Control method and system of dual-winding generator and energy storage combined power supply device

Abstract

The application discloses a control method and system of a dual-winding generator and energy storage combined power supply device, belongs to the technical field of power system control, and is based on the dual-winding generator and energy storage combined power supply device, real-time acquisition of parameters of the dual-winding generator and energy storage combined power supply device, calculation of a sudden change intensity index, logical decision-making according to the sudden change intensity index, millisecond-level improved sliding film control for adjusting disturbance, second-level deep reinforcement learning for adjusting proportional integral coefficients, realization of adaptive optimization, minute-level game for optimizing power distribution of the dual-winding generator and energy storage, realization of voltage sudden change suppression, steady-state optimization and energy efficiency management, improvement of dynamic response capability, steady-state precision and economy of the dual-winding generator and energy storage combined power supply device, and provision of technical support for efficient and stable power supply.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system control, and particularly relates to a control method and system of a dual-winding generator and energy storage combined power supply device. BACKGROUND

[0002] In modern power systems, the combination of distributed generation and energy storage technology has become a key to improving power quality and power supply reliability. However, the current control method of the energy storage and generator parallel system mainly relies on a single time scale control strategy, which has certain limitations.

[0003] Loads such as elevators, punches, piston compressors, refrigerators, and the like have periodic power changes, and the power demand fluctuates constantly during operation. The traditional control method is difficult to effectively deal with such cases. For example, the traditional PI control method relies on fixed parameters, and in the face of periodic changes in load power, it is difficult to adapt to dynamic load changes, and is prone to overshoot or oscillation; the conventional fuzzy control method relies on artificial experience to construct a rule base, and lacks online optimization capability, and when the load power fluctuates periodically, the response speed to sudden disturbances is slow; the centralized coordination control uses a global optimization algorithm, but the calculation complexity is high, and in the face of frequent adjustment requirements brought by periodic load power changes, the control delay is large, and it is difficult to meet the millisecond-level fast response requirement.

[0004] The limitations of these methods make it difficult for the power system to balance between dynamic response and steady-state accuracy, and a single control strategy is difficult to meet the needs of transient compensation and long-term energy efficiency optimization. In addition, the control objectives of the power system are complex, including voltage regulation, power distribution, and equipment protection, and traditional methods are difficult to achieve coordinated decision-making at different time scales. Moreover, the rule base of the conventional fuzzy control is fixed, and lacks adaptive evolution ability, making it difficult to adapt to changing operating conditions, thereby limiting the adaptability and robustness of the power system. SUMMARY

[0005] The present application aims to provide a control method and system of a dual-winding generator and energy storage combined power supply device based on multi-time scale coordination of a dual-winding generator and energy storage combined power supply device. Through a hierarchical and time-sharing dynamic coordination mechanism, the unity of millisecond-level transient disturbance suppression, second-level PI allocation, and minute-level power distribution optimization is realized. By breaking through the limitations of a single control strategy, an adaptive dynamic response system is constructed. And combined with intelligent algorithms, the anti-interference ability and economy of the dual-winding generator and energy storage combined power supply device are improved.

[0006] A control method for a combined power supply device of a dual-winding generator and energy storage, wherein the combined power supply device of the dual-winding generator and energy storage comprises: an energy storage battery, an energy storage converter, a dual-winding generator, and a prime mover, wherein the energy storage battery, energy storage converter, dual-winding generator, and prime mover are connected in sequence; the control method for the combined power supply device of the dual-winding generator and energy storage includes...

[0007] Step S1: Collect parameters of the dual-winding generator and energy storage combined power supply device in real time, calculate the sudden change intensity index by weighted fusion of power change rate and voltage second derivative, make logical decisions based on the sudden change intensity index, and switch control modes of different time scales.

[0008] Step S2: Determine whether the mutation intensity index is greater than the threshold. If yes, proceed to step S3; otherwise, proceed to step S4.

[0009] Step S3: Switch to millisecond-level mode, use sliding diaphragm control combined with disturbance prediction and phasor decomposition for feedforward-feedback collaborative compensation, and return to step S1;

[0010] Step S4: Determine whether the mutation duration of the mutation intensity index is greater than 1 second. If yes, proceed to step S5; otherwise, proceed to step S6.

[0011] Step S5: Switch to second-level mode, use deep reinforcement learning to adjust the proportional integral coefficient, optimize the control rules through real-time data, and return to step S1;

[0012] Step S6: Determine whether the dual-winding generator and energy storage combined power supply device periodically triggers sudden changes. If yes, proceed to step S7; otherwise, proceed to step S8.

[0013] Step S7: Switch to minute-level mode, use game theory to optimize the power allocation between the dual-winding generator and energy storage in the dual-winding generator and energy storage combined power supply device, and return to step S1;

[0014] Step S8: The dual-winding generator and energy storage combined power supply device maintain the current mode and return to step S1.

[0015] Optionally, the mutation intensity index is expressed as follows:

[0016]

[0017] in, Indicators representing mutation intensity; , representing instantaneous active power The instantaneous change; Represents the power at time t; This represents the power at time t-1; These are the weighting coefficients.

[0018] Optionally, in the millisecond level mode, the feed-forward compensation current of the feed-forward compensation is represented as follows:

[0019]

[0020] wherein, represents the feed-forward compensation current; represents the feed-forward gain; is a time constant;

[0021] The feedback compensation current of the feedback compensation is represented as follows:

[0022]

[0023] wherein, represents the feedback compensation current; and both represent the feedback gain; represents a smoothness coefficient; tanh represents a hyperbolic tangent function; sat(s) represents a saturation function, s represents a sliding surface a calculated value;

[0024] The sliding surface of the sliding mode control is represented as follows:

[0025]

[0026] wherein, represents the sliding surface of the sliding mode control; i.e. the reference voltage and the actual measured voltage represents a voltage deviation; represents a sliding surface coefficient, used to adjust the weight of the integral term.

[0027] Optionally, in the second level mode, the deep reinforcement learning is used to adjust the proportional and integral coefficients, including: defining a state space and an action space to reflect the state of the dual-winding generator and the energy storage combined power supply device in real time;

[0028] wherein, the state space includes the interactive features of the dual-winding generator and the energy storage, including: voltage deviation, voltage change rate, energy storage battery state of charge, load power, generator output power, energy storage charge and discharge power, past 10 seconds historical error mean, error variance, harmonic distortion rate, control output amplitude, environmental temperature and time stamp cycle characteristics;

[0029] The action space is defined as adjusting the proportional and integral coefficients, represented as follows:

[0030]

[0031]

[0032] wherein, represents an adjustment amount of a proportional coefficient, represents an adjustment amount of an integral coefficient.

[0033] Optionally, in the second mode, the adjusting the proportional and integral coefficients by using the deep reinforcement learning further comprises: designing a reward function, which is represented as follows:

[0034]

[0035] wherein, R represents the reward function; represents a voltage stability weight; represents an energy efficiency weight; represents an energy efficiency coefficient of the dual-winding generator and energy storage combined power supply device, represents an output power of the dual-winding generator and energy storage combined power supply device, represents an input power of the dual-winding generator and energy storage combined power supply device; represents a frequency deviation penalty term, represents a frequency deviation; represents a storage battery state of charge health penalty; represents a storage battery state of charge health penalty term.

[0036] Optionally, in the second mode, after the control rule is optimized by using the real-time data, the second mode further comprises: performing a stability check, wherein the stability check comprises: monitoring the voltage deviation and the frequency deviation in real time, and when the voltage deviation, the frequency deviation, and the optimized control rule do not satisfy a stability condition, triggering a strategy rollback, and restoring the proportional coefficient and the integral coefficient to the stable version before the adjustment; the stability condition is not satisfied, including: the voltage deviation > 10% of the reference voltage, the frequency deviation > 0.5 Hz, and the optimized control rule does not satisfy a Lyapunov stability condition.

[0037] Optionally, in the first mode, the power distribution optimization of the dual-winding generator and the energy storage of the dual-winding generator and energy storage combined power supply device by using the game theory comprises:

[0038] defining game participants and a strategy space;

[0039] establishing a participant benefit function;

[0040] establishing game constraints;

[0041] The participant benefit function is represented as follows:

[0042]

[0043] wherein, represents the dual-winding generator revenue function; represents the dual-winding generator unit power generation revenue coefficient; represents the dual-winding generator power generation cost penalty coefficient; represents the energy storage battery state of charge balancing coefficient, for penalizing the working condition deviating from the intermediate value 50%; represents the power generation power of the dual-winding generator;

[0044]

[0045] wherein, represents the energy storage revenue function; represents the energy storage discharge revenue coefficient; represents the energy storage discharge loss penalty coefficient; represents the energy storage battery state of charge target preference coefficient, for preferring to maintain the energy storage battery state of charge at 60%;

[0046] Game constraints are established, including: power balance constraint, energy storage battery state of charge dynamic constraint;

[0047] The power balance constraint is represented as follows:

[0048] ;

[0049] wherein, represents the dual-winding generator output power set value; represents the energy storage output power set value; represents the load power;

[0050] The energy storage battery state of charge dynamic constraint is represented as follows:

[0051]

[0052] wherein, represents the energy storage battery state of charge at i+1 time; represents the energy storage battery state of charge at i time; represents the energy storage rated capacity; Take 5 min.

[0053] Optionally, in the minute-level mode, the power distribution optimization of the dual-winding generator and the energy storage of the dual-winding generator and energy storage combined power supply device by using game theory further includes: distributed solving of the dual-winding generator output power and the energy storage output power by using an alternating direction multiplier method to obtain the dual-winding generator output power of the k+1th iteration and the energy storage output power of the k+1th iteration, updating the kth iteration value of the global coordination variable and the kth iteration value of the Lagrange multiplier to obtain the k+1th iteration value of the global coordination variable and the k+1th iteration value of the Lagrange multiplier.

[0054] A control system of a dual-winding generator and energy storage combined power supply device for executing the control method of the dual-winding generator and energy storage combined power supply device as described in any one of the above, comprising: a data acquisition unit, a logic decision unit and a hierarchical time domain control unit, the data acquisition unit, the logic decision unit and the hierarchical time domain control unit being connected in sequence.

[0055] Optionally, the hierarchical time domain control unit comprises: a millisecond-level sliding mode dynamic compensation module, a second-level deep reinforcement learning module and a minute-level game theory power distribution module; the millisecond-level sliding mode dynamic compensation module, the second-level deep reinforcement learning module and the minute-level game theory power distribution module are connected with the logic decision unit respectively.

[0056] The control method and system of the dual-winding generator and energy storage combined power supply device provided in the application are applied to a dual-winding generator and energy storage combined dual-winding generator and energy storage combined power supply device, adopt a hybrid triggering mechanism of event-driven and state evaluation, calculate a mutation intensity index by real-time acquisition of voltage, current, power and other parameters, and determine the power distribution of the dual-winding generator and the energy storage according to the mutation intensity index , event-driven and state evaluation switching control level is performed according to the real-time state of the dual-winding generator and energy storage combined power supply device, multi-time scale layered collaborative control under the parallel structure of the dual-winding generator and the energy storage system, the control tasks between each time scale level are not repetitive, hybrid triggering and non-preemptive arbitration are realized; the millisecond-level improved sliding film control is used for adjusting the disturbance, through the coordinated compensation of feedforward-feedback, fast disturbance response is realized, voltage mutation and harmonic disturbance can be effectively inhibited, and power quality is improved; the second-level deep reinforcement learning is used for adjusting the proportional integral coefficient, adaptive optimization is realized, no artificial rules are needed, the control parameters can be dynamically adjusted, voltage stability, energy efficiency optimization and equipment life are considered, and system robustness and long-term benefits are improved; the minute-level game is used for optimizing the power distribution of the dual-winding generator and the energy storage, the power generation cost can be reduced, the energy storage life can be prolonged, meanwhile, the state of charge of the energy storage battery can be ensured to be maintained in a safe range, and the economy and reliability of the dual-winding generator and the energy storage combined power supply device are improved; the control method of the dual-winding generator and the energy storage combined power supply device can be compatible with different load change scenarios, and is especially suitable for power generation and energy storage collaborative control in a high dynamic fluctuation environment, and the stability and reliability of the dual-winding generator and the energy storage combined power supply device are improved.

[0057] To make the above features and advantages of the application more obvious and easy to understand, the following embodiments are specifically described below, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 The structure diagram of the dual-winding generator and the energy storage combined power supply device used in the present application.

[0059] Figure 2 The flow chart of the control method of the dual-winding generator and the energy storage combined power supply device of the present application.

[0060] Figure 3 The module diagram of the control system of the dual-winding generator and the energy storage combined power supply device of the present application. DETAILED DESCRIPTION

[0061] To make the purpose and technical scheme of the embodiments of the present application more clear, the technical scheme of the embodiments of the present application will be described clearly and completely in combination with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the present application.

[0062] The present application provides a control method of a dual-winding generator and an energy storage combined power supply device, which is applied to a dual-winding generator and an energy storage combined power supply device. Please refer to Figure 1 , Figure 1The structure diagram of the double-winding generator and energy storage combined power supply device used in the application, the double-winding generator and energy storage combined power supply device comprises: an energy storage battery 1, an energy storage converter 2, a double-winding generator 3 and a prime mover 4, the energy storage battery 1, the energy storage converter 2, the double-winding generator 3 and the prime mover 4 are connected in sequence. The double-winding generator 3 is equipped with two sets of stator windings, including a control winding and a power winding (not shown in the figure); wherein the control winding is connected with the energy storage converter 2, the magnetic field strength of the double-winding generator 3 is controlled by adjusting excitation reactive power, and the power winding is connected to the load. The power output of the double-winding generator 3 is controlled by the energy storage converter 2, and the remaining power required by the load is supplemented by the battery side. The prime mover 4 drives the double-winding generator to operate.

[0063] As an example, the control method of the double-winding generator and energy storage combined power supply device provided by the application is applied to the double-winding generator and energy storage combined power supply device, which is separated from the traditional synchronous machine or single power system in system structure; after the energy storage and the generator are used in parallel, the control method of the double-winding generator and energy storage combined power supply device provided by the application is used to solve the problem that the control optimization of the single generator is difficult to solve, the single energy storage does not have this problem but the capacity of the energy storage is small and it is difficult to meet the long-time power supply demand, which optimizes the control of the single generator and increases the capacity of the single energy storage. Among them, the double-winding generator 3 is used as a controllable current source by excitation adjustment, and the energy storage battery 1 is used as a fast response voltage source.

[0064] In an embodiment of the application, please refer to Figure 2 , Figure 2 The flow chart of the control method of the double-winding generator and energy storage combined power supply device of the application, the application provides a control method of a double-winding generator and energy storage combined power supply device, comprising: steps S1-S8.

[0065] Step S1: real-time acquisition of parameters of the double-winding generator and energy storage combined power supply device, calculation of the mutation intensity index by weighted fusion of the power change rate and the second derivative of the voltage According to the mutation intensity index Logical decision is made to switch different time scale control modes;

[0066] Step S2: judging whether the mutation intensity index is greater than the threshold value, if yes, entering step S3; if no, entering step S4;

[0067] Step S3: switching to the millisecond level mode, using sliding film control combined with disturbance prediction and phasor decomposition for feedforward-feedback collaborative compensation, returning to step S1;

[0068] Step S4: judging whether the mutation intensity index whether the mutation duration is greater than 1s, if yes, go to step S5; if no, go to step S6;

[0069] Step S5: switch to the second-level mode, adjust the proportional integral (PI) coefficient by deep reinforcement learning (DQN), and return to step S1 through real-time data-driven control rule optimization;

[0070] Step S6: determine whether the double-winding generator and energy storage combined power supply device is periodically triggered mutation, if yes, go to step S7; if no, go to step S8;

[0071] Step S7: switch to the minute-level mode, and use game theory to optimize the power distribution of the double-winding generator and energy storage of the double-winding generator and energy storage combined power supply device, and return to step S1;

[0072] Step S8: the double-winding generator and energy storage combined power supply device maintains the current mode, and returns to step S1.

[0073] The control method of the double-winding generator and energy storage combined power supply device provided in the application is applied to the double-winding generator and energy storage combined power supply device, adopts a hybrid triggering mechanism of event driving and state evaluation, calculates the mutation intensity index by real-time acquisition of voltage, current, power and other parameters, switches the control level according to the real-time state of the double-winding generator and energy storage combined power supply device, and realizes multi-time scale layered collaborative control under the parallel structure of the double-winding generator and energy storage system, the control tasks between each time scale level are not repetitive, and hybrid triggering and non-preemptive arbitration are realized; the millisecond-level improved sliding film control is used for adjusting disturbance, realizes fast disturbance response through feedforward-feedback collaborative compensation, can effectively suppress voltage mutation and harmonic disturbance, and improves power quality; the second-level deep reinforcement learning is used for adjusting the proportional integral coefficient, realizes adaptive optimization, does not need artificial rules, can dynamically adjust the control parameters, takes into account voltage stability, energy efficiency optimization and equipment life, improves system robustness and long-term benefits; the minute-level game is used for optimizing the power distribution of the double-winding generator and energy storage, can reduce power generation cost, prolong the service life of the energy storage, and at the same time ensure that the state of charge of the energy storage battery is maintained within a safe range, improve the economy and reliability of the double-winding generator and energy storage combined power supply device; the control method of the double-winding generator and energy storage combined power supply device can be compatible with different load change scenarios, and is especially suitable for power generation and energy storage collaborative control in high dynamic fluctuation environment, and improves the stability and reliability of the double-winding generator and energy storage combined power supply device.

[0074] In step S1, please refer to Figure 2In step S1, parameters of the dual-winding generator and energy storage combined power supply device are collected in real time, and the power change rate and the second derivative of voltage are weighted and fused to calculate the sudden change intensity index. According to the mutation intensity index Make logical decisions and switch between different time-scale control modes.

[0075] As an example, parameters of the dual-winding generator and energy storage combined power supply device are collected in real time, including: voltage. Current Instantaneous active power Isotropic signals. Mutation intensity index By weighted and fused with the power change rate and the second derivative of voltage, this method specifically addresses the voltage surge characteristics caused by dual-winding coupling in dual-winding generators, and improves the surge intensity index. It is expressed as follows:

[0076]

[0077] in, , representing instantaneous active power The instantaneous change Represents the power at time t. This represents the power at time t-1. The weighting coefficients for balancing the effects of power surges and voltage disturbances need to be dynamically adjusted based on real-time operating conditions.

[0078] As an example, this application uses a mutation intensity index. The system determines the triggering level, implements a hybrid triggering mechanism combining event-driven and state-evaluation approaches, and dynamically adjusts the mutation intensity index. Unlike fixed-time window scanning, it has responsiveness and adaptability.

[0079] As an example, dynamically adjusting the mutation intensity index This includes: using reinforcement learning methods to dynamically optimize based on the operating status and control effect of the dual-winding generator and energy storage combined power supply device; specifically including steps S11 to S15.

[0080] Step S11: Design the state space.

[0081] As an example, a state vector S containing the operating characteristics of a dual-winding generator and energy storage combined power supply device is constructed, including: voltage deviation ΔV, power change ΔP, voltage second derivative d²V / dt², current mutation intensity index ξ, current weight coefficient λ value, control response delay, number of false triggers, energy storage battery state of charge (SOC), and load type.

[0082] Step S12: Design the motion space.

[0083] As an example, the adjustment step of setting the weight coefficient λ is set as a discrete action set A = {−0.1, 0, +0.1}; wherein, reducing the weight coefficient λ represents reducing the sensitivity to voltage disturbance, and increasing the weight coefficient λ represents enhancing the response to voltage mutation.

[0084] Step S13: design reward function R2.

[0085] As an example, a multi-objective reward function R2 considering stability, response speed and energy efficiency is constructed, represented as follows:

[0086]

[0087] wherein, represents the absolute value of voltage deviation; represents the time for the double-winding generator and energy storage combined power supply device to reach steady state; represents the number of false triggering controls; represents the degree of deviation from the ideal range; represents the weight coefficient of voltage deviation; represents the weight coefficient of the time for the double-winding generator and energy storage combined power supply device to reach steady state; represents the weight coefficient of the number of false triggering controls; represents the weight coefficient of the degree of deviation from the ideal range.

[0088] Step S14: policy optimization.

[0089] As an example, deep Q network is adopted for policy learning, represented as follows:

[0090]

[0091] wherein, represents the current state-action pair; S represents the state space value, and A represents the action space value; represents the learning rate, used to control the update step; represents the discount factor, used to weigh the immediate and long-term rewards; represents the adjusted action space value obtained state space value; represents the adjusted action space value; represents the adjusted state-action pair.

[0092] Step S15: state rollback.

[0093] As an example, if adjusting the weight coefficient λ causes the double-winding generator and energy storage combined power supply device to frequently mis-trigger or voltage fluctuation to be greater than 10%, the strategy rollback is triggered immediately to restore the action space to the stable version before adjustment; if abnormality is triggered for 3 times in succession, the action is shielded.

[0094] In step S2, please refer to the S2 step in Figure 2 , judge whether the mutation intensity index is greater than the threshold value, if yes, enter step S3; if no, enter step S4.

[0095] In an embodiment of the present application, the threshold value is set to 50.

[0096] As an example, whether the mutation intensity index is greater than the threshold value includes: Or .

[0097] In step S3, please refer to the S3 step in Figure 2 , switch to the millisecond level mode, adopt the sliding film control combined with the disturbance prediction and phasor decomposition for the feedforward-feedback collaborative compensation, and return to step S1.

[0098] As an example, the sliding film control is adopted in the millisecond level mode, which is driven according to the physical model.

[0099] As an example, after switching to the millisecond level mode, the voltage , current , instantaneous active power and other signals are synchronously collected.

[0100] As an example, for the feedforward compensation, the power disturbance is predicted through the power change rate , the reverse current is directly injected to quickly offset the power change, and the overshoot caused by the feedback delay is avoided. The feedforward compensation current of the feedforward compensation is represented as follows:

[0101]

[0102] Among them, represents the feedforward gain, which determines the compensation intensity; is the time constant, which is used to control the compensation decay speed.

[0103] As an example, the sliding film control is adopted, which is used for the control winding of the double-winding generator to adjust the excitation reactive power, controls the magnetic field intensity of the double-winding generator and the energy storage, extracts and processes the error of the voltage phasor (fundamental wave, noise, etc.), converts the error dynamic convergence process into a controllable trajectory, and thus facilitates subsequent calculation. The sliding surface of the sliding film control is represented as follows: ​​

[0104]

[0105] wherein, i.e. reference voltage and actual measured voltage the difference represents voltage deviation; represents sliding surface coefficient, used to adjust integral term weight.

[0106] As an example, feedback compensation is carried out, high-frequency noise and fundamental component are separated by dynamic phasor decomposition technology, and only effective disturbance is compensated; at the same time, power change rate is combined to predict power disturbance and real-time feedback, response speed and control accuracy real-time error adjustment are taken into account, unmodeled disturbance and uncertainty of double-winding generator and energy storage combined power supply device are eliminated, and steady-state accuracy is improved. Feedback compensation current of feedback compensation is represented as follows:

[0107]

[0108] wherein, and both represent feedback gain, used to adjust steady-state accuracy; represents threshold parameter of error boundary layer, used to control smoothness, so as to suppress chattering, the essence is to solve the problem of high-frequency chattering caused by sign function sgn in traditional sliding mode control by setting boundary layer, the smaller the boundary layer is, the narrower the boundary layer is, the more sensitive the control is but the chattering may increase, which needs to be set according to characteristics of double-winding generator and energy storage combined power supply device; tanh represents hyperbolic tangent function; sat(s) represents saturation function, s represents sliding surface obtained calculation value; wherein, saturation function sat(s) is represented as follows:

[0109]

[0110] wherein, is a set threshold value; when , the function value is ; when , the function value is .

[0111] Specifically, feedback compensation current , hyperbolic tangent function tanh is used to replace sign function sgn , smooth transition is realized within error boundary layer, i.e. , and high-frequency chattering is eliminated.

[0112] As an example, the feedforward-feedback collaborative compensation task is completed, and the step S1 is returned to wait for the next trigger. , so as to quickly suppress the voltage mutation and harmonic disturbance within 1-100 ms. The composite compensation current is expressed as follows:

[0113] .

[0114] As an example, the feedforward-feedback collaborative compensation task is completed, and the step S1 is returned to wait for the next trigger.

[0115] As an example, in the millisecond mode, the traditional sliding mode control is improved in the engineering optimization of the application of the dual-winding generator and the energy storage combined power supply device in the energy storage system, the feedforward-feedback collaborative compensation is combined with disturbance prediction and phasor decomposition, and the problem that the traditional sliding mode control is only dependent on feedback, the mutation disturbance is suppressed quickly, but the parameter change and the unmodeled disturbance are not sensitive, so that the response speed and the accuracy are difficult to be considered; meanwhile, in the feedback compensation, the combination of the hyperbolic tangent function tanh and the saturation function sat(s) is combined with the low inertia characteristics of the dual-winding generator and the energy storage combined power supply device (the energy storage responds quickly but the capacity is limited), which avoids the influence of the high-frequency chattering of the traditional sliding mode on the service life of the energy storage battery.

[0116] In step S4, please refer to the S4 step in Figure 2 , judge whether the mutation duration of the mutation intensity index is greater than 1s, if yes, enter step S5; if not, enter step S6.

[0117] As an example, whether the mutation duration of the mutation intensity index is greater than 1s, includes: And the duration is greater than 1s.

[0118] In step S5, please refer to the S5 step in Figure 2 , switch to the second mode, adjust the proportional integral coefficient by using deep reinforcement learning, and return to step S1 by driving the control rule optimization in real time.

[0119] As an example, the second mode uses deep reinforcement learning according to data driving.

[0120] As an example, the state space and the action space are defined to reflect the state of the dual-winding generator and the energy storage combined power supply device in real time, wherein the state space includes the interaction characteristics of the dual-winding generator and the energy storage instead of the general power grid parameters, contains 12-dimensional characteristics, reflects the dynamic characteristics of the dual-winding generator and the energy storage combined power supply device, including: voltage deviation , rate of voltage change , state of charge of energy storage, load power , generator output power , energy storage charge-discharge power , past 10 seconds history error mean, error variance, harmonic distortion, control output amplitude, ambient temperature, time stamp periodicity.

[0121] The action space is defined as adjusting the proportional and integral coefficients, suppressing overshoot and shortening the adjustment time, and is represented as follows:

[0122]

[0123]

[0124] wherein, represents the adjustment amount of the proportional coefficient of PI control, represents the adjustment amount of the integral coefficient of PI control.

[0125] As an example, the reward function is designed R , quantifying voltage stability, energy efficiency, equipment life, and other multi-objective to maximize long-term benefits, and is represented as follows:

[0126]

[0127] wherein, represents the voltage stability weight, used to dominate error suppression; represents the energy efficiency weight, used to encourage efficient operation of the system; represents the energy efficiency coefficient of the dual-winding generator and energy storage combined power supply device, represents the output power of the dual-winding generator and energy storage combined power supply device, represents the input power of the dual-winding generator and energy storage combined power supply device; represents the frequency offset penalty term; represents the frequency offset; represents the state of charge health penalty of the energy storage battery, represents the state of charge health penalty term, which is adapted to periodic loads (such as the periodic fluctuations of refrigerators and punchers), and solves the problem of poor adaptability of traditional deep reinforcement learning to periodic disturbances by optimizing the history error mean (past 10 seconds) optimization rule.

[0128] As an example, the network structure of deep reinforcement learning is constructed, and the Q value update formula of the network structure of deep reinforcement learning is represented as follows:

[0129]

[0130] wherein, represents the current state-action pair; represents the state space value; represents the action space value; represents the learning rate for controlling the update step; represents the discount factor for weighing immediate and long-term rewards; represents the adjusted action space value the resulting state space value; represents the adjusted action space value; represents the adjusted state-action pair.

[0131] Specifically, the single control rule modification amount is limited to ≤10% to avoid control mutations.

[0132] As an example, after optimizing the control rule, stability verification is performed, including: real-time monitoring of voltage deviation and frequency deviation , if or or the optimized control rule does not meet the Lyapunov stability condition, the strategy rollback is triggered, and the proportional coefficient and integral coefficient are restored to the stable version before adjustment, represented as follows:

[0133]

[0134] wherein, represents the proportional coefficient after triggering the strategy rollback; represents the integral coefficient after triggering the strategy rollback; represents the proportional coefficient before adjustment; represents the integral coefficient before adjustment.

[0135] Specifically, the optimized control rule meeting the Lyapunov stability condition includes: the optimized control rule meeting the parameter adjustment boundary of the Lyapunov Lyapunov ) stability condition, ensuring the convergence of the control method of the double-winding generator and energy storage combined power supply device, and combining with the strategy rollback mechanism, avoiding the instability risk commonly seen in traditional deep reinforcement learning in power systems. The Lyapunov stability condition is represented as follows:

[0136]

[0137] wherein, , are the first and second derivatives of , respectively.

[0138] As an example, real-time monitoring of voltage deviation and frequency deviation , if or If the optimized control rules do not meet the Lyapunov stability condition, a policy rollback will be triggered, restoring the proportional and integral coefficients to their stable versions before adjustment, as shown below:

[0139]

[0140] in, This represents the proportional coefficient after triggering the policy rollback; This represents the integral coefficient after triggering the policy rollback; This indicates the scaling factor before adjustment; This represents the integral coefficient before adjustment.

[0141] Furthermore, when a policy rollback is triggered, the abnormal action at that time is marked. To reduce abnormal actions, the risk level is set at "high". The Q-value weights are represented as follows:

[0142]

[0143]

[0144] in, This represents the maximum temporal difference error observed during the current training period, used to quantify anomalous actions. The degree of abnormality, The larger the value, the greater the difference between the agent's expected reward and the actual result for that action, requiring more severe punishment; i represents the experience replay pool, which represents the parameters that are expected to be used in two consecutive training iterations. They are independent, but adjacent sets of parameters There is a strong correlation between the data and the training data. Using highly correlated data often yields poor results. Experience replay can be used to randomly select a set of data from the array each time. Used for parameter updates, the experience replay pool has a capacity of 10,000 data sets and stores transfer tuples. The subscript i represents the sample index number in the experience replay pool, identifying the i-th group of state transition data randomly sampled from historical data. ; The reward function represents the state transition data for the i-th group; Represents the adjusted action space value of the i-th set of state transition data. The obtained state space values; Represents the state space value of the i-th group of state transition data; This represents the action space value of the i-th group of state transition data.

[0145] Specifically, if no abnormality occurs for 10 consecutive minutes, the complete learning process is resumed. If the same action triggers an abnormality for 3 consecutive times, the abnormal action is permanently shielded .

[0146] As an example, the optimization control rule or parameter task is completed, and the step S1 is returned to wait for the next trigger.

[0147] As an example, in the second-level mode, the control rule or parameter is dynamically optimized by deep reinforcement learning, and the logical linkage mechanism of sliding mode control and game control is combined to dynamically optimize the control parameter or instruction in the second-level time scale (1-60 seconds), so as to achieve the following goals: autonomously adjusting the proportional integral coefficient or directly generating the control instruction to suppress overshoot and shorten the adjustment time; balancing voltage stability, energy efficiency and device life loss; without human experience, the rule evolution is driven by real-time data.

[0148] In step S6, please refer to the S6 step in Figure 2 , to determine whether the double-winding generator and energy storage combined power supply device periodically triggers a mutation. If yes, go to step S7; if no, go to step S8.

[0149] As an example, the double-winding generator and energy storage combined power supply device periodically triggers a mutation, that is, the double-winding generator and energy storage combined power supply device is in steady-state operation.

[0150] In step S7, please refer to the S7 step in Figure 2 , step S7: switch to the minute-level mode, use game theory to optimize the power distribution of the double-winding generator and energy storage of the double-winding generator and energy storage combined power supply device, and return to step S1.

[0151] As an example, the minute-level mode uses game driving and alternating direction multiplier method (ADMM) for solution.

[0152] As an example, the game participants and strategy space are defined, wherein the game participants are the double-winding generator and energy storage. The strategy space includes: represents the double-winding generator power output range, represents the energy storage power output range; wherein, represents the double-winding generator output power set value; represents the minimum value of the double-winding generator output power set; represents the maximum value of the double-winding generator output power set; represents the energy storage output power set value; represents the minimum value of the energy storage output power set; represents the maximum value of the energy storage output power set.

[0153] As an example, the double-winding generator revenue function , is expressed as follows:

[0154]

[0155] wherein, represents the unit power generation benefit coefficient of the double-winding generator (yuan / kWh), used to reflect the fuel cost; represents the power generation cost penalty coefficient of the double-winding generator (yuan / kW2), used to suppress the loss of high power output; represents the storage battery state of charge balancing coefficient, used to punish the working condition deviating from the intermediate value 50%. Since the 50% storage battery state of charge enables the storage to have bidirectional regulation capability (balanced charging and discharging space), the storage battery state of charge is balanced to 50%, so that the double-winding generator can flexibly adjust the power output according to the load fluctuation; represents the power generation power of the double-winding generator.

[0156] As an example, the storage benefit function is established , is expressed as follows:

[0157]

[0158] wherein, represents the storage discharge benefit coefficient (yuan / kWh), reflecting the economic value of the storage discharge; represents the storage discharge loss penalty coefficient (yuan / kW2); represents the storage battery state of charge target preference coefficient. Since the 60% storage battery state of charge reserves more discharge margin for sudden load growth or renewable energy output decline, the storage battery state of charge is preferred to be maintained at 60%, thereby improving the power supply reliability of the double-winding generator and storage combined power supply device.

[0159] Specifically, in the game between the double-winding generator and the storage of the double-winding generator and storage combined power supply device, the storage battery state of charge is dynamically stabilized in the safe and efficient interval (50%-60%). The double-winding generator and storage combined power supply device realizes the double-target dynamic balance of the storage battery state of charge through the penalty term in the benefit function, such as the double-winding generator being preferred to be set at 50% and the storage being preferred to be set at 60%, which solves the contradiction between economy (50% bidirectional regulation) and reliability (60% margin) that the existing game method mainly adopts a single storage battery state of charge target, and takes into account economy, reliability and equipment life.

[0160] As an example, the game constraints are constructed, including: power balance constraint, storage battery state of charge dynamic constraint. The power balance constraint only includes the double-winding generator, the storage and the load, and does not interact with the power grid, and is expressed as follows:

[0161] ;

[0162] The dynamic constraint of the energy storage battery state of charge is expressed as follows:

[0163]

[0164] wherein, represents the energy storage battery state of charge at t+1; represents the energy storage battery state of charge at t; represents the energy storage rated capacity; 5 min is taken.

[0165] As an example, the alternating direction multiplier method is adopted for distributed solving, the alternating direction multiplier method is combined with the excitation characteristics of the double-winding generator (power distribution of the control winding and the power winding), the local strategy is independently iteratively optimized for the output power distribution of the double-winding generator and the energy storage, the global optimization is achieved through the coordination variable, and is expressed as follows:

[0166]

[0167]

[0168] wherein, represents the double-winding generator output power of the k+1th iteration; represents the energy storage output power of the k+1th iteration; represents a penalty coefficient, used for adjusting the weight of the local optimization and the global constraint, and the value of the penalty coefficient increases to force the participants to more strictly meet the power balance constraint; represents the kth iteration value of the global coordination variable, representing the reference value of the total power demand of the double-winding generator and the energy storage combined power supply device; represents the kth iteration value of the Lagrange multiplier, used for balancing the local optimization and the global constraint.

[0169] Further, the double-winding generator output power of the k+1th iteration and the energy storage output power of the k+1th iteration after the iterative optimization are sent to the control winding of the double-winding generator and the energy storage battery, the kth iteration value of the global coordination variable and the kth iteration value of the Lagrange multiplier are updated to obtain the k+1th iteration value of the global coordination variable and the k+1th iteration value of the Lagrange multiplier , which are expressed as follows:

[0170]

[0171]

[0172] wherein, represents the k+1th iteration value of the global coordination variable; represents the k+1th iteration value of the Lagrange multiplier. The traditional alternating direction multiplier method is used for power grid, which relies on central coordination. The application uses the alternating direction multiplier method to solve the control winding excitation regulation of the double-winding generator and the charging and discharging control of the energy storage, which is completely based on local decision-making without the intervention of the central controller, and is suitable for the decentralized scenario of micro-grid or emergency power supply.

[0173] Specifically, the alternating direction multiplier method is iterated every 5 minutes, and when is terminated to prevent infinite loop.

[0174] As an example, the power distribution optimization task is completed, and step S1 is returned to wait for the next trigger.

[0175] As an example, the minute-level mode is selected when the double-winding generator and energy storage combined power supply device is periodically triggered or in steady state, the power distribution optimization of the double-winding generator and energy storage is realized through game theory, the power distribution of the double-winding generator and energy storage is globally coordinated, so as to minimize the fuel cost of power generation, prolong the service life of energy storage, ensure that the output power of the double-winding generator and energy storage meets the load demand, and maintain the state of charge of the energy storage in the safe interval; At the same time, the control winding excitation regulation of the double-winding generator and the charging and discharging control of the energy storage are completely based on local decision-making without the intervention of the central controller.

[0176] As an example, during the control process of the double-winding generator and energy storage combined power supply device, the minute-level mode is switched to, which also includes: every 5 minutes or when the state of charge of the energy storage battery deviates from the target value ±10%, the minute-level mode is switched to for the power distribution optimization of the double-winding generator and energy storage of the double-winding generator and energy storage combined power supply device.

[0177] Specifically, the state evaluation is integrated into the whole control method of the double-winding generator and energy storage combined power supply device, which drives the double-winding generator and energy storage combined power supply device to switch the control level and optimize. The state evaluation includes: calculating the mutation intensity index , designing the reward function R , checking the stability, establishing the double-winding generator benefit function and establishing the energy storage benefit function and other key control judgments, which are implicit but continuously executed processes.

[0178] As an example, the control method of the dual-winding generator combined with the energy storage power supply device further includes fault protection. When the current of the bus of the dual-winding generator combined with the energy storage power supply device exceeds 1.2 times the rated current, the bus voltage exceeds 1.2 times the rated voltage, the voltage frequency exceeds the rated frequency ± 0.5 Hz, the battery SOC is less than 10% or greater than 100%, any of the above conditions is met, the fault protection is triggered, the dual-winding generator combined with the energy storage power supply device is hierarchically isolated and investigated to determine the fault point, and after the fault is eliminated, verification is performed to observe whether the dual-winding generator combined with the energy storage power supply device triggers the fault protection again, and to ensure the safe operation of the dual-winding generator combined with the energy storage power supply device. The fault protection belongs to an independent safety step and has the highest priority.

[0179] In the control of the dual-winding generator combined with the energy storage power supply device, the tasks among the multi-layer algorithms are clearly distinguished and do not interfere with each other. A non-preemptive priority queue scheduling is set for control: fault protection > millisecond level > second level > minute level.

[0180] In another embodiment of the present application, please refer to Figure 3 , Figure 3 As a module diagram of the control system of the dual-winding generator combined with the energy storage power supply device of the present application, the present application further provides a control system of a dual-winding generator combined with an energy storage power supply device for executing the above-mentioned control method of a dual-winding generator combined with an energy storage power supply device, comprising: a data acquisition unit 31, a logic decision unit 32 and a hierarchical time domain control unit 33, which are connected in sequence.

[0181] As an example, the data acquisition unit 31 is used to acquire the parameters of the dual-winding generator combined with the energy storage power supply device in real time.

[0182] The logic decision unit 32 is used to calculate the mutation intensity index by fusing the power change rate and the second derivative of the voltage, and to make logical decisions according to the mutation intensity index to switch different time scale control modes.

[0183] The hierarchical time domain control unit 33 is used to switch the millisecond level mode, the second level mode and the minute level mode for hierarchical time domain control.

[0184] As an example, the hierarchical time domain control unit 33 includes: a millisecond level sliding mode dynamic compensation module 331, a second level deep reinforcement learning module 332 and a minute level game theory power distribution module 333; the millisecond level sliding mode dynamic compensation module 331, the second level deep reinforcement learning module 332 and the minute level game theory power distribution module 333 are connected with the logic decision unit 32 respectively.

[0185] As an example, the millisecond-level sliding mode dynamic compensation module 331 includes an improved sliding film observer (not shown in the figure) for adopting sliding film control to combine disturbance prediction and phasor decomposition for feedforward-feedback collaborative compensation;

[0186] The second-level deep reinforcement learning module 332 includes a deep reinforcement learning rule optimizer (not shown in the figure) for adopting deep reinforcement learning to adjust the proportional integral coefficient through real-time data-driven control rule optimization.

[0187] The minute-level game theory power distribution module 333 includes a game coordinator (not shown in the figure) for adopting game theory to optimize the power distribution of the double-winding generator and the energy storage of the double-winding generator and energy storage combined power supply device.

[0188] In another embodiment of the present application, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions which, when executed by a computer, enable the computer to perform the steps of the control method of the double-winding generator and energy storage combined power supply device provided in the above embodiments.

[0189] The control method and system of the double-winding generator and energy storage combined power supply device provided by the present application are applied to the double-winding generator and energy storage combined power supply device, adopt a hybrid triggering mechanism of event-driven and state evaluation, calculate the mutation intensity index by real-time acquisition of voltage, current, power and other parameters, switch the control level according to the real-time state of the double-winding generator and energy storage combined power supply device, and perform multi-time scale layered collaborative control under the parallel structure of the double-winding generator and energy storage system, so that the control tasks between the time scale levels are not repetitive, hybrid triggering and non-preemptive arbitration are realized, the millisecond-level improved sliding film control is used to adjust the disturbance, feedforward-feedback collaborative compensation is realized, fast disturbance response is realized, voltage mutation and harmonic disturbance can be effectively suppressed, and power quality is improved, the second-level deep reinforcement learning is used to adjust the proportional integral coefficient, adaptive optimization is realized, no artificial rules are needed, control parameters can be dynamically adjusted, voltage stability, energy efficiency optimization and equipment life are considered, and system robustness and long-term benefits are improved, the minute-level game is used to optimize the power distribution of the double-winding generator and the energy storage, the power generation cost can be reduced, the energy storage life can be prolonged, the state of charge of the energy storage battery can be maintained within a safe range, the economy and reliability of the double-winding generator and energy storage combined power supply device are improved, and the control method of the double-winding generator and energy storage combined power supply device can be compatible with different load change scenarios, is especially suitable for power generation and energy storage collaborative control in a high dynamic fluctuation environment, and improves the stability and reliability of the double-winding generator and energy storage combined power supply device.

[0190] Although the present application has been disclosed with reference to the embodiments above, it is not intended to limit the present application, and any person skilled in the art can make some changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application is defined by the appended claims.

Claims

1. A control method of a dual-winding generator and energy storage combined power supply device, the dual-winding generator and energy storage combined power supply device comprising: The energy storage battery, energy storage converter, double-winding generator and prime mover are connected in sequence, and the control method of the double-winding generator and the energy storage combined power supply device comprises, Step S1: Real-time acquisition of parameters of the double-winding generator and the energy storage combined power supply device, fusion calculation of the power change rate and the second derivative of the voltage to obtain a sudden change intensity index, logical decision according to the sudden change intensity index, and switching of different time scale control modes; Step S2: Determine whether the sudden change intensity index is greater than a threshold value, if yes, go to step S3; if no, go to step S4; Step S3: Switch to the millisecond level mode, adopt sliding mode control combined with disturbance prediction and phasor decomposition for feedforward-feedback collaborative compensation, and return to step S1; Step S4: Determine whether the sudden change duration of the sudden change intensity index is greater than 1s, if yes, go to step S5; if no, go to step S6; Step S5: Switch to the second level mode, adjust the proportional integral coefficient by deep reinforcement learning, and optimize the control rule by real-time data driving, and return to step S1; Step S6: Determine whether the double-winding generator and the energy storage combined power supply device is periodically triggered by sudden change, if yes, go to step S7; if no, go to step S8; Step S7: Switch to the minute level mode, optimize the power distribution of the double-winding generator and the energy storage of the double-winding generator and the energy storage combined power supply device by game theory, and return to step S1; Step S8: The double-winding generator and the energy storage combined power supply device maintains the current mode, and returns to step S1.

2. The control method of the dual-winding generator and energy storage combined power supply device according to claim 1, characterized by, The sudden change intensity index is represented as follows: wherein, represents a mutation intensity index; represents a transient change amount of the transient active power ; represents the power at time t; represents the power at time t-1; is a weight coefficient.

3. The control method of the dual-winding generator and energy storage combined power supply device according to claim 2, characterized by, In the millisecond level mode, the feedforward compensation current of the feedforward compensation is represented as follows: wherein, represents a feedforward compensation current; represents a feedforward gain; is a time constant; represents an instantaneous active power; The feedback compensation current of the feedback compensation is represented as follows: wherein, denotes a feedback compensation current; denotes a feedback gain; denotes a feedback gain; denotes a smoothing factor; tanh denotes a hyperbolic tangent function; sat(s) denotes a saturation function, s denotes a sliding surface the resulting calculated value; The sliding mode surface of the sliding mode control is represented as follows: wherein, represents a sliding surface controlled by a sliding mode; i.e. the reference voltage and the actual measured voltage the difference between the reference voltage represents a sliding surface coefficient for adjusting the weight of the integral term; represents a time constant is 0 to t the integral of the difference between the reference voltage and the actual measured voltage at the instant of time.

4. The control method of the dual-winding generator and energy storage combined power supply device according to claim 2, characterized by, In the second level mode, adjusting the proportional integral coefficient by deep reinforcement learning includes: defining the state space and action space to reflect the state of the double-winding generator and the energy storage combined power supply device in real time; The state space includes the interaction characteristics of the double-winding generator and the energy storage, including: voltage deviation, voltage change rate, energy storage battery state of charge, load power, generator output power, energy storage charging and discharging power, past 10s historical error mean, error variance, harmonic distortion rate, control output amplitude, environmental temperature and time stamp periodic characteristics; The action space is defined as adjusting the proportional integral coefficient, represented as follows: wherein represents an adjustment amount of a proportional coefficient, represents an adjustment amount of an integral coefficient.

5. The control method of the dual-winding generator and energy storage combined power supply device according to claim 4, characterized by, In the second level mode, adjusting the proportional integral coefficient by deep reinforcement learning also includes: designing a reward function, represented as follows: wherein, R represents a reward function; represents a voltage stability weight; represents an energy efficiency weight; represents a double-winding generator and energy storage combined power supply device energy efficiency coefficient, represents a double-winding generator and energy storage combined power supply device output power, represents a double-winding generator and energy storage combined power supply device input power; represents a frequency deviation penalty term, represents a frequency deviation; represents an energy storage battery state of charge health penalty; represents an energy storage battery state of charge health penalty term; represents a voltage deviation.

6. The control method of the dual-winding generator and energy storage combined power supply device according to claim 5, wherein After optimizing the control rule by real-time data driving in the second level mode, it also includes: stability check, which includes: real-time monitoring of voltage deviation and frequency deviation, when the voltage deviation, frequency deviation and optimized control rule do not meet the stability condition, the strategy rollback is triggered, and the proportional coefficient and integral coefficient are restored to the stable version before adjustment; not meeting the stability condition includes: voltage deviation > 10% reference voltage, frequency deviation > 0.5Hz and optimized control rule not meeting Lyapunov stability condition.

7. The control method of the dual-winding generator and energy storage combined power supply device according to claim 2, characterized by, In the minute-level mode, the power distribution optimization of the dual-winding generator and the energy storage of the dual-winding generator and energy storage combined power supply device by using game theory includes: defining game participants and strategy space; establishing participant benefit function; establishing game constraints; The participant benefit function is expressed as: wherein, represents the double-winding generator benefit function; represents the double-winding generator unit power generation benefit coefficient; represents the double-winding generator power generation cost penalty coefficient; represents the energy storage battery state of charge balancing coefficient, for penalizing the working condition deviating from the intermediate value 50%; represents the power generation power of the double-winding generator; wherein, represents a storage benefit function; represents a storage discharge benefit coefficient; represents a storage discharge penalty coefficient; represents a storage battery state of charge target preference coefficient, which maintains the preference storage battery state of charge at 60%; represents a storage discharge power; The game constraints include: power balance constraint, energy storage battery state of charge dynamic constraint; The power balance constraint is expressed as: ； wherein, represents the double-winding generator output power set value; represents the energy storage output power set value; represents the load power; The energy storage battery state of charge dynamic constraint is expressed as: wherein, represents the state of charge of the energy storage battery at time i+1; represents the state of charge of the energy storage battery at time i; represents the rated capacity of the energy storage; Take 5 min.

8. The control method of the dual-winding generator and energy storage combined power supply device according to claim 4, characterized by, In the minute-level mode, the power distribution optimization of the dual-winding generator and the energy storage of the dual-winding generator and energy storage combined power supply device by using game theory also includes: using the alternating direction multiplier method to distribute the dual-winding generator output power and the energy storage output power, to obtain the dual-winding generator output power of the k+1th iteration and the energy storage output power of the k+1th iteration, and update the kth iteration value of the global coordination variable and the kth iteration value of the Lagrange multiplier to obtain the k+1th iteration value of the global coordination variable and the k+1th iteration value of the Lagrange multiplier.

9. A control system of a dual-winding generator and energy storage combined power supply device, for executing the control method of the dual-winding generator and energy storage combined power supply device according to any one of claims 1 to 8, characterized by It includes: a data acquisition unit, a logical decision unit and a hierarchical time domain control unit, the data acquisition unit, the logical decision unit and the hierarchical time domain control unit are connected in sequence.

10. The control system for a dual-winding generator and energy storage combined power supply device according to claim 9, wherein The hierarchical time domain control unit includes: a millisecond-level sliding mode dynamic compensation module, a second-level deep reinforcement learning module and a minute-level game theory power distribution module; the millisecond-level sliding mode dynamic compensation module, the second-level deep reinforcement learning module and the minute-level game theory power distribution module are connected with the logical decision unit respectively.

Images