Multi-mode hybrid energy storage cooperative control system and method for megawatt-level network-building type energy storage power station

Abstract

The invention discloses a multi-mode hybrid energy storage cooperative control method and system for a megawatt-level network-building type energy storage power station, and the method comprises the steps: S1, collecting the frequency, voltage amplitude and phase of a power grid, and the charge state and temperature of each energy storage branch, and generating real-time monitoring data; s2, determining an operation mode of the energy storage power station based on a comparison result between the frequency deviation and the voltage deviation in the real-time monitoring data and a preset threshold value; s3, based on the operation mode of the energy storage power station, performing double-layer decomposition on the power instruction through a high-pass filter and a low-pass filter to generate a super-capacitor power instruction and a lithium battery power instruction; s4, generating an internal potential amplitude and a phase angle by adopting a virtual synchronous machine algorithm; and S5, on the basis of the amplitude and the phase angle of the internal potential and the detection result of the power grid connection state, executing switching control of a network construction type control mode or a network following type control mode. The millisecond-second level multi-time scale power optimal distribution is realized; self-adaptive inertia and voltage support is carried out; and island / grid-connected seamless switching is realized.

Description

Technical Field

[0001] This invention relates to the field of new energy power system control technology, and in particular to a multimodal hybrid energy storage collaborative control system and method for a 100-megawatt grid-type energy storage power station. Background Technology

[0002] Because new energy power generation equipment generally uses power electronic interfaces and lacks the rotational inertia of traditional synchronous generators, problems such as a significant reduction in the overall inertia level of the power grid, insufficient short-circuit capacity, and weakened voltage support capabilities are becoming increasingly prominent. Traditional grid-connected energy storage systems rely on grid voltage and frequency for phase-locked control, and can only passively follow the grid, unable to actively provide support during grid disturbances. In contrast, grid-forming BESS (Brain Energy Storage System) simulates the internal potential characteristics and inertial response mechanism of synchronous generators, and can actively provide inertia support, damping characteristics, and voltage support in both grid-connected and islanded modes, becoming a key technology for ensuring the safe and stable operation of power grids with a high proportion of new energy.

[0003] However, megawatt-scale grid-connected energy storage power stations typically consist of dozens of 2.5MW / 5MWh energy storage units connected in parallel, and are equipped with supercapacitor systems with millisecond-level response for rapid power compensation. The coordinated control of such large-scale hybrid energy storage systems faces numerous technical challenges. Lithium-ion batteries have high energy density but relatively slow response times (seconds), while supercapacitors have high power density but short storage times (seconds to minutes), resulting in significant differences in their power-energy characteristics. Existing control strategies often employ centralized control architectures for single battery or supercapacitor systems, which suffer from the following key technical drawbacks: First, the lack of an effective energy management coordination mechanism among multiple energy storage units easily leads to circulating currents and unbalanced power distribution; second, grid-connected converter control strategies are incompatible with traditional grid-connected control strategies, easily causing power oscillations and transient instability during islanding / grid-connected mode switching; third, the lack of a unified coordinating controller at the site level makes it impossible to achieve optimized coordination of multiple objectives such as inertial response, primary frequency regulation, and voltage support.

[0004] Therefore, there is an urgent need to develop a multimodal hybrid energy storage collaborative control system and method for megawatt-level application scenarios. Summary of the Invention

[0005] This invention provides a multi-modal hybrid energy storage collaborative control system and method for a 100-megawatt grid-type energy storage power station to solve the above-mentioned problems in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A multi-modal hybrid energy storage collaborative control method for a 100-megawatt grid-type energy storage power station includes:

[0008] S1: Collect grid frequency, voltage amplitude, phase, and the state of charge and temperature of each energy storage branch to generate real-time monitoring data;

[0009] S2: Determine the operating mode of the energy storage power station based on the comparison results of frequency deviation and voltage deviation in real-time monitoring data with preset thresholds;

[0010] S3: Based on the operation mode of the energy storage power station, the power command is decomposed into supercapacitor power command and lithium battery power command through high-pass filter and low-pass filter.

[0011] S4: Based on the power commands of supercapacitors and lithium batteries, a virtual synchronous machine algorithm is used to generate the internal potential amplitude and phase angle;

[0012] S5: Based on the internal potential amplitude, phase angle, and grid connection status detection results, execute the switching control between grid-based control mode and grid-following control mode.

[0013] Furthermore, S1 includes:

[0014] S11: Acquire three-phase voltage, three-phase current and frequency signals of the power grid with a sampling period of 100μs;

[0015] S12: Synchronously acquire the state of charge signals of each lithium battery branch and supercapacitor branch;

[0016] S13: Collect temperature signals from each energy storage branch, synchronize them with grid signals and state of charge signals to generate real-time monitoring data.

[0017] Furthermore, S2 includes:

[0018] S21: Extract the power grid frequency deviation Δf and voltage deviation ΔU from real-time monitoring data;

[0019] S22: Compare the frequency deviation Δf with the 0.1Hz threshold, and compare the voltage deviation ΔU with the 5% rated voltage threshold;

[0020] S23: When any deviation exceeds the corresponding threshold, the operating mode is determined to be the active support mode; when neither deviation exceeds the corresponding threshold, the operating mode is determined to be the economic operation mode.

[0021] Furthermore, S3 includes:

[0022] S31: Extract the power components in the 0.1Hz to 10Hz frequency band from the total power command through a high-pass filter with a cutoff frequency of 0.1Hz, and use them as the supercapacitor power command Psc_ref;

[0023] S32: Extracts the power component in the 0 to 0.1 Hz frequency band from the total power command through a low-pass filter with a cutoff frequency of 0.1 Hz, and uses it as the lithium battery power command Pbat_ref;

[0024] S33: Model predictive control is used to perform rolling optimization of the power command, with a rolling optimization time domain of 2s.

[0025] Furthermore, the model predictive control in S33 includes:

[0026] S331: Constructing the objective function:

[0027]

[0028] Where ΔPgrid is the grid power deviation, defined as the difference between the total active power command Pgrid_ref dispatched by the grid to the energy storage power station at time t and the total active power Pgrid_act actually injected into the grid by the power station at that time; SOCdev is the deviation of the lithium battery state of charge, defined as the difference between the actual SOC of the lithium battery and its ideal operating point SOCref at time t; ΔPbat is the lithium battery power change rate, defined as the change in the output power of the lithium battery within adjacent control cycles; and λ and μ are weighting coefficients.

[0029] S332: Minimize the objective function within the 2s prediction time domain;

[0030] S333: Dynamically adjust the power commands of the supercapacitor and the lithium battery based on the solution results.

[0031] Furthermore, S4 includes:

[0032] S41: Calculate the power difference between the received power reference value Pref and the actual output power Pout;

[0033] S42: Based on formula Calculate the rate of change of the internal potential phase angle, where H is the virtual inertia. The rated angular frequency;

[0034] S43: The internal potential phase angle δ is obtained by integrating the internal potential phase angle change rate, and combined with the preset internal potential amplitude E, a virtual synchronous machine control signal is generated.

[0035] Furthermore, S5 includes:

[0036] S51: Grid-connected to islanded switching: After detecting voltage loss at the grid connection point, the grid-connected converter will be switched from grid-following PQ control to grid-connected VF control within 5ms.

[0037] S52: Island to Grid Switching: Inject a sinusoidal disturbance signal with an amplitude of 1% of the rated voltage and a frequency deviation of ±0.05Hz for phase pre-synchronization. When the phase locking error is less than 0.5°, the circuit breaker is closed, and the switching time is less than 100ms.

[0038] Furthermore, the virtual inertia H is adjusted online based on the frequency deviation:

[0039] Where H0 is the basic virtual inertia and k is the adjustment coefficient;

[0040] When |Δf| exceeds the set value, H is limited to the maximum value Hmax, where Hmax represents the maximum virtual inertia.

[0041] Furthermore, a multi-modal hybrid energy storage collaborative control system for a 100-megawatt grid-type energy storage power station includes:

[0042] The site-level coordination controller adopts a dual-CPU redundant architecture and communicates with all energy storage units via the IEC61850-GOOSE protocol, with a communication latency of less than or equal to 5ms, and is used to generate global power commands.

[0043] Multiple unit-level grid-type converters, each converter is connected to a corresponding energy storage unit, supporting dual modes of grid-type VF control and grid-connected PQ control, rated capacity 2.5MW, short-time overload capacity 1.5pu / 10s;

[0044] The multi-mode energy storage interface module includes a lithium battery branch, a supercapacitor branch, and a high-speed DC bus coupler. The coupler uses SiCMOSFETs with a switching frequency of 20kHz and a bidirectional power flow control error of less than or equal to 0.5%.

[0045] The fiber optic self-healing ring network adopts a redundant topology structure, with a single-point fault self-healing time of less than 50ms, and connects the station-level coordination controller with all unit-level network-type converters.

[0046] Compared with the prior art, the present invention has the following advantages:

[0047] This invention employs a Model Predictive Control (MPC) algorithm to design an optimized interaction mechanism between the controller and the system. The optimized controller makes optimization decisions (such as power commands allocated to lithium batteries and supercapacitors) based on the current state of the power grid and energy storage system (e.g., grid frequency, voltage, and state of charge (SOC) of each energy storage unit) to minimize the multi-objective optimization function. The optimized control framework uses Model Predictive Control (MPC), with the state space including grid frequency, voltage, and SOC, and optimization variables including the charging and discharging power of the energy storage. The objective function comprehensively considers factors such as grid power tracking deviation, SOC balance, and battery power change rate, guiding the system to achieve the optimal cooperative control strategy.

[0048] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention.

[0049] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0050] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0051] Figure 1 This is a diagram of the overall single-line architecture of the system in an embodiment of the present invention;

[0052] Figure 2 This is a block diagram of the dual-layer-dual-loop collaborative control in an embodiment of the present invention;

[0053] Figure 3 This is the virtual inertia adaptive curve in this embodiment of the invention;

[0054] Figure 4 The waveforms for islanding / grid-connected switching experiments in this embodiment of the invention are shown (CH1 - PCC voltage, CH2 - battery power, CH3 - supercapacitor power). Detailed Implementation

[0055] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0056] Example 1:

[0057] This invention provides a multi-modal hybrid energy storage collaborative control method for a 100-megawatt grid-type energy storage power station, comprising:

[0058] S1: Collect grid frequency, voltage amplitude, phase, and the state of charge and temperature of each energy storage branch to generate real-time monitoring data;

[0059] S2: Determine the operating mode of the energy storage power station based on the comparison results of frequency deviation and voltage deviation in real-time monitoring data with preset thresholds;

[0060] S3: Based on the operation mode of the energy storage power station, the power command is decomposed into supercapacitor power command and lithium battery power command through high-pass filter and low-pass filter.

[0061] S4: Based on the power commands of supercapacitors and lithium batteries, a virtual synchronous machine algorithm is used to generate the internal potential amplitude and phase angle;

[0062] S5: Based on the internal potential amplitude, phase angle, and grid connection status detection results, execute the switching control between grid-based control mode and grid-following control mode.

[0063] The following is a detailed explanation with reference to the embodiments:

[0064] This embodiment proposes as follows: Figure 2 As shown, the specific steps of the "dual-layer-dual-loop" collaborative control strategy are as follows:

[0065] Step S1: Real-time monitoring: Collect power grid frequency f, voltage amplitude U, phase θ, SOC of each branch, and temperature T, with a sampling period of 100µs;

[0066] Step S2: Determine the operating mode: If Hz or If UN is reached, the "active support mode" is triggered; otherwise, the "economic operation mode" is activated.

[0067] Step S3: Power command decomposition (two-layer allocation): High frequency layer (0.1Hz~10Hz): undertaken by supercapacitors, power command Psc_ref is obtained through a high-pass filter;

[0068] Low-frequency layer (0~0.1Hz): powered by lithium battery, power command Pbat_ref is obtained through low-pass filter;

[0069] Model predictive control (MPC) is employed, with rolling optimization over a 2-second time domain. The weight matrix Q / R is adaptively adjusted. The objective function is:

[0070]

[0071] In the formula, The grid power deviation is defined as the difference between the total active power command Pgrid_ref sent by the grid to the energy storage power station at time t and the total active power Pgrid_act actually injected into the grid by the power station at that time. The deviation of the state of charge (SOC) of a lithium battery is defined as the difference between the actual SOC of the lithium battery at time t and its ideal operating point SOCref (usually taken as 50%). The lithium battery power change rate is defined as the change in the output power of the lithium battery within adjacent control cycles. This is an adjustable positive weighting coefficient used to balance high-precision power tracking, SOC equalization, and battery power smoothing; the larger the value, the higher the optimization priority of the corresponding term. The objective function is minimized within a 2-second prediction time domain; the supercapacitor power command and lithium battery power command are dynamically adjusted based on the solution results.

[0072] Step S4: Network-based control: The Virtual Synchronous Machine (VSM) algorithm is embedded in UGPCS, and the internal potential amplitude E and phase angle δ are generated by the following equations:

[0073]

[0074] In the formula, This is a virtual inertia that can be adjusted online. The rated angular frequency is used. The internal potential phase angle δ is obtained by integrating the rate of change of the internal potential phase angle, and combined with the preset internal potential amplitude E, a virtual synchronous machine control signal is generated.

[0075] Step S5: Island / Grid Switching:

[0076] Grid-connected to islanded mode: First, SLCC detects PCC voltage loss of synchronization, and HEIM switches to VF mode within 5ms;

[0077] Islanding to grid connection: Small signal phase pre-synchronization is adopted, and a sinusoidal disturbance with an amplitude of 1%UN and a frequency of ±0.05Hz is injected to achieve phase locking error <0.5° before closing the circuit breaker, with a switching time of <100ms.

[0078] like Figure 3 As shown in the figure, the absolute value of the virtual inertia H as a function of frequency deviation is displayed as a curve. The changing regulation patterns include the base inertia H0, the linear growth range, and the limiting characteristics after reaching Hmax. This reflects the dynamic characteristics of the system adaptively providing inertia support based on the degree of grid frequency disturbance. The virtual inertia H is adjusted online according to the frequency deviation.

[0079] Where H0 is the basic virtual inertia and k is the adjustment coefficient;

[0080] When |Δf| exceeds the set value, H is limited to the maximum value Hmax.

[0081] Example 2:

[0082] This invention provides a multi-modal hybrid energy storage collaborative control system for a 100-megawatt grid-type energy storage power station, comprising:

[0083] The site-level coordination controller adopts a dual-CPU redundant architecture and communicates with all energy storage units via the IEC61850-GOOSE protocol, with a communication latency of less than or equal to 5ms, and is used to generate global power commands.

[0084] Multiple unit-level grid-type converters, each converter is connected to a corresponding energy storage unit, supporting dual modes of grid-type VF control and grid-connected PQ control, rated capacity 2.5MW, short-time overload capacity 1.5pu / 10s;

[0085] The multi-mode energy storage interface module includes a lithium battery branch, a supercapacitor branch, and a high-speed DC bus coupler. The coupler uses SiCMOSFETs with a switching frequency of 20kHz and a bidirectional power flow control error of less than or equal to 0.5%.

[0086] The fiber optic self-healing ring network adopts a redundant topology structure, with a single-point fault self-healing time of less than 50ms, and connects the station-level coordination controller with all unit-level network-type converters.

[0087] The following is a detailed explanation with reference to the embodiments:

[0088] System Architecture

[0089] Station-Level Coordinated Controller (SLCC):

[0090] Deployed in the secondary compartment of the 110kV substation, it adopts a dual-CPU redundant architecture and communicates with all units via IEC61850-GOOSE, with a communication latency of ≤5ms.

[0091] Unit-Level Grid-forming PCS (UGPCS):

[0092] Each 2.5MW / 5.015MWh energy storage unit is equipped with one unit, supporting VF (grid-connected) / PQ (grid-connected) dual modes, with a rated capacity of 2.5MW, short-term overload of 1.5pu / 10s, and support for black start.

[0093] Hybrid Energy Interface Module (HEIM):

[0094] Includes: Lithium battery branch: 2.5MW / 5.015MWh lithium iron phosphate; Supercapacitor branch: 1MW×30s (8.33kWh); High-speed DC bus coupler: using SiCMOSFET, switching frequency 20kHz, bidirectional power flow control error ≤0.5%.

[0095] Communication network: fiber optic self-healing ring network, redundant topology, single-point fault self-healing time <50ms.

[0096] Node voltage constraints: Limit the voltage amplitude of each node in the distribution network to within the allowable range; Line power constraints: Limit the power transmission of each line in the distribution network to not exceed its rated capacity; Energy storage system constraints: Limit the charging and discharging power and state of charge of the energy storage system to within the safe range.

[0097] A deep reinforcement learning framework is constructed: Deep reinforcement learning algorithms are employed to design an interaction mechanism between the agent and the environment. The agent makes decisions based on the current state of the distribution network (e.g., node voltage, photovoltaic output, energy storage state of charge, etc.) (e.g., energy storage charging and discharging power, reactive power adjustment of distributed photovoltaic systems, etc.) to maximize cumulative rewards. The deep reinforcement learning framework uses a deep Q-network (DQN) or its variant. The agent's state space includes node voltage, photovoltaic output, energy storage state of charge, etc., while its action space includes energy storage charging and discharging power and reactive power adjustment of distributed photovoltaic systems.

[0098] Design reward function: The reward function comprehensively considers factors such as node voltage deviation, active power loss, energy storage system lifespan and photovoltaic absorption rate, and guides the agent to learn the optimal voltage control strategy.

[0099] To facilitate examiners and those skilled in the art to fully understand and reproduce this invention, the following section uses the complete engineering example of the Hainan Wenchang 100MW / 200.6MWh Chang'an Energy Storage Demonstration Power Station (hereinafter referred to as "Chang'an Station") to explain the system layout, hardware selection, software process, key parameter tuning, field tests, and operational results point by point. Unless otherwise specified, all values ​​are actual measurements taken on-site or recorded in type test reports.

[0100] Overall layout of the power station: Chang'an station adopts a 35kV centralized current collection and 110kV step-up transmission scheme. For example... Figure 1 As shown, the primary equipment area is equipped with 40 sets of 2.5MW / 5.015MWh lithium iron phosphate energy storage units, with 8 sets forming 5 35kV collection lines. The 110kV step-up substation is equipped with one 120MVA double-winding main transformer (110kV / 35kV), and the high-voltage side of the main transformer is connected to the grid via the 110kV Changxia line. A 1MW×30s supercapacitor system (8.33kWh) is also installed on the DC side, connected to the 35kV Section I busbar via an independent 215kW grid-type PCS for high-frequency power pulse compensation.

[0101] Secondary systems and communications:

[0102] Site-level Coordination Controller (SLCC): Cabinet: Standard 19″ 42U, dual CPU redundancy (Intel Xeon E-2278G), real-time operating system QNX7.1.

[0103] Sampling: 4-channel 100µs synchronous sampling fiber optic board, connected to 110kV bus three-phase voltage, three-phase current, frequency, and power and SOC of 40 PCS sets.

[0104] Communication: IEC61850-9-2LE sampling value + GOOSE dual network redundancy, ring network self-healing time 18ms; point-to-point fiber optic delay with each PCS is 2.8ms~3.9ms, meeting the ≤5ms requirement.

[0105] • Algorithm: The MPC solver is based on qpOASES and predicts the step size. Control step size Weight The tuning was completed within 3 months using an offline simulation-online rolling calibration method.

[0106] Unit-level grid converter (UGPCS):

[0107] Power topology: Three-level NPC, SiCMOSFET (1200V / 600A), switching frequency 16kHz, DC bus 1.2kV.

[0108] Control board: DSP28388D+FPGAXC7K325T, with embedded VSM algorithm; virtual inertia H is adjustable online with a step size of 10ms, H0=8s, k=2.5.

[0109] Protection: 1.5 PU overload for 10 seconds, 3 PU overcurrent for 1 second; supports anti-bot boot, boot time <30 seconds.

[0110] Multimodal Energy Storage Interface Module (HEIM)

[0111] Topology: Three-port DC / DC, 1kV / 2.5MW on the lithium battery side, 750V / 1MW on the supercapacitor side, and 1.2kV on the bus side; adopts a dual active bridge (DAB) + synchronous Buck / Boost hybrid structure with a peak efficiency of 97.8%.

[0112] Control: Local 200µs current loop + 2ms energy loop, receiving Pbat and Psc commands issued by SLCC; SOC equalization strategy is to enable slow equalization (≤0.1C) when the average SOC deviation is <2%, and enable fast equalization (≤0.3C) when the deviation is >5%.

[0113] Software process:

[0114] The main program has a cycle time of 10ms, and the tasks are divided as follows:

[0115] a) Sampling and filtering (100µs);

[0116] b) State machine (grid-connected / islanding / fault ride-through);

[0117] c) MPC solution (0.8ms~1.2ms);

[0118] d) GOOSE frame encapsulation and transmission (<0.2ms).

[0119] like Figure 4 As shown, the dynamic performance of the mode switching process was verified through experimental waveforms (CH1 - PCC voltage, CH2 - battery power, CH3 - supercapacitor power). The stability of the grid connection point (PCC) voltage and the coordinated response of the battery power and supercapacitor power were demonstrated, proving that the proposed control method can achieve seamless switching in less than 100ms with minimal voltage fluctuation and no inrush current.

[0120] Island / Grid-connected handover process:

[0121] Grid-connected → Islanded: SLCC monitors PCC voltage amplitude drop >10% or frequency exceedance >±0.2Hz → within 100µs switch UGPCS to VF mode, lock out synchronous machine side circuit breaker, and switch HEIM to droop control.

[0122] Islanding to grid connection: During the pre-synchronization stage, a small sinusoidal signal of 1%UN and ±0.05Hz is injected. When the phase locking error is <0.5°, the frequency difference is <0.02Hz, and the voltage difference is <1%UN, the circuit is closed. The entire switching process takes 67ms.

[0123] Key parameter tuning method: Selection of weights λ and μ:

[0124] Offline phase: An electromagnetic transient model containing 40 UGPCS units was built using MATLAB / Simscape. NSGA-II multi-objective optimization was employed, with ΔPgridRMS, SOCdevRMS, and ΔPbatRMS as the indices. After 3000 iterations, the model was obtained. , The optimal solution of the Pareto front.

[0125] During the online phase: SLCC fine-tunes λ (within ±20%) every 5 minutes based on the average SOCdev value to ensure SOC drift <3% / day.

[0126] Virtual Inertia H Online Curve

[0127] ,when The time limit Hmax is set to 20 seconds to prevent over-adjustment of the frequency. The unit is seconds.

[0128] Field tests and results:

[0129] Test 1: 110kV three-phase ground fault (0.2s)

[0130] Before the failure, Pgrid=100MW, Qgrid=0Mvar.

[0131] During the fault, Pgrid dropped to 42MW, the supercapacitor injected 0.96MW within 1ms, and the battery climbed to 58MW within 30ms; the system frequency dropped by a maximum of 0.35Hz and recovered to ±0.05Hz in 4.1s, meeting the primary frequency regulation requirements.

[0132] Experiment 2: Smooth handover from islanded to grid-connected systems:

[0133] The islanded load is 35MW, with a power deficit of 5MW; grid connection was successful 67ms after pre-synchronization, with a peak PCC voltage fluctuation of 2.8% and no inrush current.

[0134] Experiment 3: SOC Equalization

[0135] Initially, SOCdev_max = 7%, after 8 minutes of fast balancing, it drops to 1.9%, with a balancing current of 0.25C and a temperature rise of <3℃. The interaction process between the agent and the environment is as follows:

[0136] Running data

[0137] As of March 2026, Chang'an Station had accumulated 180 days of operation, issuing 1.8 × 10^7 power commands via SLCC, with ΔPgridRMS = 0.72%, SOCdevRMS = 1.4%, and ΔPbatRMS = 0.31%. The measured equivalent inertia ranged from 9.8 s to 11.2 s, with an error of <5% compared to the theoretical value. State normalization: All features in the state space were normalized to the range [0,1] to improve training stability and convergence speed; Exploration strategy: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] - Greedy strategy: randomly selects actions with a certain probability to avoid the agent getting stuck in local optima; Reward normalization: normalizes the reward signal to avoid the reward value being too large and affecting the training process.

[0138] Scalability:

[0139] When the power plant is expanded to 200MW / 400MWh, only:

[0140] Add 40 sets of UGPCS and continue the existing communication protocol;

[0141] The number of SLCC software threads was increased from 4 to 8, the MPC step size was kept at 2 seconds, and the CPU load increased from 42% to 68%, while still meeting the real-time requirements.

[0142] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from the spirit and scope of this invention.

Claims

1. A multi-modal hybrid energy storage collaborative control method for a 100-megawatt grid-type energy storage power station, characterized in that, include: S1: Collect grid frequency, voltage amplitude, phase, and the state of charge and temperature of each energy storage branch to generate real-time monitoring data; S2: Determine the operating mode of the energy storage power station based on the comparison results of frequency deviation and voltage deviation in real-time monitoring data with preset thresholds; S3: Based on the operation mode of the energy storage power station, the power command is decomposed into supercapacitor power command and lithium battery power command through high-pass filter and low-pass filter. S4: Based on the power commands of supercapacitors and lithium batteries, a virtual synchronous machine algorithm is used to generate the internal potential amplitude and phase angle; S5: Based on the internal potential amplitude, phase angle, and grid connection status detection results, execute the switching control between grid-based control mode and grid-following control mode.

2. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 1, characterized in that, S1 includes: S11: Acquire three-phase voltage, three-phase current and frequency signals of the power grid with a sampling period of 100μs; S12: Synchronously acquire the state of charge signals of each lithium battery branch and supercapacitor branch; S13: Collect temperature signals from each energy storage branch, synchronize them with grid signals and state of charge signals to generate real-time monitoring data.

3. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 1, characterized in that, S2 include: S21: Extract the power grid frequency deviation Δf and voltage deviation ΔU from real-time monitoring data; S22: Compare the frequency deviation Δf with the 0.1Hz threshold, and compare the voltage deviation ΔU with the 5% rated voltage threshold; S23: When any deviation exceeds the corresponding threshold, the operating mode is determined to be the active support mode; when neither deviation exceeds the corresponding threshold, the operating mode is determined to be the economic operation mode.

4. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 1, characterized in that, S3 include: S31: Extract the power components in the 0.1Hz to 10Hz frequency band from the total power command through a high-pass filter with a cutoff frequency of 0.1Hz, and use them as the supercapacitor power command Psc_ref; S32: Extracts the power component in the 0 to 0.1 Hz frequency band from the total power command through a low-pass filter with a cutoff frequency of 0.1 Hz, and uses it as the lithium battery power command Pbat_ref; S33: Model predictive control is used to perform rolling optimization of the power command, with a rolling optimization time domain of 2s.

5. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 4, characterized in that, Model predictive control in S33 includes: S331: Constructing the objective function: ， Where ΔPgrid is the grid power deviation, defined as the difference between the total active power command Pgrid_ref dispatched by the grid to the energy storage power station at time t and the total active power Pgrid_act actually injected into the grid by the power station at that time; SOCdev is the deviation of the lithium battery state of charge, defined as the difference between the actual SOC of the lithium battery and its ideal operating point SOCref at time t; ΔPbat is the lithium battery power change rate, defined as the change in the output power of the lithium battery within adjacent control cycles; and λ and μ are weighting coefficients. S332: Minimize the objective function within the 2s prediction time domain; S333: Dynamically adjust the power commands of the supercapacitor and the lithium battery based on the solution results.

6. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 1, characterized in that, S4 include: S41: Calculate the power difference between the received power reference value Pref and the actual output power Pout; S42: Based on formula Calculate the rate of change of the internal potential phase angle, where H is the virtual inertia. The rated angular frequency, This is a power reference value. The actual output power of the converter; S43: The internal potential phase angle δ is obtained by integrating the internal potential phase angle change rate, and combined with the preset internal potential amplitude E, a virtual synchronous machine control signal is generated.

7. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 1, characterized in that, S5 include: S51: Grid-connected to islanded switching: After detecting voltage loss at the grid connection point, the grid-connected converter will be switched from grid-following PQ control to grid-connected VF control within 5ms. S52: Island to Grid Switching: Inject a sinusoidal disturbance signal with an amplitude of 1% of the rated voltage and a frequency deviation of ±0.05Hz for phase pre-synchronization. When the phase locking error is less than 0.5°, the circuit breaker is closed, and the switching time is less than 100ms.

8. The multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station according to claim 6, characterized in that, The virtual inertia H is adjusted online according to the frequency deviation: Where H0 is the basic virtual inertia and k is the adjustment coefficient. This is the absolute value of the frequency deviation; When |Δf| exceeds the set value, H is limited to the maximum value Hmax.

9. A system for a multi-modal hybrid energy storage collaborative control method for a 100 MW-level grid-type energy storage power station as described in any one of claims 1-8, characterized in that, include: The site-level coordination controller adopts a dual-CPU redundant architecture and communicates with all energy storage units via the IEC61850-GOOSE protocol, with a communication latency of less than or equal to 5ms, and is used to generate global power commands. Multiple unit-level grid-type converters, each converter is connected to a corresponding energy storage unit, supporting dual modes of grid-type VF control and grid-connected PQ control, rated capacity 2.5MW, short-time overload capacity 1.5pu / 10s; The multi-mode energy storage interface module includes a lithium battery branch, a supercapacitor branch, and a high-speed DC bus coupler. The coupler uses SiCMOSFETs with a switching frequency of 20kHz and a bidirectional power flow control error of less than or equal to 0.5%. The fiber optic self-healing ring network adopts a redundant topology structure, with a single-point fault self-healing time of less than 50ms, and connects the station-level coordination controller with all unit-level network-type converters.

Images