Energy storage adaptive collaborative dispatching method and system for power distribution area

Abstract

The application belongs to the technical field of power distribution network intelligent dispatching, and discloses a power distribution area energy storage adaptive collaborative dispatching method and system, S1. A space-time fluctuation correlation matrix is constructed, including obtaining cloud data and preprocessing, and constructing a space-time fluctuation correlation matrix according to the preprocessed data; S2. An edge-cloud collaborative control architecture is developed: the edge-cloud collaborative control includes two parts of the cloud and the edge side, is connected through a reliable communication network, forms a bidirectional closed-loop information flow and control flow; S3. An adaptive rolling optimization engine is designed: a double-layer optimization model is solved in a fixed interval cycle; an LSTM short-term power prediction correction module is introduced to predict the fluctuation trend in the future period of time, the application accurately predicts the photovoltaic fluctuation transmission path caused by the movement of the cloud layer through the construction of a dynamic space-time correlation matrix; through the area cluster collaborative dispatching, the energy storage charging and discharging strategy is optimized; the adaptability of the power distribution network to high-proportion renewable energy fluctuation is enhanced.

Description

Technical Field

[0001] This invention belongs to the field of intelligent dispatching technology for distribution networks, specifically relating to an adaptive and collaborative dispatching method and system for distribution transformer substations that integrates the spatiotemporal fluctuation characteristics of distributed resources, applicable to rural distribution networks with high penetration of distributed photovoltaic access. Background Technology

[0002] The primary bottleneck currently facing rural power grid energy storage dispatch lies in the difficulty of quantifying the coupling effects of spatiotemporal fluctuations. Distributed photovoltaic power output is affected by sunlight and cloud movement, exhibiting significant spatiotemporal correlation. However, traditional dispatch models typically treat each distribution area as an independent unit, failing to fully consider the cluster fluctuation transmission effect (e.g., the chain reaction of sharp power drops in adjacent distribution areas caused by cloud shading), resulting in a mismatch between energy storage charging and discharging strategies and the complex power fluctuations in reality.

[0003] Traditional scheduling models suffer from control response lag. Existing scheduling systems rely on minute-level (e.g., 15-minute) SCADA data updates, which cannot effectively capture instantaneous power fluctuations at the second level. When encountering rapid disturbances such as sudden drops in photovoltaic output (rate exceeding 20% / minute) or sudden load changes, triggering mechanisms based on fixed thresholds often react slowly, causing delays in energy storage support and significantly exacerbating the risk of voltage overruns. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive collaborative scheduling method for energy storage in distribution substations. This addresses the problems raised in the background section, where existing scheduling models typically treat each substation as an independent unit, failing to fully consider the cluster fluctuation propagation effect. This leads to a mismatch between energy storage charging and discharging strategies and actual complex power fluctuations, as well as lag in control response, significantly exacerbating the risk of voltage exceeding limits. This invention constructs a spatiotemporal correlation matrix to quantify the distributed resource fluctuation propagation effect, breaking through the traditional mindset of "isolated scheduling." It uses second-level rolling optimization technology to dynamically balance the economy and stability of system operation in real time. Furthermore, based on an innovative edge-cloud collaborative architecture, it compresses the critical control response latency to within 1.5 seconds, successfully meeting the stringent power quality requirements of rural power grids.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] On the one hand, an adaptive cooperative scheduling method for energy storage in distribution substations is provided, which includes the following steps:

[0007] S1. Constructing a spatiotemporal fluctuation correlation matrix includes acquiring cloud cluster data and preprocessing it, and constructing the spatiotemporal fluctuation correlation matrix based on the preprocessed data;

[0008] Data acquisition and preprocessing include acquiring meteorological data layers. Preprocessing involves identifying and segmenting cloud clusters in the acquired meteorological data layers and vectorizing the cloud cluster motion to obtain the cloud cluster movement vector field. The geographical boundaries of all distribution substations within the jurisdiction and the precise coordinates of photovoltaic power stations are acquired. Based on the geographical information of the substations, the distance between substations and the azimuth angle are calculated, and the trajectory of meteorological satellite cloud images is predicted.

[0009] Constructing the spatiotemporal fluctuation correlation matrix involves inputting the results of the above steps to construct a high-dimensional matrix, the spatiotemporal fluctuation correlation matrix C. ij Construction algorithm:

[0010] By combining meteorological satellite cloud image trajectory prediction with historical photovoltaic power output data from photovoltaic power distribution clusters, a spatiotemporal transmission model for photovoltaic fluctuations was established.

[0011] ;

[0012] in : Geographical coordinates of the distribution area; σ is the cloud cluster movement scale parameter (calibrated via satellite cloud image trajectory); Δt is the wave propagation delay (fitted based on historical data). Dynamic collaborative clusters are partitioned based on the correlation matrix, when...

[0013] Automatic grouping occurs when Cij > 0.7.

[0014] S2. Develop an edge-cloud collaborative control architecture:

[0015] Edge-cloud collaborative control consists of two parts: the cloud and the edge. They are connected through a reliable communication network to form a two-way closed-loop information flow and control flow.

[0016] S3. Design an adaptive rolling optimization engine: Solve the two-layer optimization model using a fixed interval as the cycle.

[0017] Upper layer (hourly level): Minimize the total operating cost of the cluster:

[0018] ;

[0019] C grid The electricity purchase and sale price (unit: yuan / kWh) is a key economic signal affecting energy storage peak-valley arbitrage strategies; P grid (t) represents the power exchanged with the grid at time t (unit: kW), with positive values ​​representing electricity purchases and negative values ​​representing electricity sales. This is a decision variable in the optimization process; C deg This is the energy storage loss cost coefficient (unit: yuan / kW), used to quantify the aging loss of batteries due to charging and discharging into economic costs, and to avoid overuse of energy storage. It represents the absolute value of energy storage charging and discharging power (unit: kW), reflecting the equipment's operating intensity measured solely by power regardless of the charging and discharging direction.

[0020] Lower layer (second level): Suppress voltage deviation and power backfeed

[0021] ;

[0022] It is the real-time voltage measurement value of the k-th node at time t;

[0023] This is the voltage reference value (rated voltage);

[0024] It is the absolute value of the voltage deviation at the k-th node;

[0025] It is the power fed back from the distribution network to the next higher level of the power grid at time t;

[0026] It is the absolute value of the power backfeed.

[0027] An LSTM short-term power prediction and correction module is introduced to predict fluctuation trends over a future period.

[0028] On the other hand, a distribution substation energy storage adaptive cooperative scheduling system is provided to implement the above-mentioned distribution substation energy storage adaptive cooperative scheduling method, characterized in that the system includes:

[0029] The spatiotemporal fluctuation correlation matrix construction module is used to acquire cloud cluster data and preprocess it, construct the spatiotemporal fluctuation correlation matrix based on the preprocessed data, and run the spatiotemporal fluctuation correlation matrix.

[0030] The edge-cloud collaborative control architecture module is used to develop and run the edge-cloud collaborative control architecture.

[0031] The Adaptive Scrolling Optimization Engine module is used to design and run the adaptive scrolling optimization engine.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. This invention constructs a dynamic spatiotemporal correlation matrix to accurately predict the transmission path of photovoltaic fluctuations caused by cloud movement, reducing the fluctuation prediction error from 21.5% in traditional methods to 6.8%. Based on this, the second-level rolling optimization and edge-cloud collaborative control shorten the voltage over-limit recovery time from 8.2 seconds to 1.3 seconds, improving the response speed by 84% and significantly improving power quality.

[0034] 2. This invention optimizes the energy storage charging and discharging strategy through coordinated scheduling of distribution network clusters, reducing the average daily number of energy storage cycles by 19% and effectively extending the equipment's lifespan. Its core feature of software algorithm upgrades eliminates the need for large-scale hardware modifications, significantly reducing deployment costs. Simultaneously, by enhancing photovoltaic absorption capacity, the average daily power generation per photovoltaic unit increases by 5.88kW, significantly improving the return on investment.

[0035] 3. This invention achieves a fundamental shift from passive response to proactive forward-looking decision-making, greatly enhancing the adaptability of the distribution network to fluctuations caused by high proportions of renewable energy. The system can automatically divide into cooperative clusters and dynamically adjust control strategies, significantly improving the reliability and resilience of the power grid operation. It provides key technical support for building a green and efficient new rural smart distribution network and is of great significance for promoting energy transition. Attached Figure Description

[0036] Figure 1 The flowchart of the adaptive collaborative scheduling method for energy storage in distribution substations provided by the present invention is shown below.

[0037] Figure 2 This is a flowchart illustrating the construction of the spatiotemporal correlation matrix in this invention.

[0038] Figure 3 This is a diagram of the edge-cloud collaborative control architecture in this invention;

[0039] Figure 4 The diagram shows the structure of the distribution substation energy storage adaptive collaborative scheduling system provided by this invention. Detailed Implementation

[0040] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0041] like Figure 1 As shown, the adaptive collaborative scheduling method for distribution radio station energy storage provided by the present invention includes the following steps:

[0042] S1. Construct a spatiotemporal fluctuation correlation matrix, including acquiring cloud cluster data and preprocessing it, and constructing the spatiotemporal fluctuation correlation matrix based on the preprocessed data;

[0043] S1. Constructing a spatiotemporal fluctuation correlation matrix includes acquiring cloud cluster data and preprocessing it, and constructing the spatiotemporal fluctuation correlation matrix based on the preprocessed data;

[0044] Data acquisition and preprocessing include acquiring meteorological data layers. Preprocessing involves identifying and segmenting cloud clusters in the acquired meteorological data layers and vectorizing the cloud cluster motion to obtain the cloud cluster movement vector field. The geographical boundaries of all distribution substations within the jurisdiction and the precise coordinates of photovoltaic power stations are acquired. Based on the geographical information of the substations, the distance between substations and the azimuth angle are calculated, and the trajectory of meteorological satellite cloud images is predicted.

[0045] Constructing the spatiotemporal fluctuation correlation matrix involves inputting the results of the above steps to construct a high-dimensional matrix, the spatiotemporal fluctuation correlation matrix. Construction algorithm:

[0046] By combining meteorological satellite cloud image trajectory prediction with historical photovoltaic power output data from photovoltaic power distribution clusters, a spatiotemporal transmission model for photovoltaic fluctuations was established.

[0047] ;

[0048] in : Geographical coordinates of the distribution area; σ is the cloud cluster movement scale parameter (calibrated via satellite cloud image trajectory); Δt is the wave propagation delay (fitted based on historical data). Dynamic collaborative clusters are partitioned based on the correlation matrix, when...

[0049] Automatic grouping when the value is >0.7.

[0050] Constructing the spatiotemporal fluctuation correlation matrix as follows Figure 2 As shown, steps S11-S17 are as follows:

[0051] S11. Input: Meteorological satellite remote sensing cloud image

[0052] The data input layer is the starting point of the process. The system accesses real-time remote sensing data from multispectral meteorological satellites (such as the Fengyun series, Himawari-8 / 9, etc.). The input is not ordinary visible light cloud images, but quantitative products containing physical information such as cloud top height, cloud thickness, and liquid water content, providing rich data dimensions and initial fields for subsequent high-precision analysis.

[0053] S12. Image Preprocessing and Feature Extraction

[0054] Raw satellite data requires a series of preprocessing steps before it can be used for analysis. This module first performs radiometric calibration, atmospheric correction, and geometric fine correction to eliminate sensor errors and observation angle distortions. Subsequently, computer vision algorithms (such as threshold segmentation and edge detection) are used to segment the cloud image, accurately identify discrete cloud entities with complete shapes, and extract key features such as their contours, areas, brightness, and temperature to prepare for tracking their movement.

[0055] S13. Calculation of cloud movement vector field

[0056] This step is crucial for calculating cloud movement. Optical flow or cross-correlation methods are used to analyze consecutive time sequences of satellite cloud images (e.g., at 10-15 minute intervals). By calculating the displacement of the same cloud feature in adjacent frames, a two-dimensional vector field with velocity magnitude and direction is generated, covering the entire observation area. This vector field visually reflects the overall movement trend and local variations of clouds in the atmosphere.

[0057] S14. Cloud Movement Trajectory Prediction

[0058] Based on the calculated instantaneous moving vector field, this module introduces steering flow data from numerical weather prediction models as the background field. Using extrapolation algorithms (such as persistence prediction and optical flow extrapolation) or simple physical models, it predicts the trajectory and morphological evolution of each identified cloud cluster within the next 0-4 hours. This provides a direct basis for predicting future fluctuations in photovoltaic power output.

[0059] S15. Geographic coordinate mapping of the distribution area

[0060] This step achieves a crucial transformation from meteorological space to power physical space. The system incorporates a high-precision Geographic Information System (GIS) layer, containing the geographical boundaries of all distribution substations within the jurisdiction and the precise coordinates of photovoltaic power plants. By overlaying the predicted cloud movement trajectory layer with the GIS layer and performing spatial correlation analysis, it is possible to determine when a specific cloud will cover (block) which substation or photovoltaic array, thereby transforming meteorological forecasts into early warnings of their impact on specific power units.

[0061] S16. Prediction of the probability of future irradiance decay

[0062] Based on cloud cover prediction results, a more refined quantitative analysis is conducted. The model not only determines "whether there is shading," but also calculates "how much shading." According to the predicted cloud type (e.g., cumulus, stratus), thickness, and coverage, combined with the atmospheric radiative transfer model, the model calculates the light intensity (irradiance) reaching the surface of the distribution area at each future time segment (e.g., every 15 minutes) and its attenuation probability and degree. The output is a time-series probability distribution function, providing input for uncertainty optimization.

[0063] S17. Constructing a dynamic spatiotemporal correlation matrix

[0064] This is the final output of the process and the essence of the entire patented method. Using the results of the above steps as input, a high-dimensional matrix is ​​constructed. The rows of this matrix represent different future time slices, and the columns represent different distributed photovoltaic nodes (or distribution substations) in the power grid. Each element in the matrix is ​​a correlation weight value, which comprehensively reflects:

[0065] Time correlation: When did the effects of cloud movement begin, how long did they last, and when did they end?

[0066] Spatial correlation: Which geographically dispersed photovoltaic nodes will be affected by the same cloud system simultaneously or sequentially, forming a spatiotemporal coupling effect.

[0067] This matrix is ​​dynamically updated and serves as the direct basis for subsequent energy storage collaborative scheduling algorithms to make forward-looking decisions and perform global optimization.

[0068] The dynamic spatiotemporal correlation matrix operates as follows:

[0069] Acquire data and preprocess it:

[0070] Meteorological data acquisition layer: The data source is high spatiotemporal resolution (e.g., once every 5 minutes, 1km×1km pixels) visible light and infrared cloud images from geostationary meteorological satellites such as the Fengyun series or Himawari-8.

[0071] Preprocessing: Cloud cluster identification and segmentation, and cloud cluster motion vectorization are performed on the meteorological data layers obtained above.

[0072] Cloud cluster identification and segmentation: Image segmentation algorithms based on deep learning models such as U-Net are used to accurately separate cloud cluster regions from satellite cloud images and filter out interference such as surface buildings and mountains.

[0073] Optical Flow: Using optical flow or cross-correlation coefficient methods, multiple frames of cloud images are continuously tracked to calculate the cloud's velocity vector v_c (km / h) and direction angle θ_c (degrees). This forms a dynamic cloud movement vector field.

[0074] Step S17 is based on the association matrix. Construction algorithm;

[0075] Correlation Matrix Construction algorithm:

[0076] By combining meteorological satellite cloud image trajectory prediction with historical photovoltaic power output data from photovoltaic power distribution clusters, a spatiotemporal transmission model for photovoltaic fluctuations was established.

[0077] ;

[0078] in : Geographic coordinates of the transformer substation; σ is the cloud cluster movement scale parameter (calibrated via satellite cloud image trajectory); Δt is the fluctuation propagation delay (fitted based on historical data); PV_i represents the photovoltaic output (power output) of the i-th transformer substation (or photovoltaic node); PV_j represents the photovoltaic output (power output) of the j-th transformer substation (or photovoltaic node); these are all time-series variables used to represent the power fluctuation of photovoltaic power generation units at different locations over time. Dynamic collaborative clusters are divided based on the correlation matrix, when... Automatic grouping when the value is >0.7.

[0079] The specific process for establishing a spatiotemporal transmission model for photovoltaic fluctuations is as follows:

[0080] The satellite cloud image data obtained from the meteorological data layer is processed by cloud cluster segmentation and motion vectorization to obtain v_c and θ_c.

[0081] Based on the geographical information of the transformer area, L_i, L_j, calculate the distance between transformer areas d_ij and the azimuth angle φ_ij.

[0082] The theoretical time delay Δt_ij is calculated based on the moving speed vector v_c, the moving direction angle θ_c, the station spacing d_ij, and the azimuth angle φ_ij.

[0083] ;

[0084] The high-frequency sampling data of photovoltaic power output in the transformer substation is cleaned and preprocessed. The correlation coefficient R_ij with time shift is calculated using the theoretical time delay Δt_ij and the cleaned and preprocessed high-frequency sampling data of photovoltaic power output in the transformer substation.

[0085] ;

[0086] Set the initial scale parameter σ

[0087] Calculate the spatial correlation degree A_ij:

[0088] ;

[0089] Calculate the final correlation degree C_ij based on the time-shifted correlation coefficient R_ij and the spatial correlation degree A_ij:

[0090] ;

[0091] The final correlation coefficient C_ij is output and iterated through all station pairs. If a pair exists (indicating there are still unprocessed station pairs), the next station pair i,j is taken and returned. The theoretical time delay Δt_ij is calculated based on the moving speed vector v_c, moving direction angle θ_c, station spacing d_ij, and azimuth angle φ_ij. If not (indicating all station pairs have been processed), a preliminary correlation matrix C is generated. After dynamically updating and optimizing the preliminary correlation matrix C, the final spatiotemporal fluctuation correlation matrix C is output.

[0092] ;

[0093] Take a new frame of satellite data and determine the cloud vector change. If the cloud vector change is less than the threshold, proceed to the step of generating the preliminary correlation matrix C. After dynamically updating and optimizing the preliminary correlation matrix C, output the final spatiotemporal fluctuation correlation matrix C. If the cloud vector change is greater than the threshold, trigger the matrix recalculation.

[0094] S2. Develop an edge-cloud collaborative control architecture, such as... Figure 3 As shown,

[0095] Edge side: FPGA computing cores deployed at the terminal of the distribution area execute millisecond-level reactive power regulation commands;

[0096] ;

[0097] Output value: This indicates the amount of reactive power that the energy storage converter (PCS) is required to generate (positive) or absorb (negative) according to calculations.

[0098] The core coefficient determines the strength of the controller's response to voltage deviations.

[0099] Input value: The actual voltage value obtained in real time by sensors (such as PT and VT) deployed at key nodes in the transformer area;

[0100] : Input value, the rated voltage that the power grid is expected to maintain during operation.

[0101] Cloud-based: Generate cluster collaboration strategies based on spatiotemporal correlation matrices and issue GOOSE commands through 5G slicing networks.

[0102] The composition, information flow, and control process of the edge-cloud collaborative control architecture:

[0103] The entire system consists of two main parts: the cloud and the edge (distribution station area). They are connected through a reliable communication network to form a two-way closed-loop information flow and control flow.

[0104] I. Edge side (radio station area)

[0105] As the direct control layer deployed on the power distribution network site, the core tasks of the edge side are "sensing" and "execution". Its hardware core is FPGA terminal hardware, and the energy storage PCS / photovoltaic inverter is the final execution mechanism.

[0106] FPGA terminal hardware:

[0107] Function: The hardware carrier for the aforementioned edge functions. FPGA (Field Programmable Gate Array), with its parallel processing, low latency, and high reliability, is perfectly suited for high-speed data acquisition, protocol conversion, and pulse transmission tasks.

[0108] Technical advantages: Compared with traditional industrial control computers, FPGA has no operating system bottleneck, strong anti-electromagnetic interference capability, and is more suitable for harsh industrial environments, ensuring the real-time performance and determinism of control.

[0109] Energy storage PCS / Photovoltaic inverter:

[0110] Function: Receives drive pulses from the FPGA and quickly and accurately adjusts its output active power (P) and reactive power (Q) to actually inject, absorb, or support power into the grid, thereby stabilizing voltage and suppressing oscillations.

[0111] Edge-side information flow and control processes include S21-S23:

[0112] S21. High-speed data acquisition (multi-channel synchronous monitoring)

[0113] Function: This is the nerve ending of the system. Through high-precision sensors and measurement units, it synchronously collects real-time data from multiple nodes within the distribution area at a sampling frequency of microseconds.

[0114] Acquisition signals include the output (P, Q) of the photovoltaic inverter, the operating status and SOC of the energy storage PCS, the voltage / current / frequency of key buses, and the load power.

[0115] Technical features: It adopts synchronous phasor measurement technology to ensure that all data have a unified timestamp, providing a high-quality spatiotemporally synchronized data source for subsequent accurate analysis.

[0116] S22. Protocol Conversion and Communication Management

[0117] Function: Serves as a "communication hub" on the edge. It is responsible for converting and encapsulating data from various industrial protocols (such as Modbus, CAN, and 104 protocol) collected by the front end into standard data formats (such as MQTT and HTTPS) suitable for long-distance transmission.

[0118] Technical features: It has data caching and disconnection resumption functions to ensure the integrity and continuity of data under network fluctuations.

[0119] S23. Control command reception and millisecond-level pulse transmission

[0120] Function: Serves as the "execution terminal" for cloud platform instructions. It receives optimized scheduling instructions issued by the cloud platform and, through the hardware parallel processing capabilities of the FPGA, parses the instructions into specific PWM pulse signals with precise timing.

[0121] Technical features: Extremely fast response speed (millisecond level), completely avoiding the latency and uncertainty caused by operating system scheduling in traditional CPU-based systems, and meeting the demanding real-time requirements for the control of power electronic equipment.

[0122] II. Cloud

[0123] As the decision-making layer of the system, the core tasks of the cloud are "analysis" and "decision-making," and its advantages lie in its powerful computing power and global perspective.

[0124] Big Data Platform (Historical and Real-Time Data Lake):

[0125] Function: The system's "memory". It receives and stores real-time status data uploaded from all edge terminals, while also archiving massive amounts of historical operational data, meteorological data, equipment model parameters, etc.

[0126] Technical features: It provides massive, multi-dimensional, and long-term data support for upper-level algorithms, and is the foundation for machine learning and deep mining.

[0127] Dynamic spatiotemporal correlation matrix operations and rolling optimization algorithm:

[0128] Function: The system's "core intelligence." This module carries the patented algorithm. Based on information from the data lake, it runs a dynamic spatiotemporal correlation matrix model in real time to predict future power fluctuation trends in the distribution area. Subsequently, it calls a rolling optimization algorithm to calculate a forward-looking scheduling strategy with the goal of optimizing the overall system's economy and stability.

[0129] Technical characteristics: It deals with complex high-dimensional optimization problems, which involve a large amount of computation, but the real-time requirements are relatively relaxed (seconds to minutes), making it suitable for completion on cloud platforms with powerful computing capabilities.

[0130] Cooperative scheduling strategy generation:

[0131] Function: Transforms the results of optimization algorithms into executable "combat instructions". It translates complex mathematical optimization results into specific and clear control commands, such as: "Command A energy storage station to start charging at rated power for 15 minutes at 10:15".

[0132] III. Two-way closed-loop information flow and control flow

[0133] Uplink data stream (edge ​​→ cloud): Real-time data from the edge is continuously uploaded to the cloud platform data lake after protocol conversion, via communication networks such as 4G / fiber optics, providing perceptual input for global optimization.

[0134] Downlink command stream (cloud → edge): The collaborative scheduling strategy generated by the cloud platform is sent as downlink commands to the FPGA terminal on the edge via the communication network. The FPGA converts these commands into drive pulses to control the execution of energy storage and photovoltaic devices.

[0135] S3. Design an adaptive rolling optimization engine: Solve the two-layer optimization model using a rolling process with a 5-second cycle.

[0136] Upper layer (hourly level): Minimize the total operating cost of the cluster:

[0137] ;

[0138] C grid The electricity purchase and sale price (unit: yuan / kWh) is a key economic signal affecting energy storage peak-valley arbitrage strategies; P grid (t) represents the power exchanged with the grid at time t (unit: kW), with positive values ​​representing electricity purchases and negative values ​​representing electricity sales. This is a decision variable in the optimization process; C deg is the energy storage loss cost coefficient (unit: yuan / kW), used to quantify the aging loss of the battery due to charging and discharging into economic cost, and to avoid overuse of energy storage; |PESS(t)| represents the absolute value of the energy storage charging and discharging power (unit: kW), reflecting the operating intensity of the equipment measured only by the power magnitude, regardless of the charging and discharging direction.

[0139] Lower layer (second level): Suppressing voltage deviation and power backfeed:

[0140] ;

[0141] It is the real-time voltage measurement value of the k-th node at time t;

[0142] This is the voltage reference value (rated voltage);

[0143] It is the absolute value of the voltage deviation at the k-th node;

[0144] It is the power fed back from the distribution network to the next higher level of the power grid at time t;

[0145] It is the absolute value of the power backfeed.

[0146] An LSTM short-term power prediction and correction module is introduced to predict the fluctuation trend in the next 30 seconds.

[0147] The timing logic for adaptive scrolling optimization is as follows:

[0148] (1) Prediction window layer: rolling update of ultra-short-term forecast

[0149] Initial phase: Perform "ultra-short-term irradiance prediction" to predict photovoltaic irradiance (or related variables, such as photovoltaic output and load) for a future period of time (e.g., tens of minutes to several hours), providing basic input for subsequent optimization.

[0150] Rolling mechanism: When time advances to the next cycle, a "new round of prediction" is started, and the prediction results of the future window are continuously updated (such as the prediction starting point moving forward with time to keep the prediction window length stable) to ensure the timeliness of the prediction data.

[0151] (2) Optimization and decision-making level: dynamic optimization based on prediction

[0152] Initial calculation: Based on the prediction results of the first stage, "rolling optimization calculation" is performed to solve the coordinated scheduling strategy of energy storage, photovoltaics and load (such as energy storage charging and discharging power, photovoltaic output regulation, etc.) through optimization algorithms (such as model predictive control, dynamic programming, etc.).

[0153] Iterative optimization: Along with the launch of the "new round of forecasting", a "new round of optimization" is carried out simultaneously - by combining the latest forecast data and the actual operation data of the feedback layer, the optimization objectives (such as cost and power smoothing) and constraints (such as energy storage capacity and grid limits) are adjusted, and the updated scheduling decisions are output.

[0154] (3) Actual scheduling layer: real-time execution of instructions

[0155] Initial execution: After the optimization layer outputs the first set of decisions, it immediately "executes the first scheduling instruction" to drive energy storage, photovoltaic and other equipment to perform regulation actions (such as energy storage charging and photovoltaic power limiting).

[0156] Dynamic updates: Whenever the optimization layer generates a "new round of optimization" results, the scheduling layer simultaneously "executes new instructions" to ensure that the equipment control strategy keeps up with the latest optimization results and achieves seamless connection between "decision-execution".

[0157] (4) Feedback and update layer: the key support for closed-loop control

[0158] Initial monitoring: While the dispatch command is being executed, “Status monitoring and feedback” is initiated to collect actual operating data (such as energy storage SOC, actual photovoltaic output, load power, etc.).

[0159] Continuous feedback: As time progresses, "new data feedback" is continuously uploaded - real-time running data is sent back to the prediction layer (to correct the prediction model and improve the accuracy of subsequent predictions) and the optimization layer (to update state variables and ensure the authenticity of optimization decisions), forming a closed loop of "actual state → prediction / optimization → decision → actual state".

[0160] (5) The temporal correlation and value of rolling optimization:

[0161] Time correlation: The tasks at each layer proceed in a cyclical manner according to the time sequence of "prediction start-up → optimization calculation → instruction execution → data feedback → new round of prediction". Within each rolling cycle, the four layers of tasks are interconnected and iterate in parallel (such as the feedback data of the previous cycle supporting the prediction optimization of the current cycle).

[0162] In another embodiment, the present invention provides an adaptive collaborative scheduling system for energy storage in distribution transformer substations, used to implement the above-mentioned adaptive collaborative scheduling method for energy storage in distribution transformer substations, such as... Figure 4 As shown, the system includes:

[0163] The spatiotemporal fluctuation correlation matrix construction module is used to acquire cloud cluster data and preprocess it, construct the spatiotemporal fluctuation correlation matrix based on the preprocessed data, and run the spatiotemporal fluctuation correlation matrix.

[0164] The edge-cloud collaborative control architecture module is used to develop and run the edge-cloud collaborative control architecture.

[0165] The Adaptive Scrolling Optimization Engine module is used to design and run the adaptive scrolling optimization engine.

[0166] This invention addresses the issue of distributed photovoltaic (PV) output being affected by cloud movement and changes in sunlight, and the strong spatiotemporal correlation of power fluctuations between adjacent PV distribution areas (e.g., after cloud cover causes a sudden drop in power in distribution area A, distribution area B follows suit within 30 seconds). This is addressed by constructing a dynamic spatiotemporal correlation matrix. This approach quantifies the transmission effect of distributed resource fluctuations, breaking through the "isolated scheduling" mindset. It addresses the issue of delayed control response in traditional scheduling by constructing an adaptive rolling optimization engine—upper layer (hourly) + lower layer (second-level)—to achieve a dynamic balance between economy and stability. Edge-cloud collaborative control compresses control response latency to within 1.5 seconds, meeting the stringent power quality requirements of rural power grids.

[0167] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0168] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0169] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0170] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0171] Those skilled in the art will 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, or a combination of computer software and electronic hardware. 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 implementations should not be considered beyond the scope of this invention.

[0172] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0173] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0174] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0175] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0176] 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, or the part that contributes to the prior art, or a part 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.

[0177] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included 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 method for adaptive and coordinated scheduling of energy storage in distribution substations, characterized in that, The method includes the following steps: S1. Construct a spatiotemporal fluctuation correlation matrix, including acquiring cloud cluster data and preprocessing it, and constructing the spatiotemporal fluctuation correlation matrix based on the preprocessed data; Data acquisition and preprocessing include acquiring meteorological data layers. Preprocessing involves identifying and segmenting cloud clusters in the acquired meteorological data layers and vectorizing the cloud cluster motion to obtain the cloud cluster movement vector field. The geographical boundaries of all distribution substations within the jurisdiction and the precise coordinates of photovoltaic power stations are acquired. Based on the geographical information of the substations, the distance between substations and the azimuth angle are calculated, and the trajectory of meteorological satellite cloud images is predicted. Constructing the spatiotemporal fluctuation correlation matrix involves inputting the results of the above steps to construct a high-dimensional matrix, the spatiotemporal fluctuation correlation matrix C. ij The algorithm for constructing the model is as follows: By using meteorological satellite cloud image trajectory prediction and historical photovoltaic power output data of the power distribution area, a spatiotemporal transmission model for photovoltaic fluctuations is established. ； in , : Geographic coordinates of the transformer area; σ is the cloud movement scale parameter; Δt is the wave propagation delay; PVi represents the photovoltaic output of the i-th transformer area; PVj represents the photovoltaic output of the j-th transformer area; S2. Develop an edge-cloud collaborative control architecture: Edge-cloud collaborative control consists of two parts: the cloud and the edge. They are connected through a reliable communication network to form a two-way closed-loop information flow and control flow. S3. Design an adaptive rolling optimization engine: Solve the two-layer optimization model using a fixed interval as the cycle. Upper layer: Minimize the total operating cost of the cluster; Lower layer: Suppress voltage deviation and power backfeed; An LSTM short-term power prediction and correction module is introduced to predict fluctuation trends over a future period.

2. The adaptive cooperative scheduling method for distribution radio station energy storage according to claim 1, characterized in that, S1. Construct the spatiotemporal fluctuation correlation matrix. The specific steps are as follows: S11. Input: Meteorological satellite remote sensing cloud image, obtain meteorological data layer; S12. Image preprocessing and feature extraction: Cloud cluster identification and segmentation are performed on the meteorological data layers obtained above to accurately identify discrete cloud cluster entities with complete shapes and extract their key features; S13. Calculation of cloud cluster movement vector field and vectorization of cloud cluster motion: Using the optical flow method or cross-correlation coefficient method, multiple frames of cloud images are continuously tracked to calculate the cloud cluster's movement velocity vector v_c and movement direction angle θ_c. This will form a dynamic cloud cluster movement vector field. The optical flow method or cross-correlation method is used to analyze the satellite cloud image sequence in consecutive time intervals. By calculating the displacement of the same cloud cluster feature in adjacent frames, a two-dimensional vector field with velocity magnitude and direction covering the entire observation area is generated. S14. Cloud movement trajectory prediction: Based on the calculated instantaneous moving vector field, the guiding airflow data of the numerical weather prediction model is introduced as the background field to predict the movement trajectory and morphological evolution trend of each identified cloud in the next 0-4 hours, providing a direct basis for predicting future photovoltaic power output fluctuations. S15. Geographic coordinate mapping of distribution areas: Based on the built-in high-precision geographic information system layer, the geographical boundaries of all distribution areas and the precise coordinates of photovoltaic power stations within the jurisdiction are obtained. Based on the geographic information L_i, L_j of the distribution areas, the distance between distribution areas d_ij and the azimuth angle ψ_ij are calculated. The predicted cloud movement trajectory layer is overlaid with the geographic information system layer and spatial correlation analysis is performed to determine when a specific cloud will cover which distribution area or photovoltaic array, thereby transforming meteorological forecasts into early warnings of the impact on specific power units. S16. Prediction of future irradiance attenuation probability: Based on the cloud type, thickness and coverage predicted in S14 and S15, and combined with the atmospheric radiative transfer model, calculate the irradiance intensity reaching the surface of the substation area in each future time segment and its attenuation probability and degree. The output is a probability distribution function in time series, providing input for uncertainty optimization. S17. Construct a dynamic spatiotemporal correlation matrix: The results of the above steps are used to construct a high-dimensional matrix. The rows of this matrix represent different time slices in the future, and the columns represent different distributed photovoltaic nodes in the power grid. Each element in the matrix is ​​a correlation weight value, which comprehensively reflects the temporal and spatial correlations. This matrix is ​​dynamically updated and serves as a direct basis for the subsequent forward-looking decision-making and global optimization of the energy storage collaborative scheduling algorithm.

3. The adaptive cooperative scheduling method for distribution radio station energy storage according to claim 1, characterized in that, The specific process for establishing a spatiotemporal transmission model for photovoltaic fluctuations is as follows: The satellite cloud image data obtained from the meteorological data layer is processed by cloud cluster segmentation and motion vectorization to obtain v_c and θ_c. Based on the geographical information of the transformer area, L_i, L_j, calculate the distance between transformer areas d_ij and the azimuth angle ψ_ij. The theoretical time delay Δt_ij is calculated based on the moving speed vector v_c, the moving direction angle θ_c, the station spacing d_ij, and the azimuth angle ψ_ij. The high-frequency sampling data of photovoltaic power output in the transformer area is cleaned and preprocessed. The correlation coefficient R_ij with time shift is calculated by using the theoretical time delay Δt_ij and the cleaned and preprocessed high-frequency sampling data of photovoltaic power output in the transformer area. Set the initial scale parameter σ; Calculate the spatial correlation degree A_ij; The final correlation degree C_ij is calculated based on the time-shifted correlation coefficient R_ij and the spatial correlation degree A_ij. The final correlation coefficient C_ij is output and iterated through all station pairs. If a pair is found, the next station pair i,j is selected and the theoretical time delay Δt_ij is calculated based on the moving speed vector v_c, moving direction angle θ_c, station spacing d_ij, and azimuth angle ψ_ij. Otherwise, a preliminary correlation matrix C is generated. The preliminary correlation matrix C is dynamically updated and optimized, and the final spatiotemporal fluctuation correlation matrix is ​​output. A new frame of satellite data is taken, and the cloud vector change is judged. If the cloud vector change is less than the threshold, the step of generating the preliminary correlation matrix C is entered. The preliminary correlation matrix C is dynamically updated and optimized, and the final spatiotemporal fluctuation correlation matrix C is output. If the cloud vector change is greater than the threshold, the matrix is ​​recalculated.

4. The adaptive cooperative scheduling method for energy storage in distribution substations according to claim 1, characterized in that, The edge side of the edge-cloud collaborative control architecture: FPGA computing cores deployed at the transformer substation terminals execute millisecond-level reactive power regulation commands. ； Q ESS Output value: This indicates the amount of reactive power that the energy storage converter is required to generate or absorb, based on calculations. K P The core coefficient determines the strength of the controller's response to voltage deviations. U meas Input value: The actual voltage value obtained in real time by sensors deployed at key nodes in the transformer area; U ref : Input value, the rated voltage that the power grid is expected to maintain during operation; The edge-cloud collaborative control architecture in the cloud: generates cluster collaboration strategies based on spatiotemporal correlation matrices and issues GOOSE commands through 5G slicing networks.

5. The adaptive cooperative scheduling method for energy storage in distribution substations according to claim 1, characterized in that, The edge-side information flow and control process of the edge-cloud collaborative control architecture includes S21-S23: S21. High-speed data acquisition; S22. Protocol Conversion and Communication Management; S23. Control command reception and millisecond-level pulse transmission.

6. The adaptive cooperative scheduling method for energy storage in distribution substations according to claim 4, characterized in that, The edge-cloud collaborative control architecture in the cloud: Based on dynamic spatiotemporal correlation matrix operation and rolling optimization algorithm, and based on the information of the data lake, the dynamic spatiotemporal correlation matrix model is run in real time to predict the future power fluctuation trend of the distribution area. The rolling optimization algorithm is called to calculate a set of forward-looking scheduling strategies with the goal of optimizing the economy and stability of the entire system.

7. The adaptive cooperative scheduling method for energy storage in distribution substations according to claim 1, characterized in that, Bidirectional closed-loop information and control flows include uplink data flow and downlink command flow. Uplink data stream: Real-time data from the edge is continuously uploaded to the cloud platform data lake after protocol conversion, via communication networks such as 4G / fiber optics, providing perceptual input for global optimization; Downlink command flow: The collaborative scheduling strategy generated by the cloud platform is sent as downlink commands to the FPGA terminal on the edge via the communication network. The FPGA converts these commands into drive pulses to control the execution of energy storage and photovoltaic equipment.

8. The adaptive cooperative scheduling method for energy storage in distribution substations according to claim 1, characterized in that, Adaptive scrolling optimization timing includes: Forecast window layer: Rolling updates for ultra-short-term forecasts; Optimization and Decision-Making Level: Dynamic Optimization Based on Prediction; Actual scheduling layer: Real-time execution of instructions; Feedback and update layer: a key support for closed-loop control; The temporal correlation and value of rolling optimization.

9. A distribution substation energy storage adaptive collaborative scheduling system, used to implement the distribution substation energy storage adaptive collaborative scheduling method as described in any one of claims 1-8, characterized in that, The system includes: The spatiotemporal fluctuation correlation matrix construction module is used to acquire cloud cluster data and preprocess it, construct the spatiotemporal fluctuation correlation matrix based on the preprocessed data, and run the spatiotemporal fluctuation correlation matrix. The edge-cloud collaborative control architecture module is used to develop and run the edge-cloud collaborative control architecture. The Adaptive Scrolling Optimization Engine module is used to design and run the adaptive scrolling optimization engine.

Images