Gravity energy storage system charging and discharging switching method based on doubly-fed motor and model prediction

Abstract

The invention relates to the technical field of large-scale gravity energy storage system control, in particular to a gravity energy storage system charging and discharging switching method based on a doubly-fed motor and model prediction, and the method comprises the steps: obtaining a smooth non-overshoot power reference trajectory based on a five-order S-shaped curve algorithm through the real-time operation data of a power grid; obtaining an optimal control instruction sequence in a future period according to the smooth non-overshoot power reference trajectory; acquiring a switching signal of the power converter based on a flux linkage directional vector control strategy by using the optimal control instruction sequence in the future time period; according to the power converter switching signal, charging and discharging switching of the gravity energy storage system is carried out, through the multi-step look-ahead optimization capability of a model prediction control algorithm and in combination with a smooth reference trajectory generation method of an S-shaped curve, power and torque abrupt changes in the mode switching process are eliminated fundamentally, and the mode switching efficiency is improved. The key technical problem in charging and discharging mode switching of the gravity energy storage system is effectively solved, and important technical support is provided for large-scale application of the gravity energy storage technology.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of large-scale gravity energy storage system control, in particular to a gravity energy storage system charging and discharging switching method based on a doubly-fed motor and model prediction. BACKGROUND

[0002] Under the "double carbon" target, renewable energy represented by wind power and photovoltaic power develops rapidly, and the randomness and volatility of its output bring great challenges to the safe and stable operation of the power grid. Energy storage system is a key means to solve this problem. With the increasing penetration rate of renewable energy, the power grid has higher requirements for the response speed, stability and flexibility of the energy storage system. Gravity energy storage, as a new type of physical energy storage method, has the advantages of long service life, high efficiency and environmental friendliness, and is particularly suitable for use with fluctuating power sources such as wind power and photovoltaic power.

[0003] However, gravity energy storage has problems such as power surge, mechanical impact, and power grid frequency disturbance during charging and discharging mode switching. Traditional switching control strategies mostly use simple ramp functions or traditional PID control, which have obvious defects and are difficult to cope with complex power grid conditions and load changes, easily leading to system oscillation or protection action, thereby affecting the reliability and service life of the energy storage system.

[0004] Therefore, developing a charging and discharging switching control method that can realize impact-free, smooth and adaptive switching is crucial for promoting the commercial application of gravity energy storage technology. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a gravity energy storage system charging and discharging switching method based on a doubly-fed motor and model prediction, which adopts a hierarchical intelligent control architecture and realizes smooth transition of charging and discharging modes through coordinated control of multiple time scales, achieving impact-free smooth switching of the charging and discharging modes of the gravity energy storage system and effectively solving problems such as power surge, mechanical stress and power grid compatibility.

[0006] To achieve the above-mentioned purpose, the present application provides a gravity energy storage system charging and discharging switching method based on a doubly-fed motor and model prediction, comprising:

[0007] Using real-time operation data of the power grid, a smooth and non-overshoot power reference trajectory is obtained based on a five-order S-shaped curve algorithm;

[0008] According to the smooth and non-overshoot power reference trajectory, an optimal control instruction sequence for the future period is obtained;

[0009] Using the optimal control instruction sequence for the future period, a power converter switching signal is obtained based on a flux-oriented vector control strategy;

[0010] According to the power converter switching signal, the gravity energy storage system charging and discharging switching is performed.

[0011] Preferably, the obtaining of the smooth non-overshoot power reference trajectory based on the five-order S-shaped curve algorithm using the real-time operation data of the power grid comprises:

[0012] The real-time frequency of the power grid and the real-time DC voltage of the power grid are collected as the real-time operation data of the power grid.

[0013] When the power grid exists mode switching, the current power value and the target power value are obtained according to the real-time operation data of the power grid.

[0014] The current power value and the target power value are used to obtain the smooth non-overshoot power reference trajectory based on the five-order S-shaped curve algorithm.

[0015] Further, the calculation formula for obtaining the smooth non-overshoot power reference trajectory based on the five-order S-shaped curve algorithm using the current power value and the target power value is as follows:

[0016]

[0017] wherein, P ref(t) is the smooth non-overshoot power reference trajectory, P start is the initial power value of power change, and P end is the target power value of power change.

[0018] Further, the obtaining of the optimal control instruction sequence of the future period according to the smooth non-overshoot power reference trajectory comprises:

[0019] A discrete state space prediction model is established.

[0020] The smooth non-overshoot power reference trajectory is used to obtain the optimal control instruction sequence of the future period according to the discrete state space prediction model.

[0021] Further, the establishment of the discrete state space prediction model comprises:

[0022] A stator-rotor loop voltage equation is established based on the two-phase synchronous rotating d-q coordinate system according to the real-time operation data of the power grid.

[0023] The calculation formula for the magnetic flux linkage and current correlation is as follows:

[0024]

[0025] wherein, Ls, Lr and Lm are respectively the stator self-inductance, the rotor self-inductance and the rotor mutual inductance, i is the current, is the magnetic flux linkage;

[0026] The corresponding DFIG electromagnetic torque is obtained according to the magnetic flux linkage and current correlation.

[0027] The calculation formula of the mechanical motion equation of the electromagnetic process of the motor and the mechanical process of the gravity energy storage is established by using the electromagnetic torque of the DFIG, and is as follows:

[0028]

[0029] Wherein, J is the total rotational inertia of the system, B is the viscous friction coefficient, T e is the electromagnetic torque of the DFIG, T L is the load torque;

[0030] The calculation formula of the active power and the reactive power of the power grid is calculated according to the real-time operation data of the power grid, and is as follows:

[0031]

[0032] Wherein, is the active power, is the reactive power;

[0033] The nonlinear continuous state space equation is established by using the stator current and the rotor current as the state variables according to the stator-rotor loop voltage equation, the flux and the current correlation, and based on the mechanical motion equation;

[0034] The calculation formula of the linear continuous state space model is obtained by using the first-order Taylor expansion of the nonlinear continuous state space equation, and is as follows:

[0035]

[0036] Wherein, is the system Jacobian matrix corresponding to the working point value, is the control Jacobian matrix corresponding to the working point value, is the disturbance input matrix Jacobian matrix corresponding to the working point value, x is the state variable vector, u is the control input vector, and d is the disturbance input vector;

[0037] The calculation formula of the discrete state space prediction model is obtained by using the linear continuous state space model based on the forward Euler method, and is as follows:

[0038]

[0039] Wherein, A d is the discrete-time system matrix, B d is the discrete-time control input matrix, B dd is the discrete-time disturbance input matrix, A c is the continuous-time system Jacobian matrix, I is the unit matrix, the dimension of which is the same as A c , T s is the system sampling time, B cFor continuous-time control input Jacobian matrix, B cd The Jacobian matrix is ​​used as the input for continuous-time perturbation.

[0040] Furthermore, the step of using stator current and rotor current as state variables to establish a nonlinear continuous state-space equation based on the stator-rotor circuit voltage equation, the correlation between flux linkage and current, and the mechanical motion equation includes:

[0041] The calculation formula for establishing the prediction step state based on the extended Kalman filter is as follows:

[0042]

[0043] in, It is the nonlinear state equation of the system. It is in The Jacobian matrix at point P is the error covariance matrix, and Q is the Jacobian matrix at point P. k It is the process noise covariance matrix;

[0044] The calculation formula for establishing the update step state based on the extended Kalman filter is as follows:

[0045]

[0046] in, It is a measurement equation. It is in Jacobian matrix at the location, It measures the noise covariance matrix. It is the Kalman gain;

[0047] Based on the predicted step state and the updated step state, and utilizing the state variables to establish the calculation formula for the nonlinear continuous state-space equation based on the mechanical motion equations, the stator-rotor circuit voltage equations, flux linkage, and current correlation are as follows:

[0048]

[0049] Where x is the state variable vector, u is the control input vector, d is the disturbance input vector, and y is the output variable vector. P is a nonlinear vector function. e This refers to the active power exchanged between the system and the power grid.

[0050] Furthermore, obtaining the optimal control command sequence for future time periods using the smooth, overshoot-free power reference trajectory based on the discrete state-space prediction model includes:

[0051] The formula for calculating the initial optimal control sequence based on the objective function is as follows:

[0052]

[0053] wherein N p is a prediction horizon, N c is a control horizon, is a predicted system output active power at time k+j given the system state at time k, is a power reference value at future time k+j, and Q and R are positive definite weight matrices;

[0054] obtaining a future state trajectory and a future output trajectory according to the discrete state-space prediction model, respectively;

[0055] wherein the future state trajectory and the future output trajectory are used to obtain a future optimal control command sequence according to an initial optimal control sequence, and the calculation formula of the future optimal control command sequence is as follows:

[0056]

[0057] wherein, is a future optimal control command sequence, k is a current time, N c is a control horizon.

[0058] Further, the calculation formula of the positive definite weight matrix Q is as follows:

[0059]

[0060] wherein Q is a positive definite weight matrix, Q0 is a reference weight, and a is an adaptive coefficient.

[0061] Further, the future optimal control command sequence is used to obtain a power converter switching signal based on a flux-oriented vector control strategy, which comprises:

[0062] obtaining a voltage optimal control amount given value according to the future optimal control command sequence;

[0063] using the voltage optimal control amount given value to obtain a composite voltage command based on a flux-oriented vector control strategy;

[0064] using the composite voltage command to obtain a power converter switching signal based on space vector modulation.

[0065] Further, the voltage optimal control amount given value is used to obtain a composite voltage command based on a flux-oriented vector control strategy, which comprises:

[0066] using the voltage optimal control amount given value to obtain a feedforward compensation term according to a motor model based on a flux-oriented vector control strategy, and the calculation formula of the feedforward compensation term is as follows:

[0067]

[0068] wherein, With is a feedforward compensation term, R r is a rotor winding resistance, L r is a rotor winding self-inductance, ω e is a synchronous electrical angular velocity, ω r is a rotor electrical angular velocity, L m is a stator-rotor mutual inductance, i dr , i qr are d-axis and q-axis components of rotor current, i ds , i qs are d-axis and q-axis components of stator current;

[0069] A calculation formula for obtaining a compound voltage command using the feedforward compensation term is as follows:

[0070]

[0071] wherein, is a compound voltage command, and is a feedforward compensation term, is a voltage given value.

[0072] Compared with the closest prior art, the present application has the beneficial effects of:

[0073] Through the multi-step look-ahead optimization capability of the model predictive control algorithm, combined with the smooth reference trajectory generation method of the S-shaped curve, the power and torque sudden changes in the mode switching process are fundamentally eliminated. Compared with the traditional ramp function method, the power fluctuation in the switching process can be greatly reduced, the maximum torque impact is significantly reduced, the stress load of the mechanical transmission system is significantly reduced, and the service life of the equipment is prolonged. The grid compatibility and support capability of the system are enhanced, the power impact on the grid is avoided, the frequency fluctuation and voltage flicker problems caused by mode switching are effectively inhibited. The adaptability and robustness of the system are improved, through the introduction of the state observer of the extended Kalman filter and the online adaptive adjustment mechanism, the unmeasured state and external disturbance of the system can be estimated in real time, and the control parameters can be dynamically adjusted according to the grid working condition. When the system faces complex working conditions such as grid frequency fluctuation, load mutation and parameter change, it can still maintain excellent switching performance, solves the problem of poor adaptability of the fixed parameter controller, effectively solves the key technical problems in the charge-discharge mode switching of the gravity energy storage system, and provides important technical support for the large-scale application of the gravity energy storage technology. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 is a flow chart of a gravity energy storage system charge-discharge switching method based on a double-fed motor and a model prediction provided by the present application;

[0075] Figure 2It is a system control flowchart of a gravity energy storage system charge-discharge switching method based on a doubly-fed motor and model prediction provided by the application

[0076] Figure 3 It is a prediction model derivation flowchart of a gravity energy storage system charge-discharge switching method based on a doubly-fed motor and model prediction provided by the application.

[0077] Figure 4 It is an MPC rolling optimization and state observer coordination principle diagram of a gravity energy storage system charge-discharge switching method based on a doubly-fed motor and model prediction provided by the application. DETAILED DESCRIPTION

[0078] The specific embodiments of the application will be further described in detail below with reference to the accompanying drawings.

[0079] In order to make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme in the embodiments of the application will be described clearly and completely below with reference to the drawings of the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0080] Embodiment 1:

[0081] The application provides a gravity energy storage system charge-discharge switching method based on a doubly-fed motor and model prediction, as shown in Figure 1 , comprising:

[0082] S1, acquiring a smooth and undamped power reference trajectory based on a five-order S-shaped curve algorithm using real-time operation data of a power grid;

[0083] S2, acquiring an optimal control instruction sequence of a future period according to the smooth and undamped power reference trajectory;

[0084] S3, acquiring a power converter switching signal based on a flux linkage oriented vector control strategy using the optimal control instruction sequence of the future period;

[0085] S4, performing gravity energy storage system charge-discharge switching according to the power converter switching signal.

[0086] S1 specifically comprises:

[0087] S1-1, collecting real-time frequency of a power grid and real-time DC voltage of a power grid as real-time operation data of a power grid;

[0088] S1-2, acquiring a current power value and a target power value according to the real-time operation data of the power grid when there is a mode switching of the power grid;

[0089] S1-3. Using the current power value and the target power value, obtain a smooth, overshoot-free power reference trajectory based on a fifth-order S-curve algorithm.

[0090] The formula for calculating S1-3 is as follows:

[0091]

[0092] Among them, P ref(t) To smooth the overshoot-free power reference trajectory, P start P is the initial power value for the power change. end The target power value for power change.

[0093] S2 specifically includes:

[0094] S2-1. Establish a discrete state-space prediction model;

[0095] S2-2. Using the smooth, overshoot-free power reference trajectory, obtain the optimal control command sequence for future time periods based on the discrete state-space prediction model.

[0096] S2-1 specifically includes:

[0097] S2-1-1. Based on the real-time operation data of the power grid, establish the stator-rotor circuit voltage equations using a two-phase synchronous rotating dq coordinate system.

[0098] S2-1-2, The formula for establishing the correlation between magnetic flux and current is as follows:

[0099]

[0100] in, These represent the stator self-inductance, rotor self-inductance, and rotor mutual inductance, respectively, where i is the current. For magnetic linkage;

[0101] S2-1-3. Obtain the corresponding DFIG electromagnetic torque based on the correlation between the magnetic flux and the current;

[0102] S2-1-4. The calculation formula for establishing the mechanical motion equations of the motor electromagnetic process and the gravity energy storage mechanical process using the DFIG electromagnetic torque is as follows:

[0103]

[0104] Where J is the total moment of inertia of the system, B is the coefficient of viscous friction, and T... e For DFIG electromagnetic torque, T L This is the load torque;

[0105] S2-1-5. The calculation formulas for the active power and reactive power of the power grid based on the real-time operation data of the power grid are as follows:

[0106]

[0107] in, Active power Reactive power;

[0108] S2-1-6. Using stator current and rotor current as state variables, establish a nonlinear continuous state space equation based on the stator-rotor circuit voltage equation, the correlation between flux linkage and current, and the mechanical motion equation.

[0109] S2-1-7. Using the aforementioned nonlinear continuous state-space equations, a first-order Taylor expansion yields the following formula for the linear continuous state-space model:

[0110]

[0111] in, The operating point value corresponding to the system Jacobian matrix. To control the operating point value corresponding to the Jacobian matrix, Let x be the operating point value corresponding to the disturbance input matrix Jacobian matrix, u be the control input vector, and d be the disturbance input vector.

[0112] S2-1-8. Using the aforementioned linear continuous state-space model and the forward Euler method, the discrete state-space equations are obtained as the calculation formula for the discrete state-space prediction model, as follows:

[0113]

[0114] Among them, A d For the discrete-time system matrix, B d B is the discrete-time control input matrix. dd Let A be the discrete-time perturbation input matrix. c Let I be the Jacobian matrix of a continuous-time system, and let I be the identity matrix, whose dimensions are the same as A. c Same, T s B is the system sampling time. c For continuous-time control input Jacobian matrix, B cd The Jacobian matrix is ​​used as the input for continuous-time perturbation.

[0115] S2-1-6 specifically includes:

[0116] S2-1-6-1, The calculation formula for establishing the prediction step state based on the extended Kalman filter is as follows:

[0117]

[0118] in, It is the nonlinear state equation of the system. It is in The Jacobian matrix at point P is the error covariance matrix, and Q is the Jacobian matrix at point P. k It is the process noise covariance matrix;

[0119] S2-1-6-2, The calculation formula for establishing the update step state based on the extended Kalman filter is as follows:

[0120]

[0121] in, It is a measurement equation. It is in Jacobian matrix at the location, It measures the noise covariance matrix. It is the Kalman gain;

[0122] S2-1-6-3. Based on the predicted step state and the updated step state, and utilizing the state variables to establish the calculation formula for the nonlinear continuous state-space equation based on the mechanical motion equations, the stator-rotor circuit voltage equations, flux linkage, and current correlation are as follows:

[0123]

[0124] Where x is the state variable vector, u is the control input vector, d is the disturbance input vector, and y is the output variable vector. P is a nonlinear vector function. e This refers to the active power exchanged between the system and the power grid.

[0125] S2-2 specifically includes:

[0126] S2-2-1. The calculation formula for establishing the initial optimal control sequence based on the objective function is as follows:

[0127]

[0128] Where, N p For prediction in the time domain, N c To control the time domain, Predict the system output active power at time k+j for time k. The power reference value is given at time k+j in the future, and both Q and R are positive definite weight matrices.

[0129] S2-2-2. Obtain the future state trajectory and the future output trajectory according to the discrete state space prediction model.

[0130] S2-2-3. The calculation formula for obtaining the optimal control command sequence for the future time period based on the initial optimal control sequence using the future state trajectory and the future output trajectory is as follows:

[0131]

[0132] in, The optimal control command sequence for the future time period, where k is the current time and N is the number of control commands. c To control the time domain.

[0133] The formula for calculating the positive definite weight matrix Q is as follows:

[0134]

[0135] Where Q is the positive definite weight matrix, Q0 is the baseline weight, and α is the adaptive coefficient.

[0136] S3 specifically includes:

[0137] S3-1. Obtain the optimal voltage control quantity setpoint based on the optimal control instruction sequence for the future time period;

[0138] S3-2. Obtain composite voltage commands based on the flux-oriented vector control strategy using the given value of the optimal voltage control quantity.

[0139] S3-3. The power converter switching signal is obtained by using the composite voltage command based on space vector modulation.

[0140] S3-2 specifically includes:

[0141] S3-2-1. Using the voltage optimal control quantity setpoint, the calculation formula for the feedforward compensation term is obtained based on the flux-oriented vector control strategy and the motor model, as follows:

[0142]

[0143] in, and R is the feedforward compensation term. r L is the rotor winding resistance. r For the rotor winding self-inductance, ω e For synchronous electric angular velocity, ω r L is the rotor electric angular velocity. m For the mutual inductance between the stator and rotor, i dr i qr These are the d-axis and q-axis components of the rotor current, i ds i qs These are the d-axis and q-axis components of the stator current, respectively.

[0144] S3-2-2, The calculation formula for obtaining the composite voltage command using the feedforward compensation term is as follows:

[0145]

[0146] in, This is a composite voltage command. and For feedforward compensation term, This is the voltage setpoint.

[0147] In this embodiment, a method for switching charge and discharge modes of a gravity energy storage system based on a doubly-fed induction generator (DFIG) and model prediction is proposed. This method utilizes the look-ahead optimization capability of model predictive control (MPC) and the smooth trajectory planning of a fifth-order S-curve, combined with the excellent controllability of a doubly-fed induction generator (DFIG), to ultimately achieve a smooth and rapid switching between charge and discharge modes of the gravity energy storage system without impact.

[0148] This embodiment constructs a complete gravity energy storage system model in a simulation environment, such as... Figure 2 As shown, the system mainly consists of the following core components:

[0149] Gravity energy storage mechanical device: includes a mass block, hoisting shaft, cable, gearbox, and drive shaft. Its core function is to convert electrical energy into and out of gravitational potential energy, and its load torque... It is a key influencing factor on the dynamic characteristics of the system.

[0150] Doubly fed induction generator (DFIG): As the energy conversion hub of the system, its rotor is coupled to mechanical devices via a drive shaft, while its stator is directly connected to the power grid. The electromagnetic torque of the DFIG... It is a key execution quantity for controlling power flow.

[0151] Power converter bank: Employing a back-to-back converter structure, it includes a rotor-side converter (RSC) and a grid-side converter (GSC). The RSC is responsible for precisely controlling the active power of the DFIG. and reactive power GSC is responsible for maintaining the DC bus voltage. Stability.

[0152] Control and sensing system: includes high-precision sensors (for measuring current, voltage, speed, position, etc.) and a digital processor that executes the core algorithm of this embodiment.

[0153] To achieve a smooth transition, this embodiment employs a method such as Figure 3 The layered intelligent control architecture is shown. This architecture consists of four layers from top to bottom, each with its own specific function yet working closely together:

[0154] Mode Decision and Reference Generation Layer (Upper Layer): This layer acts as the brain of the system, continuously monitoring power grid dispatch commands and system operating status (such as power grid frequency). DC voltage When it determines that a mode switch is needed, it generates a smooth, overshoot-free power reference trajectory based on the current power value and the target power value using a fifth-order S-curve algorithm. This provides an ideal tracking target for the lower-level controller. The trajectory generated by this algorithm ensures the continuity of acceleration, avoiding abrupt changes from the source of the command.

[0155] Model Prediction and Control Layer (Core Layer): This layer is the core embodiment of the intelligence described in this embodiment. It receives the reference trajectory sent by the upper layer. Based on a built-in predictive model that can predict the future dynamic behavior of the system, MPC solves a finite-time optimization problem in each control cycle, calculates the optimal control command sequence for a future period, thereby achieving optimal tracking of the reference trajectory while strictly satisfying various system constraints.

[0156] Motor Control and Execution Layer (Bottom Layer): This layer is responsible for faithfully executing the control commands issued by the MPC layer. It adopts a flux-oriented vector control strategy, converting the high-level control quantities calculated by the MPC into power converter switching signals of RSC and GSC, thereby directly driving the power converter and achieving precise torque and power control of the DFIG.

[0157] State Observation and Adaptation Layer (Support Layer): This layer uses advanced observation algorithms such as the Extended Kalman Filter (EKF) to estimate key states and parameters in the system that are difficult to measure directly in real time. These estimates are then fed back to the MPC layer for real-time correction of its internal prediction model, greatly enhancing the robustness and adaptability of the control system and enabling it to maintain excellent performance in the face of changes in internal parameters and external disturbances.

[0158] This four-layer architecture forms an organic whole, and through feedforward trajectory planning, feedback-based rolling optimization, precise execution control, and adaptive model updates, it jointly ensures the high smoothness and reliability of the switching process.

[0159] The performance of a model predictive controller fundamentally depends on whether its internal predictive model can accurately describe the dynamic characteristics of the system. The discrete state-space predictive model used in the MPC controller of this embodiment will be shown in detail below.

[0160] First, a continuous-time mathematical model of DFIG is established in a two-phase synchronously rotating dq coordinate system. This model integrates the electromagnetic characteristics of the motor, the mechanical motion characteristics of the rotor, and the physical characteristics of the gravitational load.

[0161] The voltage equations for the stator and rotor circuits can be expressed as:

[0162]

[0163] in, For magnetic flux linkage, let's represent voltage, current, magnetic flux linkage, and resistance respectively; subscripts are used. Represents stator and rotor components; subscript Represents the d-axis and q-axis components; For synchronous electric angular velocity, ω is the rotor's electric angular velocity.

[0164] The relationship between magnetic flux linkage and electric current is expressed as:

[0165]

[0166] in, These are stator self-inductance, rotor self-inductance, and mutual inductance, respectively.

[0167] Electromagnetic torque generated by DFIG It can be calculated from magnetic flux and current:

[0168]

[0169] Where p is the number of pole pairs of the motor.

[0170] The mechanical motion equations link the electromagnetic processes of the electric motor with the mechanical processes of gravitational energy storage, and are one of the core components of the entire system model:

[0171]

[0172] Where J is the total moment of inertia of the system (converted to the motor shaft end), including the mass of the motor rotor, transmission mechanism and gravity block; B is the viscous friction coefficient; TL is the load torque, which is proportional to the mass of the gravity block and the lifting height. It is a key variable in the mode switching process, and its sign is opposite when charging (lifting the heavy object) and discharging (lowering the heavy object).

[0173] The active power exchanged between the system and the grid in the output power equation and reactive power To control the output of the target:

[0174]

[0175] To facilitate MPC design, the above equations need to be transformed into state-space form. We choose stator current and rotor current as state variables because they directly reflect the energy state of the system and are easy to measure or observe.

[0176] Define the state variable vector x, the control input vector u, and the disturbance input vector d as follows:

[0177]

[0178] Define the output variable vector y as:

[0179]

[0180] By simultaneously solving and simplifying the voltage equation and flux linkage equation, and substituting them into the mechanical motion equation, the nonlinear continuous state-space equation of the system can be derived:

[0181]

[0182] in, It is a nonlinear vector function that describes the dynamic changes in the system state.

[0183] Since MPC is usually based on linear models for optimization to ensure real-time performance, the aforementioned nonlinear models need to be processed.

[0184] At work Nearby, for nonlinear models Performing a first-order Taylor expansion yields the linearized continuous state-space model:

[0185]

[0186] in These are the values ​​of the Jacobian matrices of the system, control, and disturbance input matrices at the operating point, respectively.

[0187] To implement this in a digital processor, the continuous model needs to be converted into a discrete model. The forward Euler method is used, with a sampling period of... The discrete state-space equations can be obtained as follows:

[0188]

[0189] in

[0190]

[0191] This discrete model forms the basis of the MPC predictor, used to predict the system state at future sampling times.

[0192] Figure 2 This is the derivation process from the physical system to the available predictive model for MPC.

[0193] The power and torque shocks during mode switching largely stem from abrupt changes in commands. While traditional ramp functions smooth the commands to some extent, their discontinuous acceleration still causes considerable shocks. One of the core innovations of this embodiment is that the upper-level decision layer uses a fifth-order S-curve algorithm to generate power reference commands, ensuring high-order smoothness of the commands from the source and laying an optimal foundation for the lower-level tracking control.

[0194] The mode decision layer continuously monitors the internal and external states of the system. While its switching trigger logic is not the core of this embodiment, it is briefly described here for completeness. Common triggering conditions include:

[0195] External dispatch instructions: Receive power setpoints from the grid energy management system (EMS) or power plant controller. .when The switching process is triggered when the sign (representing the charging / discharging direction) or magnitude undergoes a significant change across zero.

[0196] System protection signal: If a serious over-limit of the power grid frequency or abnormal equipment operation is detected, it is necessary to switch to a specific mode such as emergency support or shutdown.

[0197] Local operation strategy: Based on preset strategies, such as the "peak shaving and valley filling" strategy, the system automatically decides whether to start or stop the mode when the electricity price signal changes.

[0198] Once the switching conditions are met, the mode decision layer immediately starts, and its core task is to generate a path from the current actual power.

[0199] To target power Ideal transition trajectory .

[0200] This embodiment abandons the simple linear ramp function and uses a high-order S-curve as the reference trajectory generator. A fifth-order S-curve can ensure that the rate of change of acceleration is continuous, which makes the generated reference command extremely smooth and can fundamentally eliminate the impact excitation on the mechanical system.

[0201] The general mathematical expression for this algorithm is as follows:

[0202]

[0203] in, It is a fifth-order polynomial time scaling function defined on the interval [0, 1], and its expression is:

[0204]

[0205] here, It is a normalized time variable. This is the preset duration for the entire mode switching process. This function guarantees that at the start point t0 and the end point t1, the function... The zero-order (position), first-order (rate of change of velocity / power), and second-order (acceleration) derivatives are all zero:

[0206]

[0207] Therefore, the generated power reference command Not only are the values ​​equal at the start and end points, but their first and second derivatives are also zero, achieving true "soft start" and "soft stop," resulting in an incredibly smooth trajectory.

[0208] High-quality reference trajectory generated The data is sent to the MPC control layer in real time. The task of the MPC layer is to design the control law to drive the actual output power of the system. The MPC tracks this ideal trajectory as accurately as possible. Since the reference trajectory itself is smooth and overshoot-free, the MPC does not need to deal with the control problems caused by sudden changes in instructions. It can focus its optimization capabilities on overcoming system model mismatch and external disturbances, thereby achieving globally optimal control performance.

[0209] The following is the core execution phase of the method in this embodiment. The Model Predictive Controller (MPC) receives the smoothed reference trajectory generated by the upper layer and calculates the optimal control command through rolling optimization. At the same time, it works closely with the state observer to form an advanced control system with look-ahead and adaptive capabilities.

[0210] At each sampling time k, the core task of MPC is to solve a finite-time open-loop optimization problem. The mathematical formulation of this problem is as follows:

[0211] MPC computes the optimal control sequence by minimizing the following objective function J:

[0212]

[0213] in, To predict the time domain, To control the time domain (usually) ).

[0214] At any moment Predicted future +j The system output active power at any given time.

[0215] It is the future issued by the upper levels. +j The power reference value at any given time.

[0216] It is the predicted value for controlling the increment.

[0217] Q and R are positive definite weight matrices, which handle the tracking error and the magnitude of the control increment, respectively. The larger the Q value, the higher the tracking accuracy; the larger the R value, the smoother the control action.

[0218] The advantage of MPC lies in its ability to explicitly handle various physical constraints and operational limits of the system. These constraints are transformed into linear inequalities and incorporated into the optimization problem: Control program constraints (converter output voltage capability):

[0219] Control increment constraints (limiting the rate of change of control actions):

[0220] Status output constraints (such as self-flow protection limits):

[0221] MPC employs a "rolling time domain" strategy, and its online optimization process is as follows:

[0222] State acquisition: At time k, the current state of the system is acquired through sensors. .

[0223] Predicting the Future: Predicting the future of a system based on a discrete state-space model. Step state trajectory and output trajectory .

[0224] Solving the optimization problem: Solve the constrained quadratic programming (QP) problem above to obtain the optimal control sequence. .

[0225] Application control: Only the first control variable in the sequence. Apply to the controlled object.

[0226] Timing update: At the next sampling time k+1, repeat steps 1-4 and re-optimize based on the new measurements.

[0227] As shown in the schematic diagram, high-performance MPC heavily relies on accurate system state information. However, key states / parameters such as load torque TL cannot be directly measured. This embodiment uses an extended Kalman filter (EKF) as the state observer, and its core recursive algorithm is as follows:

[0228] Prediction step:

[0229]

[0230] in, It is the nonlinear state equation of the system. It is in The Jacobian matrix at point P is the error covariance matrix, and Qk is the process noise covariance matrix.

[0231] Update steps:

[0232]

[0233] in, It is a measurement equation. It is in Jacobian matrix at the location, It measures the noise covariance matrix. It is the Kalman gain.

[0234] To further improve system performance under complex power grid conditions, this embodiment introduces an adaptive mechanism. This mechanism dynamically adjusts the weight matrix Q in the MPC objective function. For example, when the power grid frequency deviation Δf is large, the system requires a faster and more accurate power response to support the power grid. In this case, the adaptive algorithm can increase the weight Q:

[0235]

[0236] Here, Q0 is the baseline weight, and α is the adaptive coefficient. Conversely, when the system is running smoothly, R can be appropriately increased to make the control action smoother. This dynamic adjustment strategy enables the controller to maintain optimal performance in different scenarios.

[0237] The optimal control quantity calculated by MPC This is the reference command for the rotor voltage. This command is sent to the rotor-side converter (RSC) at the motor control layer. The RSC employs a flux-oriented vector control strategy, and its inner current loop controller receives... The voltage is used as the setpoint. To improve dynamic response and disturbance rejection, a feedforward compensation term based on the motor model is introduced into the control:

[0238]

[0239] Ultimately, the voltage command of RSC is the sum of the MPC output and the feedforward compensation:

[0240]

[0241] This composite voltage command generates power converter switching signals through space vector modulation, drives the converter, and ultimately achieves precise and rapid control of the DFIG electromagnetic torque and output power.

[0242] like Figure 4 As shown, this is the closed-loop data flow between MPC, state observer, and controlled object.

[0243] Example 2

[0244] This embodiment provides a smooth charging and discharging switching method for a gravity energy storage system based on a doubly-fed induction generator (DFIG) and model predictive control, applicable to general situations, and can be implemented in the MATLAB / Simulink simulation platform. The core of this example lies in establishing a hierarchical intelligent control architecture, achieving a smooth transition between charging and discharging modes through multi-variable collaborative optimization. A specific implementation example is as follows:

[0245] First, establishing an accurate system prediction model is the foundation of MPC control. In this embodiment, a unified state-space model is established, which includes the electromagnetic characteristics of the doubly-fed motor, the mechanical transmission system, and the gravity load.

[0246] The voltage equation of the doubly fed motor in the synchronous rotating dq coordinate system is shown in formula (4), and the flux linkage equation is shown in formula (5).

[0247] The equation of motion for the machine is given in formula (7), where For electromagnetic torque, This is the torque due to gravity load.

[0248] Power equation:

[0249]

[0250] Discretizing the above equations yields the state-space model of the system:

[0251]

[0252] Among them, state variables Control input Disturbance input Output .

[0253] The generation of the reference trajectory uses a 5th-order S-curve to generate a smooth power reference command, ensuring continuous acceleration and jerk.

[0254]

[0255] in, and These represent the start and end power values ​​for switching, This is the switching time constant.

[0256] The design of the model predictive controller mainly involves solving the following optimization problem for the MPC in each sampling period:

[0257]

[0258] in, To predict the time domain, To control the time domain, and This is the weight matrix. To optimize variables.

[0259] As for the coordinated control of the doubly-fed motor, it is achieved through the control commands output by the MPC via the rotor-side converter:

[0260]

[0261] in, and For feedforward compensation term, and This is the optimized control increment for MPC output.

[0262] State observation and adaptive adjustment employ an extended Kalman filter (EKF) for state estimation:

[0263]

[0264] The weight matrix of MPC is dynamically adjusted based on the estimated state and grid frequency deviation:

[0265]

[0266] in For power grid frequency deviation, These are adaptive coefficients.

[0267] The system includes: a gravity energy storage mechanical device, a doubly fed induction generator, a back-to-back converter, a DC link capacitor, a voltage and current sensor, a speed and position detection device, and a digital signal processor that executes the above algorithm.

[0268] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0269] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0270] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0271] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0272] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for charge and discharge switching of a gravity energy storage system based on a doubly-fed electric machine and model prediction, characterized in that, The method comprises the following steps: obtaining a smooth and non-overshoot power reference trajectory based on a five-order S-shaped curve algorithm using real-time operation data of a power grid; obtaining an optimal control instruction sequence of a future period according to the smooth and non-overshoot power reference trajectory; obtaining a power converter switching signal based on a flux linkage oriented vector control strategy using the optimal control instruction sequence of the future period; performing charge-discharge switching of a gravity energy storage system according to the power converter switching signal.

2. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 1, wherein, The method for obtaining the smooth and non-overshoot power reference trajectory based on the five-order S-shaped curve algorithm using the real-time operation data of the power grid comprises the following steps: collecting real-time frequency of the power grid and real-time direct-current voltage of the power grid as real-time operation data of the power grid; when there is a mode switching of the power grid, obtaining a current power value and a target power value respectively according to the real-time operation data of the power grid; obtaining a smooth and non-overshoot power reference trajectory based on a five-order S-shaped curve algorithm using the current power value and the target power value.

3. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 2, wherein, The calculation formula for obtaining the smooth and non-overshoot power reference trajectory based on the five-order S-shaped curve algorithm using the current power value and the target power value is as follows: ； wherein P ref(t) is the smooth non-overshoot power reference trajectory, P start is the initial power value of the power change, P end is the target power value of the power change.

4. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 2, wherein, The method for obtaining the optimal control instruction sequence of the future period according to the smooth and non-overshoot power reference trajectory comprises the following steps: establishing a discrete state space prediction model; obtaining the optimal control instruction sequence of the future period according to the discrete state space prediction model using the smooth and non-overshoot power reference trajectory.

5. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 4, wherein, The method for establishing the discrete state space prediction model comprises the following steps: establishing a stator-rotor loop voltage equation based on a two-phase synchronous rotating d-q coordinate system according to the real-time operation data of the power grid; the calculation formula for establishing the correlation between the flux linkage and the current is as follows: ； wherein, Ls, Lr, and M are the stator self-inductance, the rotor self-inductance, and the mutual inductance, respectively, i is the current, is the flux linkage; obtaining a corresponding DFIG electromagnetic torque according to the correlation between the flux linkage and the current; the calculation formula for establishing a mechanical motion equation of a motor electromagnetic process and a gravity energy storage mechanical process using the DFIG electromagnetic torque is as follows: ； where J is the total moment of inertia of the system, B is the viscous friction coefficient, T e is the electromagnetic torque of the DFIG, T L is the load torque; the calculation formula for calculating the active power and the reactive power of the power grid respectively according to the real-time operation data of the power grid is as follows: ； wherein, P is the active power, Q is the reactive power; establishing a nonlinear continuous state space equation using stator current and rotor current as state variables according to the stator-rotor loop voltage equation, the correlation between the flux linkage and the current, and based on the mechanical motion equation; the calculation formula for obtaining a linear continuous state space model by first-order Taylor expansion using the nonlinear continuous state space equation is as follows: ； wherein, is the system Jacobian matrix corresponding to the working point value, is the control Jacobian matrix corresponding to the working point value, is the disturbance input matrix Jacobian matrix corresponding to the working point value, x is the state variable vector, u is the control input vector, and d is the disturbance input vector. the calculation formula for obtaining a discrete state space equation as the discrete state space prediction model based on the forward Euler method using the linear continuous state space model is as follows: ； where A d is the discrete-time system matrix, B d is the discrete-time control input matrix, B dd is the discrete-time disturbance input matrix, A c is the continuous-time system Jacobian matrix, I is the identity matrix with the same dimension as A c , T s is the system sampling time, B c is the continuous-time control input Jacobian matrix, B cd is the continuous-time disturbance input Jacobian matrix.

6. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 5, wherein, The method for establishing the nonlinear continuous state space equation using stator current and rotor current as state variables according to the stator-rotor loop voltage equation, the correlation between the flux linkage and the current, and based on the mechanical motion equation comprises the following steps: the calculation formula for establishing a prediction step state based on an extended Kalman filter is as follows: ； wherein is a nonlinear state equation of the system, is its Jacobian matrix at P is the error covariance matrix, Q k is a process noise covariance matrix; the calculation formula for establishing an update step state based on the extended Kalman filter is as follows: ； wherein is a measurement equation, is a Jacobian matrix thereof at is a measurement noise covariance matrix, is a Kalman gain;​ the calculation formula for establishing the nonlinear continuous state space equation according to the prediction step state and the update step state using the state variables for the stator-rotor loop voltage equation, the correlation between the flux linkage and the current, and based on the mechanical motion equation is as follows: ； where x is the state variable vector, u is the control input vector, d is the disturbance input vector, y is the output variable vector, is a nonlinear vector function, P e is the active power exchanged by the system with the grid.

7. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 6, wherein, The method for obtaining the optimal control instruction sequence of the future period according to the discrete state space prediction model using the smooth and non-overshoot power reference trajectory comprises the following steps: the calculation formula for establishing an initial optimal control sequence based on an objective function is as follows: ； wherein N p is the prediction horizon, N c is the control horizon, is the predicted system output active power at time k+j given the system state at time k, Q and R are positive definite weight matrices. is the power reference value at future time k+j. The future state trajectory and the future output trajectory are respectively obtained according to the discrete state space prediction model; A calculation formula of the future period optimal control instruction sequence is obtained according to the initial optimal control sequence and the future state trajectory and the future output trajectory, and the calculation formula is as follows: ； wherein, is the optimal control instruction sequence for the future period, k is the current time, N c is the control time domain.

8. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 7, wherein, A calculation formula of the positive definite weight matrix Q is as follows: ； Wherein, Q is a positive definite weight matrix, Q0 is a reference weight, and a is an adaptive coefficient.

9. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 7, wherein, The power converter switch signal is obtained based on the flux-oriented vector control strategy and the future period optimal control instruction sequence, and the obtaining includes: A voltage optimal control amount given value is obtained according to the future period optimal control instruction sequence; A composite voltage instruction is obtained based on the flux-oriented vector control strategy and the voltage optimal control amount given value, and the obtaining includes: The power converter switch signal is obtained based on space vector modulation and the composite voltage instruction.

10. The dual-fed motor and model predictive based gravitational energy storage system charge-discharge switching method of claim 9, wherein, The composite voltage instruction is obtained based on the flux-oriented vector control strategy and the voltage optimal control amount given value, and the obtaining includes: A calculation formula of a feedforward compensation item is obtained according to a motor model and the voltage optimal control amount given value based on the flux-oriented vector control strategy, and the calculation formula is as follows: ； wherein, with is a feed forward compensation term, R r is the rotor winding resistance, L r is the rotor winding self-inductance, ω e is the synchronous electrical angular velocity, ω r is the rotor electrical angular velocity, L m is the inter-winding mutual inductance between stator and rotor, i dr , i qr are the d-axis and q-axis components of the rotor current, i ds , i qs are the d-axis and q-axis components of the stator current; A calculation formula of the composite voltage instruction is obtained based on the feedforward compensation item, and the calculation formula is as follows: ； wherein, is a composite voltage command, and is a feedforward compensation term, is a voltage given value.

Images