Distributed light storage micro-grid multi-source coordinated energy scheduling method, device and server

Abstract

The application provides a kind of distributed light storage micro-grid multi-source coordinated energy scheduling method, device and server, it is related to the technical field of micro-grid energy management, comprising: with preset sampling frequency to micro-grid data acquisition processing, obtain full-dimensional operation data, and the full-dimensional operation data is locally preprocessed, and obtain the characteristic data after preprocessing;According to the characteristic data and historical operation data, perform multi-time scale prediction processing, obtain multi-time scale prediction result, and according to the multi-time scale prediction result, determine the optimization target, carry out multi-objective optimization solution processing, generate global scheduling plan;Based on preset control cycle, the global scheduling plan is corrected and handled, and real-time control instruction is generated, to complete the energy scheduling of micro-grid by executing real-time control instruction.The application can significantly reduce the comprehensive power consumption cost while ensuring power supply reliability through multi-source coordinated energy scheduling.

Description

Technical Field

[0001] This invention relates to the technical field of microgrid energy management, and in particular to a method, apparatus and server for multi-source coordinated energy dispatching of distributed photovoltaic-storage microgrids. Background Technology

[0002] With the advancement of dual-carbon goals, distributed photovoltaic-storage microgrids have become an important component of new power systems, enabling both grid-connected and islanded operation. As the core of microgrid regulation, the multi-source coordinated energy dispatch system directly affects the system's operational stability and economy. Currently, related technologies suggest that existing microgrid energy dispatch systems generally adopt a centralized cloud architecture, with data acquisition, optimization calculations, and command issuance all handled centrally by the cloud. However, since all operational data must be uploaded to the cloud for processing, communication transmission and optimization calculations are time-consuming, resulting in control command issuance delays of up to minutes. When photovoltaic power generation experiences a sudden drop in power or a sudden fluctuation in load due to weather changes, the system cannot respond within milliseconds, which can easily lead to voltage exceeding limits, frequency deviations, or even grid disconnection accidents. Furthermore, existing technologies often employ a single prediction model and a single time scale prediction method, failing to fully consider the impact of dynamic factors such as sudden weather changes and equipment degradation on prediction accuracy, resulting in significant prediction errors. In terms of optimization objectives, existing optimization models only focus on minimizing system operating costs, without incorporating the life-cycle degradation costs of energy storage batteries, carbon emission costs, and power supply reliability into the optimization framework. This leads to frequent deep charging and discharging of energy storage systems, significantly shortening their lifespan and resulting in poor long-term economic efficiency. Therefore, existing solutions cannot significantly reduce overall electricity costs while ensuring power supply reliability. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a multi-source coordinated energy dispatching method, device and server for distributed photovoltaic-storage microgrids, which can significantly reduce the overall electricity cost while ensuring power supply reliability through multi-source coordinated energy dispatching.

[0004] In a first aspect, embodiments of the present invention provide a multi-source coordinated energy dispatching method for distributed photovoltaic-storage microgrids. The method is applied to a multi-source coordinated energy dispatching system for distributed photovoltaic-storage microgrids. The method includes: collecting and processing data from the microgrid at a preset sampling frequency to obtain full-dimensional operational data; performing local preprocessing on the full-dimensional operational data to obtain preprocessed feature data; performing multi-timescale prediction processing based on the feature data and historical operational data to obtain multi-timescale prediction results; determining operating costs, energy storage health degradation costs, carbon emission costs, and system load failure rates based on the multi-timescale prediction results; performing multi-objective optimization processing with operating costs, energy storage health degradation costs, carbon emission costs, and system load failure rates as optimization objectives to generate a global dispatching plan; and performing rolling correction processing on the global dispatching plan based on a preset control cycle to generate real-time control commands, thereby completing the energy dispatching of the microgrid by executing the real-time control commands.

[0005] In one implementation, the step of performing local preprocessing on full-dimensional operational data to obtain preprocessed feature data includes: sequentially performing filtering, noise reduction, and outlier removal on the full-dimensional operational data to obtain preprocessed operational data; and performing multi-protocol conversion on the preprocessed operational data to convert device data from different communication protocols into a unified format to obtain feature data.

[0006] In one implementation, the step of performing multi-timescale forecasting processing based on feature data and historical operating data to obtain multi-timescale forecasting results includes: fusing feature data with historical operating data, meteorological data, electricity price data, and equipment parameter data to construct a time-series database; performing 24-hour short-term forecasting processing based on the time-series database to obtain the photovoltaic output baseline curve and load demand baseline curve for the next day; performing intraday 4-hour ultra-short-term forecasting processing on the photovoltaic output baseline curve and load demand baseline curve for the next day with an hourly cycle to correct the forecasting deviation; and performing rolling forecasting processing on the corrected photovoltaic output baseline curve and load demand baseline curve for the next day with a 15-minute cycle to respond to meteorological changes and obtain multi-timescale forecasting results.

[0007] In one implementation, the steps of performing short-term forecasting processing on the 24-hour period prior to the current day based on a time-series database to obtain the photovoltaic power output baseline curve and load demand baseline curve for the next day include: performing joint forecasting processing on the time-series database using an attention mechanism, a long short-term memory network model, and a random forest model to obtain an initial forecast result; and correcting the initial forecast result according to a pre-configured meteorological change correction factor and equipment attenuation correction factor to obtain the photovoltaic power output baseline curve and load demand baseline curve for the next day.

[0008] In one implementation, the steps of generating a global scheduling plan by performing multi-objective optimization with operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate as optimization objectives include: constructing a comprehensive optimization objective function with operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate as optimization objectives; determining the weight coefficients corresponding to each optimization objective based on the current operating mode; and performing rolling solution processing on the comprehensive optimization objective function based on the weight coefficients using an improved rolling time-domain control algorithm to obtain the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period, so as to generate a global scheduling plan.

[0009] In one implementation, the step of generating real-time control commands by performing rolling correction processing on the global scheduling plan based on a preset control cycle includes: correcting the global scheduling plan with a 10-minute cycle based on full-dimensional operational data and multi-time-scale prediction results to obtain a corrected scheduling plan; and generating real-time control commands corresponding to each target with a 50-millisecond control cycle based on the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period in the corrected scheduling plan. The real-time control commands include: photovoltaic inverter output adjustment commands, energy storage converter charging and discharging control commands, diesel generator start-stop commands, controllable load adjustment commands, and charging and discharging control commands.

[0010] In one implementation, after the step of generating real-time control commands, the process includes: real-time monitoring and processing of the grid voltage state and grid frequency state at a preset sampling frequency, and automatically identifying the current operating mode based on the monitoring results, wherein the current operating mode is any one of grid-connected operating mode, islanded operating mode, and grid-to-island switching mode; when the current operating mode is identified as grid-to-island switching mode, pre-synchronization seamless switching processing is performed to complete the switching between grid-connected mode and islanded mode within a preset time interval.

[0011] Secondly, embodiments of the present invention also provide a multi-source coordinated energy dispatching device for distributed photovoltaic-storage microgrids. The device is applied to a multi-source coordinated energy dispatching system for distributed photovoltaic-storage microgrids. The device includes: a data acquisition module, which acquires and processes data from the microgrid at a preset sampling frequency to obtain full-dimensional operational data, and performs local preprocessing on the full-dimensional operational data to obtain preprocessed feature data; a global dispatching module, which performs multi-timescale prediction processing based on the feature data and historical operational data to obtain multi-timescale prediction results, and determines operating costs, energy storage health degradation costs, carbon emission costs, and system load failure rates based on the multi-timescale prediction results, and performs multi-objective optimization processing with operating costs, energy storage health degradation costs, carbon emission costs, and system load failure rates as optimization objectives to generate a global dispatching plan; and an energy dispatching module, which performs rolling correction processing on the global dispatching plan based on a preset control cycle to generate real-time control commands, and completes the energy dispatching of the microgrid by executing the real-time control commands.

[0012] Thirdly, embodiments of the present invention also provide a server, including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement any of the methods provided in the first aspect.

[0013] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement any of the methods provided in the first aspect.

[0014] The embodiments of the present invention bring the following beneficial effects: This invention provides a method, apparatus, and server for multi-source coordinated energy dispatching of distributed photovoltaic-storage microgrids. The method collects and processes microgrid data at a preset sampling frequency to obtain full-dimensional operational data. This full-dimensional operational data is then preprocessed locally to obtain preprocessed feature data. Based on the feature data and historical operational data, multi-timescale prediction processing is performed to obtain multi-timescale prediction results. Based on these prediction results, operating costs, energy storage health degradation costs, carbon emission costs, and system load failure rates are determined. Multi-objective optimization is then performed using these costs as optimization objectives to generate a global dispatching plan. Finally, the global dispatching plan is rolled over based on a preset control cycle to generate real-time control commands. Execution of these real-time control commands completes the energy dispatching of the microgrid. This invention, through multi-source coordinated energy dispatching, can significantly reduce overall electricity costs while ensuring power supply reliability.

[0015] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a distributed photovoltaic-storage microgrid multi-source coordinated energy dispatching system provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a multi-source coordinated energy dispatching method for a distributed photovoltaic-storage microgrid provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a distributed photovoltaic-storage microgrid multi-source coordinated energy dispatching device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a server provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Currently, with the advancement of dual-carbon goals, distributed photovoltaic-storage microgrids have become a core component of new power systems, enabling grid-connected / islanded dual-mode operation, improving renewable energy absorption rates, and reducing electricity costs. The multi-source coordinated energy dispatch system is the intelligent brain of the microgrid, directly determining its operational stability, economy, and environmental friendliness. Existing microgrid energy dispatch technologies suffer from the following core shortcomings: Architecture bottlenecks: The centralized cloud-based optimization approach, which involves uploading all data to the cloud, results in control latency of up to minutes, failing to meet millisecond-level emergency response requirements; the two-level optimization scheme fails to decouple global optimization from real-time control, which can easily lead to voltage / frequency exceeding limits and network disconnection.

[0021] Prediction and optimization deficiencies: Single-model, single-time-scale predictions do not consider meteorological changes and equipment degradation, resulting in prediction bias exceeding 15%; optimization models only pursue the lowest operating costs and do not incorporate energy storage lifecycle degradation, carbon emissions, and power supply reliability, leading to severe energy storage losses.

[0022] Insufficient coordination and adaptation: It only achieves basic coordination of source-storage-grid, and has poor coordination capabilities for flexible loads, V2G, and backup power sources; it is prone to power imbalance during grid-connected / islanded switching, lacks multi-microgrid cluster coordination capabilities, and has poor stability under extreme operating conditions.

[0023] Based on this, the distributed photovoltaic-storage microgrid multi-source coordinated energy dispatching method, device and server provided by the present invention can significantly reduce the overall electricity cost while ensuring power supply reliability through multi-source coordinated energy dispatching.

[0024] To facilitate understanding of this embodiment, a detailed description of a multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids disclosed in this embodiment of the invention will be provided first. This method is applied to a multi-source coordinated energy dispatch system for distributed photovoltaic-storage microgrids. To facilitate understanding of the multi-source coordinated energy dispatch system for distributed photovoltaic-storage microgrids, this embodiment of the invention provides a structural schematic diagram of the multi-source coordinated energy dispatch system for distributed photovoltaic-storage microgrids, as shown below. Figure 1 As shown, it includes: terminal perception and execution layer, edge real-time control layer and cloud global optimization layer.

[0025] The terminal sensing and execution layer deploys an intelligent sensing and control unit with a sampling frequency of 100Hz to collect full-dimensional operational data such as voltage, current, power, SOC, and weather. The distributed generation unit is equipped with a grid-connected inverter with an integrated MPPT controller, and the energy storage unit is equipped with a bidirectional PCS converter. The controllable load unit has a built-in flexible load interaction scheduling subunit to support peak shifting, peak avoidance, and shift adjustment. The V2G charging pile cluster is connected to the terminal layer to realize bidirectional energy interaction between vehicles and the power grid.

[0026] The terminal perception and execution layer can provide high-precision, high-frequency basic data for real-time edge control and cloud-based predictive optimization, supporting full collaborative control of source-grid-load-storage-charging and solving the problem of insufficient multi-source collaborative capabilities.

[0027] The edge real-time control layer deploys an industrial-grade edge computing gateway to perform data filtering, noise reduction, outlier removal, and multi-protocol conversion. The local real-time controller uses a 50ms control cycle to perform millisecond-level power balancing and voltage / frequency stabilization control, and performs rolling corrections to the cloud scheduling plan every 10 minutes. The mode switching control module adopts pre-synchronous seamless switching technology, with grid-connected / islanded switching time of <10ms and no power surge. The fault self-healing control module can locate the fault area within 10ms and perform fault isolation and load tiered reduction.

[0028] The edge real-time control layer directly solves the shortcomings of existing solutions, such as high control latency and inability to achieve millisecond-level response. It also solves the problems of power imbalance during grid-connected / islanded switching and low power supply reliability under extreme operating conditions.

[0029] The cloud-based global optimization layer constructs a multi-source data fusion time-series database, storing more than 3 years of historical operating data, meteorological data, and equipment parameters; the multi-timescale prediction module performs three levels of prediction: 24 hours before the day, 4 hours during the day, and 15 minutes in real time; the multi-objective optimization scheduling module establishes an objective function that includes operating costs, energy storage SOH decay, carbon emissions, and load shedding rate, and uses an improved MPC algorithm for rolling solution, with a single optimization time of <30 seconds; the cluster collaborative scheduling module supports joint optimization and energy mutual assistance of multiple microgrids and can be connected to a virtual power plant platform.

[0030] The cloud-based global optimization layer solves the problems of low prediction accuracy and lack of consideration for the entire life cycle of energy storage in existing solutions, while also enabling multi-microgrid cluster collaboration and virtual power plant docking.

[0031] The communication network module adopts a redundant communication architecture of fiber optic and 5G. The terminal and the edge layer use industrial Ethernet fiber optic communication, and the edge and the cloud use dual links of 5G and fiber optic. It supports industrial standard protocols such as IEC61850, MQTT, Modbus, and OPCUA. It has a built-in AES256 data encryption unit, industrial firewall, and intrusion detection system.

[0032] The communication network module ensures low latency and high reliability of data transmission in the three-tier architecture, providing communication support for cloud-edge-device collaborative operation.

[0033] Therefore, this invention addresses the core architectural bottleneck of existing microgrid dispatching systems, which cannot simultaneously achieve global optimization accuracy and millisecond-level real-time response. It adopts a fully decoupled three-level architecture (cloud-edge-device) to completely separate global optimization from real-time control. The terminal layer collects comprehensive data on power generation, grid, load, and storage at a 100Hz sampling frequency and accurately executes dispatching commands. The edge layer deploys local controllers with a 50ms control cycle, achieving seamless grid-connected / islanded switching and fault self-healing at the 10ms level, and completing millisecond-level power balance and voltage / frequency stability control. In the cloud, a multi-timescale prediction model using LSTM, random forest, and attention mechanisms is constructed. Combined with dynamic correction factors, the prediction error is controlled within 5%. Simultaneously, a multi-objective optimization function encompassing the entire lifecycle degradation of energy storage is established, solved using an improved MPC algorithm, and supports multi-microgrid cluster collaboration and virtual power plant integration. Ultimately, this achieves a balance between high reliability, low latency, and full lifecycle economic efficiency in the dispatching system.

[0034] based on Figure 1 The diagram shown illustrates the structure of a distributed photovoltaic-storage microgrid multi-source coordinated energy dispatching system. This invention provides a detailed description of the multi-source coordinated energy dispatching method for distributed photovoltaic-storage microgrids. (See also...) Figure 2 The diagram shows a flowchart of a multi-source coordinated energy dispatch method for a distributed photovoltaic-storage microgrid. The method mainly includes the following steps S202 to S206: Step S202: Data acquisition and processing of the microgrid is performed at a preset sampling frequency to obtain full-dimensional operation data, and the full-dimensional operation data is preprocessed locally to obtain preprocessed feature data.

[0035] In one implementation, during microgrid operation, the system first needs to acquire accurate and comprehensive operational status information. This invention synchronously collects data from photovoltaic arrays, energy storage battery packs, various loads, backup power supplies, and V2G charging pile clusters within the microgrid at a preset sampling frequency. This acquires comprehensive operational data, including voltage, current, power, energy storage state of charge and health status, photovoltaic irradiance, ambient temperature and humidity, equipment operating status, and grid frequency. High-frequency acquisition ensures the system can capture key dynamic information such as second-level fluctuations in photovoltaic output due to cloud cover and instantaneous power changes during load addition or removal, providing a refined data foundation for subsequent real-time control and predictive optimization.

[0036] In existing technologies, raw data collected by the terminal is typically uploaded entirely to the cloud for processing. This results in a large amount of redundant data consuming communication bandwidth and significantly increasing transmission latency. To address this issue, this invention preprocesses the full-dimensional operational data locally immediately after acquisition. The preprocessing process includes: filtering the raw signal to eliminate measurement noise, performing noise reduction to extract valid signal components, and performing outlier removal to eliminate erroneous data caused by sensor malfunctions or communication errors. The processed operational data is then converted into a unified format for local real-time control.

[0037] Meanwhile, the preprocessed operational data also undergoes multi-protocol conversion. Since devices such as photovoltaic inverters, energy storage converters, load controllers, and charging piles in a microgrid may use different communication protocols, this invention converts data from different protocols into a unified standard format to obtain preprocessed feature data. This allows devices from different manufacturers to access the system uniformly, and also uploads the filtered and compressed feature data to the cloud, reducing the consumption of communication resources by invalid data. Through the above acquisition and preprocessing process, the system provides high-quality data support for millisecond-level real-time control at the edge layer and high-precision prediction and optimization at the cloud layer, which is also a prerequisite for the accurate execution of all subsequent scheduling decisions.

[0038] Step S204: Based on the feature data and historical operating data, perform multi-timescale prediction processing to obtain multi-timescale prediction results. Based on the multi-timescale prediction results, determine the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate. Perform multi-objective optimization processing with the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate as optimization objectives to generate a global scheduling plan.

[0039] In one implementation, after the feature data is uploaded to the cloud, the system integrates it with historical operating data, meteorological data, electricity price data, and equipment parameter data to build a unified time-series database, providing a sufficient data foundation for the prediction model.

[0040] Existing forecasting methods typically employ a single model and a single time scale, which cannot effectively address the uncertainties caused by sudden weather changes and equipment aging, resulting in forecast deviations exceeding 15% and causing significant discrepancies between scheduling plans and actual operating conditions. To address this, this invention adopts a multi-time-scale progressive forecasting strategy: First, a short-term 24-hour forecast is executed, predicting the photovoltaic output trend and load demand changes for the following day based on a time-series database, generating a baseline curve. Building upon this, an intraday 4-hour ultra-short-term forecast is executed on an hourly basis to correct the forecast deviation of the baseline curve. Finally, a real-time rolling forecast is executed on a 15-minute cycle to respond to and correct for immediate disturbances such as sudden weather changes. These three levels of forecasting, progressively progressive and layeredly corrected, gradually bring the forecast results closer to the actual values.

[0041] Based on this, the system needs to transform the prediction results into executable scheduling decisions. Existing optimization models typically focus solely on minimizing operating costs, neglecting the actual physical characteristics of accelerated aging of energy storage batteries during charge-discharge cycles, and failing to incorporate carbon emission costs and power supply reliability into the optimization framework. This leads to frequent deep charge-discharge cycles of energy storage, high carbon emission intensity, and difficulty in guaranteeing power supply reliability under extreme operating conditions. To address these shortcomings, this invention uses four indicators—operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate—as optimization objectives, constructing a comprehensive optimization objective function. These four indicators are interdependent; reducing operating costs may lead to excessive energy storage utilization and accelerated degradation, while excessive protection of energy storage may increase the risk of load failure. Conflicting relationships exist among the objectives, requiring trade-offs.

[0042] The system dynamically adjusts the weight coefficients of each objective based on the current operating mode (grid-connected mode focuses on economy and energy storage life, while islanded mode focuses on power supply reliability), so that the optimization objective is automatically matched with the operating conditions. Through an improved rolling time-domain control algorithm, the comprehensive objective function is solved in each control cycle based on the latest state and prediction information, and the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target and grid interaction power threshold are obtained in turn for each time period. This generates a global scheduling plan that can be executed and modified by the edge layer in subsequent steps.

[0043] Step S206: Based on the preset control cycle, the global scheduling plan is rolled over and modified to generate real-time control commands, so as to complete the energy scheduling of the microgrid by executing the real-time control commands.

[0044] In one implementation, the global scheduling plan generated in the cloud is based on a day-ahead forecast and hourly updates. This plan cannot directly address power fluctuations and sudden changes in equipment status at the second to millisecond level. If the cloud plan is executed directly as is, the system will be unable to respond promptly when photovoltaic power output drops instantaneously due to cloud obstruction or when loads are suddenly switched on or off, potentially leading to voltage exceedances and frequency deviations. Therefore, the edge layer needs to locally modify the global scheduling plan using a control cycle independent of the cloud.

[0045] Specifically, the edge layer, with a 10-minute cycle, modifies the photovoltaic output targets, energy storage charging and discharging power targets, controllable load adjustment targets, and grid interaction power thresholds for each time period in the global scheduling plan based on the real-time feedback of full-dimensional operating data from the terminal and the latest multi-time-scale prediction results output from the cloud. This ensures that the scheduling plan remains synchronized with the actual operating status. After the above modifications, the long-cycle, coarse-grained global plan in the cloud is transformed into a modified scheduling plan adapted to the local real-time operating conditions.

[0046] The revised scheduling plan still targets targets at the minute level, but terminal devices (such as photovoltaic inverters and energy storage converters) require specific level and switching signals to execute actions. Therefore, the edge layer uses a control cycle of fifty milliseconds to convert the target values ​​in the revised scheduling plan into real-time control commands that can be executed by the terminal devices. These commands include photovoltaic inverter output adjustment commands, energy storage converter charge and discharge control commands, diesel generator start and stop commands, controllable load adjustment commands, and charge and discharge control commands. These commands are then sent to the terminal perception and execution layer via the communication network.

[0047] After receiving real-time control commands, the terminal perception and execution layer parses and executes the corresponding actions by the controllers of each device. After execution, the terminal layer feeds back the actual operating data to the edge layer in real time. The edge layer compares the actual values ​​with the target values ​​and performs closed-loop adjustment to ensure power balance and voltage and frequency stability. Thus, the system completes a full scheduling closed loop from data acquisition, cloud optimization, edge correction to terminal execution.

[0048] In summary: To address the issues of high control latency and the inability to balance global optimization with real-time response, existing solutions involve uploading all data to the cloud in a centralized architecture, resulting in latency reaching the minute level. Furthermore, the two-level optimization fails to achieve decoupling between the global and local layers. In contrast, this invention adopts a fully decoupled three-level architecture of cloud-edge-device, where the cloud layer is responsible for long-cycle global optimization, while the edge layer is responsible for short-cycle millisecond-level real-time control, reducing control latency to within 100ms.

[0049] To address the issues of low prediction accuracy and poor robustness, existing solutions employ a single model and single time scale for prediction, failing to consider meteorological abrupt changes and equipment degradation, resulting in biases exceeding 15%. In contrast, this invention constructs a combined model of LSTM, random forest, and attention mechanism, incorporating meteorological abrupt change / equipment degradation correction factors to achieve multi-scale predictions at the day-ahead / intraday / real-time scales, with an average absolute error ≤5% and a renewable energy consumption rate ≥98%.

[0050] To address the issues of high energy storage losses and short lifespan, existing optimization models only pursue the lowest operating cost and do not incorporate the energy storage's lifespan degradation. In contrast, this invention establishes a multi-objective optimization function that includes the energy storage's state of equilibrium (SOH) degradation cost. By improving the MPC algorithm for rolling solution, it avoids deep charging and discharging, extending the energy storage cycle life by more than 25%.

[0051] To address the issues of insufficient multi-source coordination and low reliability under extreme operating conditions, existing solutions only achieve basic coordination between source, storage, and grid, and are prone to power imbalance during grid-connected / islanded switching. In contrast, this invention achieves full coordination between source, grid, load, storage, and charging, supports seamless grid-connected / islanded switching and fault self-healing at the 10ms level, and achieves power supply reliability of over 99.99%.

[0052] To address the issue of lack of cluster collaboration and virtual power plant integration capabilities, existing solutions rely on individual microgrids operating independently and unable to participate in grid ancillary services. In contrast, this invention deploys a cluster collaborative scheduling module in the cloud to enable energy sharing among multiple microgrids. This allows for unified integration with virtual power plants to participate in peak shaving and frequency regulation, thereby generating ancillary service revenue.

[0053] The multi-source coordinated energy dispatching method for distributed photovoltaic-storage microgrids provided in this embodiment of the invention can significantly reduce overall electricity costs while ensuring power supply reliability through multi-source coordinated energy dispatching.

[0054] This invention also provides an implementation method for multi-source coordinated energy dispatching of distributed photovoltaic-storage microgrids, as detailed in (1) to (7) below: (1) Full-dimensional data acquisition and instruction execution preparation at the terminal perception and execution layer. Technical problems to be solved: Existing technologies have incomplete data acquisition dimensions and low sampling frequency, which cannot support millisecond-level real-time control and high-precision prediction logic association. It provides basic data input for all subsequent control and optimization links of this system, which is a prerequisite for realizing full coordination of source-grid-load-storage-charging.

[0055] (1-1) Full-Scenario Unit Deployment and Data Acquisition. Deploy intelligent sensing and control units covering distributed generation, energy storage, backup power, controllable loads, and grid interaction, with a sampling frequency set to 100Hz (most existing technologies use 1-10Hz); Acquisition content includes: voltage, current, power, energy storage SOC / SOH, photovoltaic irradiance, ambient temperature and humidity, equipment operating status, grid frequency, and other full-dimensional data; Key improvement: For the first time, V2G charging piles and industrial flexible loads are included in a unified data acquisition system, achieving full-process data coverage from source to grid to load to storage to charging; Difference from existing technologies: Existing technologies only collect basic data on photovoltaics, energy storage, and the grid, without covering flexible loads and V2G, and the sampling frequency cannot meet the millisecond-level control requirements.

[0056] (1-2) Terminal equipment initialization and instruction reception preparation. The distributed generation unit starts MPPT maximum power point tracking control, and the energy storage unit completes PCS self-test and charging / discharging preparation; the controllable load unit activates the flexible load interactive scheduling subunit and enters the adjustable state; all terminal equipment establishes communication connection with the edge computing gateway and waits for scheduling instructions.

[0057] (2) Local data preprocessing and real-time response initialization at the edge real-time control layer. Technical problem to be solved: Existing technologies that upload all data to the cloud result in high control latency and cannot cope with instantaneous power fluctuations. Logical association: Implement local data processing and local response, filter unnecessary data before uploading it to the cloud, and reduce communication latency from the source.

[0058] (2-1) Local data preprocessing and protocol conversion. Filtering, noise reduction, and outlier removal are performed sequentially on all dimensions of the running data to obtain preprocessed running data. Then, multi-protocol conversion is performed on the preprocessed running data to convert device data with different communication protocols into a unified format to obtain feature data.

[0059] In one implementation, the edge computing gateway filters, reduces noise, and removes outliers from the raw data uploaded by the terminal, and performs multi-protocol conversion such as Modbus, IEC61850, and MQTT to achieve unified access for devices from different manufacturers. Key improvement: more than 90% of the data preprocessing is completed at the edge layer, and only feature data is uploaded to the cloud. Difference from existing technologies: existing technologies directly upload all raw data to the cloud for processing, resulting in high communication bandwidth consumption and high latency.

[0060] (2-2) Local real-time controller initialization. The local real-time controller is set to a 50ms control cycle (most existing technologies use 1-5 minutes), starts the millisecond-level power balance and voltage / frequency stabilization control thread; loads the basic operating parameters and safety constraint thresholds sent from the cloud; and establishes a real-time data interaction channel with the mode switching control module and the fault self-healing control module.

[0061] (3) Cloud-based global optimization layer with multi-timescale prediction and global optimization scheduling. Technical problem solved: Existing technologies have low prediction accuracy and single optimization model, and do not consider the logical correlation of energy storage decay throughout its entire life cycle. In contrast, this invention generates a globally optimal scheduling plan based on historical and real-time data and distributes it to the edge layer as a benchmark for real-time control.

[0062] (3-1) Multi-source data fusion and database construction. Feature data is fused with historical operating data, meteorological data, electricity price data and equipment parameter data to construct a time series database. Based on the time series database, short-term forecasting processing for the previous 24 hours is performed to obtain the photovoltaic power output baseline curve and load demand baseline curve for the next day.

[0063] In one implementation, an attention mechanism, a long short-term memory network model, and a random forest model can be used to jointly predict the time-series database to obtain initial prediction results. These initial prediction results are then corrected based on pre-configured meteorological change correction factors and equipment degradation correction factors to obtain the photovoltaic output baseline curve and load demand baseline curve for the next day. The data management module receives feature data uploaded from the edge layer and integrates it with historical operating data, meteorological data, electricity price data, and equipment parameters to construct a time-series database that stores more than three years of comprehensive operating data, providing data support for prediction and optimization.

[0064] (3-2) Multi-model fusion and multi-timescale forecasting. The photovoltaic power output baseline curve and load demand baseline curve for the next day are processed by intraday 4-hour ultra-short-term forecasting with an hourly cycle to correct the forecast deviation. The photovoltaic power output baseline curve and load demand baseline curve for the next day after the forecast deviation correction are processed by rolling forecasting with a 15-minute cycle to respond to meteorological changes and obtain multi-timescale forecasting results.

[0065] In one implementation, the multi-timescale forecasting module performs three levels of forecasting: 1. 24-hour short-term forecasting: generating baseline curves for photovoltaic / wind power output and load demand for the next day; 2. 4-hour ultra-short-term forecasting: updated hourly to correct short-term forecasting biases; 3. Real-time 15-minute rolling forecasting: updated every 15 minutes to address sudden weather changes and other emergencies. Key improvements: A combined model of LSTM, random forest, and attention mechanism is adopted, incorporating weather change correction factors and equipment degradation correction factors; Differences from existing technologies: Existing technologies use a single model and single timescale forecasting, without considering the impact of weather changes and equipment aging, resulting in forecast biases exceeding 15%; this solution has an average absolute error of ≤5%.

[0066] (3-3) Multi-objective optimization solution for the entire life cycle of energy storage. The operating cost, energy storage health degradation cost, carbon emission cost and system load failure rate are used as optimization objectives to construct a comprehensive optimization objective function. The weight coefficients corresponding to each optimization objective are determined according to the current operating mode. Based on the weight coefficients, the comprehensive optimization objective function is solved in a rolling manner through an improved rolling time-domain control algorithm to obtain the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target and grid interaction power threshold for each time period, so as to generate a global scheduling plan.

[0067] In one implementation, the multi-objective optimization scheduling module establishes the following objective function: minF=ω1Cop+ω2Cbat+ω3Ccarbon+ω4Rloss Where F is the comprehensive optimization target value, ω1, ω2, ω3, and ω4 are the weight coefficients of each optimization target, Cop is the system operating cost, Cbat is the SOH degradation cost of the energy storage battery, Ccarbon is the carbon emission cost, and Rloss is the system load failure rate.

[0068] The weighting coefficients are dynamically adjusted according to the operating mode: Grid connection mode (prioritizing economic efficiency and energy storage lifespan): =0.4、 =0.3、 =0.15、 =0.15 Island mode (prioritizing power supply reliability): =0.2、 =0.2、 =0.1、 =0.5 An improved MPC algorithm (i.e., a rolling time-domain control algorithm) is used for rolling solution, and the single optimization time is <30s.

[0069] Key improvements: For the first time, the energy storage SOH decay cost, carbon emission cost, and load failure rate are simultaneously included in the optimization objectives; Difference from existing technologies: Existing technologies only pursue the lowest operating cost, resulting in frequent deep charging and discharging of energy storage, shortening the lifespan by more than 30%; This solution can extend the energy storage cycle life by more than 25%.

[0070] (3-4) Global scheduling plan generation and distribution. Generate the 24-hour scheduling plan and the intraday rolling optimization plan, and distribute them to the edge real-time control layer. The plan content includes: photovoltaic output targets for each time period, energy storage charging and discharging power, controllable load adjustment, and grid interaction power threshold.

[0071] (4) Edge layer scheduling plan rolling correction and real-time control execution. Technical problem to be solved: The existing technology does not decouple global optimization and real-time control, and cannot take into account both optimization accuracy and response speed. Logical connection: Based on the global plan in the cloud, local correction is performed in combination with real-time running data to achieve the unity of global optimization and real-time response.

[0072] (4-1) Local rolling correction of the scheduling plan. The global scheduling plan is corrected in 10-minute cycles based on full-dimensional operational data and multi-time-scale prediction results to obtain the corrected scheduling plan.

[0073] In one implementation, the local real-time controller performs a rolling correction of the scheduling plan issued by the cloud every 10 minutes; the correction is based on the latest real-time running data, the 15-minute rolling prediction results, and the current system status; the key improvement is to achieve complete decoupling between global optimization and real-time control, with the cloud responsible for long-term large-scale optimization and the edge responsible for short-term small-scale correction.

[0074] (4-2) Real-time control command generation and issuance. With a control cycle of fifty milliseconds, real-time control commands corresponding to each target are generated based on the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold in each time period of the revised scheduling plan. The real-time control commands include: photovoltaic inverter output adjustment command, energy storage converter charging and discharging control command, diesel generator start and stop command, controllable load adjustment command, and charging and discharging control command.

[0075] In one implementation, the local real-time controller generates 50ms-level real-time control commands based on the revised plan; the commands include: photovoltaic inverter output adjustment, energy storage PCS charging and discharging control, diesel generator start-stop, controllable load adjustment, and V2G charging and discharging control; the commands are sent to the terminal sensing and execution layer and executed by each unit.

[0076] (4-3) Closed-loop feedback of operating status. After the terminal layer executes the command, it feeds back the actual operating data to the edge layer in real time; the edge layer compares the actual value with the target value and performs closed-loop control adjustment to ensure power balance and voltage / frequency stability.

[0077] (5) Full-mode adaptive switching and fault self-healing control. Technical problems to be solved: Existing technologies are prone to power imbalance during grid-connected / islanded switching, have slow fault handling speed, and low reliability under extreme conditions. Logical connection: Ensuring the continuous and stable operation of the system under different operating modes and fault states is the core link to improve power supply reliability.

[0078] (5-1) Real-time monitoring and identification of operating modes. The mode switching control module monitors the grid voltage and frequency status in real time, with a sampling frequency of 1kHz; it automatically identifies three operating modes: grid-connected operation, islanded operation, and grid-connected-islanded switching.

[0079] In one implementation, the grid voltage and frequency status are monitored and processed in real time at a preset sampling frequency, and the current operating mode is automatically identified based on the monitoring results. The current operating mode is any one of the following: grid-connected operation mode, islanded operation mode, and grid-to-island switching mode. When the current operating mode is identified as grid-to-island switching mode, pre-synchronization seamless switching processing is performed to complete the switching between grid-connected mode and islanded mode within a preset time interval.

[0080] (5-2) Full-mode adaptive control. Grid-connected mode: prioritizes economy, implements peak-valley electricity price arbitrage, and controls grid interaction power within ±1.5MW; Islanding mode: prioritizes reliability, with energy storage and backup power supply working together to provide voltage / frequency support, and prioritizes power supply to primary loads; Grid-connected to islanding switching: adopts pre-synchronous seamless switching technology, switching time <10ms, with no power surge or voltage fluctuation; Difference from existing technologies: Existing technologies have switching times of 100-500ms, which can easily lead to load power outages and equipment damage.

[0081] (5-3) Fault self-healing control. The fault self-healing control module monitors system equipment and power grid faults in real time. When a fault occurs, it can quickly locate the fault area within 10ms, perform fault isolation, and reduce power supply according to the importance level of the load. Priority is given to ensuring the power supply of important loads such as hospitals and data centers. System power reconfiguration is performed to restore power supply to non-faulty areas. The difference from the existing technology is that the fault handling time of the existing technology is mostly in the minute range and power reconfiguration cannot be achieved.

[0082] (6) Multi-microgrid cluster collaboration and virtual power plant docking. Technical problem to be solved: Existing technologies allow individual microgrids to operate independently, which cannot achieve regional energy mutual assistance and logical association with grid auxiliary services; to achieve global optimization of regional energy resources and improve the overall economic efficiency and grid friendliness of the system.

[0083] (6-1) Multi-microgrid cluster data interaction. The cloud-based cluster collaborative scheduling module connects to multiple adjacent microgrids within the region; it collects data such as the operating status, surplus power, and power shortage status of each microgrid in real time.

[0084] (6-2) Cluster joint optimization scheduling. Construct a multi-microgrid joint optimization scheduling model to achieve peak-valley complementarity and energy mutual assistance among microgrids; surplus renewable energy is consumed locally within the cluster to avoid curtailment of solar and wind power.

[0085] (6-3) Virtual power plant docking and ancillary service participation. Unified docking with the regional virtual power plant platform, aggregating the adjustable resources of all microgrids; responding to grid peak shaving, frequency regulation, and demand response commands, and participating in grid ancillary services; ancillary service revenue is distributed according to the regulation contribution of each microgrid; difference from existing technologies: existing technologies only support the independent operation of a single microgrid and cannot participate in grid ancillary services to obtain additional revenue.

[0086] (7) Full lifecycle visualized operation and maintenance management. Technical problem solved: Existing technology operation and maintenance management is extensive and lacks logical connection for full lifecycle equipment health monitoring: Provide operation and maintenance support for long-term stable operation of the system and reduce operation and maintenance costs.

[0087] (7-1) Real-time monitoring of system operation status. The visual operation and maintenance module displays the real-time operation status of the system, power flow, and output of each unit through a web interface; and provides real-time early warnings for voltage / frequency exceeding limits, equipment abnormalities, and other situations.

[0088] (7-2) Equipment health status monitoring. Based on equipment operation data and historical fault data, assess the health status of the equipment; predict the remaining service life of the equipment, and develop a preventive maintenance plan.

[0089] (7-3) Operational data statistical analysis and report generation. Automatically calculates key indicators such as renewable energy absorption rate, electricity cost, energy storage attenuation rate, and power supply reliability; generates daily, monthly, and annual reports to provide data support for user decision-making.

[0090] In practical applications, the above process has been verified in three typical scenarios: industrial parks, remote areas, and urban commercial districts. The specific results are as follows: Industrial park grid connection scenario: control latency ≤80ms, prediction error 4.2%, electricity cost reduction 32.6%, and energy storage life extension 28.3%.

[0091] In remote, isolated scenarios: power supply reliability reaches 99.98%, and the lifespan of hybrid energy storage is extended by 35%.

[0092] Commercial district cluster scenario: regional absorption rate of 99.5%, electricity cost reduction of 37.2%, and acquisition of grid ancillary service revenue.

[0093] In summary, this invention addresses the high control latency issue of existing centralized architectures by employing a layered processing architecture of data acquisition, preprocessing, cloud optimization, and edge correction. This allows the terminal to complete data preprocessing locally, the cloud to execute global optimization to generate a scheduling plan, and the edge layer to perform real-time control with a 50-millisecond cycle. This completely decouples global optimization from real-time control, reducing control latency and grid-connected / islanded switching time, and improving power supply reliability.

[0094] To address the problem of low prediction accuracy in existing methods, this invention constructs a time-series database by fusing multi-source data, and sequentially performs three-level progressive predictions: 24 hours before the current day, 4 hours within the day, and 15 minutes in real time. It also employs a joint prediction model combining attention mechanism, long short-term memory network, and random forest, supplemented by dynamic correction factors for meteorological abrupt changes and equipment degradation, thereby reducing the average absolute error of prediction and improving the renewable energy absorption rate.

[0095] To address the problem of severe energy storage losses caused by the single optimization objective in existing systems, this invention constructs a comprehensive optimization objective function based on four indicators: operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate. The weight coefficients of each objective are dynamically adjusted according to the grid-connected or islanded operation mode. Combined with an improved rolling time-domain control algorithm, the solution is solved in a rolling manner, avoiding deep charging and discharging of energy storage, extending cycle life, and reducing overall electricity costs.

[0096] To address the issue of insufficient scope of existing collaboration, this invention achieves unified access for diverse devices through multi-protocol conversion. Combined with full-dimensional data collection and full-mode adaptive switching, it realizes stable operation in both grid-connected and islanded modes across all scenarios, providing an architectural foundation for subsequent multi-microgrid cluster collaboration and virtual power plant docking.

[0097] Regarding the multi-source coordinated energy dispatching method for distributed photovoltaic-storage microgrids provided in the foregoing embodiments, this invention provides a multi-source coordinated energy dispatching device for distributed photovoltaic-storage microgrids. (See [link]) Figure 3 The diagram shows a multi-source coordinated energy dispatching device for a distributed photovoltaic-storage microgrid. The device includes the following components: The data acquisition module 302 acquires and processes data from the microgrid at a preset sampling frequency to obtain full-dimensional operational data, and performs local preprocessing on the full-dimensional operational data to obtain preprocessed feature data. The global scheduling module 304 performs multi-timescale prediction processing based on feature data and historical operating data to obtain multi-timescale prediction results. Based on the multi-timescale prediction results, it determines the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate. It then performs multi-objective optimization processing with the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate as optimization objectives to generate a global scheduling plan. The energy dispatch module 306 performs rolling correction processing on the global dispatch plan based on a preset control cycle and generates real-time control commands to complete the energy dispatch of the microgrid by executing the real-time control commands.

[0098] The distributed photovoltaic-storage microgrid multi-source coordinated energy dispatching device provided in this application embodiment can significantly reduce the overall electricity cost while ensuring power supply reliability through multi-source coordinated energy dispatching.

[0099] In one embodiment, when performing local preprocessing of the full-dimensional operational data to obtain preprocessed feature data, the data acquisition module 302 is further configured to: sequentially perform filtering, noise reduction, and outlier removal processing on the full-dimensional operational data to obtain preprocessed operational data; and perform multi-protocol conversion processing on the preprocessed operational data to convert device data of different communication protocols into a unified format to obtain feature data.

[0100] In one embodiment, when performing the step of multi-timescale prediction processing based on feature data and historical operating data to obtain multi-timescale prediction results, the global scheduling module 304 is further configured to: fuse feature data with historical operating data, meteorological data, electricity price data, and equipment parameter data to construct a time-series database; and based on the time-series database, perform short-term prediction processing for the previous 24 hours to obtain the photovoltaic output baseline curve and load demand baseline curve for the next day; perform intraday 4-hour ultra-short-term prediction processing on the photovoltaic output baseline curve and load demand baseline curve for the next day with an hourly cycle to correct the prediction deviation; and perform rolling prediction processing on the photovoltaic output baseline curve and load demand baseline curve for the next day after the prediction deviation correction with a 15-minute cycle to respond to and correct for sudden meteorological changes, thereby obtaining multi-timescale prediction results.

[0101] In one embodiment, when performing short-term forecasting processing based on a time-series database to obtain the photovoltaic power output baseline curve and load demand baseline curve for the next day, the aforementioned global scheduling module 304 is further configured to: perform joint forecasting processing on the time-series database using an attention mechanism, a long short-term memory network model, and a random forest model to obtain an initial forecast result; and correct the initial forecast result according to a pre-configured meteorological change correction factor and equipment attenuation correction factor to obtain the photovoltaic power output baseline curve and load demand baseline curve for the next day.

[0102] In one embodiment, during the step of performing multi-objective optimization with operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate as optimization targets to generate a global scheduling plan, the global scheduling module 304 is further configured to: construct a comprehensive optimization objective function with operating cost, energy storage health degradation cost, carbon emission cost, and system load failure rate as optimization targets; determine the weight coefficients corresponding to each optimization target according to the current operating mode; and perform rolling solution processing on the comprehensive optimization objective function based on the weight coefficients using an improved rolling time-domain control algorithm to obtain the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period, so as to generate a global scheduling plan.

[0103] In one embodiment, when performing the step of rolling correction of the global scheduling plan based on a preset control cycle to generate real-time control instructions, the energy scheduling module 306 is further configured to: correct the global scheduling plan based on full-dimensional operating data and multi-time-scale prediction results with a 10-minute cycle to obtain a corrected scheduling plan; and generate real-time control instructions corresponding to each target based on the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period in the corrected scheduling plan with a 50-millisecond control cycle. The real-time control instructions include: photovoltaic inverter output adjustment instructions, energy storage converter charging and discharging control instructions, diesel generator start / stop instructions, controllable load adjustment instructions, and charging and discharging control instructions.

[0104] In one embodiment, after generating real-time control commands, the energy dispatch module 306 is further configured to: perform real-time monitoring and processing of the grid voltage state and grid frequency state at a preset sampling frequency, and automatically identify the current operating mode based on the monitoring results, wherein the current operating mode is any one of grid-connected operating mode, islanded operating mode, and grid-to-island switching mode; when the current operating mode is identified as grid-to-island switching mode, perform pre-synchronization seamless switching processing, and complete the switching between grid-connected mode and islanded mode within a preset time interval.

[0105] The device provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0106] This invention provides a server, specifically, the server includes a processor and a storage device; the storage device stores a computer program, which, when run by the processor, executes the method described in any of the above embodiments.

[0107] Figure 4 This is a schematic diagram of the structure of a server provided in an embodiment of the present invention. The server 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43. The processor 40, the communication interface 43 and the memory 41 are connected through the bus 42. The processor 40 is used to execute executable modules, such as computer programs, stored in the memory 41.

[0108] The memory 41 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 43 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0109] Bus 42 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0110] The memory 41 is used to store programs. After receiving an execution instruction, the processor 40 executes the program. The method executed by the device for defining the flow process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0111] Processor 40 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 40 or by instructions in software form. Processor 40 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 41. The processor 40 reads the information in memory 41 and, in conjunction with its hardware, completes the steps of the above method.

[0112] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0113] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, 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 the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids, the method being applied to a multi-source coordinated energy dispatch system for distributed photovoltaic-storage microgrids, characterized in that, The method includes: Data is collected and processed from the microgrid at a preset sampling frequency to obtain full-dimensional operational data, and the full-dimensional operational data is preprocessed locally to obtain preprocessed feature data. Based on the feature data and historical operating data, multi-timescale prediction processing is performed to obtain multi-timescale prediction results. Based on the multi-timescale prediction results, operating costs, energy storage health status decay costs, carbon emission costs, and system load failure rates are determined. Multi-objective optimization is performed with the operating costs, energy storage health status decay costs, carbon emission costs, and system load failure rates as optimization objectives to generate a global scheduling plan. The global scheduling plan is rolled over based on a preset control cycle to generate real-time control commands, which are then executed to complete the energy scheduling of the microgrid.

2. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 1, characterized in that, The step of performing local preprocessing on the full-dimensional operational data to obtain preprocessed feature data includes: The full-dimensional operational data is sequentially subjected to filtering, noise reduction, and outlier removal to obtain preprocessed operational data. The preprocessed running data is subjected to multi-protocol conversion processing to convert device data with different communication protocols into a unified format to obtain the feature data.

3. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 1, characterized in that, The step of performing multi-time-scale prediction processing based on the feature data and historical operating data to obtain multi-time-scale prediction results includes: The characteristic data is fused with the historical operating data, meteorological data, electricity price data and equipment parameter data to construct a time series database. Based on the time series database, short-term forecasting processing for the previous 24 hours is performed to obtain the photovoltaic output baseline curve and load demand baseline curve for the next day. The photovoltaic power output baseline curve and the load demand baseline curve for the next day are subjected to intraday 4-hour ultra-short-term forecasting processing on an hourly basis to correct the forecast deviation. Then, the photovoltaic power output baseline curve and the load demand baseline curve for the next day after the forecast deviation correction are subjected to rolling forecasting processing on a 15-minute basis to respond to meteorological changes and obtain the multi-time scale forecasting results.

4. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 3, characterized in that, The step of performing short-term forecasting processing for the previous 24 hours based on the time-series database to obtain the photovoltaic output baseline curve and load demand baseline curve for the next day includes: The time-series database is jointly predicted using an attention mechanism, a long short-term memory network model, and a random forest model to obtain initial prediction results. The initial forecast results are corrected based on the pre-configured meteorological change correction factor and equipment attenuation correction factor to obtain the photovoltaic output baseline curve and the load demand baseline curve for the next day.

5. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 1, characterized in that, The step of performing multi-objective optimization to generate a global scheduling plan, with the operating cost, the energy storage health status degradation cost, the carbon emission cost, and the system load failure rate as optimization objectives, includes: The operating cost, the energy storage health status degradation cost, the carbon emission cost, and the system load failure rate are used as optimization targets to construct a comprehensive optimization objective function; Based on the current operating mode, the weight coefficients corresponding to each optimization objective are determined. Then, using an improved rolling time-domain control algorithm, the comprehensive optimization objective function is solved on a rolling basis using the weight coefficients to obtain the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period, so as to generate the global scheduling plan.

6. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 1, characterized in that, The step of performing rolling correction processing on the global scheduling plan based on a preset control period to generate real-time control commands includes: Using a 10-minute cycle, the global scheduling plan is modified based on the full-dimensional operational data and the multi-time-scale prediction results to obtain the modified scheduling plan. With a control cycle of fifty milliseconds, the real-time control instructions corresponding to each target are generated based on the photovoltaic output target, energy storage charging and discharging power target, controllable load adjustment target, and grid interaction power threshold for each time period in the revised scheduling plan. The real-time control instructions include: photovoltaic inverter output adjustment instructions, energy storage converter charging and discharging control instructions, diesel generator start and stop instructions, controllable load adjustment instructions, and charging and discharging control instructions.

7. The multi-source coordinated energy dispatch method for distributed photovoltaic-storage microgrids according to claim 1, characterized in that, Following the step of generating real-time control commands, the following is included: The grid voltage and frequency status are monitored and processed in real time at a preset sampling frequency, and the current operating mode is automatically identified based on the monitoring results. The current operating mode is any one of the following: grid-connected operation mode, islanded operation mode, and grid-connected to islanded switching mode. When the current operating mode is identified as the grid-connected to islanded switching mode, a pre-synchronization seamless switching process is performed to complete the switching between the grid-connected mode and the islanded mode within a preset time interval.

8. A multi-source coordinated energy dispatching device for a distributed photovoltaic-storage microgrid, the device being applied to a multi-source coordinated energy dispatching system for a distributed photovoltaic-storage microgrid, characterized in that, The device includes: The data acquisition module acquires and processes data from the microgrid at a preset sampling frequency to obtain full-dimensional operational data, and performs local preprocessing on the full-dimensional operational data to obtain preprocessed feature data. The global scheduling module performs multi-timescale prediction processing based on the feature data and historical operating data to obtain multi-timescale prediction results. Based on the multi-timescale prediction results, it determines the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate. It then performs multi-objective optimization processing with the operating cost, energy storage health status decay cost, carbon emission cost, and system load failure rate as optimization objectives to generate a global scheduling plan. The energy dispatch module performs rolling correction processing on the global dispatch plan based on a preset control cycle, generates real-time control commands, and completes the energy dispatch of the microgrid by executing the real-time control commands.

9. A server, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images