltaic inverters, energy storage converters, charging facility controllers, cold and heat source controllers, and unit controllers, and complete object identification normalization, field standardization, and physical quantity labeling based on a unified object model, point-to-table mapping rules, or semantic model. The edge feature extraction unit 104 performs pre-filtering, sliding windowing, time-frequency analysis, risk band energy calculation, standardization, and filtering on the node net active power sequence, outputting a frequency domain risk feature identifier. The local execution and frequency band routing control unit 105 receives the planned power trajectory and allowable deviation band, and selects either a planned tracking mode or a filtered smoothing mode based on the local frequency domain risk feature identifier, real-time node net active power, energy storage state of charge, and load adjustability. The equipment control interface and safety interlock unit 106 connects to the field execution equipment and performs safety interlocking, degraded control, and fault isolation in case of limit violations, communication interruptions, or equipment failures; the encrypted communication unit 107 uploads lightweight operation messages and receives planned power trajectories, control parameters, and strategy update information.

[0049] like Figure 2 and Figure 6As shown, the cloud-edge collaborative control platform 200 may include a data aggregation and time-scale alignment module 201, a risk synthesis and dynamic temporal resolution aggregation engine 202, a cross-scale rolling optimization engine 203, an instruction decomposition and feedback correction module 204, and a strategy and parameter management module 205. The data aggregation and time-scale alignment module 201 receives operation messages uploaded by each edge routing control terminal 100 and performs identity verification, integrity verification, time-scale alignment, outlier removal, and partition archiving. The risk synthesis and dynamic temporal resolution aggregation engine 202 synthesizes partition-weighted risk indicators based on the frequency domain risk characteristics of each node, and performs dynamic temporal resolution aggregation on the electricity, heat, cooling, and gas-side prediction sequences and state sequences according to the risk level and the minimum response time of the equipment. The cross-scale rolling optimization engine 203 constructs a multi-time-scale planning and scheduling model based on the dataset after dynamic temporal resolution aggregation, and jointly solves the electricity purchase and sale plan, energy storage charging and discharging plan, combined cooling, heating and power unit output, heat pump power, cold and heat storage plan, flexible load control sequence, and gas-electric coupling equipment operating status. The instruction decomposition and feedback correction module 204 is used to decompose the global optimization results into the planned power trajectory, fast resource reservation capacity, allowable deviation band and policy parameters corresponding to each edge routing control terminal 100, and to correct the optimization boundary, risk baseline and control parameters of the next rolling cycle based on the execution results, equipment availability and alarm information returned from the edge side. The policy and parameter management module 205 is used to maintain risk thresholds, historical window lengths, filtering coefficients, minimum device response time, control priorities, security boundaries and permission policies.

[0050] In S101, such as Figure 2 and Figure 6As shown, the edge routing control terminal 100 is deployed at electrical coupling points, key feeders, or tie-line locations in the distributed integrated energy system. For example, edge routing control terminals 100 can be set up for key objects such as photovoltaic inverters, energy storage converters, controllable load switches, combined cooling, heating and power units, and grid-connected tie-line, allowing them to connect to the primary circuit and equipment controller via voltage transformers, current transformers, industrial buses, or Ethernet interfaces. The unified timing unit 102 aligns the sampling times of each edge routing control terminal 100 through a Beidou module, GPS module, or IEEE1588PTP time synchronization interface, and collects three-phase voltage, three-phase current, active power, reactive power, and frequency at a sampling frequency greater than or equal to 100Hz under a unified time scale. Simultaneously, it synchronously collects the energy storage state of charge, thermal energy storage state, equipment start-up and shutdown status, and available capacity of interruptible loads. For nodes that do not have the conditions for three-phase waveform sampling, the synchronous active power data sent by the unified timing smart meter or protection and control device is used as the equivalent input. By using the above method, the data uploaded by different edge nodes have a unified time reference, which can avoid the impact of multi-node phase misalignment on subsequent frequency domain analysis and partition risk synthesis.

[0051] In S102, such as Figure 2 As shown, the multi-protocol access and semantic parsing unit 103 performs protocol conversion, point mapping, and object normalization on multi-source heterogeneous data from smart meters, protection and control devices, photovoltaic inverters, energy storage converters, charging facility controllers, cold and heat source controllers, and unit controllers. For data uploaded using different communication protocols such as Modbus, CAN, and IEC61850, the original message is first parsed, and then, according to the unified object model or point table mapping rules, the register address, original fields, and device points are converted into node objects, device objects, and physical quantity tags. Subsequently, the edge routing control terminal 100 generates the node net active power sequence based on the power side sampling values. Specifically, the instantaneous active power of node i at sampling time k is constructed according to the following formula:

[0052] in, These are the sampled values ​​of the three-phase voltage. These are the sampled values ​​of the three-phase current. Then, a polarity convention is executed (positive for generation injection, negative for consumption absorption) to generate the node net active power sequence. In this process, power generation injection is taken as positive, and power consumption absorption is taken as negative; and simultaneously, sequences of thermal power, cooling power, gas consumption, energy storage state variables, and adjustable capacity boundary variables are formed. Thus, the states of multiple energy flows such as electricity, heat, cooling, and gas are mapped to a node state sequence with a unified time scale, providing standardized input for subsequent cross-scale optimization.

[0053] In S103, such as Figure 3As shown, the edge feature extraction unit 104 performs time-frequency analysis on the node net active power sequence (P_{net,i}(k)) to generate frequency domain risk feature identifiers characterizing the node's micro-high-frequency disturbances. Specifically, the node net active power sequence (P_{net,i}(k)) is first subjected to anti-aliasing pre-filtering and sliding window processing. A short-time Fourier transform is then performed with a preset window length L and a preset step size S to obtain the time-frequency coefficients.

[0054] in, Sampling frequency, This is a Hamming window function. Subsequently, it addresses the issue of factors that easily induce voltage / frequency instability in microgrids. to Calculate the normalized energy percentage of key risk frequency bands. :

[0055] in, (Preferred) (This is a preset small quantity to prevent the denominator from being zero.)

[0056] Subsequently, the rolling average within a preset historical window was used. and standard deviation right Perform Z-score standardization to obtain Finally, the final frequency domain risk feature identifier is generated through a sigmoid bounded mapping and a first-order hysteresis filter. :

[0057] Among them, the filter coefficients Preferred .Should The continuity characterizes the intensity of high-frequency fluctuations at the current node.

[0058] This step involves performing time-frequency analysis on the node's net active power and generating a frequency domain risk characteristic identifier. This solves the problem of smoothing and aliasing of high-frequency disturbances after long-period aggregation, and achieves the beneficial effect of transforming the microscopic high-frequency spectrum structure into a risk variable between 0 and 1, so that the high-frequency physical dynamics can be directly called by the upper-level model.

[0059] In S104, such as Figure 2 and Figure 6As shown, the edge routing control terminal 100 organizes information such as node identifier, timestamp, node net active power, frequency domain risk characteristic identifier, energy storage state of charge, thermal energy storage status, available device capacity, and alarm status into a lightweight operation message through the encrypted communication unit 107, and uploads it to the cloud-edge collaborative control platform 200. This lightweight operation message can use JSON, Protobuf, or other structured message formats and be transmitted encrypted via 4G, 5G, or fiber optic networks. Because the message uploads the node's net active power and frequency domain risk characteristic identifier, rather than continuously uploading the complete high-frequency raw waveform, it reduces communication bandwidth, cloud storage, and computing pressure while preserving the key semantics of high-frequency disturbances' impact on system operation.

[0060] In S105, such as Figure 4 As shown, the risk synthesis and dynamic temporal resolution aggregation engine 202 synthesizes partitioned weighted risk indicators based on the frequency domain risk characteristics of each node. Specifically, the cloud-edge collaborative control platform 200 partitions nodes according to feeders, distribution areas, microgrids, parks, or regional energy stations, and sets node weights based on node rated capacity, network sensitivity coefficient, or criticality weight. Synthetic partition Partition-weighted risk indicators :

[0061] in, For partitioning internal node set The frequency domain risk characteristic identifier for node i. Let be the node weight of node i.

[0062] This formula allows nodes with larger rated capacity, higher network sensitivity, or greater criticality to the safe operation of the system to have a greater influence on the partition risk indicators, thus enabling the cloud-edge collaborative control platform 200 to identify the actual disturbance risk level of different partitions.

[0063] In S106, such as Figure 4 As shown, the cloud-edge collaborative control platform 200 uses zone-weighted risk indicators. Dynamic time-resolution aggregation is performed on new energy output forecasts, load forecasts, and resource status forecasts, and a common decision-making time granularity is generated by combining the minimum equipment response time. For rapidly changing variables on the power side, the power-side time granularity can be selected according to the relationship between the zone-weighted risk index and the preset risk threshold. :

[0064] When high-frequency fluctuations are severe ( When ), automatically switch to High resolution is used to capture power spikes; during plateau periods, the resolution is reduced. Minutes to save computing power.

[0065] in, The first risk threshold, This serves as the second risk threshold. When the zoning risk is high, a smaller first time granularity is selected to preserve short-term fluctuations, peaks, and rapid ramp-up characteristics in new energy output or load; when the zoning risk is low, a larger third time granularity is selected to reduce the number of optimization variables and computational burden. For slowly changing variables on the heat, cold, and gas sides, as well as slow-response equipment, the equipment response time granularity is generated by combining the preset minimum adjustment cycle of equipment j, forming a common decision-making time granularity:

[0066] Based on this, the multi-source prediction data is resampled and aggregated to generate a multi-granularity time-series dataset. This refers to the set of devices participating in regulation within partition r. This shared decision-making time granularity prevents heat pumps, boilers, thermal storage tanks, chillers, or gas-electric coupling equipment from being assigned rapid regulation commands exceeding their physical response capabilities.

[0067] In S107, such as Figure 4 As shown, the cross-scale rolling optimization engine 203 constructs a multi-timescale planning and scheduling model based on the aggregated multi-granularity time-series dataset. This model can be a mixed-integer nonlinear programming model, or it can be transformed into a mixed-integer linear programming, second-order cone programming, or model predictive control form according to engineering needs. In scenarios involving coupled resources of electricity, heat, cooling, and gas, the multi-timescale planning and scheduling model satisfies the following multi-energy flow balance relationship.

[0068] The power balance is:

[0069] The thermal power balance is:

[0070] The cooling power balance is:

[0071] Gas volume balance is:

[0072] In the power balance formula, This represents the electrical power input from the external power grid to the distributed integrated energy system at decision time k. If the system feeds back power to the external power grid, it can also be represented as a negative value according to the preset positive and negative sign rules. Indicates photovoltaic power generation capacity; Indicates wind power generation capacity; This represents the electrical power output of the combined cooling, heating and power unit or gas turbine at decision time k; This represents the discharge power of the electrochemical energy storage system at decision time k; Indicates the system's electrical load demand; This indicates the charging power of the electrochemical energy storage system; This indicates the electrical power consumed by the heat pump during operation; This indicates the electrical power consumed by an electric chiller or electrically driven refrigeration equipment. This indicates the amount of interruptible load reduction, curtailed wind and solar power, or other power adjustments that need to be included in the power balance. The formula states that external grid input, wind and solar power output, combined heat and power (CHP) generation, and energy storage discharge collectively meet the demands for electrical load, energy storage charging, heat pump electricity consumption, electric cooling electricity consumption, and reduced or curtailed power.

[0073] In the heat power balance formula, This represents the thermal power output of the combined cooling, heating and power unit at decision time k; This indicates the thermal power output of the boiler; This indicates the heat output power of the heat pump; This represents the heat power released by the heat storage device at decision time k; This indicates the system's heat load demand; This represents the heat power absorbed or stored by the thermal storage device at decision time k. The formula indicates that combined heating, boiler heating, heat pump heating, and thermal storage release collectively meet the heat load and the charging requirements of the thermal storage device.

[0074] In the cold power balance formula, This represents the cooling power output by the electric chiller or electrically driven cold source equipment at decision time k; This represents the cooling power released by the cold storage device at decision time k; This indicates the system's cooling load requirement; This represents the cooling power stored by the cold storage device at decision time k. The formula indicates that the cooling supplied by the electric refrigeration equipment and the cooling released by the cold storage device together meet the cooling load and the charging requirements of the cold storage device.

[0075] In the gas balance formula, This represents the amount of gas input into the system from the external gas pipeline network at decision time k; This indicates the amount of gas consumed by a combined cooling, heating and power unit or a gas turbine. This indicates the amount of gas consumed by the boiler; This indicates the amount of gas consumed by other gas-fired equipment or other gas loads. The formula states that the amount of gas input from the external gas pipeline network is used to meet the gas demand of combined heat and power (CHP) units, boilers, and other gas loads.

[0076] Furthermore, frequency domain risk characteristics are embedded as weights for the planning smoothing term, constraint boundary corrections, and risk relaxation penalty coefficients into the aforementioned multi-timescale planning and scheduling model. The objective function can be expressed as:

[0077] At the same time, the node plan change boundary was Dynamic tightening:

[0078] in, To optimize the set of discrete decision moments in the time domain, for At any decision moment in the process; The number of the node, edge routing control terminal, or adjustable resource participating in the rolling optimization; For decision-making moments The grid interaction costs arising from purchasing electricity from the external grid, selling electricity to the external grid, or power exchange at the grid connection point; the cost items indicated by the ellipsis are conventional comprehensive energy dispatch costs such as fuel costs, equipment operation and maintenance costs, operating costs of combined cooling, heating and power equipment, operating costs of heat pumps, operating costs of cold and heat storage, and flexible load dispatch costs; For decision-making moments Lifetime loss costs or operational degradation costs resulting from energy storage charging and discharging, charging facility regulation, supercapacitors, flywheel energy storage, or interruptible load operation; The base weights for the planned smoothing term are used to set the basic penalty intensity for the smoothness of the planned power trajectory; The basic weighting of risk-based penalties; For nodes At the moment of decision The frequency domain risk feature identifier, whose value range is preferably from 0 to 1, is used to characterize the high-frequency power disturbance risk intensity of the node within the current time window, which is mapped by the energy proportion of the preset risk frequency band. For nodes At the moment of decision Planned net active power obtained through optimization by the cloud-edge collaborative control platform; For nodes The planned net active power at the previous decision-making time; It is a quadratic smoothing penalty term for the change in planned net active power at adjacent decision times, used to suppress drastic jumps in the planned power trajectory; For nodes At the moment of decision Risk slack variables are used to maintain the feasibility of the optimization model under high-risk or conflicting constraint scenarios, and ; For nodes At the moment of decision Basic plan change boundaries without considering frequency domain risks; To preset the risk tightening coefficient, a value of 0.5 is preferred, which is used to control the frequency domain risk characteristic identification. The tightening intensity of the boundary of planned changes.

[0079] When a node experiences severe high-frequency fluctuations at its underlying layer, As the value approaches 1, the model will increase the strength of the planning smoothing term and risk relaxation penalty term for that node, and simultaneously tighten the planning change boundary of the node, thus shaping the planned output trajectory of that node in macro-scheduling. It is smoother, reserving sufficient dynamic margin for underlying resources to quickly cope with high-frequency disturbances.

[0080] Through the aforementioned rolling optimization, the cloud-edge collaborative control platform 200 obtains the planned power trajectory, rapid resource reservation capacity, and allowable deviation band. Subsequently, in S108, as... Figure 6 As shown, the instruction decomposition and feedback correction module 204 decomposes the planned power trajectory, fast resource reservation capacity, and allowable deviation band and sends them to the corresponding edge routing control terminal 100. The planned power trajectory received by each edge routing control terminal 100 is used to determine the low-frequency power tracking target, the fast resource reservation capacity is used to constrain the callable range of energy storage, electric vehicle charging facilities, supercapacitors, flywheel energy storage, or interruptible loads, and the allowable deviation band is used to determine whether the actual local operation exceeds the acceptable range.

[0081] In S109, such as Figure 5 As shown, the local execution and frequency band routing control unit 105 switches between planned tracking mode and filtered smoothing mode based on local frequency domain risk characteristic identifiers and power deviations. The edge routing control terminal 100 acquires the planned power trajectory based on the local frequency domain risk characteristic identifiers. Real-time node net active power ), planned power trajectory and allowable deviation band The power deviation is calculated using the following formula: ΔPi(t)=

[0082] When the local frequency domain risk feature is identified The power deviation ΔPi(t) is less than or equal to a preset safety threshold, and the power deviation ΔPi(t) is located within the allowable deviation band. During this period, execute the aforementioned plan tracking mode; When the local frequency domain risk feature is identified The power deviation is greater than the preset safety threshold, or the power deviation ΔPi(t) exceeds the allowable deviation band. When the time comes, switch to the filtering and smoothing mode; In the filtering and smoothing mode, based on the local frequency domain risk feature identifier The cutoff frequency of the low-pass filter is dynamically adjusted based on the power deviation. Calculate the high-frequency compensation power; In the filtering and smoothing mode, the edge routing control terminal 100 dynamically adjusts the cutoff frequency of the low-pass filter based on the local frequency domain risk characteristic identifier, for example:

[0083] Calculate the high-frequency compensation power that needs to be handled by local fast-response resources. :

[0084] risk The larger, The smaller the value, the stronger the filter's ability to retain high-frequency errors.

[0085] The above-mentioned high-frequency compensation power Priority routing is given to energy storage converters, electric vehicle charging facilities, supercapacitors, flywheel energy storage, and interruptible loads; low-frequency energy deviations are reserved for the cloud-edge collaborative control platform 200 to handle in the next rolling optimization cycle. Therefore, Figure 5 The state machine and frequency band routing control shown can form a control structure of "local isolation of high-frequency disturbances and upper-level coordination of low-frequency energy", avoiding the long-term undertaking of low-frequency energy regulation tasks by rapid resources, reducing energy storage cycle losses and the number of ineffective actions of flexible loads.

[0086] In S110, such as Figure 5 and Figure 6 As shown, the edge routing control terminal 100 continuously transmits back the node's net active power, compensation power, energy storage state of charge change, number of equipment actions, thermal energy storage margin, plan tracking error, and fault or alarm information after execution. The cloud-edge collaborative control platform 200 updates the risk baseline statistics based on the above-transmitted results, enabling the standardized benchmark for subsequent frequency domain risk characteristic identification to adaptively correct with changes in operating conditions. It updates the equipment availability boundary based on energy storage state of charge change, number of equipment actions, and alarm status. It updates the rapid resource reservation capacity requirement based on actual compensation power and plan tracking error; and corrects the optimization boundary for the next rolling cycle based on the equipment availability boundary, risk baseline, and rapid resource requirement. When communication is interrupted or the upper-layer plan is temporarily unavailable, the edge routing control terminal 100 enters degraded operation based on the most recent valid planned power trajectory, the current frequency domain risk characteristic identification, and local security constraints, and maintains the local security boundary through the equipment control interface and the security interlocking unit 106. It reconnects to the cloud-edge collaborative closed loop after communication is restored.

[0087] As can be seen from the above process, this embodiment does not simply upload the high-frequency data from the field to the cloud, nor does it perform local smoothing control solely on the edge side. Instead, it achieves this through... Figure 3 The frequency domain risk feature extraction shown maps high-frequency disturbances at nodes to frequency domain risk feature identifiers; through Figure 4 The risk-driven dynamic temporal resolution aggregation shown transforms micro-risks into criteria for selecting the temporal granularity of the upper-level model and for correcting optimization boundaries; through Figure 5 The edge state machine and frequency band routing control shown enable the edge routing control terminal to switch between planned tracking mode and filtered smoothing mode based on the same risk identifier; then through... Figure 6 The cloud-edge closed-loop communication, command issuance, and feedback correction timing shown enable the execution results to reverse-correct the risk baseline, equipment availability boundary, and optimization boundary for the next rolling cycle. This preserves the risk semantics of high-frequency disturbances without continuously uploading the original high-frequency waveforms, and integrates them throughout the entire process of "edge perception—risk extraction—cloud aggregation—cross-scale optimization—edge execution—feedback correction." This improves the operational stability, scheduling feasibility, and lifecycle economy of distributed integrated energy systems in scenarios with high proportions of new energy access, multi-energy flow coupling, and significant differences in equipment response time.

[0088] like Figure 7 As shown, this application proposes a distributed integrated energy cross-scale collaborative control system 10, comprising: The acquisition unit 11 is used to deploy multiple edge routing control terminals at the electrical coupling points, key feeders or tie lines of the distributed integrated energy system, and to acquire node physical quantities under unified time synchronization. The first generation unit 12 is used to perform protocol conversion, point mapping and object normalization on the collected multi-source heterogeneous data, construct a node state sequence with a unified time scale, and generate a node net active power sequence based on the power side sampling value. The second generation unit 13 is used to perform time-frequency analysis on the net active power sequence of the above nodes and generate frequency domain risk feature identifiers that characterize the micro high-frequency disturbances of the nodes. Upload unit 14 is used to organize the node identifier, timestamp, the above-mentioned node net active power sequence, the above-mentioned frequency domain risk characteristic identifier and equipment status quantity into a lightweight operation message and upload it to the cloud-edge collaborative control platform. Synthesis unit 15 is used by the cloud-edge collaborative control platform to synthesize a partition-weighted risk index based on the frequency domain risk characteristics of each node. The third generation unit 16 is used to perform dynamic time resolution aggregation on the new energy output forecast, load forecast and resource status forecast based on the above-mentioned partition weighted risk indicators, and generate common decision time granularity in combination with the minimum response time of the equipment. Optimization unit 17 is used to construct a multi-timescale planning and scheduling model based on the aggregated multi-granularity time series dataset. The above-mentioned frequency domain risk feature identifier is embedded into the above-mentioned multi-timescale planning and scheduling model as the weight of the planning smoothing term, the correction amount of the constraint boundary and the risk relaxation penalty coefficient, and rolling optimization is performed to obtain the planned power trajectory, the fast resource reservation capacity and the allowable deviation band. The sending unit 18 is used to send the above-mentioned planned power trajectory, the above-mentioned rapid resource reserved capacity and the above-mentioned allowable deviation band to the corresponding edge routing control terminal. The switching unit 19 is used for the edge routing control terminal to switch between the planned tracking mode and the filter smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, and to route the high-frequency compensation power to the fast resource execution in the filter smoothing mode. The control unit 20 is used to transmit the node operation results after execution back to the cloud-edge collaborative control platform, and the cloud-edge collaborative control platform updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle accordingly.

[0089] Understandably, the system proposed in the second aspect is capable of performing any of the methods described in the first aspect.

[0090] In summary, this invention performs time-frequency analysis on the net active power of nodes at the edge, refining the high-frequency waveform data that would otherwise require extensive uploading into frequency-domain risk feature identifiers. This reduces communication bandwidth and cloud computing pressure while preserving the semantics of high-frequency disturbance risks. The invention uploads these frequency-domain risk feature identifiers to the cloud-edge collaborative control platform and synthesizes a partitioned weighted risk index, enabling upper-layer scheduling to identify the actual fluctuation risks of different nodes and regions, avoiding coarse optimization based solely on low-frequency average power. Based on the partitioned weighted risk index, this invention performs dynamic temporal resolution aggregation, using finer time granularity to preserve power spikes and ramp-up characteristics during high-risk periods and coarser time granularity to reduce model size during low-risk periods, achieving a balance between information preservation and solution efficiency. Furthermore, this invention embeds the frequency-domain risk feature identifiers into the planning smoothing term, constraint boundary correction term, and risk relaxation penalty term of the multi-timescale planning and scheduling model, allowing high-risk nodes to automatically obtain smoother planned power trajectories and reserving more dynamic adjustment margins, thus improving the physical executability of the scheduling results. This invention switches between a planned tracking mode and a filtered smoothing mode via an edge routing control terminal, prioritizing the allocation of high-frequency compensation power to rapid resources such as energy storage, electric vehicle charging facilities, and interruptible loads. This achieves local isolation of high-frequency disturbances and cloud-based coordination of low-frequency energy, reducing ineffective actions and lifespan loss of rapid resources. By transmitting execution results back to the system, this invention continuously updates risk baseline statistics, equipment availability boundaries, reserved capacity requirements for rapid resources, and optimization boundaries for the next rolling cycle. This forms a closed-loop control system from edge perception, cloud optimization, edge execution to feedback correction, improving the operational stability, scheduling robustness, and economy of distributed integrated energy systems in scenarios with high proportions of new energy access and multi-energy flow coupling.

[0091] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A distributed integrated energy cross-scale collaborative control method, characterized in that, include: Multiple edge routing control terminals are deployed at electrical coupling points, key feeders, or tie lines in the distributed integrated energy system, and node physical quantities are collected under unified time synchronization. The collected multi-source heterogeneous data undergoes protocol conversion, point mapping, and object normalization to construct a node state sequence with a unified time scale, and generates a node net active power sequence based on power-side sampled values. Time-frequency analysis is performed on the net active power sequence of the node to generate frequency domain risk feature identifiers characterizing the node's micro-high frequency disturbances; The node identifier, timestamp, net active power sequence of the node, frequency domain risk characteristic identifier, and device status quantity are organized into a lightweight operation message and uploaded to the cloud-edge collaborative control platform. The cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node; Based on the aforementioned partition-weighted risk indicators, dynamic time-resolution aggregation is performed on the forecasts of new energy output, load, and resource status, and a common decision-making time granularity is generated by combining the minimum response time of the equipment. A multi-timescale planning and scheduling model is constructed based on the aggregated multi-granularity time series dataset. The frequency domain risk feature identifier is embedded as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty coefficient into the multi-timescale planning and scheduling model. Rolling optimization is performed to obtain the planned power trajectory, the fast resource reservation capacity, and the allowable deviation band. The planned power trajectory, the rapid resource reservation capacity, and the allowable deviation band are sent to the corresponding edge routing control terminal. The edge routing control terminal switches between the planned tracking mode and the filtered smoothing mode based on the local frequency domain risk feature identifier and power deviation, and routes the high-frequency compensation power to the fast resource execution in the filtered smoothing mode; The node execution results are sent back to the cloud-edge collaborative control platform, which then updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle.

2. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The acquisition of node physical quantities under unified time synchronization includes: The sampling times of each edge routing control terminal are aligned using a unified timing unit; Three-phase voltage, three-phase current, active power, reactive power, and frequency are collected at a sampling frequency greater than or equal to 100Hz. Synchronously collect data on energy storage status of charge, thermal energy storage status, equipment start / stop status, and available capacity of interruptible loads; For nodes that do not have the conditions for three-phase waveform sampling, the synchronous active power data sent by smart meters or protection and control devices with unified time synchronization is used as the equivalent input.

3. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The process of generating the node net active power sequence based on power-side sampled values ​​includes: The instantaneous active power of node i at sampling time k is constructed according to the following formula: ； in, , , These are the sampled values ​​of the three-phase voltage at node i at sampling time k. , , These are the sampled values ​​of the three-phase current at node i at sampling time k; The instantaneous active power is subjected to a polarity convention to obtain the node net active power Pnet,i(k), where generation injection is positive and consumption absorption is negative; Simultaneously, a sequence of thermal power, cooling power, gas consumption, energy storage state parameters, and adjustable capacity boundary parameters is generated.

4. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The step of performing time-frequency analysis on the net active power sequence of the node to generate frequency domain risk feature identifiers characterizing the node's micro-high-frequency disturbances includes: The node net active power sequence Pnet,i(k) is subjected to anti-aliasing pre-filtering and sliding window processing; Perform a short-time Fourier transform with a preset window length L and a preset step size S to obtain the time-frequency coefficients Xi(m,f); For each time frame m, extract the normalized energy percentage Ei(m) of the risk band from 0.1Hz to 2Hz; The normalized energy percentage Ei(m) is standardized using the rolling mean μi and rolling standard deviation σi within a preset historical window to obtain the standardized risk value Zi(m). The frequency domain risk feature identifier of node i is generated through bounded mapping and first-order hysteresis filtering. The frequency domain risk feature identifier Used to characterize the degree of high-frequency disturbance risk of node i within the corresponding time frame.

5. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The cloud-edge collaborative control platform synthesizes a partition-weighted risk index based on the frequency domain risk characteristics of each node, including: The nodes are divided into zones according to feeders, transformer areas, microgrids, and park or regional energy stations; Node weights are set based on node rated capacity, network sensitivity coefficient, or criticality weight. ; The partition-weighted risk index for partition r is synthesized using the following formula. : ； in, For partitioning internal node set The frequency domain risk characteristic identifier for node i. Let be the node weight of node i; The lightweight operation message includes at least the node identifier, timestamp, node net active power, frequency domain risk characteristic identifier, energy storage state of charge, thermal energy storage state, equipment available capacity, and alarm status.

6. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The step of performing dynamic time-resolution aggregation on new energy output forecast, load forecast, and resource status forecast based on the partition-weighted risk index, and generating a common decision-making time granularity by combining the minimum equipment response time, includes: The power-side time granularity is selected based on the relationship between the partition-weighted risk index and the preset risk threshold for rapidly changing variables on the power side. When the partition-weighted risk index is greater than the first risk threshold, the first time granularity is selected; When the partition-weighted risk index is less than or equal to the first risk threshold and greater than the second risk threshold, the second time granularity is selected. When the partition-weighted risk index is less than or equal to the second risk threshold, the third time granularity is selected; Wherein, the first time granularity is smaller than the second time granularity, and the second time granularity is smaller than the third time granularity; For slow-changing variables and slow-response equipment on the heat, cold, and air sides, the equipment response time granularity is generated by combining the equipment's preset minimum adjustment cycle. The power-side time granularity and the equipment response time granularity are coordinated to generate the common decision-making time granularity.

7. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The multi-timescale planning and scheduling model includes electrical power balance constraints, thermal power balance constraints, cold power balance constraints, and gas volume balance constraints. In scenarios involving coupled resources of electricity, heat, cold, and gas, the multi-timescale planning and scheduling model simultaneously satisfies the electrical power balance constraint, the thermal power balance constraint, the cold power balance constraint, and the gas volume balance constraint. In scenarios involving only power-side resources, the multi-timescale planning and scheduling model retains only the power balance constraints and power equipment constraints; The constraints on the power equipment include at least the energy storage state of charge constraints, energy storage charging and discharging power constraints, grid connection point switching power constraints, equipment ramping constraints, and interruptible load dispatch constraints.

8. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The step of embedding the frequency domain risk feature identifier as the weight of the planning smoothing term, the constraint boundary correction amount, and the risk relaxation penalty coefficient into the multi-time-scale planning and scheduling model includes: Construct an objective function that includes electricity purchase and sale costs, fuel costs, operation and maintenance costs, planning smoothing costs, and risk relaxation penalty costs; Identify the frequency domain risk characteristics of node i at decision time t. As a weight for the planned smoothing term, the changes in planned net active power corresponding to high-risk nodes are subject to stronger constraints. Set risk slack variables and identify the frequency domain risk features. As the aforementioned risk relaxation variable Penalty coefficient; The boundary of node plan changes should be corrected according to the following formula: ； in, Let be the planned net active power of node i at decision time k. Let be the planned net active power of node i at the previous decision time. Let i be the basic plan change boundary. To preset the risk tightening coefficient, For risk slack variables.

9. The distributed integrated energy cross-scale collaborative control method according to claim 1, characterized in that, The edge routing control terminal switches between planned tracking mode and filtered smoothing mode based on the local frequency domain risk characteristic identifier and power deviation, including: Obtain local frequency domain risk feature identifier Real-time node net active power Planned power trajectory and allowable deviation band ; Based on the real-time node net active power With the planned power trajectory Calculate the power deviation ΔPi(t); When the local frequency domain risk feature is identified The power deviation ΔPi(t) is less than or equal to a preset safety threshold, and the power deviation ΔPi(t) is located within the allowable deviation band. During this period, the plan tracking mode is executed; When the local frequency domain risk feature is identified The power deviation is greater than the preset safety threshold, or the power deviation ΔPi(t) exceeds the allowable deviation band. When the time comes, switch to the filtering and smoothing mode; In the filtering and smoothing mode, based on the local frequency domain risk feature identifier The cutoff frequency of the low-pass filter is dynamically adjusted based on the power deviation. Calculate the high-frequency compensation power; The high-frequency compensation power is preferentially routed to energy storage converters, electric vehicle charging facilities and interruptible loads; It continuously transmits back the node's net active power, compensation power, energy storage state of charge change, number of equipment actions, thermal energy storage margin, plan tracking error, and fault or alarm information after execution. When communication is interrupted or the upper-layer plan is temporarily unavailable, it enters degraded operation based on the most recent valid planned power trajectory, the current frequency domain risk characteristic identifier, and local security constraints.

10. A distributed integrated energy cross-scale collaborative control system, characterized in that, include: The acquisition unit is used to deploy multiple edge routing control terminals at electrical coupling points, key feeders or tie lines of a distributed integrated energy system, and to acquire node physical quantities under unified time synchronization. The first generation unit is used to perform protocol conversion, point mapping and object normalization on the collected multi-source heterogeneous data, construct a node state sequence with a unified time scale, and generate a node net active power sequence based on the power side sampling value. The second generation unit is used to perform time-frequency analysis on the net active power sequence of the node and generate frequency domain risk feature identifiers that characterize the node's micro high-frequency disturbances. The upload unit is used to organize the node identifier, timestamp, the node's net active power sequence, the frequency domain risk characteristic identifier, and the device status quantity into a lightweight operation message and upload it to the cloud-edge collaborative control platform. A synthesis unit is used by the cloud-edge collaborative control platform to synthesize a partition-weighted risk index based on the frequency domain risk characteristics of each node; The third generation unit is used to perform dynamic time resolution aggregation on the new energy output forecast, load forecast and resource status forecast based on the partition weighted risk index, and generate a common decision time granularity in combination with the minimum response time of the equipment. The optimization unit is used to construct a multi-timescale planning and scheduling model based on the aggregated multi-granularity time series dataset. The frequency domain risk feature identifier is embedded into the multi-timescale planning and scheduling model as the weight of the planning smoothing term, the correction amount of the constraint boundary, and the risk relaxation penalty coefficient. Rolling optimization is performed to obtain the planned power trajectory, the fast resource reservation capacity, and the allowable deviation band. The distribution unit is used to distribute the planned power trajectory, the rapid resource reservation capacity, and the allowable deviation band to the corresponding edge routing control terminal. The switching unit is used for the edge routing control terminal to switch between the planned tracking mode and the filtered smoothing mode based on the local frequency domain risk feature identifier and power deviation, and to route the high-frequency compensation power to the fast resource execution in the filtered smoothing mode; The control unit is used to transmit the node operation results after execution back to the cloud-edge collaborative control platform, and the cloud-edge collaborative control platform updates the risk baseline statistics, device availability boundaries, rapid resource reservation capacity requirements, and optimization boundaries for the next rolling cycle accordingly.

Images