A multi-type power supply collaborative frequency modulation method based on model predictive control

Abstract

The application discloses a kind of based on model predictive control's multi-type power supply collaborative frequency modulation method, comprising the following steps: constructing frequency dynamic model, and conversion is the state space model of discrete system-On-line acquisition system frequency and system real-time topology, power supply parameter and load parameter-Construction prediction model-Optimization model is established-Utilize mathematical programming solver to solve quadratic programming problem, obtain a group of optimal control sequence, the first element is the active power variation reference value of each power supply at current time-Select the first element in sequence to act on system, update grid state, at the next sampling time, with the latest measurement value as initial condition, re-predict system output and optimization solution.The application uses the above based on model predictive control's multi-type power supply collaborative frequency modulation method, so as to realize the optimal allocation of frequency modulation power between multi-type power supply, guarantee the frequency stability of power system.

Description

Technical Field

[0001] This invention relates to a power system frequency stability control technology, and more particularly to a multi-type power source coordinated frequency regulation method based on model predictive control. Background Technology

[0002] Frequency stability is crucial to vital aspects of a power system, including generation, transmission, transformation, distribution, and consumption. Deterioration in frequency characteristics can even lead to system collapse and widespread blackouts. With the ongoing energy revolution in my country, new energy sources such as wind and solar power are connected to the grid via power electronic devices. These new energy sources exhibit weak or no inertia in response to grid frequency changes, resulting in a decrease in the overall system inertia level and seriously threatening frequency security. Furthermore, my country is building a large-scale, long-distance interconnected system primarily based on ultra-high-voltage (UHV) transmission lines. When large-capacity UHV transmission lines are disconnected due to accidents, insufficient active power at the receiving end can cause a significant frequency drop, potentially leading to frequency collapse after disturbance. Therefore, research on frequency stability control in high-proportion new energy power systems is of great significance.

[0003] To address this, the government has issued relevant standards requiring new energy power plants to provide a certain frequency regulation capability. However, different frequency regulation resources have varying control performance and response speeds. Wind and solar power are characterized by volatility and intermittency, making control and regulation more difficult. Energy storage, on the other hand, possesses precise control and rapid response characteristics, meeting frequency regulation needs in various scenarios. To further improve the frequency regulation performance of the power grid, wind, solar, and energy storage can be configured on the power source side to participate in primary frequency regulation.

[0004] Current research on the joint participation of multiple power sources in frequency regulation mostly adopts the traditional PI control method, calculating the active power reference values ​​for virtual inertial control and droop control based on the frequency change rate and frequency deviation, respectively. This method is simple in principle and easy to implement. However, actual operating wind turbines, photovoltaic systems, and energy storage systems have constraints such as power reserve limits and energy storage state of charge (SoC). Traditional methods are difficult to effectively handle multiple constraints simultaneously, and the results are not optimal. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-type power source coordinated frequency regulation method based on model predictive control, which can achieve optimal allocation of frequency regulation power among multiple types of power sources and ensure the frequency stability of the power system.

[0006] To achieve the above objectives, this invention provides a multi-type power supply coordinated frequency modulation method based on model predictive control, comprising the following steps:

[0007] S1: For heterogeneous power sources participating in primary frequency regulation in the system, a frequency dynamic model is constructed based on the change of its active power. The frequency domain model is used to characterize the dynamic response characteristics of wind, solar and energy storage. Then, the forward Euler method is applied to transform the continuous-time model into the state-space model of the discrete system.

[0008] S2: Online acquisition of the system frequency measured in real time by the measurement device, as well as the system topology, power parameters and load parameters collected by the SCADA system;

[0009] S3: Based on the real-time topology, power parameters, and load parameters of the system, a prediction model is built with the measured values ​​as the initial conditions to predict the system frequency deviation and the changes in active power of wind turbines, photovoltaic systems, and energy storage systems within a finite time domain in the future.

[0010] S4: Establish an optimization model with the objective functions of minimizing frequency deviation and minimizing economic cost, taking into account the constraints that different types of frequency regulation power sources need to meet during actual operation of the power system.

[0011] S5: Use a mathematical programming solver to solve a quadratic programming problem and obtain a set of optimal control sequences. The first element is the reference value of the change in active power of each power source at the current moment.

[0012] S6: Select the first element in the sequence and apply it to the system to update the power grid state. At the next sampling time, use the latest measurement value as the initial condition to re-predict the system output and solve it.

[0013] Preferably, the system in step S1 includes an AC power grid, wind turbine generators, photovoltaic systems, energy storage systems, converters, transformers, and loads.

[0014] Preferably, step S1 specifically includes the following steps:

[0015] S11, The expression for the dynamic equation of the power grid frequency is:

[0016]

[0017] In the formula, f is the system frequency; H is the power grid inertia time constant; f n It is the rated frequency; P G It is the total active power output of all conventional generating units; P L It is the load power; P W It is the active power output of the wind turbine; P p It is the active power output of the photovoltaic system; P B It is the active power output of the energy storage system; D f It is the load adjustment coefficient; P LN It is the rated load power;

[0018] S12. From the perspective of wind turbines, photovoltaic systems, and energy storage systems as a whole, the total active power imbalance of the power system is as follows:

[0019] P unb =P G -P L (2)

[0020] S13. Since different types of frequency-modulated power supplies have different active power response processes after receiving frequency modulation commands, it is necessary to consider the dynamic response characteristics of wind turbines, photovoltaic systems, and energy storage systems to fully utilize their control effects. Considering the response characteristics of the control and communication systems, an inertial element is used for description. Based on the inverse Laplace transform principle, the actual active power output in the time domain can be calculated from the input frequency modulation command.

[0021]

[0022]

[0023]

[0024]

[0025] In the formula, T W1 and T W2 T p T B These are the response time constants of wind turbines, photovoltaic systems, and energy storage systems, respectively; P W ref P p ref P B ref These are the active power reference values ​​for wind, solar, and energy storage power sources, i.e., the size of the commands they need to track;

[0026] S14. Control its state of charge:

[0027]

[0028] In the formula, k+1 is the next sampling time; k is the current sampling time; T s It is the sampling period; E B It is the rated capacity of the energy storage system; the operator Δ is the difference between the current sampling time k and the previous sampling time k-1;

[0029] S15. To obtain the dynamic model of the discrete-time system with frequency, the forward Euler method is used to discretize formulas (1)-(7). At the same time, considering the introduction of integral energy to reduce or eliminate the static error of the system, the incremental form of the state-space model is adopted:

[0030]

[0031] In the formula, Δx(k) is the state variable matrix; Δu(k) is the control variable matrix; Δr(k) is the disturbance variable matrix; y(k) is the output variable matrix; A, B, D, and C are the corresponding matrix coefficients;

[0032] The state variable matrix is ​​represented as follows:

[0033] Δx(k)=[Δf(k) ΔP W mid (k) ΔP W (k) ΔP p (k) ΔP B (k) ΔSoC(k)] T (9)

[0034] The control variable matrix is ​​represented as follows:

[0035] Δu(k)=[ΔP W ref ΔP p ref ΔP B ref ] T (10)

[0036] The interference variable matrix is ​​represented as follows:

[0037] Δr(k)=ΔP unb (k) (11)

[0038] The output variable matrix is ​​represented as follows:

[0039] y(k)=[f(k) P W mid (k) P W (k) P p (k) P B (k) SoC(k)] T (12)

[0040] The state matrix coefficients A, control matrix coefficients B, disturbance matrix coefficients D, and output matrix coefficients C are respectively:

[0041]

[0042]

[0043]

[0044]

[0045] Preferably, step S3 specifically includes the following steps:

[0046] S31. Assumption: Control time domain T c No more than the prediction time domain T p Outside the control time domain, the control variables remain unchanged, i.e., Δu(k+i)=0, i=T c ,T c +1,...,T p -1; and the disturbance received by the system remains unchanged after time k, i.e., Δr(k+i)=0, i=1,2,...,T p -1;

[0047] S32. At the current sampling time k, the measured value is x(k). Calculate Δx(k) = x(k) - x(k-1) and use it as the starting point. Predict the future dynamics of the system according to formula (8). Define T. p Step prediction output vector and T c The control variables for each step are as follows:

[0048] Y(k)=[y(k+1|k) y(k+2|k) ... y(k+T p |k)] T (17)

[0049] ΔU(k)=[Δu(k) Δu(k+1) ... Δu(k+T c -1)] T (18)

[0050] Therefore, the output of the system in the time domain is calculated using the following formula:

[0051] Y(k)=S x Δx(k)+S u ΔU(k)+S d Δr(k)+τy(k) (19)

[0052] In the formula:

[0053]

[0054]

[0055]

[0056] τ=[II ... I] T (twenty three).

[0057] Preferably, step S4 specifically includes the following steps:

[0058] S41. Using a quadratic function of power offset and SoC offset to describe frequency modulation cost, the objective function for optimal overall benefit is defined as:

[0059]

[0060] In the formula, f ref This is the system frequency reference value; SoC ref α is the reference value for the state of charge; α is the frequency regulation effect weighting coefficient; β1, β2 and β3 are the cost coefficients describing the power offset of wind turbines, photovoltaic systems and energy storage systems, respectively; β4 is the cost coefficient describing the SoC offset of energy storage systems.

[0061] S42. Wind turbines and photovoltaic systems participate in frequency regulation through reserved reserve capacity. The limited reserve capacity determines the range of their output power increment, resulting in:

[0062] 0≤ΔP W (k+i|k)≤ΔP W max (25)

[0063] 0≤ΔP p (k+i|k)≤ΔP p max (26)

[0064] In the formula, ΔP(k+i|k) represents the predicted value of the change in active power from k+i at sampling time k; ΔP W max and ΔP p max These are the maximum active power reserves for wind turbines and photovoltaic systems, respectively.

[0065] S43. Assuming the energy storage system does not generate power at the initial sampling time, its active power variation is limited by the rated power. At the same time, the SoC should be kept within a certain range to reduce the adverse impact on the lifespan of the energy storage system.

[0066] -P BN ≤ΔP B (k+i|k)≤P BN (27)

[0067] SoC min ≤SoC(k+i|k)≤SoC max (28)

[0068] In the formula, P BN The rated power of the energy storage system is given by SoC(k+i|k); SoC(k+i|k) represents the predicted SoC of the energy storage system at sampling time k with respect to time k+i; SoC min and SoC max These are the set lower and upper limits for the SoC, respectively.

[0069] Preferably, step S5 specifically includes the following steps:

[0070] S51. Transform formulas (24)-(28) into a standard quadratic programming problem with constraints:

[0071] minJ=||Y(k)-R(k)|| Q 2 +||ΔU(k)|| R 2

[0072] =ΔU(k) T HΔU(k)-G(k) T ΔU(k) ​​(29)

[0073] In the formula, Q and R are weight coefficient matrices composed of α, β1, β2, β3 and β4; R(k) is the defined T p Step reference vector, by f ref and SoC ref Composition; the meanings of the remaining matrices and vectors are as follows:

[0074] E(k)=R(k)-S x Δx(k)-S d Δr(k)-τy(k) (30)

[0075] H = S u T QS u +R (31)

[0076] G(k)=2S u T QE(k) (32)

[0077] Constraints (25)-(27) are transformed into:

[0078]

[0079] In the formula:

[0080] Δu min =[0 0 -P B ] T (34)

[0081] Δu max =[ΔP W max ΔP p max P B ] T (35)

[0082]

[0083] S52. Similarly, the constraints of the energy storage system SoC can be simplified to:

[0084]

[0085] In the formula:

[0086] Y max (k)=[y max (k+1) ... y max (k+T p )] T (38)

[0087] Y min (k)=[y min (k+1) ... y min (k+T p )] T (39)

[0088] Thus, by using YALMILP and GROBI to solve the optimization problem, the control sequence [Δu(k) ... Δu(k+T)] is obtained. c The first element in the result, Δu(k), is the reference value ΔP for the change in active power of the wind-solar-storage power source at the current moment. W ref (k), ΔP p ref (k) and ΔP B ref (k).

[0089] Preferably, in step S6, the wind turbine, photovoltaic system and energy storage system track the reference value calculated in S5 to make up for the power shortage in the power grid and support the stability of the system frequency.

[0090] Therefore, the above-mentioned multi-type power supply coordinated frequency modulation method based on model predictive control has the following beneficial effects:

[0091] Based on the frequency dynamic equation of a power system with heterogeneous power sources participating in frequency regulation, the system can predict future system frequency deviation and changes in active power based on current measurements. Under constraints, it solves the objective function that balances optimal frequency regulation effect and minimum frequency regulation cost, obtains the optimal reference value of active power changes of each type of power source, and applies it to the system. As the sampling time progresses, rolling optimization is performed to achieve the optimal allocation of frequency regulation power among multiple types of power sources, thus ensuring the stability of the power system frequency.

[0092] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0093] Figure 1This is a flowchart of the present invention;

[0094] Figure 2 This is a structural block diagram of the present invention;

[0095] Figure 3 The diagram shows the improved IEEE 3-machine 9-node system topology in two embodiments of the present invention.

[0096] Figure 4 The frequency response curves of the three schemes described in Example 1 after the system is subjected to a disturbance are shown.

[0097] Figure 5 This is a frequency response curve diagram of Embodiment 2 of the present invention under different control objectives;

[0098] Figure 6 This is a graph showing the active power curves of a wind turbine under different control objectives according to Embodiment 2 of the present invention.

[0099] Figure 7 This is a graph showing the active power curves of a photovoltaic system under different control objectives according to Embodiment 2 of the present invention.

[0100] Figure 8 This is a graph showing the active power curves of the energy storage system under different control objectives according to Embodiment 2 of the present invention. Detailed Implementation

[0101] The present invention will be further described below with reference to the accompanying drawings. It should be noted that this embodiment is based on the present technical solution and provides detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to this embodiment.

[0102] Figure 1 This is a flowchart of the present invention; Figure 2 This is a structural block diagram of the present invention, as shown below. Figure 1 and Figure 2 As shown, the present invention includes the following steps:

[0103] S1: For heterogeneous power sources participating in primary frequency regulation in the system, a frequency dynamic model is constructed based on the change of its active power. The frequency domain model is used to characterize the dynamic response characteristics of wind, solar and energy storage. Then, the forward Euler method is applied to transform the continuous-time model into the state-space model of the discrete system.

[0104] Preferably, the system in step S1 includes an AC power grid, wind turbine generators, photovoltaic systems, energy storage systems, converters, transformers, and loads.

[0105] Preferably, step S1 specifically includes the following steps:

[0106] S11, The expression for the dynamic equation of the power grid frequency is:

[0107]

[0108] In the formula, f is the system frequency; H is the power grid inertia time constant; f n It is the rated frequency; P G It is the total active power output of all conventional generating units; P L It is the load power; P W It is the active power output of the wind turbine; P p It is the active power output of the photovoltaic system; P B It is the active power output of the energy storage system; D f It is the load adjustment coefficient; P LN It is the rated load power;

[0109] S12, due to P G and P L Neither of them is convenient to measure directly, so they are combined into one term, denoted as P. unb From the perspective of wind turbines, photovoltaic systems, and energy storage systems as a whole, the total active power imbalance of the power system is as follows:

[0110] P unb =P G -P L (2)

[0111] S13. Since different types of frequency-modulated power supplies have different active power response processes after receiving frequency modulation commands, it is necessary to consider the dynamic response characteristics of wind turbines, photovoltaic systems, and energy storage systems to fully utilize their control effects. Considering the response characteristics of the control and communication systems, an inertial element is used for description. Based on the inverse Laplace transform principle, the actual active power output in the time domain can be calculated from the input frequency modulation command.

[0112]

[0113]

[0114]

[0115]

[0116] In the formula, T W1 and T W2 T p T B These are the response time constants of wind turbines, photovoltaic systems, and energy storage systems, respectively; P W ref P p ref P B refThese are the active power reference values ​​for wind, solar, and energy storage power sources, i.e., the size of the commands they need to track;

[0117] S14. Overcharging / discharging of energy storage systems can easily lead to problems such as accelerated aging and lifespan reduction; therefore, it is necessary to control their state of charge.

[0118]

[0119] In the formula, k+1 is the next sampling time; k is the current sampling time; T s It is the sampling period; E B It is the rated capacity of the energy storage system; the operator Δ is the difference between the current sampling time k and the previous sampling time k-1;

[0120] S15. To obtain the dynamic model of the discrete-time system with frequency, the forward Euler method is used to discretize formulas (1)-(7). At the same time, considering the introduction of integral energy to reduce or eliminate the static error of the system, the incremental form of the state-space model is adopted:

[0121]

[0122] In the formula, Δx(k) is the state variable matrix; Δu(k) is the control variable matrix; Δr(k) is the disturbance variable matrix; y(k) is the output variable matrix; A, B, D, and C are the corresponding matrix coefficients;

[0123] The state variable matrix is ​​represented as follows:

[0124] Δx(k)=[Δf(k) ΔP W mid (k) ΔP W (k) ΔP p (k) ΔP B (k) ΔSoC(k)] T (9)

[0125] The control variable matrix is ​​represented as follows:

[0126] Δu(k)=[ΔP W ref ΔP p ref ΔP B ref ] T (10)

[0127] The interference variable matrix is ​​represented as follows:

[0128] Δr(k)=ΔP unb (k) (11)

[0129] The output variable matrix is ​​represented as follows:

[0130] y(k)=[f(k) P W mid (k) P W (k) P p (k) P B (k) SoC(k)] T (12)

[0131] The state matrix coefficients A, control matrix coefficients B, disturbance matrix coefficients D, and output matrix coefficients C are respectively:

[0132]

[0133]

[0134]

[0135]

[0136] S2: Online acquisition of the system frequency measured in real time by the measuring device and the system topology, power parameters and load parameters collected by the SCADA system; in step S2, the frequency f of the AC bus is measured, and the active power of the wind-solar-storage power source, the system unbalanced active power and the state of charge of the energy storage system are detected according to the system topology.

[0137] S3: Based on the real-time topology, power parameters, and load parameters of the system, a prediction model is built with the measured values ​​as the initial conditions to predict the system frequency deviation and the changes in active power of wind turbines, photovoltaic systems, and energy storage systems within a finite time domain in the future.

[0138] Preferably, step S3 specifically includes the following steps:

[0139] S31. Assumption: Control time domain T c No more than the prediction time domain T p Outside the control time domain, the control variables remain unchanged, i.e., Δu(k+i)=0, i=T c ,T c +1,...,T p -1; and the disturbance received by the system remains unchanged after time k, i.e., Δr(k+i)=0, i=1,2,...,T p -1;

[0140] S32. At the current sampling time k, the measured value is x(k). Calculate Δx(k) = x(k) - x(k-1) and use it as the starting point. Predict the future dynamics of the system according to formula (8). Define T. p Step prediction output vector and T c The control variables for each step are as follows:

[0141] Y(k)=[y(k+1|k) y(k+2|k) ... y(k+T p |k)] T (17)

[0142] ΔU(k)=[Δu(k) Δu(k+1) ... Δu(k+T c -1)] T (18)

[0143] Therefore, the output of the system in the time domain is calculated using the following formula:

[0144] Y(k)=S x Δx(k)+S u ΔU(k)+S d Δr(k)+τy(k) (19)

[0145] In the formula:

[0146]

[0147]

[0148]

[0149] τ=[II ... I] T (twenty three).

[0150] S4: Establish an optimization model with the objective functions of minimizing frequency deviation and minimizing economic cost, taking into account the constraints that different types of frequency regulation power sources need to meet during actual operation of the power system.

[0151] Step S4 clarifies the mathematical description of the optimization problem. The selection of the objective function reflects the performance requirements of the system. If the goal is to suppress frequency fluctuations while minimizing economic costs, a trade-off must be made between control performance and control costs. The economic losses of wind turbines, photovoltaic systems, and energy storage systems participating in frequency regulation are mainly equipment losses caused by power increases or decreases, and the frequency regulation cost is typically described using a quadratic function of power offset. Furthermore, a large SoC offset during the operation of energy storage systems leads to accelerated aging and lifespan degradation; therefore, the frequency regulation cost must be increased by adding SoC offset.

[0152] Therefore, step S4 specifically includes the following steps:

[0153] S41. Using a quadratic function of power offset and SoC offset to describe frequency modulation cost, the objective function for optimal overall benefit is defined as:

[0154]

[0155] In the formula, f refThis is the system frequency reference value; SoC ref α is the reference value for the state of charge; α is the frequency regulation effect weighting coefficient; β1, β2 and β3 are the cost coefficients describing the power offset of wind turbines, photovoltaic systems and energy storage systems, respectively; β4 is the cost coefficient describing the SoC offset of energy storage systems.

[0156] S42. To ensure the stable and reasonable operation of the power system during primary frequency regulation, the objective function must satisfy a series of constraints (the upper and lower power limits and SoC operating range constraints for wind, solar, and energy storage systems). Wind turbines and photovoltaic systems participate in frequency regulation through reserved reserve capacity. The limited reserve capacity determines the range of their output power increments, resulting in:

[0157] 0≤ΔP W (k+i|k)≤ΔP W max (25)

[0158] 0≤ΔP p (k+i|k)≤ΔP p max (26)

[0159] In the formula, ΔP(k+i|k) represents the predicted value of the change in active power from k+i at sampling time k; ΔP W max and ΔP p max These are the maximum active power reserves for wind turbines and photovoltaic systems, respectively.

[0160] S43. Assuming the energy storage system does not generate power at the initial sampling time, its active power variation is limited by the rated power. At the same time, the SoC should be kept within a certain range to reduce the adverse impact on the lifespan of the energy storage system.

[0161] -P BN ≤ΔP B (k+i|k)≤P BN (27)

[0162] SoC min ≤SoC(k+i|k)≤SoC max (28)

[0163] In the formula, P BN The rated power of the energy storage system is given by SoC(k+i|k); SoC(k+i|k) represents the predicted SoC of the energy storage system at sampling time k with respect to time k+i; SoC min and SoC max These are the set lower and upper limits for the SoC, respectively.

[0164] S5: Use a mathematical programming solver to solve a quadratic programming problem and obtain a set of optimal control sequences. The first element is the reference value of the change in active power of each power source at the current moment.

[0165] Preferably, step S5 specifically includes the following steps:

[0166] S51. Transform formulas (24)-(28) into a standard quadratic programming problem with constraints:

[0167] minJ=||Y(k)-R(k)|| Q 2 +||ΔU(k)|| R 2

[0168] =ΔU(k) T HΔU(k)-G(k) T ΔU(k) ​​(29)

[0169] In the formula, Q and R are weight coefficient matrices composed of α, β1, β2, β3 and β4; R(k) is the defined T p Step reference vector, by f ref and SoC ref Composition; the meanings of the remaining matrices and vectors are as follows:

[0170] E(k)=R(k)-S x Δx(k)-S d Δr(k)-τy(k) (30)

[0171] H = S u T QS u +R (31)

[0172] G(k)=2S u T QE(k) (32)

[0173] Constraints (25)-(27) are transformed into:

[0174]

[0175] In the formula:

[0176] Δu min =[0 0 -P B ] T (34)

[0177] Δu max =[ΔP W max ΔP p maxP B ] T (35)

[0178]

[0179] S52. The constraints of the energy storage system SoC can also be transformed similarly. For greater generality, the results are given when all output variables are restricted. Similarly, the constraints of the energy storage system SoC are simplified as follows:

[0180]

[0181] In the formula:

[0182] Y max (k)=[y max (k+1) ... y max (k+T p )] T (38)

[0183] Y min (k)=[y min (k+1) ... y min (k+T p )] T (39)

[0184] Thus, by using YALMILP and GROBI to solve the optimization problem, the control sequence [Δu(k) ... Δu(k+T)] is obtained. c The first element in the result, Δu(k), is the reference value ΔP for the change in active power of the wind-solar-storage power source at the current moment. W ref (k), ΔP p ref (k) and ΔP B ref (k).

[0185] S6: Select the first element in the sequence to act on the system, update the power grid state, and at the next sampling time, use the latest measurement value as the initial condition to re-predict the system output and solve it (i.e., the control object increases or decreases its output with the first element of the optimized solution as the reference value, and performs finite-time rolling optimization as the "current time" moves forward).

[0186] Preferably, in step S6, the wind turbine, photovoltaic system and energy storage system track the reference value calculated in S5 to make up for the power shortage in the power grid and support the stability of the system frequency.

[0187] The method of the present invention will be described below through two specific embodiments:

[0188] Figure 3 The diagram shows the improved IEEE 3-machine 9-node system topology in two embodiments of the present invention, as follows: Figure 3 As shown, the system includes a hydropower unit G1, a thermal power unit G2, and a wind turbine, photovoltaic system, and energy storage system that are all connected to the AC bus Bus 3. The steps of a multi-type power source coordinated frequency regulation method based on model predictive control for this system are as follows:

[0189] 1. At 20s, the load at Load 1 on AC bus 5 suddenly increases by 20MW to simulate the frequency dynamic process under disturbance.

[0190] 2. For hydropower, thermal power, wind power, solar power and energy storage power sources participating in the primary frequency regulation in the system, a frequency dynamic model is constructed based on the change of their active power. Considering that hydropower units and thermal power units will spontaneously respond under the action of speed governors, wind power units, photovoltaic systems and energy storage systems are taken as control objects. The dynamic response characteristics of wind power units are characterized by multiplying two first-order inertial elements as shown in formulas (3) and (4). The photovoltaic system and energy storage system are described by the first-order inertial elements of formulas (5) and (6) respectively. Then, the forward Euler method is applied to transform the continuous time model into the state space model of the discrete system.

[0191] 3. The AC bus frequency measured in real time by the online measuring device is obtained, the real-time system topology collected by the SCADA system is obtained online, and the power supply parameters and load parameters are obtained.

[0192] 4. Based on the system topology and parameters, a prediction model is built using the measured values ​​as initial conditions to predict the system frequency deviation and the changes in active power of wind turbines, photovoltaic systems and energy storage systems within a finite time domain in the future.

[0193] 5. Establish an optimization model with the minimum frequency deviation and minimum economic cost represented by formula (24) as the control objectives. Consider the constraints that different types of frequency regulation power sources need to meet during actual operation of the power system, covering formulas (25)-(28).

[0194] 6. Use MATLAB+yalmip+gurobi to solve the quadratic programming problem and obtain a set of optimal control sequences. The first element is the reference value of the change in active power of each power source at the current moment.

[0195] 7. Select the first element in the sequence and apply it to the system. At the next sampling time, use the latest measurement value as the initial condition to re-predict the system output and solve the optimization problem.

[0196] 8. Calculate the frequency characteristic values ​​of the power system during the simulation period, such as the frequency minimum point f. nadir Steady-state frequency deviation Δf ∞ .

[0197] Example 1 is used to verify the effectiveness of the method described in this invention. Example 1 is the method proposed in this invention, where hydropower, thermal power, wind power, solar power, and energy storage provide frequency support, and wind turbines, photovoltaic systems, and energy storage systems employ model predictive control. Example 2 reduces the number of frequency-regulated power sources, with only hydropower unit G1 and thermal power unit G2 participating. Example 3 changes the control method; wind, solar, and energy storage power sources adopt traditional PI control, i.e., virtual inertia control combined with droop control, collectively referred to as integrated inertia control.

[0198] Figure 4 The frequency response curves of the three schemes described in Example 1 after the system is subjected to a disturbance are shown below. Figure 4 As shown, under the same disturbance intensity, Schemes 1 and 3, by incorporating wind, solar, and energy storage, provide both frequency regulation backup and rapid response to frequency changes, thus effectively increasing f. nadir However, Option 1 can predict T. p The internal frequency situation changes, and the frequency modulation power supply output plan is rationally arranged to avoid situations where the frequency modulation demand cannot be met. Therefore, compared with Scheme 3, nadir and Δf ∞ Significant improvement. Therefore, the method proposed in this invention is applicable to the coordinated control of multiple types of power supplies and exhibits high performance.

[0199] Table 1 shows the frequency characteristic values ​​of each scheme in Example 1.

[0200]

[0201] Example 2 is used to verify the flexibility of the method described in this invention. The objective function of Scheme 1 is the optimal comprehensive benefit represented by formula (24). The objective function of Scheme 4 is the optimal frequency modulation effect, as shown in formula (40). The objective function of Scheme 5 is the optimal frequency modulation cost, expressed as formula (41).

[0202]

[0203]

[0204] Figure 5 This is a frequency response curve diagram of Embodiment 2 of the present invention under different control objectives. Figure 6 This is a graph showing the active power curves of a wind turbine under different control objectives according to Embodiment 2 of the present invention. Figure 7 This is a graph showing the active power curves of a photovoltaic system under different control objectives according to Embodiment 2 of the present invention. Figure 8 The following are active power curves of the energy storage system under different control objectives according to Embodiment 2 of the present invention, as shown in the figure. Figures 5-8As shown in the figure, it should be noted that, as shown in formula (2), the active power imbalance perceived by the wind turbine, photovoltaic system and energy storage system is the difference between the power of all conventional units and the load power. After the disturbance, the conventional units are slow to respond due to limitations such as ramp rate. Therefore, after the load surges for 20 seconds, the wind, solar and energy storage power sources will output power first to compensate for the active power deficit and gradually decrease as the output of hydro and thermal power sources increases. When the output of hydro and thermal power sources decreases for 35 seconds, the output of wind, solar and energy storage power sources will increase again. The trend is as follows. Figure 6-8 As shown.

[0205] It can be seen that Scheme 4 minimizes frequency deviation as much as possible, so that the wind-solar-storage power source can provide the maximum active power support. nadir With Δf ∞ The optimal frequency characteristic values ​​are 49.897Hz and 0.037Hz, respectively, making them the best among the three schemes. Conversely, Scheme 5 prioritizes the lowest economic cost, resulting in the lowest output from wind, solar, and energy storage compared to Scheme 4. The energy storage system, due to its lower cost coefficient compared to wind turbines and solar systems, generates more output. Scheme 1 combines the advantages of the two schemes, balancing frequency deviation reduction with lower economic costs. Its frequency curve improvement is superior to Scheme 5, and its cost for participating in frequency regulation is lower than Scheme 4. Therefore, the method proposed in this invention is highly flexible, allowing the objective function to be set according to actual needs to achieve the desired active frequency support or economic cost optimization.

[0206] Therefore, the present invention adopts the above-mentioned multi-type power source coordinated frequency regulation method based on model predictive control, which can realize the optimal allocation of frequency regulation power among multiple types of power sources and ensure the frequency stability of the power system.

[0207] 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 them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-type power supply coordinated frequency modulation method based on model predictive control, characterized in that: Includes the following steps: S1: For heterogeneous power sources participating in primary frequency regulation in the system, a frequency dynamic model is constructed based on the change of its active power. The frequency domain model is used to characterize the dynamic response characteristics of wind, solar and energy storage. Then, the forward Euler method is applied to transform the continuous-time model into the state-space model of the discrete system. S2: Online acquisition of the system frequency measured in real time by the measurement device, as well as the system topology, power parameters and load parameters collected by the SCADA system; S3: Based on the real-time topology, power parameters, and load parameters of the system, a prediction model is built with the measured values ​​as the initial conditions to predict the system frequency deviation and the changes in active power of wind turbines, photovoltaic systems, and energy storage systems within a finite time domain in the future. S4: Establish an optimization model with the objective functions of minimizing frequency deviation and minimizing economic cost, taking into account the constraints that different types of frequency regulation power sources need to meet during actual operation of the power system. S5: Use a mathematical programming solver to solve a quadratic programming problem and obtain a set of optimal control sequences. The first element is the reference value of the change in active power of each power source at the current moment. S6: Select the first element in the sequence and apply it to the system to update the power grid state. At the next sampling time, use the latest measurement value as the initial condition to re-predict the system output and solve it. Step S1 specifically includes the following steps: S11, The expression for the dynamic equation of the power grid frequency is: (1) In the formula, It is the system frequency; It is the inertial time constant of the power grid; It is the rated frequency; It is the total active power output of all conventional generating units; It is the load power; This refers to the active power output of the wind turbine generator; It is the active power output of the photovoltaic system; It is the active power output of the energy storage system; It is the load adjustment coefficient; It is the rated load power; S12. From the perspective of wind turbines, photovoltaic systems, and energy storage systems as a whole, the total active power imbalance of the power system is as follows: (2) S13. Since different types of frequency-modulated power supplies have different active power response processes after receiving frequency modulation commands, it is necessary to consider the dynamic response characteristics of wind turbines, photovoltaic systems, and energy storage systems to fully utilize their control effects. Considering the response characteristics of the control and communication systems, an inertial element is used for description. Based on the inverse Laplace transform principle, the actual active power output in the time domain can be calculated from the input frequency modulation command. (3) (4) (5) (6) In the formula, and , , These are the response time constants for wind turbines, photovoltaic systems, and energy storage systems, respectively. , , These are the active power reference values ​​for wind, solar, and energy storage power sources, i.e., the size of the commands they need to track; S14. Control its state of charge: (7) In the formula, For the next sampling time; This is the current sampling time; It is the sampling period; It is the rated capacity of the energy storage system; operator It is the current sampling time. Compared with the previous sampling time The difference; S15. To obtain the dynamic model of the discrete-time system with frequency, the forward Euler method is used to discretize formulas (1)-(7). At the same time, considering the introduction of integral energy to reduce or eliminate the static error of the system, the incremental form of the state-space model is adopted: (8) In the formula, The state variable matrix; For the control variable matrix; The matrix represents the interference variables; To output the variable matrix; These are the corresponding matrix coefficients; The state variable matrix is ​​represented as follows: (9) The control variable matrix is ​​represented as follows: (10) The interference variable matrix is ​​represented as follows: (11) The output variable matrix is ​​represented as follows: (12) State matrix coefficients Control matrix coefficients Interference matrix coefficients and output matrix coefficients They are respectively: (13) (14) (15) (16)； Step S3 specifically includes the following steps: S31. Assumption: Control Time Domain No more than the prediction time domain Outside the control time domain, the control variables remain unchanged, i.e. , Moreover, the interference the system experiences is in It remains unchanged after time, that is , ; S32, at the current sampling time The measured value is Calculate And taking this as a starting point, predict the future dynamics of the system according to formula (8), and define Step prediction output vector and The control variables for each step are as follows: (17) (18) Therefore, the output of the system in the time domain is calculated using the following formula: (19) In the formula: (20) (21) (22) (23)； Step S4 specifically includes the following steps: S41, a quadratic function of power offset and Offset describes the cost of frequency modulation, and the objective function for achieving the best overall benefit is defined as: (24) In the formula, This is the system frequency reference value; This is a reference value for the state of charge; This is the weighting coefficient for the frequency modulation effect; , and These are cost coefficients describing the power offset of wind turbines, photovoltaic systems, and energy storage systems, respectively. To describe energy storage systems The cost factor of the offset; S42. Wind turbines and photovoltaic systems participate in frequency regulation through reserved reserve capacity. The limited reserve capacity determines the range of their output power increment, resulting in: (25) (26) In the formula, Indicates the sampling time right The predicted value of the change in active power; and These are the maximum active power reserves for wind turbines and photovoltaic systems, respectively. S43. Assuming the energy storage system does not generate power at the initial sampling time, its active power change is constrained by the rated power. Simultaneously, it should be... To maintain within a certain range to reduce adverse effects on the lifespan of the energy storage system: (27) (28) In the formula, This refers to the rated power of the energy storage system. Indicates the sampling time right Real-time energy storage system The predicted value; and They are set respectively Lower limit and upper limit.

2. The multi-type power supply coordinated frequency modulation method based on model predictive control according to claim 1, characterized in that: The system in step S1 includes an AC power grid, wind turbines, photovoltaic systems, energy storage systems, converters, transformers, and loads.

3. The multi-type power supply coordinated frequency modulation method based on model predictive control according to claim 2, characterized in that: Step S5 specifically includes the following steps: S51. Transform formulas (24)-(28) into a standard quadratic programming problem with constraints: (29) In the formula, and It is by , , , and The weight coefficient matrix is ​​formed; For definition Step reference vector, by and Composition; the meanings of the remaining matrices and vectors are as follows: (30) (31) (32) Constraints (25)-(27) are transformed into: (33) In the formula: (34) (35) (36) S52, similarly, energy storage systems The constraints are simplified to: (37) In the formula: (38) (39) Thus, by using YALMILP and GROUP BY to solve the optimization problem, the control sequence was obtained. The first element in the result This is the reference value for the change in active power of wind, solar, and energy storage power sources at the current moment. , and .

4. The multi-type power supply coordinated frequency modulation method based on model predictive control according to claim 3, characterized in that: In step S6, the wind turbine, photovoltaic system and energy storage system track the reference value calculated in S5 to make up for the power shortage in the power grid and support the stability of the system frequency.

Images