Multi-functional dynamic voltage restorer based on particle swarm optimization

Abstract

The application discloses a multifunctional dynamic voltage recovery device based on particle swarm optimization, and relates to the field of voltage recovery devices. The current multifunctional device is limited to the combination of original functions and fault current limiting functions, and the effective working time proportion is very small, so the utilization rate of the equipment needs to be improved. The application comprises a direct current source, a PWM inverter, an LC filter, a series capacitor C and a transformer T, and the control of the device is realized through the combination of particle swarm optimization and model predictive control (MPC) based on the model, so that the optimization control of the PWM inverter is realized. The MF-DVR is based on vector control, can complete the consumption of new energy, can adjust the power quality of the power grid, realizes the regulation and control of active power, greatly improves the working time proportion, and improves the utilization rate of the equipment.

Description

Technical Field

[0001] This invention relates to the field of voltage recovery devices, and more particularly to a multifunctional dynamic voltage recovery device based on particle swarm optimization. Background Technology

[0002] Currently, with the increasing proportion of new energy sources and the rise of nonlinear and asymmetrical loads in power systems, power quality issues are becoming increasingly prominent, with voltage sags becoming the most significant problem. Dynamic voltage restorers (DVRs) have become the most economical and effective dynamic compensation devices due to their high efficiency, high reliability, and high speed.

[0003] Currently, domestic and international research on power systems mainly focuses on: grid voltage sag detection and parameter extraction, DVR compensation strategies, maximum output power of DVRs, energy recovery strategies for energy storage-type DVRs, DVR compensation-based flexible switching technology, and DVR control technology. However, the above research primarily concentrates on optimizing the functions and control strategies of DVR equipment. Under conditions of good grid operation, DVR equipment often remains idle, resulting in low utilization and limited functionality. Therefore, domestic and international experts and scholars have expanded the functions of series-connected devices such as DVRs. One scholar introduced a dynamic voltage restorer with fault current limiting (FCL-DVR), which can both adjust the voltage quality of the grid and limit short-circuit current. The design of the FCL-DVR system mainly involves improving the control mode limitations and increasing the flow rate of the current-limiting branch. FCL-DVR structures mainly include filter-based inductor current limiting and bridge-type current limiting. The aforementioned multi-functional devices are limited to the combination of existing functions and fault current limiting functions, and their effective operating time is relatively small, necessitating improvements in equipment utilization. Summary of the Invention

[0004] The technical problem to be solved and the technical task proposed by this invention is to improve and refine existing technical solutions, and to provide a multifunctional dynamic voltage recovery device based on particle swarm optimization, in order to improve equipment utilization. To this end, this invention adopts the following technical solution.

[0005] The multifunctional dynamic voltage recovery device (MF-DVR) based on particle swarm optimization includes a DC source, a PWM inverter, an LC filter, a series capacitor C, and a transformer T. The DC side of the device is used to connect to a renewable energy source, and the DC-DC converter provides energy to the DC-side voltage. The DC-DC converter enables DC voltage regulation and maximum power point tracking of the renewable energy source. The AC terminal of the PWM inverter is connected to the series capacitor C via an LC filter. The LC filter consists of an L... f Filter inductor and C f The filter capacitors are composed of various types of MF-DVRs. As an initial decomposition for establishing particle swarm optimization, the MF-DVR includes a P-type MF-DVR capable of scheduling active power, a combined P&Q type MF-DVR capable of scheduling both active and reactive power, and a Q-type MF-DVR capable of scheduling only reactive power. Then, a fitness function based on particle swarm optimization is established as the objective function. Through iteration of velocity and position, the optimal solution is obtained and passed to the device-level power control based on model predictive control (MPC). Device and power control serve as the primary control, combining the device-level and system-level power references P0 obtained from the optimal solution for islanding and grid-connected operations obtained through particle swarm optimization. ref and Q ref The PWM inverter is controlled through a cost function that balances the overall system. By combining particle swarm optimization with model predictive control (MPC), the MF-DVR, based on vector control, can both integrate new energy sources and adjust the power quality of the grid, thereby regulating active power and significantly increasing the proportion of operating time and improving equipment utilization.

[0006] As a preferred technical approach: In Model Predictive Control (MPC) under grid-connected mode, the voltage and frequency of the photovoltaic substation are determined by the main power grid. In the topology of the MF-DVR, Kirchhoff's laws can be used to obtain...

[0007]

[0008] Where V g This represents the rigid grid voltage, since the load voltage of the MF-DVR is determined by V. g Strictly fixed, therefore the active and reactive power outputs are calculated using the following formula:

[0009]

[0010]

[0011] Where ∧ represents the complex conjugate operation in the αβ orthogonal coordinate system;

[0012] By discretizing and using the power value at the current time k, the active and reactive power at the next time step (k+2) is predicted. The sampling period is...

[0013]

[0014]

[0015] To better control the active and reactive power of the MF-DVR, the designed delay compensation function is as follows:

[0016] J PQ =(P ref -P(k+2)) 2 +w PQ (Q ref -Q(k+2)) 2

[0017] Where ω PQ These are the weighting coefficients. The equation is solved in each sampling period to select the minimum result from all alternative voltage vectors. This effectively implements model predictive control (MPC) by solving the equation in each sampling period to select the minimum result from all alternative voltage vectors.

[0018] As a preferred technical approach, the fitness function is:

[0019]

[0020] Each term in the fitness function represents the total operating cost for each MF-DVR type. x (P x P ) indicates production (P) x P The cost of the xth P-type MF-DVR at kW. Similarly, production kW and The cost of combining a kVar P&Q type MF-DVR Representative production The cost of a Q-type MF-DVR is only at kVar. P N PQ and N Q These represent the total number of P-type units only, combined P&Q units, and Q-type units only, respectively. The fitness function of particle swarm optimization is established as the objective function, with the aim of minimizing the total operating cost of the power grid, i.e., minimizing the objective function value.

[0021] As a preferred technical approach: Considering the power generation and losses of new energy sources, the overall cost function formula can be written as:

[0022]

[0023]

[0024] Where Pload and Q load Active and reactive power load demand, P loss and Q loss There are active and reactive power losses, P grid and Q grid These represent the active and reactive power of a large power grid operating in grid-connected mode. The goal is to achieve optimal balance control.

[0025] As a preferred technique, the iterative equation for particle swarm optimization is as follows:

[0026]

[0027]

[0028] Where iter represents the number of iterations, v represents the velocity, x represents the position, λ represents the inertia weight, C1 and C2 represent the learning rate, and p best and g best These represent the optimal positions for the user and the entire system, respectively.

[0029] Although the speed is dynamically adjusted, it should still remain within a pre-specified range. It is recommended that the speed be uniform across the entire dimension, as shown below.

[0030]

[0031]

[0032] Where x max and x min These represent the maximum and minimum values ​​recorded by all particles, respectively. N is the number of intervals for adjusting the change. Based on randomly generated initial parameters, the position and velocity of the particles are updated by evaluating a fitness function, which can be set as a cost function plus a penalty variable. Then, the number of iterations is increased, and the fitness function is evaluated again to update p. best and g best The process continues until the number of iterations is exhausted or the control objective is achieved, finally yielding the optimal solution. This implements iterative particle swarm optimization for nonlinear problems.

[0033] Beneficial effects: By combining particle swarm optimization with model predictive control (MPC), optimized control of the PWM inverter is achieved. Based on vector control, the MF-DVR can not only absorb new energy sources but also adjust the power quality of the grid and regulate active power, greatly increasing the proportion of working time and improving equipment utilization. Attached Figure Description

[0034] Figure 1 This is a topology diagram of the MF-DVR of the present invention.

[0035] Figure 2 This is a schematic diagram of the circuit equivalent of the present invention.

[0036] Figure 3 This is the system active power flow diagram of the present invention.

[0037] Figure 4 This is a block diagram of the dynamic voltage recovery control at the device level for the MF-DVR of this invention.

[0038] Figure 5 This is a schematic diagram of the MF-DVR system-level algorithm control in this invention. Detailed Implementation

[0039] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] The topology of MF-DVR is as follows Figure 1 As shown, the DC source consists of new energy sources and their DC-DC converters. New energy sources such as photovoltaics and wind turbines are converted into DC power by DC / DC or AC / DC converters, effectively improving the absorption of new energy and the utilization rate of equipment. A single-phase full-bridge PWM inverter is connected to an LC filter (L...). f C f These are the filter inductor and filter capacitor, respectively. The series capacitor C is connected to the transformer. C can withstand a certain compensation voltage, thereby reducing the output voltage amplitude of the inverter.

[0041] Figure 2 For the equivalent circuit model of MF-DVR, U inv For the voltage at the inverter port of the new energy source; U cf U c U L These are the inverter output voltage, series coupling capacitor voltage, and load voltage, respectively. G2 is the power grid. The MF-DVR output voltage satisfies the following:

[0042] Figure 3 This is the active power flow diagram of the system according to the present invention. Assume α is the load voltage. and grid G2 voltage included angle It is the load voltage and load current The included angle, depending on the value of α, allows the active power flow to be divided into the following modes: 1) Mode 1. The MF-DVR absorbs active power, and both the load active power and the MF-DVR active power are provided by the grid; 2) Mode 2. The MF-DVR output active power is 0, and the load active power is entirely provided by the grid; 3) Mode 3. The MF-DVR outputs active power to the load, absorbing renewable energy; 4) Mode 4. The power grid has no active power output; the load receives active power from the MF-DVR; 5) Mode 5. In addition to supplying power to the load, the MF-DVR transmits excess active power to the grid.

[0043] Figure 4 This is a block diagram of dynamic voltage recovery control at the MF-DVR device level, employing dual-loop control of voltage and current. The outer loop primarily implements U... DVR Zero steady-state error tracking Controlled at the system level ( Figure 5 (As shown) P obtained through the particle swarm optimization algorithm ref Q ref This is achieved through droop control. Obtain the reference voltage of the DVR The current I is obtained after PI regulation. Lf Reference value Then, the control enters the inner current loop. To eliminate DC voltage bias, this patent calculates a sinusoidal reference for the series capacitor. The difference between the voltage and the sampled voltage UC is used for PI regulation to eliminate the DC component generated by the outer loop (power droop loop) during dynamic regulation.

[0044] like Figure 1-5 As shown, a multifunctional dynamic voltage recovery device (MF-DVR) based on particle swarm optimization is presented. The MF-DVR's topology mainly consists of a DC source, a PWM inverter, an LC filter, a series capacitor C, and a transformer T. The DC side of the MF-DVR is used to connect to new energy sources such as photovoltaics and wind turbines, and the DC-DC converter provides energy to the DC side voltage, effectively promoting the local consumption of new energy and improving the device's utilization rate. The DC-DC converter can achieve DC voltage regulation and maximum power point tracking for new energy sources. The AC terminal of the PWM inverter is connected to the series capacitor C via an LC filter. The LC filter consists of an L... f Filter inductor and C f The filter capacitor is composed of series capacitors, which can withstand part of the compensation voltage and reduce the output voltage amplitude of the inverter.

[0045] In Model Predictive Control (MPC) based on grid-connected mode, the voltage and frequency of the photovoltaic substation are determined by the main power grid. In the topology of the MF-DVR, Kirchhoff's laws can be used to obtain...

[0046]

[0047] Where V g This represents the rigid grid voltage, since the load voltage of the MF-DVR is determined by V. g Since the power output is strictly fixed, the active and reactive power outputs can be calculated using the following formula:

[0048]

[0049]

[0050] Where ∧ represents the complex conjugate operation in the αβ orthogonal coordinate system;

[0051] By discretizing and using the power value at the current time k, the active and reactive power at the next time step (k+2) is predicted. The sampling period is...

[0052]

[0053]

[0054] To better control the active and reactive power of the MF-DVR, the designed delay compensation function is as follows:

[0055] J PQ =(P ref -P(k+2)) 2 +w PQ (Q ref -Q(k+2)) 2

[0056] Where ω PQ These are the weighting coefficients, and the equation will be solved in each sampling period to select the minimum result from all alternative voltage vectors.

[0057] For the purpose of optimizing system power, MF-DVR includes P-type which can dispatch active power, joint P&Q type MF-DVR which can dispatch both active and reactive power, and Q-type which can only dispatch reactive power. By performing particle swarm initialization decomposition, the total system cost can be minimized as an indicator of power management.

[0058] The fitness function of particle swarm optimization is established as the objective function, and the fitness function (objective function) is designed as follows:

[0059]

[0060] Each term in the objective function represents the total operating cost for each MF-DVR type. x (P x P ) indicates production (P) x P The cost of the xth P-type MF-DVR at kW. Similarly, production kW and The cost of combining a kVar P&Q type MF-DVR Representative production The cost of a Q-type MF-DVR is only at kVar. P N PQ and N Q These represent the total number of P-type units only, combined P&Q-type units, and Q-type units only, respectively. The goal is to minimize the total operating cost of the power grid, i.e., to minimize the objective function value.

[0061] Taking into account the power generation and losses of new energy sources, the overall balance formula can be written as:

[0062]

[0063]

[0064] Where P load and Q load Active and reactive power load demand, P loss and Q loss There are active and reactive power losses, P grid and Q grid These are the active and reactive power of a large power grid operating in grid-connected mode.

[0065] This invention applies the particle swarm optimization algorithm to solve complex nonlinear problems.

[0066]

[0067]

[0068] Where iter represents the number of iterations, v represents the velocity, x represents the position, λ represents the inertia weight, C1 and C2 represent the learning rate, and p best and g best These represent the optimal positions for themselves and the global position, respectively.

[0069] Although the speed is dynamically adjusted, it should still remain within a pre-specified range. It is recommended that the speed be uniform across the entire dimension, as shown below.

[0070]

[0071]

[0072] Where x max and x min These represent the maximum and minimum values ​​recorded by all particles, respectively, and N is the number of intervals for adjusting the change. Based on randomly generated initial parameters, the position and velocity of the particles are updated by evaluating a fitness function, which can be set as a cost function plus a penalty variable. Then, the number of iterations is increased, and the fitness function is evaluated again to update p. best and gbest The process continues until the number of iterations is exhausted or the control objective is achieved, at which point the optimal solution is obtained.

[0073] During grid connection, appropriately manipulating power parameters in the droop controller will impart a secondary control effect at the equipment level. As shown in the following formula, P ref and Q re f is equivalent to the classic quadratic control formula, shifting the original droop control curve upwards by Δ. f and Δ E .

[0074]

[0075]

[0076] Therefore, if Δ f and Δ E Adding a positive value to the voltage drop characteristic will compensate for the frequency and voltage deviations inherited from the voltage drop. In grid-connected mode, P ref and Q ref Based on the system-level particle swarm optimization algorithm, the aim is to eliminate frequency and voltage deviations and achieve economical power distribution among distributed power sources.

[0077] Taking frequency as an example, the following steps are used to determine the droop control reference value in detail.

[0078] S1) In order to achieve the secondary control effect, f needs to be restored to f ref To achieve this effect, P ref It is necessary to closely track the P value of the MF-DVR unit specified in the above formula, such that (-mP+mP) ref The value is zero.

[0079] S2) In a steady-state power grid, since frequency is a global variable, the above equation should apply to all distributed generation sources with different droop coefficients, i.e.:

[0080] m1(P1-P ref1 )=m2(P2-P ref2 )=...=m n (P n -P refn ).

[0081] In addition to achieving the effect of secondary control, it is also necessary to minimize the total cost represented in the objective function. This can be achieved by using system-level optimization and following the overall equilibrium formula.

[0082] S4) The feasible operating area for the alternative optimal solution pool must cover the local load power demand and power line capacity constraints, as well as some margin for potential common load sharing.

[0083] For simplicity, the area structures are assumed to be known for the control scheme verification. In actual planning, these structures are difficult to clearly visualize but can be operated by experienced operators based on information from smart meters. Furthermore, nonlinear issues may arise due to physical limitations of power supplies, converters, or wiring. This problem can be effectively solved using particle swarm optimization algorithms.

[0084] like Figure 5 The illustrated MF-DVR algorithm control diagram shows that it consists of two control layers: a device level for islanded and grid-connected operation, and a system level. This is achieved by setting a power reference, i.e., P... ref and Q ref These two levels are interconnected. The top-level system-level power management scheme generates power dispatch commands. Then, the bottom-level device-level power control scheme receives and executes the power dispatch commands, controlling the inverter to stabilize the grid connected to the MF-DVR. From the perspective of the hierarchical control architecture, the system-level power optimization strategy based on the particle swarm optimization (PSO) algorithm simultaneously plays the role of both level two and level three control, solving optimization problems, achieving power flow optimization, and restoring the deviated frequency / voltage.

[0085] Device-level power control based on model predictive control (MPC) serves as the primary control mechanism, executing power dispatch commands by controlling the power converter. Specifically, predictive models under different modes generate predicted state variables for future instants based on new measurements of electrical signals. All generated predictions are then evaluated in a cost function, where the optimal switching state with the minimum cost is sent to the power converter.

[0086] Specifically, device-level control is implemented locally, while system-level control in the Grid Central Controller (MGCC) is implemented globally. For example, the MGCC is used to coordinate the MF-DVR during startup or to manage energy flow in a distributed secondary control microgrid. One of the key advantages of droop control is that it does not require additional communication.

[0087] In power grids with hierarchical control structures, higher levels of control and more stringent communication requirements are needed to make complex microgrid systems more reliable, visible, and interactive. From a system performance perspective, the increased communication requirements can mitigate voltage deviations, improve the inherent shortcomings of droop control, thereby optimizing active and reactive power and improving power quality at the device level.

[0088] above Figure 1-5The multifunctional dynamic voltage recovery device based on particle swarm optimization shown is a specific embodiment of the present invention, which has demonstrated the outstanding substantive features and significant progress of the present invention. According to actual use needs, equivalent modifications in shape, structure, etc. can be made to it under the guidance of the present invention, all of which are within the protection scope of this solution.

Claims

1. A multifunctional dynamic voltage recovery device based on particle swarm optimization, characterized in that: The multi-functional dynamic voltage recovery device, or MF-DVR for short, includes a DC source, a PWM inverter, an LC filter, and a series capacitor. and transformer The DC side of the device is used to connect to new energy sources, and the DC / DC converter provides energy to the DC side voltage. The DC / DC converter realizes DC voltage regulation and maximum power point tracking of the new energy source. The AC terminal of the PWM inverter is connected to a series capacitor via an LC filter. C LC filter is made of Filter inductor and The filter capacitors are composed of various types of MF-DVRs, including P-type MF-DVRs that can dispatch active power, combined P&Q type MF-DVRs that can dispatch both active and reactive power, and Q-type MF-DVRs that can only dispatch reactive power. These are used to optimize the system. A particle swarm optimization fitness function is established as the objective function. Through iteration on velocity and position, the optimal solution is obtained. This optimal solution serves as the power reference at both the device and system levels. and The power is then passed to the device-level power control based on model predictive control (MPC). The MPC predicts the active and reactive power for the next moment based on the current power value and solves for it in each sampling period using a delay compensation function to select the optimal voltage vector for controlling the PWM inverter. The device-level power control serves as the primary control mechanism, combining device-level and system-level power references optimized by particle swarm optimization for islanded and grid-connected operations. and The control of the PWM inverter is achieved through a cost function that balances the overall performance.

2. The multifunctional dynamic voltage recovery device based on particle swarm optimization according to claim 1, characterized in that: In Model Predictive Control (MPC) based on grid-connected mode, the voltage and frequency of the photovoltaic substation are determined by the main power grid. In the topology of the MF-DVR, Kirchhoff's laws are used to obtain... in This represents the rigid grid voltage, since the load voltage of the MF-DVR is determined by... Strictly fixed, therefore the active and reactive power outputs are calculated using the following formula: Where ∧ represents the complex conjugate operation in the αβ orthogonal coordinate system; By discretizing and using the power value at the current time k, predict the active and reactive power at the next time step (k+2): To better control the active and reactive power of the MF-DVR, the designed delay compensation function is as follows: in These are weighting coefficients, and the delay compensation function equation will be solved in each sampling period to select the minimum result from all alternative voltage vectors.

3. The multifunctional dynamic voltage recovery device based on particle swarm optimization according to claim 2, characterized in that: The fitness function is: Each term in the fitness function represents the total operating cost for each MF-DVR type; Indicates production kW hour The cost of a single P-type MF-DVR; similarly. production kW and The cost of a combined P&Q MF-DVR at kW Representative production The cost of a Q-type MF-DVR at kW; , and These represent the total number of P-type only, combined P&Q type, and Q-type only units, respectively.

4. The multifunctional dynamic voltage recovery device based on particle swarm optimization according to claim 3, characterized in that: Considering the power generation and losses of new energy sources, the overall cost function formula for balance can be written as: in and It is the demand for both active and reactive power loads. and It refers to active and reactive power losses. and These are the active and reactive power of a large power grid operating in grid-connected mode.

5. The multifunctional dynamic voltage recovery device based on particle swarm optimization according to claim 4, characterized in that: The iterative equations for particle swarm optimization are as follows: in Indicates the number of iterations. Indicates speed, Indicates location, Indicates inertia weight, and Indicates the learning rate, and These represent the optimal positions for the current user and the entire system, respectively. Although the speed is dynamically adjusted, it should still remain within a pre-specified range, and the speed should be uniform across the entire dimension, as shown below. in and These are the maximum and minimum values ​​recorded by all particles, respectively. N is the number of intervals for adjusting the change. Based on randomly generated initial parameters, the position and velocity of particles are updated by judging the fitness function, which is set as the cost function plus the penalty variable. Then, the number of iterations is increased, and the fitness function is evaluated again to update... and The process continues until the number of iterations is exhausted or the control objective is achieved, at which point the optimal solution is obtained.

Images