A method and system for optical power storage allocation control

Abstract

The application provides a kind of light storage power distribution control method and system, calibrates photovoltaic operation data, energy storage operation data, bus operation data, obtains homologous calibration data;Voltage safety constraint domain is generated based on homologous calibration data;Based on homologous calibration data and voltage safety constraint domain, predict photovoltaic output value and energy storage state of charge value;Based on the prediction result and voltage safety constraint domain, three-level multi-objective optimization is executed, photovoltaic power instruction and energy storage power instruction are obtained and issued to execution unit, so that execution unit carries out power distribution according to photovoltaic power instruction and the energy storage power instruction.Power distribution error is minimized, and bus voltage is most stable.

Description

Technical Field

[0001] This application relates to the field of high-proportion photovoltaic grid-connected collaborative control technology, and in particular to a photovoltaic power distribution control method and system. Background Technology

[0002] Under the "dual carbon" target, the output of large-capacity centralized photovoltaic power plants fluctuates randomly, which can easily lead to power fluctuations, voltage overruns and high curtailment rates, restricting the consumption of new energy and grid security. Therefore, it is urgent to optimize power allocation.

[0003] Currently, conventional techniques often employ intelligent optimization algorithms such as particle swarm optimization and genetic algorithms to optimize power allocation schemes for photovoltaic-storage systems. When dealing with multi-constraint, non-convex power allocation models, these algorithms typically use a weighted summation approach, transforming multiple objectives such as curtailment rate, power allocation accuracy, and bus voltage offset into a single objective for solution. However, when power fluctuations are severe, these algorithms struggle to simultaneously balance convergence speed and computational diversity, resulting in a power allocation scheme that fails to simultaneously achieve the optimal balance between minimizing curtailment rate, minimizing power allocation error, and maximizing bus voltage stability. Summary of the Invention

[0004] Based on the shortcomings of the prior art, this application provides a photovoltaic power allocation control method and system to solve the problem that the obtained power allocation command cannot simultaneously achieve the three objectives of minimizing the curtailment rate, minimizing the power allocation error, and maximizing the stability of the bus voltage.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] The first aspect of this application provides a photovoltaic-storage power distribution control method, applied to a photovoltaic-storage power distribution control system, comprising:

[0007] Collect photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data;

[0008] The photovoltaic operation data, the energy storage operation data, and the bus operation data are calibrated to obtain homogeneous calibration data;

[0009] A voltage safety constraint domain is generated based on the aforementioned homogeneous calibration data;

[0010] Based on the same source calibration data and the voltage safety constraint domain, the photovoltaic output value and the energy storage state of charge value are predicted, and the prediction results are obtained.

[0011] Based on the prediction results and the voltage safety constraint domain, a three-level multi-objective optimization is performed to obtain photovoltaic power commands and energy storage power commands; wherein, the objectives of the multi-objective optimization are to minimize the curtailment rate, minimize the power allocation error, and maximize the bus voltage stability while satisfying the voltage safety constraint domain.

[0012] The photovoltaic power command and the energy storage power command are sent to the execution unit so that the execution unit can allocate power according to the photovoltaic power command and the energy storage power command.

[0013] Optionally, the step of calibrating the photovoltaic operation data, the energy storage operation data, and the bus operation data to obtain homogeneous calibration data includes:

[0014] A unified clock synchronization mechanism is used to add timestamps to the photovoltaic operation data, the energy storage operation data, and the bus operation data respectively, resulting in timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data.

[0015] The time-series deviation of the timestamped photovoltaic data, the timestamped energy storage data, and the timestamped bus data is calibrated using a software phase compensation algorithm to obtain the homogeneous calibration data.

[0016] Optionally, generating the voltage safety constraint domain based on the homogeneous calibration data includes:

[0017] An electrical topology model is established with photovoltaic units, energy storage units, busbars and grid connection points as network nodes and collector lines and transformer windings as topology branches.

[0018] Power flow calculations are performed based on the electrical topology model and the real-time operating parameters in the same-source calibration data to obtain the voltage safety constraint domain; the voltage safety constraint domain includes bus voltage constraints, branch power constraints, and energy storage state of charge constraints.

[0019] Optionally, the prediction of photovoltaic output and energy storage state of charge based on the same source calibration data and the voltage safety constraint domain, to obtain the prediction results, includes:

[0020] Historical irradiance sequence, ambient temperature sequence, historical power sequence, and energy storage state of charge sequence are extracted from the homogeneous calibration data;

[0021] Obtain the weather forecast sequence, and generate a prediction input sequence based on the historical irradiance sequence, the ambient temperature sequence, the historical power sequence, the energy storage state of charge sequence, and the weather forecast sequence;

[0022] The predicted input sequence is input into the spatiotemporal prediction model; the spatiotemporal prediction model includes a temporal convolutional network, a self-attention mechanism, and an adaptive fusion layer.

[0023] The short-term fluctuation features of the predicted input sequence are extracted through the temporal convolutional network to obtain a short-term feature vector;

[0024] The long-term dependency features of the predicted input sequence are extracted through the self-attention mechanism to obtain a long-term feature vector;

[0025] The short-term feature vector and the long-term feature vector are weighted and fused through the adaptive fusion layer to obtain the initial prediction result;

[0026] Based on the voltage safety constraint domain, the photovoltaic output value and energy storage state of charge value in the initial prediction result are boundary-corrected to obtain the prediction result.

[0027] Optionally, the step of performing three-level multi-objective optimization based on the prediction results and the voltage safety constraint domain to obtain photovoltaic power commands and energy storage power commands includes:

[0028] At the day-ahead level, the prediction results and the voltage safety constraint domain are optimized using an optimization algorithm to obtain the day-ahead plan; the day-ahead plan includes the day-ahead photovoltaic reference output curve and the day-ahead energy storage charge and discharge plan;

[0029] At the intraday level, the photovoltaic operation data and energy storage state of charge value at the current moment are extracted from the same source calibration data. Based on the day-ahead plan and combined with the voltage safety constraint domain, the plan is rolled over and corrected using the optimization algorithm to obtain the intraday correction plan. The intraday correction plan includes the corrected photovoltaic output plan and energy storage charge and discharge plan.

[0030] At the real-time layer, the actual grid-connected power value and the actual bus voltage value are collected, and the intraday correction plan is finely adjusted in a closed loop using the actual grid-connected power value and the actual bus voltage value to obtain the photovoltaic power command and the energy storage power command.

[0031] Optionally, the step of performing closed-loop fine-tuning of the intraday correction plan using the actual grid-connected power value and the actual bus voltage value to obtain photovoltaic power command and energy storage power command includes:

[0032] The actual grid-connected power value is compared with the power reference value corresponding to the intraday correction plan to obtain the total power deviation.

[0033] The actual bus voltage value is compared with the voltage safety constraint domain to obtain the bus voltage deviation;

[0034] The total power deviation and the bus voltage deviation are finely adjusted in a closed loop using an optimization algorithm to obtain the photovoltaic power command and the energy storage power command.

[0035] Optionally, the collection of photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data includes:

[0036] Photovoltaic operation data, energy storage operation data, and bus operation data are collected through the first physically isolated communication link;

[0037] Scheduling instruction data is received through a second physically isolated communication link; wherein, there is no common electrical ground and no clock bus coupling between the first and second physically isolated communication links.

[0038] A second aspect of this application provides a photovoltaic power distribution control system, comprising:

[0039] Dual redundant isolated communication units are used to collect photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data;

[0040] The time-series homogeneous sensing unit is used to perform data calibration on the photovoltaic operation data, the energy storage operation data, and the bus operation data to obtain homogeneous calibration data;

[0041] A photovoltaic-storage-bus electrical coupling identification unit is used to generate a voltage safety constraint domain based on the homogeneous calibration data.

[0042] A multi-timescale intelligent power allocation unit is used to predict photovoltaic output and energy storage state of charge based on the same source calibration data and the voltage safety constraint domain, and obtain prediction results; it is also used to perform three-level multi-objective optimization based on the prediction results and the voltage safety constraint domain to obtain photovoltaic power command and energy storage power command; wherein, the objectives of the multi-objective optimization are to minimize the curtailment rate, minimize the power allocation error, and maximize the stability of the bus voltage while satisfying the voltage safety constraint domain.

[0043] A synchronous instruction execution unit is used to send the photovoltaic power instruction and the energy storage power instruction to the execution unit so that the execution unit can perform power allocation according to the photovoltaic power instruction and the energy storage power instruction.

[0044] Optionally, the temporal co-origin sensing unit is specifically used for:

[0045] A unified clock synchronization mechanism is used to add timestamps to the photovoltaic operation data, the energy storage operation data, and the bus operation data respectively, resulting in timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data.

[0046] The time-series deviation of the timestamped photovoltaic data, the timestamped energy storage data, and the timestamped bus data is calibrated using a software phase compensation algorithm to obtain the homogeneous calibration data.

[0047] Optionally, the photovoltaic-storage-bus electrical coupling identification unit is specifically used for:

[0048] An electrical topology model is established with photovoltaic units, energy storage units, busbars and grid connection points as network nodes and collector lines and transformer windings as topology branches.

[0049] Power flow calculations are performed based on the electrical topology model and the real-time operating parameters in the same-source calibration data to obtain the voltage safety constraint domain; the voltage safety constraint domain includes bus voltage constraints, branch power constraints, and energy storage state of charge constraints.

[0050] This application provides a photovoltaic (PV) power allocation control method and system. The method involves calibrating PV operation data, energy storage operation data, and bus operation data to obtain homogeneous calibration data. A voltage safety constraint domain is generated based on this data. Based on the homogeneous calibration data and the voltage safety constraint domain, PV output and energy storage state of charge (SOC) values ​​are predicted to obtain prediction results. Based on the prediction results and the voltage safety constraint domain, a three-level multi-objective optimization is performed to obtain PV power commands and energy storage power commands. The objectives of the multi-objective optimization are to minimize the curtailment rate, minimize power allocation error, and maximize bus voltage stability. The PV power commands and energy storage power commands are then sent to the execution unit. By performing time-series calibration on multi-source data to generate homogeneous calibration data, and performing a three-level multi-objective optimization based on the voltage safety constraint domain and prediction results, the optimization task is decomposed at different time scales, effectively balancing convergence speed and computational diversity. This overcomes the shortcomings of traditional weighted summation methods in simultaneously achieving optimal multi-objective optimization, thus achieving a synergistic optimization of minimizing the curtailment rate, minimizing power allocation error, and maximizing bus voltage stability. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0052] Figure 1 A flowchart of a power allocation control method for photovoltaic storage provided in this application embodiment;

[0053] Figure 2 A schematic diagram comparing the total light waste curves before and after optimized control, provided for an embodiment of this application;

[0054] Figure 3 A functional module and data interaction block diagram of a power distribution control system provided in this application embodiment;

[0055] Figure 4 This is a schematic diagram of the architecture of a photovoltaic power distribution control system provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0058] like Figure 1 As shown in the figure, this application provides a flowchart of a photovoltaic-storage power allocation control method. This method can be applied to a photovoltaic-storage power allocation control system, which includes a dual-redundant isolated communication unit, a timing-based sensing unit, a photovoltaic-storage-bus electrical coupling identification unit, a multi-time-scale intelligent power allocation unit, and a synchronization command execution unit. Specifically, it includes steps S101 to S106:

[0059] It should be noted that the power distribution control system provided in this embodiment is applied to a large-scale photovoltaic power plant. This power plant has an installed capacity of 1GW of photovoltaic power, supported by a 100MW / 200MWh electrochemical energy storage system. The collection side includes 40 35kV photovoltaic collector lines and 4 35kV / 220kV main transformers, and is equipped with multiple 35kV busbars and a 220kV grid-connected busbar. Before the upgrade, the original system of this power plant had prominent problems such as large power distribution deviations and significant fluctuations in grid-connected power. After adopting the system of this embodiment, precise power distribution between energy storage, photovoltaics, and the busbar can be achieved, ensuring the safe and stable operation of the system under all operating conditions.

[0060] In this embodiment, the power distribution control system adopts industrial-grade standardized edge control hardware as its physical carrier. Its core is equipped with the AIMB-708 high-performance industrial motherboard, which possesses strong computing power, high reliability, and abundant expansion interfaces, enabling long-term stable operation in the complex field environment of photovoltaic power plants. Its specific hardware configuration is as follows:

[0061] In terms of core processing and basic configuration, the motherboard uses an LGA1700 socket, supports Intel 12th / 13th generation Core series processors, and is equipped with an Intel® H610E PCH chipset; it supports dual-channel DDR4 3200 industrial-grade memory with a maximum capacity of 64GB; it is equipped with one M.2 PCIe Gen3 solid-state storage interface and four SATA3.0 high-speed storage interfaces to meet the storage and backup needs of operating systems, models, historical data, and policy logs; it also has PCIe×16, PCIe×4, and PCI expansion slots, which can be used to expand functional modules as needed.

[0062] In terms of communication interface configuration, the motherboard integrates two independent Gigabit Ethernet interfaces. LAN1 uses an Intel® I219-V controller, supporting speeds of 10 / 100 / 1000Mbps, while LAN2 uses an Intel® I226-V controller, supporting speeds of 10 / 100 / 1000 / 2500Mbps. These two ports can achieve physically isolated communication, used for field data acquisition, command issuance, and interaction with the dispatch platform, respectively. Simultaneously, the motherboard is equipped with multiple serial ports. COM1, COM4 through COM6 are RS-232 interfaces, and COM3 can be configured for RS-232 / 422 / 485 adaptive mode, compatible with diverse devices such as field inverters, energy storage converters, transformer substations, and measurement and control meters. It also features multiple USB 3.2 Gen1 and USB 2.0 interfaces for debugging, external device connection, and data export.

[0063] In terms of display and reliability hardware, the motherboard integrates a CPU core graphics controller, supports dual display output of HDMI and VGA, and supports a maximum resolution of 4K; it has a built-in hardware watchdog timer with multi-level timeout reset capability, which can automatically reset when the system is abnormal; the hardware operating temperature range is 0℃ to 60℃, which fully meets the harsh environmental requirements of the power plant control room for long-term uninterrupted operation.

[0064] In terms of power supply and structural design, the system adopts a standard industrial power supply mode, supports wide voltage input, has a standardized overall size and compact structure, strong anti-electromagnetic interference capability, and can be directly installed in the substation cabinet of the power station.

[0065] Based on the aforementioned hardware platform, the system adopts a modular software architecture and is developed based on the Ubuntu embedded operating system. All functional modules are designed in an integrated manner around the goals of optical-storage collaborative optimization, precise power allocation, and minimization of curtailment rate. Each unit interacts with the other through standardized data interfaces.

[0066] S101: Collects photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data.

[0067] Among them, photovoltaic operation data, energy storage operation data, bus operation data and dispatch command data are collected through dual redundant isolated communication units.

[0068] Optionally, in another embodiment of this application, the specific implementation of step S101 includes processes A1 to A2.

[0069] A1: Collect photovoltaic operation data, energy storage operation data, and bus operation data through the first physically isolated communication link.

[0070] Specifically, the first physically isolated communication link is one of the two independent Ethernet channels configured in the dual-redundant isolated communication unit. This link is dedicated to connecting all on-site photovoltaic subarrays, energy storage converters, transformer substations, combiner boxes, and measurement and control meters, collecting multi-source operational data in real time, including photovoltaic output, energy storage state of charge (SOC), energy storage charging and discharging power, bus voltage, bus current, branch power flow, and grid connection point power. This link uses an independent physical interface and transmission medium to ensure that the data acquisition process is not interfered with by other communication services.

[0071] A2: Receive scheduling instruction data through the second physically isolated communication link.

[0072] Among them, the first physically isolated communication link and the second physically isolated communication link have no common electrical ground and no clock bus coupling.

[0073] It should be noted that the second physically isolated communication link is also another independent Ethernet channel configured in the dual-redundant isolated communication unit. This link is specifically used to interact with the upper-level dispatch center or station control management layer, and is responsible for receiving dispatch command data such as the total active power command and voltage regulation command issued by the dispatch center. It can also be used to upload the system's own operating status and grid connection information and to issue power commands.

[0074] Understandably, the two communication links are completely physically isolated in terms of electrical interfaces, transmission cables, grounding systems, and clock sources. There is no electrical grounding connection between them, nor is there clock bus coupling. This physically eliminates the risk of electromagnetic interference crosstalk and the spread of single-point faults, fully meeting the stringent requirements of power systems for communication security and reliability.

[0075] S102: Perform data calibration on photovoltaic operation data, energy storage operation data, and bus operation data to obtain homogeneous calibration data.

[0076] Among them, the photovoltaic operation data, energy storage operation data, and bus operation data are calibrated through the time-series homogeneous sensing unit.

[0077] Understandably, since photovoltaic operation data, energy storage operation data, and bus operation data come from different acquisition terminals and communication links, there are inherent timing deviations between the data sources, such as inconsistent sampling times and asynchronous transmission delays. The role of the timing-consistent sensing unit is to eliminate these deviations, unify all data under the same time base, and provide timing-consistent data input for subsequent topology identification, power flow calculation, and power optimization.

[0078] Optionally, in another embodiment of this application, the specific implementation of step S102 includes processes B1 to B2.

[0079] B1: A unified clock synchronization mechanism is used to add timestamps to photovoltaic operation data, energy storage operation data, and bus operation data respectively, resulting in timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data.

[0080] Among them, the time-series co-source sensing unit has a built-in high-precision time synchronization module, which receives standard clock signals from the Global Positioning System or the BeiDou satellite system to provide a unified time reference.

[0081] Understandably, each communication link of the dual-redundant isolated communication unit independently performs standard time synchronization and local timekeeping functions. When data packets uploaded by each acquisition terminal arrive, the time-series co-source sensing unit immediately marks each frame of data with a precise arrival timestamp, thereby obtaining raw data with a unified clock identifier.

[0082] B2: Timing deviation calibration is performed on time-stamped photovoltaic data, time-stamped energy storage data, and time-stamped bus data using a software phase compensation algorithm to obtain homogeneous calibration data.

[0083] Specifically, relying solely on timestamps is insufficient to completely eliminate timing discrepancies between different data sources, as each acquisition channel inherently possesses different hardware, software processing, and network transmission delays. The timing synchronization sensing unit further utilizes a built-in software phase compensation algorithm to accurately measure and dynamically compensate for the inherent phase delay of each channel. This algorithm calculates the delay of each data channel relative to a unified clock reference and performs phase correction on the timestamped data, thereby achieving timing synchronization of multi-source data at the logical level. After this calibration process, the synchronization error of photovoltaic data, energy storage data, and bus data can be controlled within 15ms, effectively eliminating the impact of timing discrepancies on the accuracy of power calculation and optimization decisions.

[0084] S103: Generate voltage safety constraint domain based on homogeneous calibration data.

[0085] This process is performed by the photovoltaic-storage-bus electrical coupling identification unit. This unit is responsible for establishing the electrical topology model of the photovoltaic power plant and determining the safety boundaries of system operation through real-time power flow calculations, providing clear constraints for subsequent power optimization allocation.

[0086] Optionally, in another embodiment of this application, the specific implementation of step S103 includes processes C1 to C2.

[0087] C1: Establish an electrical topology model with photovoltaic units, energy storage units, busbars and grid connection points as network nodes, and collector lines and transformer windings as topology branches.

[0088] Specifically, based on the extracted node and branch parameters, an electrical topology model reflecting the actual physical connections of the power plant is constructed. This model defines photovoltaic units, energy storage units, busbars at all levels, and grid connection points as network nodes, and collector lines and transformer windings as topological branches connecting each node.

[0089] It should be noted that, to improve modeling efficiency, the photovoltaic-storage-bus electrical coupling identification unit has a built-in standardized topology template library for photovoltaic power plants, enabling rapid parameter import and automatic modeling. Simultaneously, this unit also possesses online adaptive topology identification capabilities, which can correct the topology structure in real time based on the on-site switch status, circuit breaker opening and closing information, and power flow distribution changes, thereby adapting to the dynamic changes in the power plant's operating mode.

[0090] C2: Power flow calculation is performed based on the real-time operating parameters in the electrical topology model and the same-source calibration data to obtain the voltage safety constraint domain.

[0091] The voltage safety constraint domain includes bus voltage constraints, branch power constraints, and energy storage state of charge constraints.

[0092] Understandably, after completing the electrical topology modeling, power flow calculations are performed based on real-time operating parameters from the same-source calibration data to solve for the voltage and power distribution under the current operating state. Based on the power flow calculation results, and combined with the equipment's rated parameters and safe operating procedures, a multi-dimensional voltage safety constraint domain is generated. This constraint domain specifically includes: photovoltaic active power output between 0 and the real-time maximum generateable power; the allowable operating range of voltage for each bus level, with bus voltage deviation ≤ ±5%; energy storage charging and discharging power not exceeding the equipment's rated limits; the maximum allowable transmission power of each collector line and transformer branch; and the safe operating range of the energy storage's state of charge, typically set between 20% and 90%; and the total active power of the entire station following the requirements of the superior dispatch instructions. These constraints collectively constitute the basic boundary of the feasible domain for the power optimization allocation problem.

[0093] S104: Based on the same source calibration data and voltage safety constraint domain, the photovoltaic output value and the energy storage state of charge value are predicted, and the prediction results are obtained.

[0094] The prediction result is a set of quantitative data output after predicting the photovoltaic power output and energy storage state of charge for future periods.

[0095] Optionally, by jointly extracting temporal fluctuation features and spatial correlation features through a temporal convolution and self-attention fusion prediction model (TCN-Self-Attention-Fusion, TSAF) within a multi-timescale intelligent power distribution unit, high-precision prediction of photovoltaic output and energy storage SOC for future periods can be achieved with a prediction error ≤3.5%.

[0096] Optionally, in another embodiment of this application, the specific implementation of step S104 includes processes D1 to D7.

[0097] D1: Extract historical irradiance sequence, ambient temperature sequence, historical power sequence, and energy storage state of charge sequence from the same source calibration data.

[0098] Specifically, historical irradiance sequences reflecting changes in illumination conditions, environmental temperature sequences reflecting environmental thermal conditions, historical power sequences reflecting the actual power generation capacity of photovoltaics, and historical state of charge sequences reflecting the energy storage status are extracted from the same source calibration data.

[0099] D2: Obtain the weather forecast sequence and generate the prediction input sequence based on the historical irradiance sequence, ambient temperature sequence, historical power sequence, energy storage state of charge sequence and weather forecast sequence.

[0100] It should be noted that, in addition to historical data collected on-site, external weather forecast information also needs to be incorporated. The multi-timescale intelligent power allocation unit obtains future irradiance and temperature forecast data provided by numerical weather prediction through a communication interface, forming a weather forecast sequence. Then, the above five sequences are aligned and concatenated along the time dimension to form a prediction input sequence containing multi-dimensional features, which serves as the input tensor for the subsequent deep inference model.

[0101] D3: Input the predicted input sequence into the spatiotemporal prediction model.

[0102] The spatiotemporal prediction model includes a temporal convolutional network, a self-attention mechanism, and an adaptive fusion layer.

[0103] D4: Extract short-term fluctuation features of the predicted input sequence through a temporal convolutional network to obtain a short-term feature vector.

[0104] Specifically, the temporal convolutional network employs a one-dimensional convolutional kernel that is lightweighted to suit the characteristics of photovoltaic power fluctuations, performing causal convolution operations on the predicted input sequence. This network can effectively capture short-term fluctuation features such as power spikes and ramp rates on short timescales, and encode these local temporal information into short-term feature vectors containing rich details.

[0105] D5: Extract long-term dependency features of the predicted input sequence through a self-attention mechanism to obtain a long-term feature vector.

[0106] The self-attention mechanism globally models the dependency between any two points in the predicted input sequence. By calculating the attention weight matrix, this mechanism can effectively capture long-term temporal correlation characteristics at the hourly or even daily level, such as global patterns like day-night cycles and weather trends, and encode these long-distance dependency information into long-term feature vectors.

[0107] D6: The short-term and long-term feature vectors are weighted and fused through an adaptive fusion layer to obtain the initial prediction result.

[0108] Understandably, the adaptive fusion layer doesn't simply concatenate or average the two types of feature vectors. Instead, it uses a learnable gating network to dynamically adjust the fusion weights of short-term and long-term features based on the temporal characteristics of the predicted input sequence. This mechanism allows the model to adaptively emphasize local details or global trends under different weather conditions and time periods, thus outputting an initial prediction result that integrates the advantages of both. Specifically, this initial prediction result includes the photovoltaic maximum power generation prediction curve and the energy storage SOC operation trajectory curve for 96 future time points (corresponding to the next 24 hours, one point every 15 minutes).

[0109] D7: Based on the voltage safety constraint domain, the photovoltaic output value and energy storage state of charge value in the initial prediction results are corrected by boundary adjustment to obtain the prediction results.

[0110] It should be noted that since the initial prediction results are purely data-driven fitting outputs, there may be individual prediction points that exceed the actual operating boundaries of the physical equipment. To address this, the model incorporates a boundary correction layer at the output. This layer uses the voltage safety constraint domain as a hard constraint condition to trim and correct any photovoltaic output values ​​or energy storage SOC values ​​that exceed the limits in the initial prediction results. This ensures that the final output prediction results strictly meet the equipment's rated parameters and safe operating procedures, thus providing physically feasible inputs for subsequent optimization algorithms.

[0111] S105: Perform three-level multi-objective optimization based on the prediction results and voltage safety constraint domain to obtain photovoltaic power command and energy storage power command.

[0112] The objectives of the multi-objective optimization are to minimize the curtailment rate, the power allocation error, and the bus voltage stability, while satisfying the voltage safety constraint domain.

[0113] Optionally, a three-level multi-objective optimization is performed based on the prediction results and the voltage safety constraint domain by a multi-timescale intelligent power allocation unit.

[0114] Optionally, in another embodiment of this application, the specific implementation of step S105 includes processes E1 to E3.

[0115] E1: At the day-ahead level, the prediction results and voltage safety constraint domain are optimized through optimization algorithms to obtain the day-ahead plan.

[0116] The day-ahead plan includes the day-ahead photovoltaic benchmark output curve and the day-ahead energy storage charging and discharging plan.

[0117] Optionally, the optimization algorithm includes, but is not limited to, the Improved Multi-Strategy Symbiotic Organisms Search Algorithm (MS-SBO), a multi-objective optimization algorithm that introduces a multi-strategy global mutation perturbation mechanism on the basis of the traditional symbiotic organism optimization algorithm to enhance population diversity and effectively avoid local optimum traps.

[0118] Understandably, the day-ahead layer uses the TSAF model's predicted maximum PV power output curve for the next day at 96 points and the predicted SOC trajectory of energy storage as its basic input. It uses the voltage safety constraint domain as the boundary of the feasible region and a 1-hour time resolution to perform global multi-objective optimization of the total power allocation for the next 24 hours using the MS-SBO multi-objective optimization algorithm. The optimization result of this layer is the day-ahead plan, specifically including the day-ahead baseline output curves for each PV subarray and the day-ahead charge / discharge plans for each energy storage unit.

[0119] E2: In the intraday layer, the photovoltaic operation data and energy storage state of charge value at the current moment are extracted from the same source calibration data. Based on the day-ahead plan and combined with the voltage safety constraint domain, the intraday correction plan is obtained by rolling correction through optimization algorithm.

[0120] The intraday revision plan includes revised photovoltaic power output plans and energy storage charging and discharging plans.

[0121] Specifically, the intra-day layer operates on a 5-minute control cycle, with a rolling optimization time domain of one hour in the future. At the start of each control cycle, the intra-day layer first obtains the latest actual photovoltaic output, actual energy storage SOC value, and bus voltage from the same-source calibration data. Then, using the power allocation scheme for the corresponding time period in the day-ahead plan as a benchmark, and combining the latest system state and voltage safety constraints, it performs local rolling optimization using the MS-SBO multi-objective optimization algorithm. This layer's optimization aims to correct the accumulated errors caused by prediction deviations in the day-ahead plan online, generating a corrected photovoltaic output plan and energy storage charge / discharge plan, ensuring that the control strategy closely tracks changes in actual operating conditions.

[0122] E3: In the real-time layer, the actual grid-connected power value and the actual bus voltage value are collected, and the intraday correction plan is finely adjusted in a closed loop using the actual grid-connected power value and the actual bus voltage value to obtain the photovoltaic power command and the energy storage power command.

[0123] Understandably, the real-time layer is the final stage in the three-level multi-objective optimization strategy, and its role is to compensate for ultra-short-term stochastic fluctuations that neither day-ahead forecasts nor intraday rolling corrections can cover. Since the day-ahead layer formulates a global plan based on meteorological forecast data, and the intraday layer performs rolling corrections every 5 minutes, neither can cope with second- or even millisecond-level sudden changes in sunlight or load disturbances. Therefore, the real-time layer uses a higher sampling frequency to acquire the actual power at the grid connection point and the actual bus voltage in real time as feedback signals, and compares them with the reference values ​​in the current intraday correction plan. Based on the deviation obtained from the comparison, the power command is fine-tuned online.

[0124] It should be noted that this layer does not rerun the complete optimization algorithm, but instead adopts a lightweight closed-loop correction strategy to maintain optimal power allocation while ensuring response speed. Through the intervention of the real-time layer, the impact of prediction errors and random disturbances can be further eliminated on top of day-ahead global planning and intraday rolling corrections. The final output can be directly sent to each execution unit as photovoltaic power commands and energy storage power commands.

[0125] Optionally, in another embodiment of this application, the intraday correction plan is finely adjusted in process E3 by using the actual grid connection point power value and the actual bus voltage value to obtain specific implementations of photovoltaic power command and energy storage power command, including processes F1 to F3.

[0126] F1: Compare the actual grid-connected power value with the power reference value corresponding to the intraday correction plan to obtain the total power deviation.

[0127] The real-time layer collects the actual power value of the grid-connected point in real time with a millisecond sampling period, and calculates the difference between it and the grid-connected point power reference value given at the current moment in the intraday correction plan to obtain the total power deviation signal.

[0128] F2: Compare the actual bus voltage value with the voltage safety constraint domain to obtain the bus voltage deviation.

[0129] Specifically, the real-time layer simultaneously acquires the actual voltage amplitude of each busbar in real time and compares it with the upper and lower limits of the allowable busbar voltage in the voltage safety constraint domain. If the actual voltage is within the allowable range, the voltage deviation is zero; if the actual voltage exceeds the allowable range, the deviation exceeding the upper or lower limit is calculated to obtain the busbar voltage deviation signal.

[0130] F3: Through optimization algorithms, the total power deviation and bus voltage deviation are finely adjusted in a closed loop to obtain photovoltaic power commands and energy storage power commands.

[0131] Understandably, the real-time layer uses the total power deviation signal and the bus voltage deviation signal as dual inputs, feeding them into a lightweight, fast optimization solver. This solver aims to eliminate these two deviations, making minor adjustments to the photovoltaic output and energy storage power commands at the current moment in the intraday correction plan, under voltage safety constraints and equipment response rate constraints. This adjustment is completed with a millisecond-level response speed, aiming to quickly smooth out random power fluctuations ranging from seconds to milliseconds and maintain bus voltage stability. After closed-loop fine-tuning by the real-time layer, the final execution power commands are generated and issued to each photovoltaic subarray and each energy storage unit.

[0132] S106: Send the photovoltaic power command and energy storage power command to the execution unit so that the execution unit can allocate power according to the photovoltaic power command and energy storage power command.

[0133] Among them, the synchronous command execution unit is responsible for sending photovoltaic power commands and energy storage power commands to the execution unit. This unit adopts a synchronous broadcast command mechanism to send photovoltaic power commands and energy storage power commands to all photovoltaic inverters, energy storage converters and other execution units on site at the same time, ensuring that all equipment in the station performs power adjustment actions in a unified sequence.

[0134] Specifically, the synchronous instruction execution unit is equipped with a high-precision synchronous timer. When the instruction is issued, it simultaneously sends a power instruction carrying a precise execution time tag to each execution unit via the second physically isolated communication link in the dual-redundant isolated communication unit, using multicast or broadcast methods. After receiving the instruction, each execution unit performs power adjustment at the same time according to the time tag in the instruction, thereby ensuring that the instruction execution time difference between units is no more than 10ms, effectively avoiding instantaneous power surges or imbalances caused by inconsistent execution times.

[0135] Furthermore, after the command is issued, the synchronous command execution unit is also responsible for collecting feedback data such as the actual output value, operating status, and alarm information from each execution unit. This feedback data is transmitted back in real time to the photovoltaic-storage-bus electrical coupling identification unit and the multi-timescale intelligent power allocation unit to update the electrical topology model, refresh the power flow calculation results, correct the input state of the prediction model, and calibrate the constraints and objective function of the optimization algorithm. Thus, the entire system forms a complete local autonomous closed-loop control link of "sensing-prediction-optimization-deployment-execution-feedback". All functions of the entire system are completed independently on the edge-side embedded hardware platform, without relying on cloud servers for real-time calculation and decision-making. Even in extreme conditions where communication with the dispatch center is interrupted, the system can still operate independently and stably based on local data and preset strategies.

[0136] Figure 2 This is a schematic diagram comparing the total light waste curves before and after optimized control, as provided in an embodiment of this application. Figure 2 As shown, the horizontal axis represents time, and the vertical axis represents the curtailed power. The curves respectively display the total curtailment curve before optimization and the total curtailment curve after optimization using this system. Figure 2 As can be seen, after adopting the system of this application, the power of light curtailment is significantly lower than that before optimization in all time periods, and the curtailment rate can be reduced by no less than 1 percentage point compared with that before optimization, which fully verifies the effectiveness of this system in suppressing light curtailment.

[0137] Tests have shown that, using the method described in this application, the power distribution error of the entire station can be controlled within ±1.0%, the curtailment rate can be reduced by no less than 1 percentage point compared to before optimization, and the overall system control response delay does not exceed 200ms. This fully meets the comprehensive requirements for power distribution accuracy, reliability, and speed in high-proportion photovoltaic grid-connected scenarios.

[0138] It should be noted that, Figure 3 This is a block diagram illustrating the functional modules and data interaction of the power distribution control system provided in an embodiment of this application. Figure 3 As shown, the system includes a dual-redundant isolated communication unit, an edge collaborative computing core unit, a time-series co-source sensing unit, a multi-timescale intelligent power allocation unit (i.e., including a TSAF spatiotemporal joint prediction unit, an MS-SBO multi-objective optimization unit, and a three-level timescale power allocation unit), a photovoltaic-storage-bus electrical coupling identification unit, and a synchronization command execution unit. The dual-redundant isolated communication unit interacts with external photovoltaic equipment, energy storage equipment, the bus, and the dispatch platform, respectively, and sends the collected data to the time-series co-source sensing unit. After completing data calibration, the time-series co-source sensing unit transmits the calibration data to the photovoltaic-storage-bus electrical coupling identification unit and the multi-timescale intelligent power allocation unit. The photovoltaic-storage-bus electrical coupling identification unit generates a voltage safety constraint domain and feeds it back to the multi-timescale intelligent power allocation unit. After completing three-level optimization, the multi-timescale intelligent power allocation unit transmits the power command to the synchronization command execution unit. The synchronization command execution unit issues the command to the execution device and sends the execution feedback back to the photovoltaic-storage-bus electrical coupling identification unit and the multi-timescale intelligent power allocation unit, forming a complete closed loop.

[0139] It should be noted that, based on the processes shown in S101-S106 above, this embodiment can achieve the following beneficial effects:

[0140] 1. All functions of the system are implemented independently on the edge hardware platform, without relying on cloud servers for computation and decision-making. Control commands are generated and executed locally. Even in extreme situations where communication is interrupted, the system can still maintain stable operation by relying on local resources, with a response latency of no more than 200ms. This completely eliminates the inherent communication latency and single point of failure risks of centralized control at the architectural level.

[0141] 2. The TSAF model adopts a network structure that integrates temporal convolutional networks and self-attention mechanisms to jointly extract the temporal fluctuation characteristics and spatial correlation characteristics of photovoltaic power output. The prediction error can be controlled within 3.5%, which can provide stable and accurate data input for power optimization allocation.

[0142] 3. The MS-SBO multi-objective optimization algorithm takes the minimum curtailment rate, the lowest power allocation error, and the most stable bus voltage as joint optimization objectives. It performs global optimization through a multi-strategy global mutation perturbation mechanism, effectively avoiding local optima and achieving multi-objective collaborative optimization. The curtailment rate can be reduced by no less than 1 percentage point compared with that before optimization.

[0143] 4. Through a unified clock synchronization mechanism and software phase compensation algorithm, the timing alignment and phase calibration of multi-source data from photovoltaics, energy storage, and the bus are achieved, with a data synchronization error of no more than 15ms. Based on this data, the power distribution error of the entire station can be controlled within ±1.0%, and the fluctuation amplitude of grid-connected power can be reduced by more than 30%.

[0144] 5. It adopts a three-level control strategy of day-ahead global planning, intraday rolling correction, and real-time deviation correction, which can adaptively cope with various operating conditions such as light fluctuations, changes in dispatching commands and changes in energy storage SOC constraints, and maintain a stable and optimal power distribution state under all operating conditions.

[0145] like Figure 4 As shown in the figure, this application provides a schematic diagram of the architecture of a photovoltaic power distribution control system. The photovoltaic power distribution control system includes a dual redundant isolated communication unit 100, a timing co-source sensing unit 200, a photovoltaic-storage-bus electrical coupling identification unit 300, a multi-time-scale intelligent power distribution unit 400, and a synchronous instruction execution unit 500.

[0146] The dual-redundant isolated communication unit 100 is used to collect photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data.

[0147] The dual-redundant isolated communication unit 100 is specifically used for: collecting photovoltaic operation data, energy storage operation data and bus operation data through the first physically isolated communication link; and receiving scheduling instruction data through the second physically isolated communication link; wherein, there is no common electrical ground and no clock bus coupling between the first physically isolated communication link and the second physically isolated communication link.

[0148] The time-series homogeneous sensing unit 200 is used to perform data calibration on photovoltaic operation data, energy storage operation data, and bus operation data to obtain homogeneous calibration data.

[0149] The timing-based sensing unit 200 is specifically used to: add timestamps to photovoltaic operation data, energy storage operation data, and bus operation data using a unified clock synchronization mechanism to obtain timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data; and perform timing deviation calibration on the timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data using a software phase compensation algorithm to obtain time-based calibration data.

[0150] The photovoltaic-storage-bus electrical coupling identification unit 300 is used to generate a voltage safety constraint domain based on the same source calibration data.

[0151] The photovoltaic-storage-bus electrical coupling identification unit 300 is specifically used for: establishing an electrical topology model with photovoltaic units, energy storage units, buses and grid connection points as network nodes, and collector lines and transformer windings as topology branches; performing power flow calculations based on the electrical topology model and real-time operating parameters in the same-source calibration data to obtain the voltage safety constraint domain; the voltage safety constraint domain includes bus voltage constraints, branch power constraints and energy storage state of charge constraints.

[0152] The multi-timescale intelligent power allocation unit 400 is used to predict the photovoltaic output value and the energy storage state of charge value based on the same source calibration data and the voltage safety constraint domain, and obtain the prediction results; it is also used to perform three-level multi-objective optimization based on the prediction results and the voltage safety constraint domain to obtain the photovoltaic power command and the energy storage power command; wherein, the objectives of the multi-objective optimization are to minimize the curtailment rate, minimize the power allocation error, and maximize the stability of the bus voltage under the premise of satisfying the voltage safety constraint domain.

[0153] Multi-timescale intelligent power distribution unit 400 includes:

[0154] The acquisition unit is used to extract historical irradiance sequences, ambient temperature sequences, historical power sequences, and energy storage state of charge sequences from the same source calibration data.

[0155] The generation unit is used to acquire the weather forecast sequence and generate the prediction input sequence based on the historical irradiance sequence, ambient temperature sequence, historical power sequence, energy storage state of charge sequence and weather forecast sequence.

[0156] The input unit is used to input the predicted input sequence into the spatiotemporal prediction model; the spatiotemporal prediction model includes a temporal convolutional network, a self-attention mechanism, and an adaptive fusion layer.

[0157] The first extraction unit is used to extract short-term fluctuation features of the predicted input sequence through a temporal convolutional network to obtain a short-term feature vector.

[0158] The second extraction unit is used to extract long-term dependency features of the predicted input sequence through a self-attention mechanism to obtain a long-term feature vector.

[0159] The fusion unit is used to perform weighted fusion of short-term and long-term feature vectors through an adaptive fusion layer to obtain the initial prediction result.

[0160] The correction unit is used to perform boundary correction on the photovoltaic output value and energy storage state of charge value in the initial prediction result according to the voltage safety constraint domain, so as to obtain the prediction result.

[0161] Multi-timescale intelligent power distribution unit 400 includes:

[0162] The optimization unit is used to optimize the prediction results and voltage safety constraint domain through optimization algorithms at the day-ahead level to obtain the day-ahead plan; the day-ahead plan includes the day-ahead photovoltaic reference output curve and the day-ahead energy storage charge and discharge plan.

[0163] The rolling correction unit is used to extract the current photovoltaic operation data and energy storage state of charge value from the same source calibration data at the intraday layer, and to perform rolling correction based on the day-ahead plan and combined with the voltage safety constraint domain through optimization algorithm to obtain the intraday correction plan; the intraday correction plan includes the corrected photovoltaic power output plan and energy storage charge and discharge plan.

[0164] The fine-tuning unit is used to collect the actual grid-connected point power value and the actual bus voltage value at the real-time layer, and to perform closed-loop fine-tuning of the intraday correction plan based on the actual grid-connected point power value and the actual bus voltage value to obtain photovoltaic power command and energy storage power command.

[0165] The fine-tuning unit is specifically used to: compare the actual grid-connected power value with the power reference value corresponding to the intraday correction plan to obtain the total power deviation;

[0166] The actual bus voltage value is compared with the voltage safety constraint domain to obtain the bus voltage deviation;

[0167] By optimizing the algorithm, closed-loop fine-tuning of the total power deviation and bus voltage deviation is performed to obtain photovoltaic power command and energy storage power command.

[0168] The synchronous instruction execution unit 500 is used to send photovoltaic power instructions and energy storage power instructions to the execution unit so that the execution unit can perform power allocation according to the photovoltaic power instructions and energy storage power instructions.

[0169] In summary, time-series calibration of multi-source data generates homogeneous calibration data, and based on the voltage safety constraint domain, a three-level multi-objective optimization is performed. The optimization task is decomposed at different time scales, effectively balancing convergence speed and computational diversity. This overcomes the shortcomings of traditional weighted summation methods in achieving multi-objective synergistic optimization, and achieves synergistic optimization with minimum curtailment rate, minimum power allocation error, and most stable bus voltage.

[0170] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0171] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A light storage power distribution control method characterized by, Applications in photovoltaic-storage power distribution control systems include: Collect photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data; The photovoltaic operation data, the energy storage operation data, and the bus operation data are calibrated to obtain homogeneous calibration data; A voltage safety constraint domain is generated based on the aforementioned homogeneous calibration data; Based on the same source calibration data and the voltage safety constraint domain, the photovoltaic output value and the energy storage state of charge value are predicted, and the prediction results are obtained. Based on the prediction results and the voltage safety constraint domain, a three-level multi-objective optimization is performed to obtain photovoltaic power commands and energy storage power commands; wherein, the objectives of the multi-objective optimization are to minimize the curtailment rate, minimize the power allocation error, and maximize the bus voltage stability while satisfying the voltage safety constraint domain. The photovoltaic power command and the energy storage power command are sent to the execution unit so that the execution unit can allocate power according to the photovoltaic power command and the energy storage power command.

2. The method according to claim 1, characterized in that, The process of calibrating the photovoltaic operation data, the energy storage operation data, and the bus operation data to obtain homogeneous calibration data includes: A unified clock synchronization mechanism is used to add timestamps to the photovoltaic operation data, the energy storage operation data, and the bus operation data respectively, resulting in timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data. The time-series deviation of the timestamped photovoltaic data, the timestamped energy storage data, and the timestamped bus data is calibrated using a software phase compensation algorithm to obtain the homogeneous calibration data.

3. The method according to claim 1, characterized in that, The generation of the voltage safety constraint domain based on the homogeneous calibration data includes: An electrical topology model is established with photovoltaic units, energy storage units, busbars and grid connection points as network nodes and collector lines and transformer windings as topology branches. Power flow calculations are performed based on the electrical topology model and the real-time operating parameters in the same-source calibration data to obtain the voltage safety constraint domain; the voltage safety constraint domain includes bus voltage constraints, branch power constraints, and energy storage state of charge constraints.

4. The method according to claim 1, characterized in that, The prediction of photovoltaic power output and energy storage state of charge based on the same source calibration data and the voltage safety constraint domain yields prediction results, including: Historical irradiance sequence, ambient temperature sequence, historical power sequence, and energy storage state of charge sequence are extracted from the homogeneous calibration data; Obtain the weather forecast sequence, and generate a prediction input sequence based on the historical irradiance sequence, the ambient temperature sequence, the historical power sequence, the energy storage state of charge sequence, and the weather forecast sequence; The predicted input sequence is input into the spatiotemporal prediction model; the spatiotemporal prediction model includes a temporal convolutional network, a self-attention mechanism, and an adaptive fusion layer. The short-term fluctuation features of the predicted input sequence are extracted through the temporal convolutional network to obtain a short-term feature vector; The long-term dependency features of the predicted input sequence are extracted through the self-attention mechanism to obtain a long-term feature vector; The short-term feature vector and the long-term feature vector are weighted and fused through the adaptive fusion layer to obtain the initial prediction result; Based on the voltage safety constraint domain, the photovoltaic output value and energy storage state of charge value in the initial prediction result are boundary-corrected to obtain the prediction result.

5. The method according to claim 1, characterized in that, The process of performing a three-level multi-objective optimization based on the prediction results and the voltage safety constraint domain to obtain photovoltaic power commands and energy storage power commands includes: At the day-ahead level, the prediction results and the voltage safety constraint domain are optimized using an optimization algorithm to obtain the day-ahead plan; the day-ahead plan includes the day-ahead photovoltaic reference output curve and the day-ahead energy storage charge and discharge plan; At the intraday level, the photovoltaic operation data and energy storage state of charge value at the current moment are extracted from the same source calibration data. Based on the day-ahead plan and combined with the voltage safety constraint domain, the plan is rolled over and corrected using the optimization algorithm to obtain the intraday correction plan. The intraday correction plan includes the corrected photovoltaic output plan and energy storage charge and discharge plan. At the real-time layer, the actual grid-connected power value and the actual bus voltage value are collected, and the intraday correction plan is finely adjusted in a closed loop using the actual grid-connected power value and the actual bus voltage value to obtain the photovoltaic power command and the energy storage power command.

6. The method according to claim 5, characterized in that, The process of fine-tuning the intraday correction plan using actual grid-connected power and actual bus voltage values ​​to obtain photovoltaic power commands and energy storage power commands includes: The actual grid-connected power value is compared with the power reference value corresponding to the intraday correction plan to obtain the total power deviation. The actual bus voltage value is compared with the voltage safety constraint domain to obtain the bus voltage deviation; The total power deviation and the bus voltage deviation are finely adjusted in a closed loop using an optimization algorithm to obtain the photovoltaic power command and the energy storage power command.

7. The method according to claim 1, characterized in that, The collected photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data include: Photovoltaic operation data, energy storage operation data, and bus operation data are collected through the first physically isolated communication link; Scheduling instruction data is received through a second physically isolated communication link; wherein, there is no common electrical ground and no clock bus coupling between the first and second physically isolated communication links.

8. A photovoltaic-storage power distribution control system, characterized in that, include: Dual redundant isolated communication units are used to collect photovoltaic operation data, energy storage operation data, bus operation data, and dispatch command data; The time-series homogeneous sensing unit is used to perform data calibration on the photovoltaic operation data, the energy storage operation data, and the bus operation data to obtain homogeneous calibration data; A photovoltaic-storage-bus electrical coupling identification unit is used to generate a voltage safety constraint domain based on the homogeneous calibration data. A multi-timescale intelligent power allocation unit is used to predict photovoltaic output and energy storage state of charge based on the same source calibration data and the voltage safety constraint domain, and obtain prediction results; it is also used to perform three-level multi-objective optimization based on the prediction results and the voltage safety constraint domain to obtain photovoltaic power command and energy storage power command; wherein, the objectives of the multi-objective optimization are to minimize the curtailment rate, minimize the power allocation error, and maximize the stability of the bus voltage while satisfying the voltage safety constraint domain. A synchronous instruction execution unit is used to send the photovoltaic power instruction and the energy storage power instruction to the execution unit so that the execution unit can perform power allocation according to the photovoltaic power instruction and the energy storage power instruction.

9. The system according to claim 8, characterized in that, The temporal homogeneous sensing unit is specifically used for: A unified clock synchronization mechanism is used to add timestamps to the photovoltaic operation data, the energy storage operation data, and the bus operation data respectively, resulting in timestamped photovoltaic data, timestamped energy storage data, and timestamped bus data. The time-series deviation of the timestamped photovoltaic data, the timestamped energy storage data, and the timestamped bus data is calibrated using a software phase compensation algorithm to obtain the homogeneous calibration data.

10. The system according to claim 8, characterized in that, The photovoltaic-storage-bus electrical coupling identification unit is specifically used for: An electrical topology model is established with photovoltaic units, energy storage units, busbars and grid connection points as network nodes and collector lines and transformer windings as topology branches. Power flow calculations are performed based on the electrical topology model and the real-time operating parameters in the same-source calibration data to obtain the voltage safety constraint domain; the voltage safety constraint domain includes bus voltage constraints, branch power constraints, and energy storage state of charge constraints.

Images