Cloud-edge layered collaborative energy storage EMS dual-mode energy scheduling and automatic configuration method

Abstract

The application discloses a cloud-edge layered cooperative energy storage EMS dual-mode energy scheduling and automatic configuration method. Through constructing a layered cooperative architecture of cloud global optimization and edge local autonomy, designing economic steady-state scheduling and dynamic emergency scheduling dual operation modes, combining big data research and adaptive algorithm, the application realizes automatic switching of energy storage EMS scheduling mode, automatic configuration of operation parameters, collaborative optimization and scheduling of cluster resources, greatly improves the operation stability, real-time regulation and control and economic benefits of the energy storage system, and adapts to the multi-scene, dynamic and intelligent regulation and control requirements of the new power system.

Description

Technical Field

[0001] This invention belongs to the field of energy storage management and grid dispatching technology, specifically involving a cloud-edge hierarchical collaborative dual-mode energy dispatching and automatic configuration method for energy storage EMS. It is applicable to the fine-grained energy dispatching, adaptive parameter configuration, and multi-scenario stable operation control of distributed energy storage clusters, photovoltaic-storage integrated power stations, integrated energy systems in industrial parks, and energy storage on the distribution network side. Background Technology

[0002] With the continuous advancement of the construction of new power systems and the large-scale grid connection of intermittent new energy sources such as wind power and photovoltaics, distributed energy storage devices, with their flexible charging and discharging regulation capabilities, have become core equipment for smoothing fluctuations in new energy output, peak shaving and valley filling, supporting the voltage stability of the distribution network, and improving the clean energy consumption rate. The energy storage management system (EMS), as the core control unit of the energy storage system, undertakes key functions such as data acquisition, status monitoring, energy dispatching, parameter configuration, and fault diagnosis. Its control performance directly determines the operating efficiency, safety, stability, and economic benefits of the energy storage system.

[0003] Current energy storage EMS (Energy Management System) dispatching technologies generally adopt a single centralized dispatching architecture, with the cloud-based master station uniformly handling the dispatching calculations and command issuance for all network energy storage resources. While this architecture is feasible for small-scale, centralized energy storage power plants, it suffers from numerous technical shortcomings in large-scale distributed energy storage clusters and multi-regional heterogeneous energy storage scenarios. Firstly, centralized cloud-based dispatching suffers from long data transmission links and high latency, making it unsuitable for dynamic scenarios such as real-time power fluctuations and load surges in the distribution network. Real-time control responses are delayed, easily leading to problems such as voltage exceeding limits, power imbalance, and overcharging / over-discharging of energy storage. Secondly, existing EMS systems mostly employ a single dispatching operation mode, which cannot adapt to differentiated scenarios such as daily steady-state operation of the power grid, significant fluctuations in new energy sources, grid fault emergencies, and electricity market transactions. The dispatching strategies lack versatility and adaptability.

[0004] Meanwhile, existing energy storage EMS configuration methods are mostly manual and static. Maintenance personnel must manually set core parameters such as energy storage charging and discharging thresholds, scheduling weights, protection parameters, and coordination strategies based on experience, which has significant drawbacks. On the one hand, manual configuration relies on maintenance experience, resulting in low parameter matching and an inability to adaptively adjust to dynamic changes in load characteristics, renewable energy output, and grid operating status, leading to low energy storage resource utilization and poor scheduling economy. On the other hand, in large-scale energy storage cluster scenarios, the large number of devices and complex topologies make manual configuration labor-intensive and inefficient, prone to parameter configuration errors and inconsistencies, leading to cluster coordination failures and equipment malfunctions.

[0005] Furthermore, existing cloud-edge collaborative energy storage control technologies primarily focus on data interaction and task migration, failing to achieve deep integration of dual-mode scheduling under a layered architecture. The collaborative logic between cloud-based global optimization and edge-based local autonomy is fragmented, resulting in abrupt switching between steady-state and dynamic emergency scheduling. Moreover, a systematic automatic configuration mechanism is lacking, hindering the integrated adaptive configuration of scheduling modes, equipment parameters, and collaborative strategies. Specifically, this manifests as follows: during grid steady-state operation, optimal economic scheduling of the entire network's energy storage resources cannot be achieved; during grid dynamic disturbances, edge nodes cannot respond quickly locally, relying excessively on cloud commands, leading to insufficient real-time control; and during scenario switching, scheduling parameters require manual modification, failing to automatically adapt to scenario changes, severely restricting the large-scale and intelligent application of large-scale distributed energy storage clusters.

[0006] In summary, existing technologies suffer from bottlenecks such as a single scheduling architecture, poor mode adaptability, low level of configuration intelligence, insufficient cloud-edge collaboration, imbalance between steady-state and dynamic scenario regulation, and low resource utilization. There is an urgent need for an energy storage EMS regulation method that can achieve cloud-edge hierarchical collaboration, dual-mode adaptive scheduling, and automatic configuration across all scenarios to address the many shortcomings of existing technologies. Summary of the Invention

[0007] The purpose of this invention is to address the technical problems of existing energy storage EMS scheduling architectures, such as rigidity, single operating modes, insufficient intelligent configuration, low cloud-edge collaboration efficiency, and poor adaptability to multiple scenarios. This invention provides a cloud-edge layered collaborative dual-mode energy scheduling and automatic configuration method for energy storage EMS. By constructing a layered collaborative architecture with global optimization in the cloud and local autonomy at the edge, this invention designs dual operating modes of economic steady-state scheduling and dynamic emergency scheduling. Combined with big data analysis and adaptive algorithms, it achieves automatic switching of energy storage EMS scheduling modes, automatic configuration of operating parameters, and collaborative optimization scheduling of cluster resources. This significantly improves the operational stability, real-time control, and economic benefits of energy storage systems, adapting to the multi-scenario, dynamic, and intelligent control needs of new power systems.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A cloud-edge hierarchical collaborative energy storage EMS dual-mode energy scheduling and automatic configuration method includes six core steps: cloud-edge hierarchical collaborative architecture construction, EMS dual scheduling mode modeling, scene identification and mode adaptive switching, hierarchical collaborative energy scheduling, full parameter automatic configuration, and closed-loop correction of operating status. The specific steps are as follows:

[0010] S1. Build a cloud-edge layered collaborative energy storage EMS system architecture. The architecture is divided into three layers: cloud scheduling layer, edge collaboration layer, and energy storage device layer, to realize layered data interaction and hierarchical scheduling and control.

[0011] The cloud-based scheduling layer deploys a global energy management platform, responsible for the statistics of energy storage resources across the entire network, load and new energy output forecasting, global optimization scheduling calculations, scheduling strategy distribution, scenario situation analysis, and unified parameter distribution, achieving global economic optimal control of cross-regional energy storage clusters. The edge collaboration layer deploys multiple edge EMS nodes, each edge node corresponding to and managing a distributed energy storage cluster within a region, responsible for real-time local data acquisition, local status analysis, instant power control, cloud command parsing and execution, and local parameter adaptive correction, achieving rapid autonomous control of energy storage devices within the region. The energy storage device layer includes various distributed energy storage devices, photovoltaic units, load equipment, and monitoring sensors, responsible for executing charging and discharging commands, uploading real-time operating data, and completing underlying energy interaction and status feedback.

[0012] The cloud scheduling layer and the edge collaboration layer employ dual-link communication of 5G and fiber optics to achieve high-volume, low-latency data interaction. The edge collaboration layer and the energy storage device layer combine industrial Ethernet and LoRa wireless communication to ensure millisecond-level transmission of real-time commands. Simultaneously, a layered data interaction protocol is constructed, with the cloud sending global scheduling targets, baseline parameters, and scene switching commands to edge nodes; and edge nodes uploading local load data, energy storage status data, equipment fault information, and local scheduling execution results to the cloud, forming a two-way closed-loop data interaction link.

[0013] S2. Construct an EMS dual-mode energy dispatch model, including an economic steady-state dispatch mode and a dynamic emergency dispatch mode, to adapt to different power grid operation scenarios.

[0014] This invention designs a dual independent scheduling mode, which can adaptively switch according to the grid operating status, the fluctuation range of new energy output, the load change rate, and the equipment operating status. The modeling methods for the two modes are as follows:

[0015] S21. Economic steady-state dispatch mode modeling: This mode is suitable for conventional scenarios such as steady-state operation of the power grid, stable output of new energy sources, and smooth load fluctuations. It constructs a global optimization dispatch model with multiple optimization objectives, namely maximizing the economic benefits of the entire network energy storage operation, maximizing the clean energy consumption rate, and minimizing network losses.

[0016] The objective function specifically includes: maximizing energy storage peak-valley arbitrage profits, minimizing renewable energy curtailment losses, minimizing grid transmission losses, and minimizing energy storage charging and discharging losses. Constraints include: upper and lower limits of the energy storage unit's state of charge (SOC), maximum charging and discharging power constraints, distribution network voltage amplitude constraints, line transmission power constraints, load supply and demand balance constraints, and renewable energy output absorption constraints.

[0017] The economic steady-state dispatch mode is dominated by the cloud dispatch layer. Based on the day-ahead forecast data for the next 24 hours and the intraday rolling forecast data for 4 hours, the improved multi-objective particle swarm algorithm is used to solve the global optimal solution, generating the day-ahead dispatch plan and intraday rolling optimization instructions for edge nodes in each region, so as to realize the overall optimization and allocation of energy storage resources across the entire network.

[0018] The cloud-based economic steady-state multi-objective optimization scheduling algorithm is as follows:

[0019] 1) Applicable conditions;

[0020] Scene recognition output M run =steady, cloud-based global overall calculation, outputs the optimal total scheduling power for each edge partition;

[0021] 2) Decision variables;

[0022] P ess,i,t P represents the power output of the i-th energy storage unit during time period t. ess >0 charging, P ess <0 discharge.

[0023] 3) Multi-objective function;

[0024] Objective 1 (Minimize arbitrage costs and maximize equivalent returns)

[0025] C t I(p ess,i,t >0)-p ess,i,t dis C t I( pess,i,t <0);

[0026] Objective 2 (Minimize waste light loss)

[0027] 2= (p pv,t +p ess,i,t -p load,t ,0);

[0028] Objective 3 (Minimize network loss)

[0029] 3= ;

[0030] Weighted fusion total cost function;

[0031] F total = + + , =0.5, =0.3, =0.2, ;

[0032] 4) All constraints;

[0033] Power constraints;

[0034] -P dis,max ≤P ess,i,t ≤P ch,max ;

[0035] SOC timing constraints;

[0036] ;

[0037] SOC min =0.2≤SOC i,t ≤SOC max =0.9;

[0038] Regional power balance constraints:

[0039] P pv,t + P ess,i,t =P load,t +P grid,t ;

[0040] 5) Improve the MOPSO iterative calculation steps;

[0041] Step 1 Initialization: Population size Npop=60, maximum iteration Gmax=20; initialize particle position and velocity, cut out-of-bounds power to the constraint boundary, calculate initial Ftotal to generate Pareto non-dominated solution set.

[0042] Step 2 Particle Update Formula (Linear Decrease of Inertial Weight)

[0043] V k+1 = V k +c1r1(P best -X k )+c2r2(G best -X k );

[0044] X k+1 =X k +V k+1 ;

[0045] = start - , start =0.9, end =0.4;

[0046] In the formula: c1=c2=1.49, r1, r2 A random number, where g is the current iteration number.

[0047] Step 3 Constraint Correction: When the updated power exceeds the SOC / power boundary, a forced limit is applied, and the total cost is recalculated.

[0048] Step 4: Update the optimal value: Compare the costs of the old and new particles to update the individual optimal Pbest; select the most economically optimal particle from the Pareto solution set as the global optimal Gbest.

[0049] Step 5 Convergence Criterion: The rate of change of Gbest cost is <10 for 15 consecutive iterations. -4 Terminate the iteration.

[0050] Step 6: Output the optimal total power P for each edge partition. * area,t It is then sent to the edge EMS.

[0051] 6) Edge secondary power allocator algorithm;

[0052] After calculating the total power of the edge receiving zone, the power is distributed equally according to the state weight of each individual energy storage unit:

[0053] ;

[0054] Among them, W k The weight of the kth energy storage unit is dynamically generated based on real-time SOC and equipment health SOH, enabling balanced charging and discharging of the energy storage cluster.

[0055] S22. Dynamic Emergency Dispatch Mode Modeling: This mode is adapted to dynamic disturbance scenarios such as large fluctuations in new energy sources, sudden increases and decreases in load, grid voltage / frequency exceeding limits, local line overload, and minor equipment faults. With the core objectives of rapid recovery of grid operation status, rapid suppression of power fluctuations, and optimal system operation stability, a local rapid response dispatch model is constructed.

[0056] The objective functions specifically include: minimizing grid frequency deviation, minimizing node voltage deviation, minimizing power fluctuation amplitude, and minimizing disturbance recovery time. Constraints include: dynamic power regulation response time constraints, energy storage fast charging and discharging power constraints, equipment safe operation constraints under fault conditions, and grid emergency voltage and frequency regulation constraints.

[0057] The dynamic emergency dispatch mode is led by the edge collaboration layer. It does not need to wait for cloud instructions. Based on millisecond-level operation data collected locally in real time, it uses a fast gradient descent algorithm to complete the instantaneous dispatch solution and adjust the charging and discharging power of energy storage devices in the region in real time, so as to realize local rapid autonomous control in disturbance scenarios.

[0058] The edge dynamic emergency adaptive gradient descent scheduling algorithm is as follows:

[0059] 1) Applicable conditions;

[0060] Scene recognition output M run =dynamic, independent local operation at the edge, shielding steady-state scheduling commands from the cloud, and smoothing voltage and frequency disturbances in milliseconds.

[0061] 2) Decision variables;

[0062] P edge (t); emergency regulation power of energy storage in the edge area.

[0063] 3) Loss optimization function;

[0064] ;

[0065] in, Voltage regulation weight, Frequency adjustment weight, K u k f Power-voltage / frequency coupling coefficient.

[0066] 4) Constraints;

[0067] Reuse the energy storage power and SOC constraints of module 2; add calculation time constraints; single solution time ≤ 20ms.

[0068] 5) Complete iterative calculation steps;

[0069] Step 1: Calculate the disturbance intensity and adaptively update the control parameters;

[0070] ;

[0071] Step 2: Solve for the gradient of the loss function;

[0072] ;

[0073] Step 3: Iterate through gradient descent to update the power;

[0074] ;

[0075] Step 4 Boundary trimming: Limit Pnew to the allowable charge and discharge range of energy storage, and verify the upper and lower limits of SOC.

[0076] Step 5 Convergence Criterion: Loss change in a single iteration < 10 -6 The calculation stops when the number of iterations reaches 20.

[0077] Step 6: Cloud-Edge Collaborative Linkage: Edge Upload Disturbance Intensity D dist The disturbance type is sent to the cloud; the cloud coordinates the energy storage in neighboring areas to issue coordinated adjustment commands, and after the disturbance is eliminated, the execution power smoothly drops back, preparing to switch back to steady-state mode.

[0078] S3. Construct a multi-dimensional scene recognition mechanism to achieve adaptive switching between dual scheduling modes;

[0079] A scenario discrimination index system is constructed based on grid operation parameters, renewable energy output parameters, load parameters, and energy storage operation status parameters. The indicators include: renewable energy output volatility, load change rate, grid voltage deviation rate, grid frequency deviation, line load factor, and energy storage SOC change rate. Steady-state and dynamic thresholds for each indicator are preset, and automatic scenario identification and scheduling mode switching are achieved through real-time data analysis.

[0080] When all monitoring indicators are within the steady-state threshold range, the power grid is determined to be in a normal steady-state scenario, and the system automatically switches to the economic steady-state dispatch mode, with global optimized dispatch executed by the cloud. When any monitoring indicator exceeds the steady-state threshold and enters the dynamic threshold range, the power grid is determined to be in a dynamic disturbance scenario, and the system immediately cuts off the cloud-based steady-state dispatch command and automatically switches to the dynamic emergency dispatch mode, with local rapid control executed by edge nodes. When the disturbance is eliminated, all indicators return to the steady-state threshold range, and this continues for a preset duration, the system automatically switches back to the economic steady-state dispatch mode and restores the global economic optimal dispatch.

[0081] At the same time, a mode switching anti-jitter mechanism is set up to avoid frequent erroneous mode switching caused by short-term parameter fluctuations, thus ensuring the stability of scheduling operations.

[0082] The multi-dimensional scene recognition and stabilization mode switching algorithm is as follows:

[0083] 1) Algorithm input and output:

[0084] Input: New energy volatility F PV Load change rate F L Voltage deviation ΔU, frequency deviation Δf, line load factor R line SOC change rate R SOC Current time t now Disturbance timing buffer t abn Previous operating mode M last

[0085] Output: Current running scheduling mode M run(steady state / dynamic emergency)

[0086] 2) Definition of the judgment threshold;

[0087] F pv,sta =0.05,F l,sta =0.08,U sta =0.03;

[0088] f sta =0.05,R line,sta =0.9,R SOC,sta =0.02;

[0089] Image stabilization duration: T anti =30s

[0090] 3) Calculation of disturbance markers;

[0091] Flag dis = 1 F pv >F pv,sta F L >F L,sta |ΔU|>U sta |Δf|>f sta R line >R line,sta

[0092] Flag dis = 0 All indicators fall within the steady-state range;

[0093] 4) Anti-shake mode switching logic;

[0094] If Mlast=steady;

[0095] When Flag dis =1: If t abn =0, assign the value t abn =t now If t now -t abn ≥T anti M run =dynamic,t abn =0,

[0096] When Flag dis =0: Reset t abn =0, M run =steady.

[0097] If Mlast=dynamic;

[0098] When Flag dis =1: If t abn =0, assign the value t abn =t now If t now -t abn ≥T anti M run =steady, t abn =0,

[0099] When Flag dis =0: Reset t abn =0, M run =dynamic.

[0100] 5) Supplementary engineering constraints;

[0101] 1. Data sampling period is 1 second;

[0102] 2. The power smoothing coefficient is increased by 0.9~1.0 during mode switching to suppress the step impact of energy storage power;

[0103] 3. Disturbance timing data is cached when the device is powered off and automatically resumes timing upon restarting.

[0104] S4. Based on the cloud-edge layered collaborative architecture, execute dual-mode layered energy scheduling;

[0105] S41. Cloud-Edge Layered Collaborative Scheduling under Economic Steady-State Mode: The cloud scheduling layer completes the aggregation of network-wide data, load and renewable energy output forecasting, and global multi-objective optimization, generating optimal power scheduling instructions, energy storage charging and discharging plans, and clean energy consumption ratios for each edge region, and distributing them to each edge node. After receiving the global scheduling instructions from the cloud, the edge nodes, based on the SOC status, equipment capacity, and health parameters of each energy storage device in their region, perform secondary fine-tuning of the cloud instructions, breaking down the total regional scheduling power into charging and discharging instructions for individual energy storage devices, and distributing them to the underlying energy storage devices for execution. Simultaneously, the edge nodes monitor the local execution effect in real time and upload the operating data to the cloud in real time. The cloud performs daily rolling optimization based on the feedback data, corrects the scheduling plan, and achieves layered collaborative steady-state scheduling of "global cloud optimization + precise edge implementation".

[0106] S42. Cloud-Edge Layered Collaborative Scheduling in Dynamic Emergency Mode: After a scenario switch, edge nodes prioritize local autonomous scheduling, ignoring routine cloud-based scheduling commands. Based on real-time local disturbance data, they quickly calculate the emergency charging and discharging regulation power of energy storage within the region, instantly smoothing power fluctuations and correcting voltage and frequency deviations to ensure real-time grid stability. Simultaneously with local control, edge nodes upload disturbance type, disturbance amplitude, and local control data to the cloud in real time. Upon receiving the data, the cloud completes a comprehensive network situation assessment, coordinates energy storage resources in adjacent areas, and issues collaborative control commands to assist the disturbed area in quickly restoring steady state, achieving layered emergency scheduling of "rapid local response at the edge + global collaborative support from the cloud." After the disturbance is eliminated, cloud-edge collaboration smoothly transitions scheduling commands, avoiding sudden power fluctuations caused by mode switching.

[0107] S5. Enables full-parameter adaptive automatic configuration of the energy storage EMS;

[0108] Based on the dynamic changes in scheduling mode, operating scenario, and equipment status, an automatic parameter configuration model is constructed to achieve fully automatic adaptive configuration of EMS scheduling parameters, equipment operating parameters, collaborative control parameters, and protection threshold parameters, replacing the traditional manual static configuration. Specifically, it includes three parts: automatic configuration of steady-state parameters, automatic configuration of dynamic parameters, and correction of scenario linkage parameters.

[0109] S51. Automatic configuration of steady-state parameters: Under the economic steady-state dispatch mode, the cloud automatically optimizes and configures global parameters such as the upper and lower limits of energy storage charging and discharging SOC, peak-valley arbitrage time threshold, network dispatch weight, network loss optimization coefficient, and new energy consumption priority based on the characteristics of the entire network load, the output pattern of new energy sources, and the characteristics of electricity price periods. The edge nodes automatically correct local parameters such as the charging and discharging rate, standby threshold, and dispatch response accuracy of a single energy storage device based on the characteristics of the local equipment, so as to achieve optimal matching of steady-state scenario parameters.

[0110] S52. Automatic configuration of dynamic parameters: In dynamic emergency dispatch mode, edge nodes automatically and adaptively configure dynamic control parameters such as emergency adjustment power threshold, response delay parameters, voltage and frequency regulation weights, fault protection thresholds, and power adjustment step size according to the type of disturbance (power fluctuation, voltage over-limit, frequency deviation, line overload) and the intensity of the disturbance. This ensures accurate adaptation of control parameters under different disturbance scenarios and improves emergency control efficiency.

[0111] S53, Scene Linkage Parameter Correction: After each scheduling mode switch and scene switch, the system uses machine learning algorithms to iteratively optimize the parameter configuration model based on the scheduling effect, control error, and equipment wear data of this operation. This automatically corrects various benchmark parameters and threshold parameters, achieving continuous optimization of parameter configuration and adapting to the dynamic changes of long-term operating scenarios.

[0112] S6. Construct a closed-loop correction mechanism for operational status to ensure continuous optimization of scheduling and configuration accuracy;

[0113] The system collects real-time data on dual-mode scheduling operation, parameter configuration execution, power grid operation status, and energy storage device loss. It then constructs a scheduling performance evaluation index system, including: scheduling power deviation rate, voltage and frequency stability, energy storage resource utilization rate, economic benefit efficiency, equipment loss rate, and scenario switching accuracy. Based on these evaluation indicators, the system continuously assesses the suitability of the current scheduling strategy and configuration parameters. If any indicators fail to meet standards or control errors exceed thresholds, a closed-loop correction process is automatically triggered. The global scheduling model and parameters are re-optimized in the cloud, and edge nodes correct their local control strategies, achieving a fully closed-loop intelligent control system encompassing "scheduling-configuration-operation-evaluation-correction."

[0114] The technical solution of the present invention has the following beneficial technical effects:

[0115] 1. This invention constructs a cloud-edge layered collaborative architecture that features global optimization in the cloud and local autonomy at the edge. This breaks through the bottlenecks of high latency, slow response, and poor adaptability inherent in traditional centralized scheduling architectures, achieving a deep integration of global economic optimization and rapid local response. The cloud coordinates network-wide resources for economical scheduling, while the edge, based on local scenarios, achieves millisecond-level emergency control. The clear division of labor and tight collaboration significantly improve the overall control performance of large-scale distributed energy storage clusters.

[0116] 2. This invention innovatively designs a dual dispatch mode of economic steady-state and dynamic emergency, which can adaptively switch according to the power grid operation scenario, solving the defect of the existing EMS single dispatch mode being unable to adapt to multiple operation scenarios. In the steady-state scenario, it maximizes the economic benefits of energy storage and the renewable energy absorption rate, while in the dynamic disturbance scenario, it quickly smooths out fluctuations and ensures the stability of the power grid, taking into account both the economy and safety of energy storage operation, and adapting to the complex dynamic operation scenarios of new power systems.

[0117] 3. This invention achieves fully automatic adaptive configuration of all parameters of the energy storage EMS, completely replacing the traditional manual static configuration method and solving the problems of low efficiency, large error, and poor adaptability of manual configuration. It can optimize parameters in real time according to dynamic changes in load, new energy sources, and grid status, significantly improving parameter matching accuracy, effectively increasing the utilization rate of energy storage resources, reducing equipment operating losses, and reducing manual operation and maintenance costs.

[0118] 4. This invention has established a comprehensive scene recognition, mode anti-shaking switching and closed-loop correction mechanism to avoid frequent erroneous switching of scheduling modes and failure of control parameters. Through continuous iterative optimization of scheduling models and configuration parameters, it enables the EMS system to achieve long-term self-adaptation, self-optimization and self-maintenance, and significantly improves the intelligent operation level of energy storage clusters.

[0119] 5. The technical solution of this invention has strong versatility and wide applicability. It can be adapted to various scenarios such as centralized energy storage power stations, distributed energy storage clusters, photovoltaic-storage integrated systems, and integrated energy systems in industrial parks. It is compatible with various types of energy storage equipment such as lithium batteries, flywheel energy storage, and supercapacitors, and has extremely strong engineering implementation value and promotion prospects. Attached Figure Description

[0120] Figure 1 This is a flowchart of the cloud-edge hierarchical collaborative energy storage EMS dual-mode energy scheduling and automatic configuration method provided by the present invention. Detailed Implementation

[0121] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0122] This embodiment provides a cloud-edge hierarchical collaborative dual-mode energy dispatch and automatic configuration method for energy storage EMS, applied to a distributed energy storage cluster control scenario in an industrial park. This park includes three sets of distributed energy storage units, a 2MW distributed photovoltaic unit, and industrial adjustable loads, connected to a 10kV distribution network. The method of this invention is used to achieve intelligent dispatch and automatic configuration. The specific implementation steps are as follows:

[0123] The first step is to build a cloud-edge layered collaborative EMS architecture. A global energy management platform is deployed in the cloud to access energy storage, photovoltaic, load, and grid operation data from the entire park. One edge EMS node is deployed within the park to manage three groups of distributed energy storage units. The underlying device layer completes real-time data acquisition and command execution for energy storage, photovoltaic, and load, and establishes a 5G+fiber dual communication link to achieve bidirectional low-latency data interaction between the cloud and the edge.

[0124] The second step is to initialize the parameters of the dual-mode scheduling model. The cloud initializes the multi-objective weights, constraint thresholds, and optimization algorithm parameters of the economic steady-state scheduling model; the edge nodes initialize the basic parameters of the dynamic emergency scheduling model, such as the disturbance discrimination threshold, response delay, and power adjustment step size, to complete the model initialization configuration.

[0125] The third step is real-time identification of multi-dimensional scenarios. The system collects six core indicators in real time: photovoltaic power output fluctuation rate, industrial load change rate, 10kV bus voltage deviation, grid frequency deviation, and line load rate, and analyzes the operating scenario in real time.

[0126] During regular weekdays, photovoltaic (PV) output is stable, industrial load fluctuations are small, and all monitoring indicators are within the steady-state threshold range. The system determines this as a steady-state scenario and automatically switches to the economic steady-state dispatch mode. Based on day-ahead forecast data, the cloud-based system uses an improved multi-objective particle swarm optimization algorithm to generate a day-ahead energy storage charging and discharging plan, aiming to minimize electricity costs in the industrial park and maximize PV absorption. During peak PV periods in the daytime, energy storage is scheduled to charge to absorb excess PV power; during peak grid periods in the evening, energy storage is scheduled to discharge to achieve peak-valley arbitrage; and during off-peak periods, energy storage is on standby for power supply. The cloud-based system sends global dispatch commands to edge nodes. These edge nodes, based on the SOC status of the three energy storage units, finely allocate the charging and discharging power of each device to ensure balanced operation of each energy storage unit. Simultaneously, they automatically configure steady-state operating parameters, including SOC operating range, charging and discharging rate, and arbitrage period thresholds.

[0127] When scenarios such as cloud cover causing a sudden drop in photovoltaic (PV) output or a sudden start-up of industrial loads in the park occur, and the PV output volatility and load change rate exceed the steady-state threshold, the system identifies this as a dynamic disturbance scenario and automatically switches to dynamic emergency dispatch mode. Edge nodes immediately initiate local autonomous control, calculating emergency energy storage discharge power in milliseconds without cloud commands. This rapidly compensates for PV output shortfalls, smooths power fluctuations caused by sudden load changes, and corrects bus voltage deviations. Simultaneously, edge nodes upload disturbance data to the cloud, which coordinates surrounding power grid resources and issues coordinated control commands to assist the park's power grid in quickly restoring steady state. During this process, edge nodes automatically adapt to the disturbance intensity, dynamically adjusting dynamic parameters such as emergency adjustment thresholds, voltage regulation weights, and power adjustment step sizes to achieve adaptive parameter configuration.

[0128] The fourth step is smooth mode switching and anti-shake control. A 30-second anti-shake judgment period is set. Mode switching is only performed when the disturbance index continuously exceeds the steady-state threshold for 30 seconds. After the disturbance is eliminated, if the index remains stable for more than 30 seconds, it automatically switches back to the steady-state scheduling mode to avoid frequent mode switching caused by short-term fluctuations.

[0129] The fifth step is closed-loop optimization and correction. The system compiles daily scheduling and operation data, evaluates core indicators such as energy storage utilization rate, photovoltaic absorption rate, electricity consumption economy, and voltage stability, and iteratively optimizes the scheduling model weights and parameter configuration rules based on operation error data, automatically correcting subsequent operation parameters to achieve continuous optimization.

[0130] Through actual operation verification, after applying the method of the present invention in this embodiment, the photovoltaic absorption rate of the park increased by 9.2%, the energy storage resource utilization rate increased by 11.5%, the peak-valley arbitrage income of the park increased by 13.3%, the voltage fluctuation amplitude of the power grid decreased by 78%, and no manual parameter configuration operation was required, achieving fully intelligent automatic operation and maintenance.

[0131] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A cloud-edge hierarchical collaborative dual-mode energy dispatch and automatic configuration method for energy storage EMS, characterized in that the method... include: S1. Build a cloud-edge hierarchical collaborative energy storage EMS system architecture, including a cloud scheduling layer, an edge collaboration layer, and an energy storage device layer, for hierarchical data interaction and graded scheduling and control. S2. Construct a dual-mode energy dispatch model for EMS, including an economic steady-state dispatch mode and a dynamic emergency dispatch mode; S3. Construct a multi-dimensional scene recognition mechanism to achieve adaptive switching of dual scheduling modes based on power grid operation parameters, new energy output parameters, load parameters and energy storage operation status parameters; S4. Based on a cloud-edge layered collaborative architecture, dual-mode layered energy scheduling is implemented, wherein: Under the steady-state economic scheduling mode, the cloud scheduling layer performs global optimization scheduling, and the edge collaboration layer performs secondary fine-grained allocation. In dynamic emergency dispatch mode, the edge collaboration layer performs local autonomous control, while the cloud dispatch layer performs global collaborative support. S5. Perform adaptive automatic configuration of all parameters of the energy storage EMS; S6. Construct a closed-loop correction mechanism for operational status, and analyze the adaptability of scheduling strategies and configuration parameters in real time based on scheduling effect evaluation indicators, and automatically trigger the closed-loop correction process.

2. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 1, characterized in that, The cloud scheduling layer in step S1 is equipped with a global energy management platform, which is used for the statistics of energy storage resources across the entire network, load and new energy output prediction, global optimization scheduling calculation, scheduling strategy distribution, scenario situation analysis and unified parameter distribution. The edge collaboration layer is deployed with several edge EMS nodes. Each edge EMS node manages a distributed energy storage cluster within a region and is used for real-time local data acquisition, local status assessment, instant power regulation, cloud command parsing and execution, and local parameter adaptive correction. The energy storage device layer includes distributed energy storage devices, photovoltaic units, load equipment, and monitoring sensors, which are used to execute charging and discharging commands and upload real-time operating data.

3. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 1, characterized in that, The economic steady-state scheduling mode in step S2 takes maximizing the peak-valley arbitrage revenue of energy storage, minimizing the loss of new energy curtailment, and minimizing the transmission loss of the power grid as multiple objective functions to construct a global optimization scheduling model. The cloud scheduling layer uses an improved multi-objective particle swarm optimization algorithm to solve the global optimal solution based on the day-ahead forecast data and the intraday rolling forecast data, generating the day-ahead scheduling plan and intraday rolling optimization instructions for the edge nodes of each region.

4. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 3, characterized in that, The dynamic emergency dispatch mode in step S2 takes the rapid recovery of the power grid operation status, the rapid smoothing of power fluctuations and the optimal system operation stability as its core objectives. It constructs a local fast response dispatch model, and the edge collaboration layer uses the fast gradient descent algorithm to complete the instantaneous dispatch solution based on the millisecond-level operation data collected locally in real time, and adjusts the charging and discharging power of energy storage devices in the region in real time.

5. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 3, characterized in that, The multi-objective function is expressed as follows: = + + ； in: The profit from energy storage peak-valley arbitrage is represented as follows: C t I(p ess,i,t >0)-p ess,i,t dis C t I( pess,i,t <0); The losses from curtailment of renewable energy are represented as follows: = (p pv,t +p ess,i,t -p load,t ,0); For power grid transmission losses, it is expressed as follows: = ； , , These are the weighting coefficients, and ; The constraints include: Power constraints: -P dis,max ≤P ess,i,t ≤P ch,max ; SOC timing constraints: ； SOC min =0.2≤SOC i,t ≤SOC max =0.9; Regional power balance constraints: P pv,t + P ess,i,t =P load,t +P grid,t 。 6. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 3, characterized in that, The improved multi-objective particle swarm optimization algorithm specifically includes the following steps: a. Initialization Let the population size be N. pop =60, maximum iteration G max =20; Initialize particle position and velocity, trim out-of-bounds power to the constraint boundary, and calculate initial F. total Generate a set of Pareto non-dominated solutions; b. Particle Update Update particles using the following formula: V k+1 = V k +c1r1(P best -X k )+c2r2(G best -X k ); X k+1 =X k +V k+1 ; = start - , start =0.9, end =0.4； Where c1=c2=1.49, r1,r2 A random number, where g is the current iteration number; c. Constraint Modification After particle updates, if the power exceeds the SOC / power boundary, forced limiting is applied, and the total cost is recalculated. d. Optimal value update Compare the costs of new and old particles to update the individual optimal P. best The most economically optimal particle is selected from the Pareto solution set as the globally optimal G. best ; e. Convergence criterion 15 consecutive iterations of G best Cost change rate <10 -4 Terminate iteration; f. Output the optimal total power P for each edge zone. * area,t It is then sent to the edge EMS.

7. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 4, characterized in that, The local rapid response scheduling model constructed by the dynamic emergency dispatch mode takes minimizing grid frequency deviation, minimizing node voltage deviation, minimizing power fluctuation amplitude, and minimizing disturbance recovery time as objective functions. The constraints include dynamic power regulation response time constraints, energy storage fast charging and discharging power constraints, equipment safe operation constraints under fault conditions, and grid emergency voltage and frequency regulation constraints.

8. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 1, characterized in that, The multi-dimensional scene recognition mechanism in step S3 is based on a six-dimensional discrimination index system, including the fluctuation rate of new energy output, the rate of load change, the grid voltage deviation rate, the grid frequency deviation value, the line load rate, and the rate of change of energy storage state of charge. The steady-state threshold and dynamic threshold of each index are preset, and the automatic scene recognition and scheduling mode switching are realized through real-time data analysis.

9. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 1, characterized in that, Step S5 specifically includes: S51, Automatic configuration of steady-state parameters; Under the economic steady-state dispatch mode, the cloud automatically optimizes the configuration of global parameters such as the upper and lower limits of SOC for energy storage charging and discharging, the threshold of peak-valley arbitrage period, the dispatch weight of the whole network, the network loss optimization coefficient, and the priority of new energy consumption, based on the characteristics of the entire network load, the output pattern of new energy, and the characteristics of electricity price during the time period. The edge nodes automatically correct the local parameters of a single energy storage device based on the characteristics of the local equipment, so as to achieve the optimal matching of parameters in the steady-state scenario. S52, Automatic configuration of dynamic parameters; In dynamic emergency dispatch mode, edge nodes automatically and adaptively configure dynamic control parameters, including emergency adjustment power threshold, response delay parameters, voltage and frequency regulation weights, fault protection threshold, and power adjustment step size, according to the type and intensity of disturbance. This ensures accurate adaptation of control parameters under different disturbance scenarios and improves emergency control efficiency. S53, Scene linkage parameter correction; After each scheduling mode switch and scenario switch, based on the scheduling effect, control error, and equipment wear data of this operation, the parameter configuration model is iteratively optimized through machine learning algorithms. Various benchmark parameters and threshold parameters are automatically corrected to achieve continuous optimization of parameter configuration and adapt to the dynamic changes of long-term operation scenarios.

10. The cloud-edge hierarchical collaborative energy storage EMS dual-mode energy dispatch and automatic configuration method according to claim 1, characterized in that, The closed-loop correction mechanism for the operating state constructed in step S6 specifically includes: Real-time collection of dual-mode scheduling operation data, parameter configuration execution data, power grid operation status data, and energy storage device loss data; construction of a scheduling effect evaluation index system, including scheduling power deviation rate, voltage and frequency stability, energy storage resource utilization rate, economic benefit efficiency, equipment loss rate, and scenario switching accuracy; when any index fails to meet the standard or the control error exceeds the threshold, a closed-loop correction process is automatically triggered, the global scheduling model and parameters are re-optimized by the cloud, and the edge nodes correct the local control strategy.

Images