Optimized fuzzy control energy management method of dual three-phase permanent magnet synchronous generator system

Abstract

The invention discloses an optimized fuzzy control energy management method for a dual three-phase permanent magnet synchronous generator system, and the method comprises the steps: carrying out the distribution and coordination of the output power Pgen of a dual three-phase generator, the output power Pbat of a lithium battery, the output power Psc of a super capacitor, and the demand power Preq of the dual three-phase permanent magnet synchronous generator system through a fuzzy logic algorithm; when the output power Pgen of the dual three-phase generator is smaller than the required power Preq of the system, the required power Preq of the system and the charge state SOCbat-sc of the hybrid energy storage system are input into a fuzzy control unit for defuzzification processing, and the output power Pgen of the dual three-phase generator, the output power Pbat of the lithium battery and the output power Psc of the super capacitor are output; therefore, system power distribution is adjusted, and generator output meets working condition requirements. The energy consumption ratio of the system is effectively reduced, and the fuel economy of the system is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of generator energy management, and in particular relates to an optimized fuzzy control energy management method for a dual three-phase permanent magnet synchronous generator system. Background Art

[0002] In the new era, China's national defense needs, industrial energy structure adjustment, and changes in consumption concepts have promoted the increasing number of advanced equipment and green energy vehicles, and my country's demand for high-performance DC power generator systems has become increasingly urgent. Modern electrical systems integrate power generation units, energy distribution networks, energy storage devices, and power loads for intelligent perception management. Efficient generator systems have an irreplaceable position in situations with space weight restrictions, diversified equipment loads, and high endurance requirements. Multi-phase permanent magnet motors, with their advantages of a large number of windings, good electromagnetic properties, and compact structure, are gradually replacing traditional three-phase motors with less control freedom and poor fault tolerance in generator systems. Under the same output voltage and power level, multiple sets of motor windings can greatly reduce the current flowing through each phase winding, reduce the generator system's dependence on the performance of hardware devices such as the withstand voltage and current indicators of power devices and current sensor parameters, effectively reduce the heating of copper windings, and relieve the burden on the cooling system, which has a positive significance for improving the operating efficiency of the generator system.

[0003] Although permanent magnet synchronous generator technology has made great progress, there are still some urgent problems to be solved in energy management and practical applications: permanent magnet synchronous generators often need to be used in conjunction with energy storage systems (such as battery packs) to cope with load changes or fluctuations in power demand. In independent power supply or emergency backup systems, the load volatility can be very large, and it is necessary to intelligently distribute electrical energy according to the actual load conditions to ensure efficient collaboration between generators and energy storage devices. How to deal with this change has become a problem that the power generation system must solve.

[0004] In this case, the generator alone cannot provide stable power output, which highlights the importance of energy management system. EMS can not only optimize the working status of the generator, but also ensure the stable operation of the power grid under the condition of power generation fluctuations by adjusting the energy storage system and load distribution.

[0005] However, EMS needs to balance multiple objectives, such as maximizing power generation efficiency, optimizing energy storage systems, frequency and voltage stability, and minimizing power losses. This multi-objective optimization problem is extremely complex, involving dynamic load changes, energy fluctuations, and equipment health management. Existing control algorithms often lack optimization capabilities when dealing with such complex problems, making it difficult to make optimal decisions in real time. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides an optimized fuzzy control energy management method for a dual three-phase permanent magnet synchronous generator system.

[0007] The present invention achieves the above technical objectives through the following technical means.

[0008] Optimized fuzzy control energy management method for dual three-phase permanent magnet synchronous generator system:

[0009] Build a dual three-phase permanent magnet synchronous generator system, including a power generation system, a hybrid energy storage system and other loads, the power generation system, the hybrid energy storage system and other loads are all connected to the DC bus, the power generation system includes a dual three-phase generator and a rectifier module thereof, the hybrid energy storage system contains a lithium battery and a super capacitor, and the lithium battery and the super capacitor are connected in parallel;

[0010] Get the output power P of the dual three-phase generator gen , the output of the hybrid energy storage system, the required power P of the dual three-phase permanent magnet synchronous generator system req and is coordinated through fuzzy logic algorithm. The output of the hybrid energy storage system includes the output power P of the lithium battery. bat 、The output power P of the supercapacitor sc , the fuzzy logic algorithm allocation coordination is specifically:

[0011] When the output power of the dual three-phase generator P gen Greater than the system's required power P req When the charge state SOC of the hybrid energy storage system is determined bat-sc Is it greater than the limit discharge power SOC(L)? If it is less than SOC(L), the hybrid energy storage system stops supplying power. Otherwise, further determine SOC bat-sc Is it greater than the limited charging power SOC(H)? When it is greater than SOC(H), the hybrid energy storage system is prohibited from charging, where SOC(L)<SOC(H);

[0012] When the output power of the dual three-phase generator P gen Less than the system's required power P req When the system power requirement P req , SOC bat-sc The fuzzy control unit of the fuzzy logic algorithm is input for defuzzification and the output power P of the dual three-phase generator is output. gen , output power P of lithium battery bat And the output power P of the supercapacitor sc , thereby adjusting the system power distribution so that the generator output meets the operating requirements.

[0013] Furthermore, a triangle mapping method of fast screening rules is used for defuzzification processing. The triangle mapping method of fast screening rules uses an index structure to store and query the mapping relationship between the rules of the fuzzy logic algorithm and the triangles. During the defuzzification process, only the centroid of the triangle area corresponding to the screened rules is calculated.

[0014] Furthermore, the index structure is as follows: a KD-Tree tree is established, which uses the input fuzzy value of each rule as the node coordinate and constructs the rule itself as the node content; when there is an input fuzzy value, the KD-Tree is used to quickly locate all rules that may match the input fuzzy value, and the rules are screened according to the fit threshold: rules with a fit less than 5% are considered unmatched.

[0015] Furthermore, the construction process of the triangle mapping relationship is as follows: a mapping relationship is established between each rule and its output membership function, specifically, a reverse mapping relationship from rule to triangle is established, that is, each triangle corresponds to one or more rules for use in reverse query; based on the filtered rule ID, the vertex coordinates and membership values ​​of the triangle area corresponding to the rule are queried, and the weighted summation is performed using the center of gravity and area of ​​the corresponding triangle area to obtain the precise value of the defuzzification.

[0016] Furthermore, a multi-dimensional search method is used to screen the triangular area corresponding to the rule: the multi-dimensional input value is used as the node content of the KD-Tree to realize multi-dimensional search.

[0017] Furthermore, after the defuzzification process, the PSO particle swarm optimization algorithm is used to optimize the membership function set in the fuzzy logic algorithm rules.

[0018] Furthermore, the top and bottom points of each membership function in the fuzzy logic algorithm are adjusted to select the required power P of the system. req , Hybrid Energy Storage System SOC bat-sc And the output power P of the dual three-phase generator gen The parameters in the membership function are optimized.

[0019] Furthermore, the annealing algorithm is used to optimize the inertia weight w, the individual optimal acceleration factor c1 and the global optimal acceleration factor c2 in the iterative update formula of the PSO particle swarm optimization algorithm. Specifically:

[0020] Define a fitness function to evaluate the performance of the current parameter combination;

[0021] Initialize the parameter range, w is in the range of [0, 1], c1 and c2 are in the range of [0, 3]; set the initial temperature to 100, and set the cooling rate value to 0.95;

[0022] After entering the iteration cycle, the PSO is run using the randomly generated new inertia weight, individual optimal acceleration factor, and global optimal acceleration factor to calculate the fitness; whether to accept the randomly generated new parameters is determined based on the fitness change: if the new fitness is better, the new parameters are directly accepted; if the new fitness is worse, the new parameters are determined with probability P. Among them, f current is the current fitness, f new is the new fitness and T is the current temperature.

[0023] Furthermore, the “historical memory inversion disturbance” strategy is adopted to prevent the annealing algorithm from being troubled by the local optimal solution. Specifically:

[0024] Memory of historical solutions: Record historical solutions in each iteration, including the location of the current solution, the current temperature, and the acceptance probability;

[0025] Introduce inversion perturbation: When the temperature drops to a certain level, multiple historical solutions are considered in reverse, and new solutions are generated in the form of reverse perturbation; the certain level is determined based on the initial temperature and empirical values;

[0026] Introduce the "historical solution spacing" threshold: Calculate the distance between the current solution and the historical solution. If the distance is less than the "historical solution spacing" threshold, force a larger reverse perturbation to try to break out of this cycle mode.

[0027] Multi-stage inversion: different inversion strategies are introduced in the early, middle and late stages. The reverse perturbation of the current solution is used in the early and middle stages, and the mean reverse perturbation of the historical trajectory is used in the late stage.

[0028] Furthermore, a nonlinear multi-stage dynamic perturbation strategy is used to improve the adaptability of the annealing algorithm, specifically:

[0029] Dynamic adjustment of disturbance amplitude: Divide the disturbance into multiple stages, and use different disturbance strategies in each stage: use larger random disturbance in the early stage, use smaller target-oriented disturbance in the middle stage, and introduce larger random disturbance again in the later stage for global search. The larger random disturbance, i.e., the disturbance temperature, is 5% of the maximum temperature, and the smaller target-oriented disturbance, i.e., the disturbance temperature, is 1% of the maximum temperature.

[0030] Initialization settings: Initialize w, c1, c2 to [0.7, 1.5, 1.5], the initial temperature to 100, the temperature decay rate to 0.95, and the maximum number of iterations to 100;

[0031] Specific operations at each iteration:

[0032] Parameter perturbation: When the number of iterations is a multiple of 10, the “historical memory inversion perturbation” strategy is used to generate a new solution: a historical solution is randomly selected, and the direction between the current solution and the historical solution is calculated. If they are in the same direction, a new solution is generated in the opposite direction. Otherwise, the standard adaptive perturbation strategy is used: the normrnd function is used to perform random perturbations on the basis of the current solution, and the perturbation amplitude is proportional to the current temperature.

[0033] Update the current solution and the global optimal solution: Calculate the acceptance probability based on the temperature of the current annealing algorithm and the difference between the new solution and its target value; if the new solution is better than the current solution, accept the new solution; otherwise, decide whether to accept the new solution based on the probability P; if the target value of the current solution is better than the global optimal solution, update the global optimal solution;

[0034] Dynamic temperature adjustment: every 10 iterations, the frequency of accepting new solutions is calculated. If the frequency is lower than the set empirical threshold, the temperature drop rate is slowed down; if the current solution is not closer to the target solution than the solution of the previous iteration in the next iteration, the temperature drop rate is accelerated; if no better solution is found after 10 consecutive iterations, the temperature drop rate is accelerated to achieve rapid convergence.

[0035] The beneficial effects of the present invention are as follows: the present invention uses a fuzzy logic algorithm to calculate the output power P of the dual three-phase generator. gen , output power P of lithium battery bat 、The output power P of the supercapacitor sc , the required power P of the dual three-phase permanent magnet synchronous generator system req For distribution coordination, when the output power P of the dual three-phase generator gen Less than the system's required power P req When the system power requirement P req , Hybrid Energy Storage System SOC bat-sc As the input parameter, it is input into the fuzzy control unit of the fuzzy logic algorithm for defuzzification, and the output power P of the dual three-phase generator is output. gen , output power P of lithium battery bat And the output power P of the supercapacitor sc ; In the defuzzification process, the triangular mapping method of fast screening rules is adopted; after the defuzzification process, the PSO particle swarm optimization algorithm is used to optimize the membership function of the fuzzy logic algorithm: adjust the vertices and bottom points of each membership function in the fuzzy logic algorithm, and select the required power P of the system. req , Hybrid Energy Storage System SOC bat-sc And the output power P of the dual three-phase generator genThe parameters in the membership function are optimized, and the annealing algorithm is further used to optimize the inertia weight w, the individual optimal acceleration factor c1 and the global optimal acceleration factor c2 in the iterative update formula of the PSO particle swarm optimization algorithm; the "historical memory inversion perturbation" strategy is also used to prevent the annealing algorithm from being troubled by the local optimal solution, and the nonlinear multi-stage dynamic perturbation strategy is used to improve the adaptive ability of the annealing algorithm. The present invention can improve the convergence speed of the algorithm while maintaining the global search capability through the combined optimization of the dual strategies of historical memory inversion perturbation and nonlinear multi-stage dynamic perturbation, and effectively solve the problems of premature convergence and local optimality in traditional simulated annealing algorithms. The method of the present invention can significantly improve the fuel economy and system stability of the dual three-phase permanent magnet synchronous generator system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a structural diagram of a dual three-phase permanent magnet synchronous generator system containing a battery-supercapacitor according to the present invention;

[0037] Figure 2 This is a structural diagram of the power generation system of the present invention;

[0038] Figure 3 This is the fuzzy control principle diagram of the present invention;

[0039] Figure 4 The power requirement P of the system of the present invention req Parameter design diagram;

[0040] Figure 5 It is a schematic diagram of the design of the residual state SOC parameters of the hybrid energy storage system of the present invention;

[0041] Figure 6 is the output power P of the generator of the present invention gen Parameter design diagram;

[0042] Figure 7 This is a local enlarged view of the 0.1s voltage in the embodiment of the present invention;

[0043] Figure 8 This is a partial enlarged view of the 0.2s voltage in the embodiment of the present invention;

[0044] Fig. 9 This is a partial enlarged view of the 0.3s voltage in the embodiment of the present invention;

[0045] Fig.10 This is a partial enlarged view of the 0.4s voltage in the embodiment of the present invention;

[0046] Fig.11 This is the fuel consumption contour diagram of the generator described in the present invention. DETAILED DESCRIPTION

[0047] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0048] like Figure 1 As shown, a schematic structural diagram of the dual three-phase permanent magnet synchronous generator system of the present invention includes a power generation system, a hybrid energy storage system and other loads, the power generation system, the hybrid energy storage system and other loads are all connected to a DC bus, wherein the hybrid energy storage system contains a lithium battery and a super capacitor, and the lithium battery and the super capacitor are connected in parallel.

[0049] The optimized fuzzy control energy management method of the dual three-phase permanent magnet synchronous generator system of the present invention specifically includes the following contents:

[0050] (1) Power generation system

[0051] According to the DC grid system model and load size, a power generation system is established, including a dual three-phase generator and its rectifier module. The dual three-phase generator adopts a stator structure with neutral points isolated from each other and two sets of windings directly 30° apart; the rectifier module adopts dual vector control rectification, such as Figure 2 shown.

[0052] The multi-phase motor dual vector VSD decoupling method is used to decompose the physical quantities of the dual three-phase generator into a new static six-dimensional space coordinate system αβz1z2o1o2. Since αβ, z1z2, and o1o2 are orthogonal to each other, the six-dimensional space coordinate system can be divided into three sub-planes. Since z1z2 does not participate in energy conversion, only the αβ sub-plane is rotated and changed into a rotating dq axis, so the voltage equation of this dual three-phase generator is:

[0053]

[0054] Among them, u d and u q are d and q axis voltages respectively; u z1 and u z2 are the voltages of the z1 and z2 sub-planes respectively; L d , L q 、i d and i q are the d-axis and q-axis inductance and current respectively; L z1 , L z2 、i z1 and i z2 are the inductance and current of sub-plane z1 and z2 respectively; R s , and ω e are the stator resistance, permanent magnet flux amplitude and rotor electrical angular velocity respectively.

[0055] In the rotating coordinate system, the active power output by the dual three-phase generator is:

[0056] P s =3(u d i d +u q i q )

[0057] The electromagnetic torque equation of the dual three-phase generator is:

[0058]

[0059] Where p is the number of motor pole pairs;

[0060] Since the d-axis and q-axis inductances of the surface mounted permanent magnet synchronous motor are equal, that is, L d =L q , the electromagnetic power output by the dual three-phase generator is:

[0061]

[0062] Among them, ω m is the mechanical angular velocity of the rotor.

[0063] (2) Hybrid Energy Storage System

[0064] ① Considering the voltage drop, polarization voltage, polarization internal resistance and other issues caused by the ohmic internal resistance of the lithium battery, the mathematical expression of the equivalent circuit model when the lithium battery is charged is determined as:

[0065]

[0066] The mathematical expression of the equivalent circuit model during discharge is:

[0067]

[0068] The SOC of a lithium battery is expressed as:

[0069]

[0070] Where U bat Represents the terminal voltage of the lithium battery, E0 represents the open circuit voltage of the lithium battery, R b represents the internal resistance of the lithium battery, i represents the actual current of the lithium battery, K represents the resistance of the lithium battery after polarization, q represents the actual capacity of the lithium battery, Q represents the total capacity ... * represents the current of the lithium battery after low-pass filtering, A represents the amplitude of the lithium battery voltage in the exponential region, B is the inverse of the time constant, SOC ini It is the initial value of the lithium battery state of charge.

[0071] ②The supercapacitor adopts the RC model, and the mathematical expression of its terminal voltage Usc is:

[0072] U=UU=IR

[0073] SC CRS

[0074] The state of charge (SOC) of a supercapacitor is defined as:

[0075]

[0076] Where U max , U min are the maximum and minimum allowable voltages of supercapacitors, Q t is the charge stored in the supercapacitor at time t, Q N is the maximum charge that the supercapacitor can store, U SC is the terminal voltage of the supercapacitor;

[0077] The maximum energy output E of the supercapacitor is:

[0078]

[0079] Where C is the capacitance of the supercapacitor.

[0080] (3) Obtain the output of the dual three-phase generators, the output of the hybrid energy storage system, and the required power of the dual three-phase permanent magnet synchronous generator system, and allocate and coordinate them through a fuzzy logic algorithm.

[0081] The fuzzy logic algorithm is loaded into the fuzzy logic controller, and the input of the fuzzy logic controller is: the required power P of the dual three-phase permanent magnet synchronous generator system req , the SOC of the lithium battery and the SOC of the supercapacitor (the sum of the two is the SOC of the hybrid energy storage system bat-sc ), the output of the fuzzy logic controller is: the output power P of the dual three-phase generator gen , output power P of lithium battery bat And the output power P of the supercapacitor sc .

[0082] The process of fuzzy logic algorithm allocation coordination is as follows: In the dual three-phase permanent magnet synchronous generator system containing battery-supercapacitor, when the output power P of the dual three-phase generator is gen Greater than the system's required power P req When the SOC of the hybrid energy storage system is determined bat-sc Is it greater than the limit discharge power SOC (L)? If it is less than SOC (L), the hybrid energy storage system stops supplying power. Otherwise, further determine the hybrid energy storage system SOC bat-scIs it greater than the limited charging power SOC(H)? If it is greater than SOC(H), the hybrid energy storage system is prohibited from charging, and SOC(L)<SOC(H); when the output power P of the dual three-phase generator is gen Less than the system's required power P req When the system power requirement P req , Hybrid Energy Storage System SOC bat-sc As the input parameter, it is input into the fuzzy control unit of the fuzzy logic algorithm for defuzzification, and the output power P of the dual three-phase generator is output. gen , output power P of lithium battery bat And the output power P of the supercapacitor sc , thereby adjusting the system power distribution so that the generator output meets the working conditions. Figure 3 shown.

[0083] 1) Rules of fuzzy logic algorithm

[0084] Set the system's required power P req , the output power of the hybrid energy storage system (including the output power of the lithium battery P bat 、The output power P of the supercapacitor sc ) and the output power P of the dual three-phase generator gen There are five membership function sets, namely VL (very small), L (small), M (medium), H (large), VH (very large), and the system demand power P req , the output power of the hybrid energy storage system and the output power of the dual three-phase generator P gen The fuzzy domains are [0, 120] (kw), [0, 12] (kw), and [20, 180] (kw). State of charge SOC of hybrid energy storage system bat-sc The membership function sets are L (small), M (medium), and H (large), and the fuzzy domain is [0, 100] (%).

[0085] The following two tables are the rule tables of the fuzzy control unit. Table 1 is the output power P of the generator. gen Table 2 shows the rules for the output power of the hybrid energy storage system.

[0086] Table 1 Generator output power P gen Rules

[0087]

[0088] Table 2 Rules for output power of hybrid energy storage system

[0089]

[0090] 2) In the defuzzification process, the traditional Mamdani fuzzy logic algorithm uses a centroid algorithm, which determines the final output value by calculating the centroid position of the fuzzy output set (fuzzy region) on a two-dimensional plane. This process can be imagined as finding the centroid of an irregularly shaped two-dimensional figure. The two-dimensional figure is composed of multiple fuzzy membership functions superimposed to form a complex fuzzy set region.

[0091] The present invention adopts a triangular mapping method of fast screening rules to perform defuzzification processing. The triangular mapping method of fast screening rules uses an efficient index structure to store and query the rules of the fuzzy logic algorithm and the triangle mapping relationship, so that during the defuzzification process, only the center of gravity calculation is performed on the triangular area corresponding to the screened rules.

[0092] In order to speed up the rule screening process, an efficient index structure is created for fuzzy rules: a KD-Tree is established. KD-Tree is a data structure for fast query in multidimensional space. It uses the input fuzzy value of each rule as the node coordinates and the rule itself as the node content. When there is an input fuzzy value, KD-Tree is used to quickly locate all rules that may match the input fuzzy value. According to the fit threshold, rules with a fit of less than 5% are considered unmatched rules, and only qualified rules are retained for subsequent calculations. After rule screening, each rule may correspond to one or more output membership functions.

[0093] Construction of triangle mapping relationship: Establish a mapping relationship between each rule and its output membership function, specifically, establish a reverse mapping relationship from rule to triangle, that is, each triangle corresponds to one or more rules for use in reverse query; according to the currently filtered rule ID, quickly query the vertex coordinates and membership value of the triangle area corresponding to the rule, use the center of gravity and area of ​​these triangle areas for weighted summation, and finally get the accurate value of defuzzification. The specific process of constructing the triangle mapping relationship is:

[0094] The output membership function of the complex shape corresponding to each rule is segmented into several triangles, and the complex shape is wrapped with an envelope or convex hull to calculate the centroid of each triangle: given the three vertices of a triangle (x1, y1), (x2, y2), and (x3, y3), the horizontal and vertical coordinates of the centroid G are:

[0095]

[0096] The calculation formula of triangle area A is:

[0097]

[0098] Then the horizontal coordinate G of the centroid of each trianglex The area A is solved according to the formula to obtain the centroid ordinate of the output membership function:

[0099]

[0100] Among them, A i is the area of ​​the ith triangle, G x,i is the abscissa of the centroid of the ith triangle;

[0101] The calculation formula of the horizontal coordinate of the center of gravity of the output membership function is:

[0102]

[0103] In order to further improve the efficiency of screening triangular areas corresponding to rules in the defuzzification process, a multi-dimensional search method is adopted: since each triangular area is not only related to the input value, but may also involve multiple input conditions, the multi-dimensional input value can be used as the node content of the KD-Tree to achieve multi-dimensional search, and fast screening can be performed according to the range of the multi-dimensional input value during the search.

[0104] 3) After solving the defuzzification problem of the fuzzy algorithm, the PSO particle swarm optimization algorithm is used to optimize the membership function of the fuzzy logic algorithm.

[0105] 3.1) The shape of the membership function also has a great influence on the final optimization result. A good membership function shape can greatly speed up the convergence of the algorithm. The optimal membership function shape is selected through PSO optimization. The membership function shape usually uses three function shapes: triangular membership function, trapezoidal membership function and Gaussian membership function. Specifically:

[0106] The triangle membership function is:

[0107]

[0108] Among them, a, b, and c are the lower left endpoint, vertex, and lower right endpoint of the triangle membership function respectively.

[0109] The trapezoidal membership function is:

[0110]

[0111] Among them, a, b, c and d are the lower left endpoint, upper left vertex, upper right vertex and lower right endpoint of the trapezoidal membership function respectively.

[0112] The Gaussian membership function is:

[0113]

[0114] Where c is the center of the Gaussian function and σ is the standard deviation.

[0115] For the convenience of calculation, the center of the Gaussian function is defined as the vertex in the above membership function calculation result, and the point where μ(x)≤0.01 is defined as the bottom of the membership function.

[0116] Define the minimum control system performance index (MSE):

[0117]

[0118] Among them, yi is the expected output, is the system output, N is the number of samples, and θ is the shape parameter of the membership function.

[0119] The following are the specific steps for optimizing the selection:

[0120] Initialize particles. Each particle contains a type parameter (indicating the type of membership function: 1 = triangle, 2 = trapezoid, 3 = Gaussian). The type of particle and the corresponding parameters are generated by the random function rand. The parameters are used to describe the shape of the membership function. The initialization speed of the particle is randomly generated between -1 and 1.

[0121] Set the function evaluate_membership_function to calculate the fitness value of each particle. The fitness value reflects the quality of each particle's solution, that is, the performance of the particle in the current solution space; the lower the fitness value, the better the solution.

[0122] Update individual optimal solution: For each particle, compare the current fitness value with its historical optimal fitness value; if the current fitness value is better than the historical optimal, update the individual optimal solution of the particle; the individual optimal solution of the particle includes the optimal fitness value and the corresponding particle position.

[0123] Update the global optimal solution: among all particles, find the current optimal fitness value and determine whether this solution is better than the global optimal solution; if the currently found solution is better than the global optimal solution, update the global optimal solution; the global optimal solution is the best solution among all current particles, which includes the optimal fitness value and the corresponding particle parameters.

[0124] The number of iterations is set to 100, and the algorithm stops when the maximum number of iterations is reached. In each iteration of the algorithm, particles adjust their speed and position according to the individual optimal solution and the global optimal solution, and gradually converge to the global optimal solution.

[0125] 3.2) Particle swarm optimization fuzzy control has three key points, namely, the selection of PSO parameters, the determination of PSO objective function, and the co-simulation of PSO embedded in the fuzzy control strategy. The selection of PSO parameters in this invention mainly adjusts the top and bottom points of each membership function in the fuzzy logic algorithm, selects the required power P of the system, and req , Hybrid Energy Storage System SOC bat-sc And the output power P of the generator gen The membership function of is optimized. The following parameters are optimized respectively:

[0126] The system's required power P req The distance between the M function vertex and the H function vertex in the membership function set is a1 (a1 is also the distance between the M function vertex and the L function vertex), the distance between the H function and the VH function vertex is a2 (a2 is also the distance between the L function vertex and the VL function vertex), the horizontal coordinate distance between the M function vertex and the bottom point is a3 (the distances from the two bottom points of the M function to the vertex are equal), the horizontal coordinate distance between the L / H function vertex and the bottom point is a4 (the distances from the two bottom points to the vertex are equal), and the horizontal coordinate distance between the VL / VH function vertex and the bottom point is a5. See Figure 4 .

[0127] Hybrid Energy Storage System SOC bat-sc The distance b1 between the vertices of the M function and the L / H function in the membership function, the horizontal coordinate distance b2 between the vertex and the bottom point of the M function in the hybrid energy storage system residual state SOC membership function (the distances from the two bottom points to the vertex are equal), and the distance b3 between the bottom point of the M function and the bottom point of the L function in the hybrid energy storage system residual state SOC membership function (b3 is also the distance between the bottom point of the M function and the bottom point of the H function). See Figure 5 .

[0128] Generator output power P gen In the membership function, the distance between the M function vertex and the H function vertex is c1 (c1 is also the distance between the M function vertex and the L function vertex), the distance between the H function and the VH function vertex is c2 (c2 is also the distance between the L function vertex and the VL function vertex), the horizontal coordinate distance between the M function vertex and the bottom point is c3 (the distances from the two bottom points to the vertex are equal), and the horizontal coordinate distance between the H / L function vertex and the bottom point is c4 (the distances from the two bottom points to the vertex are equal). See Figure 6 .

[0129] Optimize the selected PSO parameters:

[0130] Input: System power requirement P req , the remaining state SOC of the hybrid energy storage system, the output power P of the generator gen ;

[0131] Output: a1-a5, b1-b3, c1-c4;

[0132] Initialize the particle position and velocity. First, for each particle (i represents the index of the particle), initialize its position x in each dimension. id and speed v id ,The initial position and velocity of the particle are randomly generated within the allowed range. In the main loop, the algorithm iterates continuously, updating the velocity and position of the particle,until the maximum number of iterations or the minimum error threshold is reached.

[0133] Calculate the fitness value. Calculate the current fitness value of each particle. The fitness value is used to evaluate the quality of the particle's position. Usually, the fitness value is related to the objective function. Update the historical optimal solution. If the current fitness value of the particle is better than its best fitness value in history, set the current fitness value to the particle's historical optimal value P. id,best .

[0134] For each dimension of each particle, iterative update is performed according to the formula:

[0135]

[0136] Among them, w is the inertia weight, c1 is the individual optimal acceleration factor, c2 is the global optimal acceleration factor, r1 and r2 are random numbers between [0, 1]. is the velocity vector of the particle in the dth dimension at the kth iteration; is the position vector of the particle in the dth dimension at the kth iteration; is the optimal position of the particle in the dth dimension in the kth iteration; is the historical optimal position of the particle in the dth dimension in k iterations.

[0137] ① In the PSO algorithm, inertia weight, individual optimal acceleration factor and global optimal acceleration factor are all key parameters, which jointly affect the convergence speed and global search ability of the algorithm. Therefore, when performing PSO optimization, it is very important to reasonably set and adjust these parameters. The present invention uses annealing algorithm to optimize the parameters in the PSO algorithm, which can improve the convergence speed of the algorithm and the quality of the solution. The specific method is introduced below:

[0138] First, a fitness function needs to be defined to evaluate the performance of the current parameter combination. The present invention uses the equivalent fuel consumption of the generator as the fitness function.

[0139] Initialize the parameter range, w is in the range of [0, 1], c1 and c2 are in the range of [0, 3]; Initial temperature: set a higher initial temperature of 100 to allow larger parameter changes; Cooling rate: set a cooling rate value of 0.95 to control the temperature reduction rate. After entering the iteration cycle, use the randomly generated new inertia weight, individual optimal acceleration factor and global optimal acceleration factor to run PSO and calculate the fitness; decide whether to accept the randomly generated new parameters based on the fitness change: if the new fitness is better, directly accept the new parameters; if the new fitness is worse, accept the new parameters with a certain probability (based on the current temperature and fitness difference), and the probability calculation formula is:

[0140]

[0141] Among them, f current is the current fitness, f new is the new fitness and T is the current temperature.

[0142] ② Most annealing algorithms will converge to the local optimal solution prematurely due to the local optimal problem. In order to eliminate the problem of local optimal solution at the root, this paper proposes a new memory mechanism and reverse search strategy, called "historical memory inversion perturbation", which is mainly achieved through the following process:

[0143] Memory of historical solutions: Record historical solutions in each iteration, including the location of the current solution, current temperature, and acceptance probability;

[0144] Introduce inversion perturbation: When the temperature drops to a certain level (determined based on the initial temperature and empirical values, for example, when the initial temperature is 100 and drops to 30, it is considered satisfied), reversely consider multiple historical solutions and generate new solutions in the form of inverse perturbation; specifically, the inversion perturbation can generate a new solution in a different direction from the historical solution based on the trajectory information of the historical solution;

[0145] Introduce the "historical solution spacing" threshold: Calculate the distance between the current solution and the historical solution. If the distance is less than the "historical solution spacing" threshold, force a larger reverse perturbation to try to break out of this cycle mode.

[0146] Multi-stage inversion: Different inversion strategies are introduced in different stages (early, middle and late stages). In the early and middle stages, the reverse perturbation of the current solution can be used, while in the late stage, the mean reverse perturbation of the historical trajectory can be used.

[0147] ③ In addition, the annealing algorithm has another disadvantage. Most simulated annealing algorithms will gradually reduce the disturbance amplitude according to the current temperature, but usually this disturbance is reduced uniformly or linearly. This uniform or linear reduction strategy weakens the algorithm's adaptive ability. Therefore, the present invention proposes a "nonlinear multi-stage dynamic disturbance strategy", the specific scheme is as follows:

[0148] Dynamic adjustment of disturbance amplitude: Divide the disturbance into multiple stages, and use different disturbance strategies in each stage: use larger random disturbances (for example, the disturbance temperature is 5% of the maximum temperature) in the early stage, use smaller target-oriented disturbances (for example, the disturbance temperature is 1% of the maximum temperature) in the middle stage, and introduce larger random disturbances (for example, the disturbance temperature is 5% of the maximum temperature) again in the later stage for global search. This method can avoid converging to the local optimal solution too quickly in the early stage, and can also jump out of the local optimal solution by introducing larger disturbances in the later stage.

[0149] The detailed steps are as follows:

[0150] Initialization settings:

[0151] Initialized to [0.7, 1.5, 1.5], corresponding to w, c1, and c2 of PSO respectively;

[0152] The initial temperature is 100;

[0153] The temperature decay rate is 0.95;

[0154] The maximum number of iterations is 100.

[0155] Specific operations at each iteration:

[0156] Parameter perturbation: When the number of iterations is a multiple of 10, the “historical memory inversion perturbation” strategy is used to generate a new solution: randomly select a historical solution, calculate the direction between the current solution and the historical solution, and if they are in the same direction, generate a new solution in the opposite direction. Otherwise, use the standard adaptive perturbation strategy: use the normrnd function to perform random perturbations on the current solution, and the perturbation amplitude is proportional to the current temperature.

[0157] Update the current solution and the global optimal solution:

[0158] Calculate the acceptance probability based on the temperature of the current annealing algorithm and the difference between the new solution and its target value;

[0159] If the new solution is better than the current solution, then accept the new solution; otherwise, decide whether to accept the new solution based on a certain probability;

[0160] If the objective value of the current solution is better than the global optimal solution, the global optimal solution is updated.

[0161] Dynamic temperature adjustment: Every 10 iterations, the frequency of accepting new solutions is calculated. If the frequency is too low (lower than the set empirical threshold), the temperature drop rate is slowed down; if the current solution is not closer to the target solution than the solution of the previous iteration in the next iteration, the temperature drop rate (i.e., cooling rate) is accelerated.

[0162] When the temperature is high, the probability of accepting a poor solution (the difference between the current solution and the solution of the previous iteration and the target value is large) is higher, thus conducting more extensive exploration;

[0163] When the temperature is low, the probability of accepting a poor solution is low, converging to a local area for fine search.

[0164] Temperature adjustment based on the number of consecutive non-improvements: If no better solution is found after 10 consecutive iterations, the cooling rate is accelerated to achieve faster convergence.

[0165] By combining the optimization of the dual strategies of historical memory inversion perturbation and nonlinear multi-stage dynamic perturbation, the convergence speed of the algorithm can be improved while maintaining the global search capability, effectively solving the problems of premature convergence and local optimality existing in the traditional simulated annealing algorithm.

[0166] (4) A new set of membership functions is obtained through hybrid PSO optimization, and then the new set of membership functions is added to the fuzzy logic controller; by running dual three-phase generators, the dual three-phase permanent magnet synchronous generator system has good fault tolerance and high power density, and has better performance in terms of bus voltage overshoot and recovery time. Combined with the hybrid energy storage system, it has a significant effect on bus voltage control, effectively improves bus voltage fluctuations, and improves system stability. The optimized energy management strategy concentrates more working points in the optimal energy consumption range of the generator, while reducing the working points in the high fuel consumption circle. The fuzzy control energy management strategy optimized by the PSO particle swarm optimization algorithm is effective in controlling fuel consumption and effectively improves fuel economy.

[0167] Example

[0168] Taking the 270V DC bus voltage of an aircraft as an example, at 0-0.1s, the aircraft is in the ground standby stage, and the system power required is about 14.98kW. The system power supply mainly depends on the generator. After optimization, the bus voltage is suppressed by the supercapacitor, and the overshoot is about 20%, which is about 30% lower than before optimization; at 0.1-0.2s, the aircraft is in the taxiing stage, and the power required by the system increases to about 51.03kW. The DC bus voltage fluctuates as follows: Figure 7As shown in the figure, after optimization, the bus voltage fluctuates by 4V and the recovery time is 15ms, which are 4% and 50% higher than before optimization, respectively. After the voltage recovers, it stabilizes within 270.2V±0.1V, which meets the normal working voltage requirements. At 0.2-0.3s, the aircraft is in the climbing stage, and the system power demand increases to about 87.48kW. The DC bus voltage fluctuates as shown in the figure. Figure 8 As shown in the figure, after optimization, the bus voltage fluctuates by 3V and the recovery time is 10ms, which are 2.7% and 40% higher than before optimization, respectively. After the voltage recovers, it stabilizes within 270.2V±0.1V, which meets the normal working voltage requirements. At 0.3-0.4s, the aircraft is in the cruising stage, and the power required by the system reaches the maximum, which is about 111.78kW. The DC bus voltage fluctuates as shown in the figure. Fig. 9 As shown in the figure, after optimization, the bus voltage fluctuates by 2V and the recovery time is 5ms, which are 1.12% and 66.7% higher than before optimization, respectively. After the voltage recovers, it stabilizes within 270.2V±0.1V, which meets the normal working voltage requirements. At 0.4s-0.5s, the aircraft is in the descending taxiing stage, and the power required by the system is reduced to 26.73kW. The DC bus voltage fluctuates as shown in the figure. Fig.10 As shown, after optimization, the bus voltage fluctuates by 0.5V and the recovery time is 3ms, which are respectively improved by 0.23% and 10% compared with before optimization. After the voltage recovers, it stabilizes within 270.1V±0.1V, which meets the normal working voltage requirements.

[0169] Refer to the flight mission of a certain aircraft, adjust the parameters, and fit the optimal working point curve, such as Fig.11 As shown in the figure, more working points are concentrated in the generator's optimal energy consumption range, while the working points in the high fuel consumption circle are reduced. The flight mission of a civil aircraft is about 3 hours, saving a total of about 5.81% of fuel consumption. The results fully verify that the fuzzy control energy management strategy optimized by the PSO particle swarm optimization algorithm is effective in controlling fuel consumption and effectively improves the economy of aircraft fuel.

[0170] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. The optimized fuzzy control energy management method of dual three-phase permanent magnet synchronous generator system is characterized by: Build a dual three-phase permanent magnet synchronous generator system, including a power generation system, a hybrid energy storage system and other loads, the power generation system, the hybrid energy storage system and other loads are all connected to the DC bus, the power generation system includes a dual three-phase generator and a rectifier module thereof, the hybrid energy storage system contains a lithium battery and a super capacitor, and the lithium battery and the super capacitor are connected in parallel; Get the output power P of the dual three-phase generator gen , the output of the hybrid energy storage system, the required power P of the dual three-phase permanent magnet synchronous generator system req and is coordinated through fuzzy logic algorithm. The output of the hybrid energy storage system includes the output power P of the lithium battery. bat 、The output power P of the supercapacitor sc , the fuzzy logic algorithm allocation coordination is specifically: When the output power of the dual three-phase generator P gen Greater than the system's required power P req When the charge state SOC of the hybrid energy storage system is determined bat-sc Is it greater than the limit discharge power SOC(L)? If it is less than SOC(L), the hybrid energy storage system stops supplying power. Otherwise, further determine SOC bat-sc Is it greater than the limited charging power SOC(H)? When it is greater than SOC(H), the hybrid energy storage system is prohibited from charging, where SOC(L)<SOC(H); When the output power of the dual three-phase generator P gen Less than the system's required power P req When the system power requirement P req , SOC bat-sc The fuzzy control unit of the fuzzy logic algorithm is input for defuzzification and the output power P of the dual three-phase generator is output. gen , output power P of lithium battery bat And the output power P of the supercapacitor sc , thereby adjusting the system power distribution so that the generator output meets the operating requirements.

2. The optimized fuzzy control energy management method according to claim 1, characterized in that: A triangle mapping method of fast screening rules is used for defuzzification processing. The triangle mapping method of fast screening rules uses an index structure to store and query the mapping relationship between the rules of the fuzzy logic algorithm and the triangle. During the defuzzification process, the centroid calculation is only performed on the triangle area corresponding to the screened rules.

3. The optimized fuzzy control energy management method according to claim 2, characterized in that: The index structure is as follows: a KD-Tree is established, which uses the input fuzzy value of each rule as the node coordinate and the rule itself as the node content; when there is an input fuzzy value, the KD-Tree is used to quickly locate all rules that may match the input fuzzy value, and the rules are screened according to the fit threshold: rules with a fit less than 5% are considered unmatched.

4. The optimized fuzzy control energy management method according to claim 3 is characterized in that: The construction process of the triangle mapping relationship is as follows: a mapping relationship is established between each rule and its output membership function, specifically, a reverse mapping relationship from rule to triangle is established, that is, each triangle corresponds to one or more rules for use in reverse query; according to the screened rule ID, the vertex coordinates and membership value of the triangle area corresponding to the rule are queried, and the weighted summation is performed using the center of gravity and area of ​​the corresponding triangle area to obtain the precise value of the defuzzification.

5. The optimized fuzzy control energy management method according to claim 4 is characterized in that: The multi-dimensional search method is used to screen the triangular area corresponding to the rule: the multi-dimensional input value is used as the node content of the KD-Tree to realize multi-dimensional search.

6. The optimized fuzzy control energy management method according to claim 2, characterized in that: After defuzzification, the PSO particle swarm optimization algorithm is used to optimize the membership function set in the fuzzy logic algorithm rules.

7. The optimized fuzzy control energy management method according to claim 6, characterized in that: Adjust the vertices and bases of each membership function in the fuzzy logic algorithm to select the required power P for the system req , Hybrid Energy Storage System SOC bat-sc And the output power P of the dual three-phase generator gen The parameters in the membership function are optimized.

8. The optimized fuzzy control energy management method according to claim 7, characterized in that: The annealing algorithm is used to optimize the inertia weight w, the individual optimal acceleration factor c1 and the global optimal acceleration factor c2 in the iterative update formula of the PSO particle swarm optimization algorithm. Specifically: Define a fitness function to evaluate the performance of the current parameter combination; Initialize the parameter range, w is in the range of [0, 1], c1 and c2 are in the range of [0, 3]; set the initial temperature to 100, and set the cooling rate value to 0.95; After entering the iteration cycle, the PSO is run using the randomly generated new inertia weight, individual optimal acceleration factor, and global optimal acceleration factor to calculate the fitness; whether to accept the randomly generated new parameters is determined based on the fitness change: if the new fitness is better, the new parameters are directly accepted; if the new fitness is worse, the new parameters are determined with probability P. Among them, f current is the current fitness, f new is the new fitness and T is the current temperature.

9. The optimized fuzzy control energy management method according to claim 8, characterized in that: The "historical memory inversion disturbance" strategy is used to prevent the annealing algorithm from being troubled by the local optimal solution. Specifically: Memory of historical solutions: Record historical solutions in each iteration, including the location of the current solution, the current temperature, and the acceptance probability; Introduce inversion perturbation: When the temperature drops to a certain level, multiple historical solutions are considered in reverse, and new solutions are generated in the form of reverse perturbation; the certain level is determined based on the initial temperature and empirical values; Introduce the "historical solution gap" threshold: calculate the distance between the current solution and the historical solution. If the distance is less than the "historical solution gap" threshold, a larger reverse perturbation is forced to be introduced to try to jump out of this cycle mode; Multi-stage inversion: different inversion strategies are introduced in the early, middle and late stages. The reverse perturbation of the current solution is used in the early and middle stages, and the mean reverse perturbation of the historical trajectory is used in the late stage.

10. The optimized fuzzy control energy management method according to claim 8, characterized in that: A nonlinear multi-stage dynamic perturbation strategy is used to improve the adaptive ability of the annealing algorithm. Specifically: Dynamic adjustment of disturbance amplitude: Divide the disturbance into multiple stages, and use different disturbance strategies in each stage: use larger random disturbance in the early stage, use smaller target-oriented disturbance in the middle stage, and introduce larger random disturbance again in the later stage for global search. The larger random disturbance, i.e., the disturbance temperature, is 5% of the maximum temperature, and the smaller target-oriented disturbance, i.e., the disturbance temperature, is 1% of the maximum temperature. Initialization settings: Initialize w, c1, c2 to [0.7, 1.5, 1.5], the initial temperature to 100, the temperature decay rate to 0.95, and the maximum number of iterations to 100; Specific operations at each iteration: Parameter perturbation: When the number of iterations is a multiple of 10, the "historical memory inversion perturbation" strategy is used to generate a new solution: randomly select a historical solution, calculate the direction between the current solution and the historical solution, and if they are in the same direction, generate a new solution in the opposite direction. Otherwise, use the standard adaptive perturbation strategy: use the normrnd function to perform random perturbations on the basis of the current solution, and the perturbation amplitude is proportional to the current temperature; Update the current solution and the global optimal solution: Calculate the acceptance probability based on the temperature of the current annealing algorithm and the difference between the new solution and its target value; if the new solution is better than the current solution, accept the new solution; otherwise, decide whether to accept the new solution based on the probability P; if the target value of the current solution is better than the global optimal solution, update the global optimal solution; Dynamic temperature adjustment: every 10 iterations, the frequency of accepting new solutions is calculated. If the frequency is lower than the set empirical threshold, the temperature drop rate is slowed down; if the current solution is not closer to the target solution than the solution of the previous iteration in the next iteration, the temperature drop rate is accelerated; if no better solution is found after 10 consecutive iterations, the temperature drop rate is accelerated to achieve rapid convergence.

Images