A multi-objective optimization method, device, equipment and medium

Abstract

The present invention discloses a multi-objective optimization method, device, equipment and medium. The method includes: obtaining a population and a preset optimization objective; initializing the population according to the preset optimization objective to obtain an initial solution set; obtaining a preset reference coefficient, and classifying the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set; training a preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain a first preferred offspring; and performing environmental selection according to the initial solution set and the first preferred offspring to obtain a target preferred solution set. The embodiment of the present invention introduces an expert knowledge and a multi-objective optimization method obtained by a generative adversarial network, effectively solving the multi-objective optimization problem. Especially when facing the multi-objective optimization problem in a hybrid power control system, it effectively realizes multi-objective optimization, making the control of the hybrid power control system more reasonable.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a multi-objective optimization method, device, equipment and medium. Background Art

[0002] In industrial production and life, many problems are composed of multiple conflicting and influencing objectives. People often encounter optimization problems of making multiple objectives as optimal as possible simultaneously in a given area, that is, multi-objective optimization problems. In the design of a hybrid control system, the main objective of the designer is to reduce fuel consumption. However, in addition to fuel consumption, factors such as driving experience, noise, vibration and harshness also need to be considered. Existing methods often only consider the hybrid control system with fuel consumption as a single objective, and do not consider the multi-objective optimization problem of the hybrid control system. Summary of the Invention

[0003] The present invention aims to at least solve one of the technical problems existing in the prior art. For this purpose, the present invention provides a multi-objective optimization method, device, equipment and medium, which can effectively solve the multi-objective optimization problem, especially when facing the multi-objective optimization problem in a hybrid control system, and effectively achieve multi-objective optimization.

[0004] The multi-objective optimization method according to the first aspect embodiment of the present invention includes: obtaining a population and a preset optimization objective;

[0005] Initializing the population according to the preset optimization objective to obtain an initial solution set;

[0006] Obtaining a preset reference coefficient, and classifying the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set;

[0007] Training a preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain a first preferred offspring;

[0008] Performing environmental selection according to the initial solution set and the first preferred offspring to obtain a target preferred solution set.

[0009] The multi-objective optimization method according to the embodiment of the present invention has at least the following beneficial effects: The multi-objective optimization method introduced by the present invention embodiment with expert knowledge and generative adversarial network effectively solves the multi-objective optimization problem. Especially when facing the multi-objective optimization problem in a hybrid control system, it effectively achieves multi-objective optimization, making the control of the hybrid control system more reasonable.

[0010] According to some embodiments of the present invention, it further includes:

[0011] Perform crossover and mutation operations on the target preference solution set to obtain an initial offspring generation;

[0012] Perform environmental selection based on the initial offspring generation and the target preference solution set to obtain a first updated target preference solution set.

[0013] According to some embodiments of the present invention, it further includes:

[0014] Obtain a second preference solution set and a second non-preference solution set according to the first updated target preference solution set and the preset reference coefficient;

[0015] Train the preset generative adversarial network according to the second preference solution set and the second non-preference solution set to obtain a second preference offspring generation;

[0016] Perform environmental selection based on the second preference offspring generation and the first updated target preference solution set to obtain a second updated target preference solution set.

[0017] According to some embodiments of the present invention, the initializing the population according to the preset optimization objective to obtain an initial solution set includes:

[0018] Obtain the constraint conditions of the preset optimization objective and the upper and lower bounds of the preset optimization objective according to the preset optimization objective;

[0019] Initialize the population according to the constraint conditions and the upper and lower bounds to obtain an initial solution set.

[0020] According to some embodiments of the present invention, the obtaining the preset reference coefficient and performing preference classification processing on the initial solution set according to the preset reference coefficient to obtain a first preference solution set and a first non-preference solution set includes:

[0021] Obtain a reference vector according to the preset reference coefficient;

[0022] Perform preference classification processing on the initial solution set according to the reference vector and a preset radius to obtain a first preference solution set and a first non-preference solution set.

[0023] According to some embodiments of the present invention, the performing environmental selection based on the initial solution set and the first preference offspring generation to obtain a target preference solution set includes:

[0024] Merge the initial solution set and the first preference offspring generation to obtain individuals to be processed;

[0025] Among the individuals to be processed, use the individuals whose distance relationship with the reference vector meets the preset conditions as dominant individuals;

[0026] Obtain a target preference solution set based on the obtained dominant individuals.

[0027] According to some embodiments of the present invention, among all the individuals to be processed, taking the individuals whose distance relationship with the reference vector meets a preset condition as dominant individuals includes:

[0028] Obtaining the distance between the individual to be processed and the reference vector, and selecting a preset number of individuals in ascending order of the distance as the dominant individuals.

[0029] A multi-objective optimization device according to an embodiment of the second aspect of the present invention includes:

[0030] An acquisition module: used to acquire a population and preset optimization objectives;

[0031] A first processing module: used to initialize the population according to the preset optimization objectives to obtain an initial solution set;

[0032] A second processing module: used to obtain a preset reference coefficient, and classify the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set;

[0033] A third processing module: used to train a preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain a first preferred offspring;

[0034] A target preference solution set generation module: used to perform environmental selection according to the initial solution set and the first preferred offspring to obtain a target preference solution set.

[0035] A computer device according to an embodiment of the third aspect of the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method according to any one of the embodiments of the first aspect of the present invention is implemented.

[0036] A storage medium according to an embodiment of the fourth aspect of the present invention is a computer-readable storage medium storing computer-executable instructions for executing the method according to any one of the embodiments of the first aspect of the present invention.

[0037] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described below in conjunction with the drawings and embodiments, where:

[0039] Figure 1 is a flowchart of a multi-objective optimization method provided by an embodiment of the present invention;

[0040] Figure 2 It is a flowchart of another multi - objective optimization method provided by an embodiment of the present invention;

[0041] Figure 3 It is a flowchart of another multi - objective optimization method provided by an embodiment of the present invention;

[0042] Figure 4 It is a flowchart of another multi - objective optimization method provided by an embodiment of the present invention;

[0043] Figure 5 It is a flowchart of another multi - objective optimization method provided by an embodiment of the present invention;

[0044] Figure 6 It is a schematic diagram of dividing the preference region according to the reference coefficient provided by Example 1 of the present invention;

[0045] Figure 7 It is a flowchart of the multi - objective optimization method provided by Example 2 of the present invention. Detailed implementation manners

[0046] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

[0047] In the description of the present invention, it should be understood that for the orientation description, such as the orientation or positional relationship indicated by up, down, front, back, left, right, etc., is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.

[0048] In the description of the present invention, the meaning of several is more than one, the meaning of multiple is more than two, greater than, less than, exceeding, etc. are understood as not including the present number, and above, below, within, etc. are understood as including the present number. If the first and the second are described only for the purpose of distinguishing technical features, they should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or the sequence relationship of the indicated technical features.

[0049] In the description of the present invention, unless otherwise clearly defined, words such as setting, installing, connecting, etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0050] In the description of the present invention, the description referring to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0051] In the related art, in the design of a hybrid control system, the main goal of the designer is to reduce fuel consumption. Now most of the research work focuses on reducing fuel consumption. In order to reduce energy consumption, some traditional and classical single-objective optimization methods such as the dynamic programming method have been adopted in the design of the power control model and have been widely used in practice. However, in addition to fuel consumption, the performance of the actual power control system also needs to consider the influence of other aspects. Such as driving experience, noise, vibration, and harshness, etc.

[0052] Based on this, the embodiments of the present invention propose a multi-objective optimization method, device, equipment, and medium. Considering the influence of other aspects on the performance of the power control system, the embodiments of the present invention introduce expert knowledge and a multi-objective optimization method obtained by a generative adversarial network, effectively solving the multi-objective optimization problem. Especially when facing problems such as noise problems, emission problems, operation times problems, and battery voltage problems in a hybrid control system, through the multi-objective optimization method of the embodiments of the present invention, the solution set obtained makes the operation times of the hybrid control system reduced, the battery life longer, the emissions less, and the noise smaller.

[0053] The embodiments of the present invention provide a multi-objective optimization method, device, equipment, and medium, which will be specifically described through the following embodiments. First, the multi-objective optimization method in the embodiments of the present invention will be described.

[0054] Referring to Figure 1 , the multi-objective optimization method according to the embodiments of the present invention includes, but is not limited to, steps S110 to S150.

[0055] Step S110, obtaining a population and a preset optimization target;

[0056] Step S120, initializing the population according to the preset optimization target to obtain an initial solution set;

[0057] Step S130, obtaining a preset reference coefficient, and classifying the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set;

[0058] Step S140: Train the preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain the first preferred offspring;

[0059] Step S150: Perform environmental selection based on the initial solution set and the first preferred offspring to obtain the target preferred solution set.

[0060] In step S110, the preset optimization objective can be set by itself. In the embodiments of the present invention, the preset optimization objective includes at least one of reducing the number of operations, having a longer battery life, emitting less, and producing less noise.

[0061] In step S120, initialize the population according to the preset optimization objective to obtain the initial solution set. For example, randomly initialize 100 individuals that meet the preset optimization objective.

[0062] It can be understood that the "individual" referred to herein can also be called a "solution", and a set of individuals is called a solution set.

[0063] In steps S130 to S140, generally, for a multi-objective optimization problem, an engineer will provide a reference coefficient according to experience or test experiments. Specifically, the reference coefficient divides the space into a preferred region and a non-preferred region. The individuals located in the preferred region form the first preferred solution set, and the individuals located in the non-preferred region form the first non-preferred solution set.

[0064] It can be understood that classifying the initial solution set into adversarial samples by introducing a preset reference coefficient is more conducive to the training of the generative adversarial network than the method of randomly generating adversarial samples according to the solution set in the prior art. Training the generative adversarial network based on the first preferred solution set and the first non-preferred solution set, compared with the prior art, the combination of the reference coefficient and the generative adversarial network can guide the initial solution set to produce and search for the preferred solution set in the preferred region, thereby improving the search efficiency and increasing the accuracy rate instead of blindly iteratively searching for the possible optimal solution in the entire target space, which helps to improve the global search ability and convergence speed of the algorithm.

[0065] In step S150, specifically, merge the initial solution set and the first preferred offspring, and excellent individuals can be selected therefrom to obtain the target preferred solution set. The selection method can be the roulette wheel method or the tournament method in the prior art, and the embodiments of the present invention do not make specific limitations.

[0066] It can be understood that through the above steps, the embodiments of the present invention obtain the target preference solution set based on the combination of the reference coefficient and the generative adversarial network. When the preset optimization objectives are the noise problem, emission problem, operation times problem, and battery voltage problem in the hybrid control system, the target preference solution set obtained by the multi-objective optimization method of the embodiments of the present invention can reduce the operation times of the hybrid control system, prolong the battery life, reduce emissions, and reduce noise.

[0067] It can be understood that the target solution set can be used as the initial solution set and input into the multi-objective optimization algorithm multiple times for offspring generation and environmental selection to obtain a more reasonable and accurate target preference solution set.

[0068] In some embodiments, referring to Figure 2 , according to the multi-objective optimization method of the embodiments of the present invention, it further includes but is not limited to steps S210 to S220.

[0069] Step S210, performing crossover and mutation processing on the target preference solution set to obtain the initial offspring;

[0070] Step S220, performing environmental selection according to the initial offspring and the target preference solution set to obtain the first updated target preference solution set.

[0071] Specifically, the initial offspring is obtained through the crossover and mutation operations of the evolutionary algorithm, thereby maintaining the diversity of the population in the algorithm to a certain extent and better avoiding the occurrence of premature phenomenon. The first updated target preference solution set is obtained according to the initial offspring and the target preference solution set, making the solution set obtained by the multi-objective optimization method more diverse.

[0072] It can be understood that the first updated target preference solution set obtained can also be updated through the following steps:

[0073] Obtaining the second preference solution set and the second non-preference solution set according to the first updated target preference solution set and the preset reference coefficient;

[0074] Training the preset generative adversarial network according to the second preference solution set and the second non-preference solution set to obtain the second preference offspring;

[0075] Performing environmental selection according to the second preference offspring and the first updated target preference solution set to obtain the second updated target preference solution set.

[0076] Through the above steps, the first updated target preference solution set obtained is updated again, and the obtained second updated target preference solution set more conforms to the preset optimization objective, that is, the obtained solution set is more accurate.

[0077] In some embodiments, referring to Figure 3, step S120 includes but is not limited to steps S310 to S320.

[0078] Step S310, obtaining the constraint conditions of the preset optimization objective and the upper and lower bounds of the preset optimization objective according to the preset optimization objective;

[0079] Step S320, initializing the population according to the constraint conditions and the upper and lower bounds to obtain an initial solution set.

[0080] Specifically, randomly initialize the population to generate individuals that satisfy the constraint conditions and the upper and lower bounds, and further obtain the initial solution set. So that the target preference solution set generated according to the initial solution set conforms to the corresponding optimization problem.

[0081] In some embodiments, referring to Figure 4 , step S130 includes but is not limited to steps S410 to S420.

[0082] Step S410, obtaining a reference vector according to the preset reference coefficient;

[0083] Step S420, performing preference classification processing on the initial solution set according to the reference vector and the preset radius to obtain a first preference solution set and a first non-preference solution set.

[0084] Specifically, there is a reference point corresponding to the preset reference coefficient, which can form a reference vector in space, and geometrically divide the target space into a preference region and a non-preference region. The region within the reference vector and the preset radius is the preference region, and other regions are non-preference regions. The individuals located in the preference region form the first preference solution set, and the individuals located in the non-preference region form the first non-preference solution set.

[0085] In some embodiments, referring to Figure 5 , further includes but is not limited to steps S510 to S530.

[0086] Step S510, merging the initial solution set and the first preference offspring to obtain the individuals to be processed;

[0087] Step S520, among the individuals to be processed, taking the individuals whose distance relationship with the reference vector meets the preset conditions as dominant individuals;

[0088] Step S530, obtaining the target preference solution set based on the obtained dominant individuals.

[0089] Specifically, the individuals within the preference region among the individuals to be processed are all preference individuals, but not all of them are dominant individuals. Only the individuals whose distance relationship with the reference vector meets the preset conditions can be used as dominant individuals.

[0090] Specifically, by the following steps, select the dominant individuals:

[0091] Among the individuals in the preference region, 100 individuals are randomly selected as dominant individuals.

[0092] Specifically, the dominant individuals can also be selected through the following steps:

[0093] Obtain the distances between the individuals to be processed and the reference vector, and select a preset number of individuals as dominant individuals in ascending order of the distances.

[0094] For example: Select 100 individuals as dominant individuals in ascending order of the distances between the individuals to be processed and the reference vector.

[0095] The multi-objective optimization method of the present invention will be illustrated below through two actual examples.

[0096] Example 1. Specifically, the design of the power control system of a hybrid vehicle usually needs to consider some parameters and rules. Specifically, the parameters are defined as shown in Table 1.

[0097] Table 1

[0098]

[0099] The specific rules for hybrid control are as follows:

[0100] Rule 1: If the speed is lower than Voff, turn off the internal combustion engine.

[0101] Rule 2: If the state of charge of the battery is greater than SOCmax, turn off (Note: SOCmax is not the maximum state of charge of the battery. Here, it only represents the state of charge level of the battery, and its state value makes the battery stop charging.).

[0102] Rule 3: If the state of charge of the battery is less than SOCmin, turn off the internal combustion engine.

[0103] Rule 4: If the internal combustion engine is in the on state and the speed is less than V1, perform Operation 1.

[0104] Rule 5: If the internal combustion engine is in the on state and the speed is between V1 and V2, perform Operation 2.

[0105] Rule 6: If the internal combustion engine is in the on state and the speed is greater than V2, perform Operation 3.

[0106] Note: Operations 1, 2, and 3 correspond to the torque values defined in Table 1 in the actual process.

[0107] The multi-objective optimization problem of the hybrid system is as follows:

[0108] Objective 1, battery stress (BS) battery voltage:

[0109] The battery life is crucial for the HEV control system, and the operation strategy must consider the impact of control actions on the battery. However, the battery life model has a certain complexity and is difficult to calibrate. In this example, under the simulation platform conditions, only the charge / discharge state BS current (t) and the battery temperature BS temp (t) are considered, as follows:

[0110] BS = ∑ t (BS current (t) + c T *BS temp (t)) (1)

[0111] BS current (t) = ΔI(t) 2 (2)

[0112] BS temp (t) = ΔT(t) 2 (3)

[0113] Among them, the battery state BS current (t) represents the battery deviation value at time t, and the temperature represents the temperature deviation value at time t, and their values are all within the safe operation range. c T represents the temperature standard intensity parameter value, and the current battery state value and temperature are both generated from the simulation software platform.

[0114] Objective 2, Operation changes operation count (OPC):

[0115] The opening and closing of the internal combustion engine can often be perceived by the driver or passengers, so it is inevitable to reduce the operation count in the control system. In this example, the operation count of the internal combustion engine switch within a certain time range is calculated to determine this objective, and the value of OPC is optimized and reduced.

[0116] OPC = ∑ k H(|t k - t t-1 | < 60s) (4)

[0117] Among them, k is the parameter recording the change of the internal combustion engine switch, and t k represents the occurrence of the k-th switch change event within t. When the boolean variable x is true if and only if, the value of the step equation H(x) is 1.

[0118] Objective 3, Emission emissions:

[0119]

[0120] Emission = ∑ t ΘICE (t)*S cat (t) (6)

[0121] Since the emissions of other gases do not reach the operating temperature conditions, this target only considers the emissions of CO2 that are most closely related to fuel consumption. S cat (t) represents the state of the catalytic exhaust purification system at time t, T cat (t) is its temperature at time t, The maximum temperature of the exhaust system during a limited operation.

[0122] Objective 4, Noise:

[0123] Noise = ∑ t Θ Noise (t)*(N ICE (t) - N Rolling (t)) 2 (7)

[0124] The generation of vehicle noise is affected by many factors. The two main sources are rolling noise and the noise generated by the ICE. However, the former is unavoidable during driving at a certain speed, and the latter is mainly affected by the control strategy. The noise target is modeled using the measurement data of the above two noises, where this target is valid only when the ICE noise exceeds the rolling noise.

[0125] Population initialization: The population is initialized according to the constraint conditions and upper and lower bounds of the above objectives, and an initial solution set is obtained. Specifically, 100 individuals that satisfy the constraint conditions and upper and lower bounds are randomly initialized.

[0126] Offspring generation based on the generative adversarial network includes:

[0127] A. Classify the population: According to the reference coefficients provided by experts or engineers, a reference vector is formed and the target space is divided into a preferred region and a non-preferred region in a geometric sense, as Figure 6 shown; the initial solution set is divided into preferred individuals and non-preferred individuals. Specifically, the 50 individuals with the shortest distance to the reference vector are preferred individuals, and the others are non-preferred individuals.

[0128] B. Train the generative adversarial network (GAN): Specifically, in the GAN, the preferred individual sub-population is regarded as the real sample, and the sub-population in the non-preferred region is regarded as the fake sample, that is, the two classified sub-populations are used to train the generative adversarial network.

[0129] C. Offspring generation: Use the Generator in the trained GAN model to generate offspring individuals.

[0130] Environmental selection: Specifically, the generated offspring and parents are combined, and 100 excellent individuals are selected from them. An excellent individual is defined as the top 100 individuals sorted by the smallest distance from the individual to the reference vector.

[0131] Determine whether the termination condition is satisfied: Specifically, the termination condition is the number of population cycle iterations.

[0132] Output the target preference solution set.

[0133] Example 2, referring to Figure 7 Then in Example 2, the method includes:

[0134] Population initialization: Specifically, 100 individuals that satisfy the constraint conditions and upper and lower bounds are randomly initialized, that is, the initial solution set.

[0135] Determine whether the current population cycle iteration count satisfies the preset condition: Specifically, the preset condition is that the current population cycle iteration count is 1, 5, 10, 15 until the cycle ends, and the total population cycle iteration count can be 100.

[0136] When the preset condition is not satisfied, the initial solution set is subjected to crossover and mutation processing to obtain initial offspring and update the target preference solution set according to the initial offspring.

[0137] When the preset condition is satisfied, offspring generation based on the generative adversarial network is performed to obtain preference offspring and update the target preference solution set according to the preference offspring.

[0138] When the termination condition is satisfied, the output target preference solution set is the optimal preference solution set. Specifically, the termination condition is the total number of population cycle iterations.

[0139] An embodiment of the present invention provides a multi-objective optimization device, which includes but is not limited to:

[0140] Acquisition module: used to acquire the population and preset optimization objectives;

[0141] First processing module: used to initialize the population according to the preset optimization objective to obtain the initial solution set;

[0142] Second processing module: used to obtain the preset reference coefficient, and classify the initial solution set according to the preset reference coefficient to obtain the first preference solution set and the first non-preference solution set;

[0143] Third processing module: used to train the preset generative adversarial network according to the first preference solution set and the first non-preference solution set to obtain the first preference offspring;

[0144] Target preference solution set generation module: used to perform environmental selection according to the initial solution set and the first preference offspring to obtain the target preference solution set.

[0145] The device according to the embodiment of the present invention obtains a target preference solution set based on the combination of a reference coefficient and a generative adversarial network. If the preset optimization objectives are noise problems, emission problems, operation times problems, and battery voltage problems in a hybrid control system, through the multi-objective optimization method according to the embodiment of the present invention, the obtained target preference solution set can reduce the operation times of the hybrid control system, make the battery life longer, reduce emissions, and reduce noise.

[0146] In one embodiment, an acquisition module is configured to acquire a population and a preset optimization objective, where the preset optimization objective can be set by itself. In the embodiment of the present invention, the preset optimization objective includes at least one of reducing operation times, making the battery life longer, reducing emissions, and reducing noise.

[0147] The first processing module initializes the population according to the preset optimization objective to obtain an initial solution set. For example, 100 individuals that meet the preset optimization objective are randomly initialized. It can be understood that the "individuals" referred to herein can also be called "solutions", and a set of individuals is called a solution set.

[0148] The second processing module is configured to obtain a preset reference coefficient. Generally, for a multi-objective optimization problem, an engineer will provide a reference coefficient based on experience or test experiments. Specifically, the reference coefficient divides the space into a preference region and a non-preference region. Individuals located in the preference region form a first preference solution set, and individuals located in the non-preference region form a first non-preference solution set.

[0149] The third processing module is configured to train a preset generative adversarial network according to the first preference solution set and the first non-preference solution set to obtain a first preference offspring. It can be understood that classifying the initial solution set into adversarial samples by introducing a preset reference coefficient is more conducive to the training of the generative adversarial network than the method of randomly generating adversarial samples according to the solution set in the prior art. Training the generative adversarial network based on the first preference solution set and the first non-preference solution set can, compared with the prior art, guide the initial solution set to produce and search for the preference solution set in the preference region by combining the reference coefficient and the generative adversarial network, thereby improving the search efficiency and increasing the accuracy, rather than blindly iteratively searching for the possible optimal solution in the entire target space, which helps to improve the global search ability and convergence speed of the algorithm.

[0150] In the target preference solution set generation module, specifically, the initial solution set and the first preference offspring are combined, and excellent individuals can be selected from them to obtain the target preference solution set. The selection method can be the roulette method or the tournament method in the prior art, and the embodiment of the present invention does not make specific limitations.

[0151] It can be understood that through the above steps, the embodiments of the present invention obtain the target preference solution set based on the combination of the reference coefficient and the generative adversarial network. When the preset optimization objectives are noise problems, emission problems, operation times problems, and battery voltage problems in the hybrid control system, the target preference solution set obtained through the multi-objective optimization method of the embodiments of the present invention can reduce the operation times of the hybrid control system, prolong the battery life, reduce emissions, and reduce noise.

[0152] It can be understood that the target solution set can be used as the initial solution set and input into the multi-objective optimization algorithm multiple times for offspring generation and environmental selection to obtain a more reasonable and accurate target preference solution set.

[0153] Among them, the specific implementation steps of a multi-objective optimization device refer to the above multi-objective optimization method and will not be elaborated here.

[0154] The embodiments of the present invention also provide a computer device, including: at least one processor, and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the methods in any of the above method embodiments.

[0155] In addition, an embodiment of the present invention also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by one or more control processors, and the one or more control processors execute the methods in the above method embodiments, for example, execute the method steps S110 to S150 described above Figure 1 in, Figure 2 the method steps S210 to S220 in, Figure 3 the method steps S310 to S320 in, Figure 4 the method steps S410 to S420 in, Figure 5 the method steps S510 to S530 in.

[0156] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network nodes. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0157] Those of ordinary skill in the art will understand that all or some of the steps and systems disclosed above in the methods can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or can be implemented as hardware, or can be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer-readable storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer-readable storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. The computer-readable storage medium includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0158] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the relevant technical field, various changes can be made without departing from the spirit of the present invention. In addition, the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

Claims

1. A multi-objective optimization method for a hybrid vehicle, characterized in that Including: Obtaining a population and a preset optimization objective; wherein, the population is a solution set of a power control system of a hybrid vehicle, and the preset optimization objective includes at least one of longer battery life, fewer operation times, less emissions, and less noise; Initializing the population according to the preset optimization objective to obtain an initial solution set; Obtaining a preset reference coefficient, and classifying the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set; Training a preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain a first preferred offspring; Performing environmental selection according to the initial solution set and the first preferred offspring to obtain a target preferred solution set; wherein, the target preferred solution set is used to represent the design parameters and hybrid control rules of the power control system.

2. The method according to claim 1, characterized in that Also including: Performing crossover and mutation processing on the target preferred solution set to obtain an initial offspring; Performing environmental selection according to the initial offspring and the target preferred solution set to obtain a first updated target preferred solution set.

3. The method according to claim 2, wherein, Also including: Obtaining a second preferred solution set and a second non-preferred solution set according to the first updated target preferred solution set and the preset reference coefficient; Training the preset generative adversarial network according to the second preferred solution set and the second non-preferred solution set to obtain a second preferred offspring; Performing environmental selection according to the second preferred offspring and the first updated target preferred solution set to obtain a second updated target preferred solution set.

4. The method according to claim 1, characterized in that, The initializing the population according to the preset optimization objective to obtain an initial solution set includes: Obtaining the constraint conditions of the preset optimization objective and the upper and lower bounds of the preset optimization objective according to the preset optimization objective; Initializing the population according to the constraint conditions and the upper and lower bounds to obtain an initial solution set.

5. The method according to claim 1, wherein The obtaining a preset reference coefficient, and classifying the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set includes: Obtaining a reference vector according to the preset reference coefficient; Classifying the initial solution set according to the reference vector and a preset radius to obtain a first preferred solution set and a first non-preferred solution set.

6. The method according to claim 5, characterized in that, The performing environmental selection according to the initial solution set and the first preferred offspring to obtain a target preferred solution set includes: Merging the initial solution set and the first preferred offspring to obtain individuals to be processed; Among the individuals to be processed, taking the individuals whose distance relationship with the reference vector meets the preset conditions as dominant individuals; Obtaining a target preferred solution set based on the obtained dominant individuals.

7. The method according to claim 6, characterized in that, The taking the individuals whose distance relationship with the reference vector meets the preset conditions as dominant individuals among all the individuals to be processed includes: Obtaining the distance between the individuals to be processed and the reference vector, and selecting a preset number of individuals in ascending order of the distance as the dominant individuals.

8. A multi-objective optimization device for a hybrid vehicle, characterized in that, Including: Acquisition module: used to acquire a population and a preset optimization objective; wherein, the population is a solution set of the power control system of a hybrid vehicle, and the preset optimization objective includes at least one of longer battery life, fewer operation times, less emissions, and less noise; First processing module: used to initialize the population according to the preset optimization objective to obtain an initial solution set; Second processing module: used to obtain a preset reference coefficient, and classify the initial solution set according to the preset reference coefficient to obtain a first preferred solution set and a first non-preferred solution set; Third processing module: used to train a preset generative adversarial network according to the first preferred solution set and the first non-preferred solution set to obtain a first preferred offspring; Target preferred solution set generation module: used to perform environmental selection according to the initial solution set and the first preferred offspring to obtain a target preferred solution set; wherein, the target preferred solution set is used to represent the design parameters and hybrid control rules of the power control system.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.

10. A storage medium, the storage medium being a computer-readable storage medium, characterized in that, Stores computer-executable instructions for executing the method according to any one of claims 1 to 7.

Images