WO2024119654A1 - 同步发电机输出电流的预测方法、装置、设备和存储介质 - Google Patents

同步发电机输出电流的预测方法、装置、设备和存储介质 Download PDF

Info

Publication number
WO2024119654A1
WO2024119654A1 PCT/CN2023/081309 CN2023081309W WO2024119654A1 WO 2024119654 A1 WO2024119654 A1 WO 2024119654A1 CN 2023081309 W CN2023081309 W CN 2023081309W WO 2024119654 A1 WO2024119654 A1 WO 2024119654A1
Authority
WO
WIPO (PCT)
Prior art keywords
initial
synchronous generator
axis current
model
predicted
Prior art date
Application number
PCT/CN2023/081309
Other languages
English (en)
French (fr)
Inventor
李鹏
黄文琦
侯佳萱
曹尚
戴珍
梁凌宇
白昱阳
张焕明
Original Assignee
南方电网数字电网研究院有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南方电网数字电网研究院有限公司 filed Critical 南方电网数字电网研究院有限公司
Publication of WO2024119654A1 publication Critical patent/WO2024119654A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • G06F17/13—Differential equations

Definitions

  • the present application relates to the field of power grid technology, and in particular to a method, device, equipment and storage medium for predicting output current of a synchronous generator.
  • the analysis of generators in the power grid is generally achieved by establishing an analysis model based on a neural network algorithm.
  • the accuracy of the neural network algorithm is limited by the sample size. If the sample size is insufficient, the accuracy rate is likely to be low.
  • the present application provides a method for predicting output current of a synchronous generator.
  • the method comprises:
  • the working state data includes d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator;
  • the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model
  • the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the processing of the operating state data by the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current includes:
  • the rotor operation equation group and the generator winding flux differential equation group are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • solving the rotor operation equations and the generator winding flux differential equations to obtain the initial predicted d-axis current and the initial predicted q-axis current includes:
  • the discretized rotor operation equations and the generator winding flux differential equations are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the neural network model includes a cascaded first long short-term memory network model, a first fully connected layer, a Dropout layer and a second fully connected layer.
  • the method before inputting the working status data into a preset hybrid drive model, the method further includes:
  • the initial hybrid drive model includes a cascaded initial synchronous generator physical model and an initial neural network model
  • the initial hybrid driving model is subjected to first-stage model training and second-stage model training to obtain the preset hybrid driving model;
  • the first stage model training includes: fixing the parameters of the initial synchronous generator physical model unchanged, training the initial neural network model using the training sample set until the error of the initial neural network model is less than a first threshold, so as to obtain a candidate neural network model;
  • the second stage training process includes: using the training sample set to jointly train the initial synchronous generator physical model and the candidate neural network model until the error of the initial hybrid drive model is less than a second threshold, wherein the second threshold is less than the first threshold.
  • the using the training sample set to jointly train the initial synchronous generator physical model and the candidate neural network model includes:
  • the initial synchronous generator physical model is trained using the training sample set at a first learning rate, and the candidate neural network model is trained at a second learning rate.
  • obtaining a training sample set includes:
  • the working data of the generator when it is disturbed is collected at a preset frequency as the training sample.
  • the present application also provides a device for predicting output current of a synchronous generator.
  • the device comprises:
  • a working state data acquisition module used to acquire the current working state data of the synchronous generator, wherein the working state data includes a d-axis voltage, a q-axis voltage, an excitation voltage, and a mechanical power input to the synchronous generator;
  • a prediction module is used to input the working status data into a preset hybrid drive model, wherein the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model; process the working status data through the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current; and correct the initial predicted d-axis current and the initial predicted q-axis current through the neural network model to obtain a predicted d-axis current and a predicted q-axis current of the synchronous generator.
  • the present application further provides a computer device.
  • the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
  • the working state data includes d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator;
  • the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model
  • the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the present application further provides a computer-readable storage medium.
  • the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • the working state data includes the d-axis voltage, the q-axis voltage, the excitation voltage and the mechanical power input to the synchronous generator;
  • the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model
  • the working state data is processed by the synchronous generator physical model to obtain an initial Predicted d-axis current and initial predicted q-axis current;
  • the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the present application further provides a computer program product.
  • the computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • the working state data includes d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator;
  • the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model
  • the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the above-mentioned prediction method, device, computer equipment, storage medium and computer program product of the output current of the synchronous generator are obtained by obtaining the current working state data of the synchronous generator, the working state data including the d-axis voltage, the q-axis voltage, the excitation voltage and the mechanical power input to the synchronous generator; the working state data is input into a preset hybrid drive model, the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model; the working state data is processed by the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current; the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the present application obtains a prediction result with high accuracy by first predicting the working state data of the generator through the synchronous generator physical model, obtaining the initial predicted d-axis current and the initial predicted q-axis current, and then correcting them through the neural network model, thereby improving the accuracy of the prediction result.
  • the initial predicted d-axis current and the initial predicted q-axis current obtained are very close to the actual d-axis current and q-axis current, resulting in a significant reduction in the number of samples required during the training of the neural network model.
  • FIG1 is a diagram showing an application environment of a method for predicting output current of a synchronous generator in one embodiment
  • FIG2 is a schematic flow chart of a method for predicting output current of a synchronous generator in one embodiment
  • FIG3 is a structural block diagram of a device for predicting output current of a synchronous generator in one embodiment
  • FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment.
  • the method for predicting the output current of a synchronous generator can be applied in an application environment as shown in FIG. 1 .
  • the synchronous generator 102 communicates with the computer device 104 through a network.
  • the data storage system can store data that the computer device 104 needs to process.
  • the data storage system can be integrated on the computer device 104, or it can be placed on a cloud or other network device.
  • the computer device 104 can obtain the working status data of the synchronous generator 102 and the nameplate value of the synchronous generator 102, and the nameplate value includes various parameters of the synchronous generator 102.
  • the computer device 104 can scan the nameplate of the synchronous generator 102 or the user inputs through an input device, so that the computer device 104 obtains the nameplate value of the synchronous generator 102.
  • the computer device 104 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc.
  • Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc.
  • a method for predicting the output current of a synchronous generator is provided, which is described by taking the method applied to the computer device in FIG1 as an example, and includes the following steps:
  • Step 210 obtaining current working state data of the synchronous generator, wherein the working state data includes d-axis voltage, q-axis voltage, excitation voltage, and mechanical power input to the synchronous generator;
  • the computer device can obtain the current (time point) working state data of the synchronous generator, and the working state data includes d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator. Specifically, the computer device can directly or indirectly obtain the current (time point) working state data of the synchronous generator through various measuring instruments.
  • Step 220 input the working state data into a preset hybrid drive model.
  • the model includes a cascaded synchronous generator physical model and a neural network model;
  • Step 230 processing the working state data by using the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current;
  • Step 240 Correct the initial predicted d-axis current and the initial predicted q-axis current by using the neural network model to obtain a predicted d-axis current and a predicted q-axis current of the synchronous generator.
  • the working status data is input into a preset hybrid drive model, which includes a cascaded synchronous generator physical model and a neural network model.
  • a preset hybrid drive model which includes a cascaded synchronous generator physical model and a neural network model.
  • the d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator are input into the synchronous generator physical model, and the input d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator are processed by the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current.
  • the physical model of the synchronous generator includes a rotor motion equation group and a synchronous generator winding flux differential equation group, wherein the rotor motion equation group is as follows:
  • ⁇ is the generator power angle
  • ⁇ represents the motor speed
  • ⁇ s is the synchronous speed of the power system
  • Tj is the inertia constant of the generator
  • Pe represents the electromagnetic power of the generator set
  • Pm represents the mechanical power of the generator set.
  • T′ represents the open-circuit transient time constant of each winding of the synchronous generator
  • T′′ represents the open-circuit sub-transient time constant of each winding
  • E′ is the transient electromotive force
  • E′′ is the sub-transient electromotive force
  • Vf represents the synchronous generator terminal excitation voltage
  • I represents the synchronous generator stator current
  • X′ and X′′ are the synchronous motor transient and sub-transient conditions, respectively.
  • reactance Xl represents the motor leakage reactance related to saturation
  • Std represents the flux saturation degree of the d-axis synchronous generator
  • Stq represents the flux saturation degree of the q-axis synchronous generator.
  • the neural network is a pre-trained neural network model.
  • the initial predicted d-axis current and the initial predicted q-axis current, or the predicted d-axis current and q-axis current, are corrected by a neural network.
  • the neural network model includes a cascaded first long short-term memory network model, a first fully connected layer, a Dropout layer and a second fully connected layer.
  • the initial predicted d-axis current and the initial predicted q-axis current pass through the first long short-term memory network model, the first fully connected layer, the Dropout layer and the second fully connected layer in sequence, that is, the initial predicted d-axis current and the initial predicted q-axis current pass through the first long short-term memory network model, the output result of the first long short-term memory network model is input into the first fully connected layer, the output result of the first fully connected layer is input into the Dropout layer, the output result of the Dropout layer is input into the second fully connected layer, and finally the corrected predicted d-axis current and q-axis current are obtained.
  • the neural network model includes a cascaded multi-layer first long short-term memory network model, a first fully connected layer, a Dropout layer and a second fully connected layer.
  • the initial predicted d-axis current and the initial predicted q-axis current pass through multiple layers of the first long short-term memory network model (exemplarily, the first long short-term memory network model can be 3 layers), the first fully connected layer, the Dropout layer, and the second fully connected layer in sequence, that is, the initial predicted d-axis current and the initial predicted q-axis current pass through each layer of the first long short-term memory network model, and the output result of the last layer of the first long short-term memory network model is input into the first fully connected layer, the output result of the first fully connected layer is input into the Dropout layer, and the output result of the Dropout layer is input into the second fully connected layer, and finally the corrected predicted d-axis current and q-axis current are obtained.
  • the first long short-term memory network model can be 3 layers
  • the computer device can obtain the working state data of the synchronous generator in real time or at a fixed time, and input the working state data obtained each time into the preset hybrid drive model, so as to continuously obtain the working state data of the synchronous generator.
  • the above-mentioned method for predicting the output current of the synchronous generator is by obtaining the working state data of the synchronous generator at the current time point, the working state data including the d-axis voltage, the q-axis voltage, the excitation voltage and the mechanical power input to the synchronous generator; inputting the working state data into a preset hybrid drive model, the preset hybrid drive model including a cascaded synchronous generator physical model and a neural network model; processing the working state data through the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current; correcting the initial predicted d-axis current and the initial predicted q-axis current through the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the present application obtains a prediction result with high accuracy by first predicting the working state data of the generator through the synchronous generator physical model to obtain the initial predicted d-axis current and the initial predicted q-axis current, and then correcting them through the neural network model, thereby improving the accuracy of the prediction result.
  • the step of processing the operating state data by using the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current comprises:
  • the rotor operation equation group and the generator winding flux differential equation group are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the process of processing the working state data by using the synchronous generator physical model to obtain the initial predicted d-axis current and the initial predicted q-axis current may include:
  • the d-axis voltage, q-axis voltage, excitation voltage and mechanical power are input into the rotor operation equation group and the generator winding flux differential equation group in the physical model of the synchronous generator, and the rotor operation equation group and the generator winding flux differential equation group are as shown in the previous embodiment.
  • the solution can be performed to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • solving the rotor operation equation group and the generator winding flux differential equation group to obtain the initial predicted d-axis current and the initial predicted q-axis current includes:
  • the discretized rotor operation equations and the generator winding flux differential equations are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the implicit trapezoidal integration method is used for discretization to ensure the stability of the model under the sampling frequency of the synchronized phasor measurement unit (PMU).
  • PMU synchronized phasor measurement unit
  • ⁇ t represents the time step
  • k represents the kth time point
  • u represents the input variable, such as the excitation voltage, etc.
  • p represents the set of dynamic parameters of the synchronous generator
  • f represents the differential equation group of the electromagnetic transient state of the synchronous motor
  • h represents all output algebraic equations.
  • the discretization process can refer to the prior art.
  • Fj is the jth equation among l discrete equations
  • xi is the ith state quantity among m state quantities
  • x0 is the initial value at time t0 .
  • the rotor operation equations and the generator winding flux differential equations can also be solved directly by using the above solution equations.
  • the method before inputting the working status data into a preset hybrid drive model, the method further includes:
  • the initial hybrid drive model includes a cascaded initial synchronous generator physical model and an initial neural network model
  • the initial hybrid driving model is trained in the first stage and the second stage based on the training sample set.
  • the first stage model training includes: fixing the parameters of the initial synchronous generator physical model unchanged, training the initial neural network model using the training sample set until the error of the initial neural network model is less than a first threshold, so as to obtain a candidate neural network model;
  • the second stage training process includes: using the training sample set to jointly train the initial synchronous generator physical model and the candidate neural network model until the error of the initial hybrid drive model is less than a second threshold, wherein the second threshold is less than the first threshold.
  • the computer device can collect the working data of the generator when it is disturbed at a preset frequency as the training sample.
  • the preset frequency can be any frequency between 50 and 100 Hz.
  • the working state data (i.e., variables) of the synchronous generator are collected, including the d-axis voltage, q-axis voltage, excitation voltage (or excitation current), and generator input mechanical power, as well as the d-axis current and q-axis current actually output by the synchronous generator.
  • sample data are obtained, each of which includes: d-axis voltage, q-axis voltage, excitation voltage (or excitation current), and generator input mechanical power.
  • the parameter set of the synchronous generator physical model is obtained by obtaining the nameplate values of the generator.
  • the synchronous generator physical model is initialized according to the parameter set of the synchronous generator physical model, and the neural network model is initialized, and then the initialized synchronous generator physical model and the initialized neural network model are spliced to obtain an initial hybrid drive model.
  • the initial hybrid drive model is subjected to the first stage model training and the second stage model training based on the training sample set to obtain the preset hybrid drive model.
  • the first stage model training includes: fixing the parameters of the initial synchronous generator physical model unchanged, training the initial neural network model using the training sample set, that is, inputting the training sample into the initial synchronous generator physical model, obtaining the initial predicted d-axis current and the initial predicted q-axis current, and then using the initial predicted d-axis current and the initial predicted q-axis current as the input of the initial neural network model to realize the training of the initial neural network model.
  • the neural network model finally obtained is used as the candidate neural network model.
  • the initial neural network model includes a cascaded first initial long short-term memory network model, a first initial fully connected layer, an initial Dropout layer, and a second initial fully connected layer.
  • the initial predicted d-axis current and the initial predicted q-axis current pass through the first initial long short-term memory network model, the first initial fully connected layer, the initial Dropout layer and the second initial fully connected layer in sequence, that is, the initial predicted d-axis current and the initial predicted q-axis current pass through the first initial long short-term memory network model, the output result of the first initial long short-term memory network model is input into the first initial fully connected layer, the output result of the first initial fully connected layer is input into the initial Dropout layer, the output result of the initial Dropout layer is input into the second initial fully connected layer, and finally the corrected predicted d-axis current and q-axis current are obtained.
  • the neural network model includes a cascaded multi-layer first initial long short-term memory network model, a first initial fully connected layer, an initial Dropout layer and a second initial fully connected layer.
  • the initial predicted d-axis current and the initial predicted q-axis current pass through multiple layers of the first initial long short-term memory network model (exemplarily, the first initial long short-term memory network model can be 3 layers), the first initial fully connected layer, the initial Dropout layer, and the second initial fully connected layer in sequence, that is, the initial predicted d-axis current and the initial predicted q-axis current pass through each layer of the first initial long short-term memory network model, and the output result of the last layer of the first initial long short-term memory network model is input into the first initial fully connected layer, the output result of the first initial fully connected layer is input into the initial Dropout layer, and the output result of the initial Dropout layer is input into the second initial fully connected layer, and finally the corrected predicted d-axis current and q-axis current are obtained.
  • the first initial long short-term memory network model exemplarily, the first initial long short-term memory network model can be 3 layers
  • the first initial fully connected layer exemplarily, the initial Dropout layer, and the second
  • the error between the corrected predicted d-axis current and q-axis current and the d-axis current and q-axis current actually output by the synchronous generator in the training sample is calculated. If the error is greater than or equal to the first error, the parameters in the initial neural network model are updated, and the initial predicted d-axis current and the initial predicted q-axis current are used as inputs of the initial neural network model again. This process is repeated until the error of the initial neural network model is less than the first error. At this time, the first stage of training is terminated, and the initial neural network model with an error less than the first error is used as a candidate neural network model.
  • the second stage of training includes: using the training sample set to jointly train the initial synchronous generator physical model and the candidate neural network model until the error of the initial hybrid drive model is less than the second threshold, wherein the second threshold is less than the first threshold. That is, the parameters in the initial synchronous generator physical model and the candidate neural network model are updated simultaneously, the above sequence process is repeated, and the error between the output result and the d-axis current and q-axis current calculation of the synchronous generator actual output is calculated until the error is less than the second threshold. In this way, a trained hybrid drive model is obtained.
  • the training sample set is used to jointly train the initial synchronous generator physical model and the candidate neural network models include:
  • the initial synchronous generator physical model is trained using the training sample set at a first learning rate, and the candidate neural network model is trained at a second learning rate.
  • the first learning rate and the second learning rate can be the same or different.
  • the parameters of the initial synchronous generator physical model are updated by the first learning rate to train the initial synchronous generator physical model, and the parameters of the candidate neural network model are updated by the second learning rate to train the candidate neural network model.
  • the hybrid drive model trained in the above manner can maintain physically consistent results (i.e., high accuracy, close to 100%), while narrowing the parameter search range (in the first stage, the parameter search range of the initial neural network model is narrowed, and in the second stage, the parameter search range of the initial synchronous generator physical model and the candidate neural network model is narrowed), which can improve the training efficiency. Since the output result of the synchronous generator physical model is relatively high, the dependence on data in the entire training process is relatively small, that is, only a small number of training samples are needed to complete the training.
  • steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
  • the embodiment of the present application also provides a device for predicting the output current of a synchronous generator for implementing the method for predicting the output current of a synchronous generator involved above.
  • the implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more devices for predicting the output current of a synchronous generator provided below can refer to the limitations of the method for predicting the output current of a synchronous generator above, and will not be repeated here.
  • a device for predicting output current of a synchronous generator comprising:
  • the working state data acquisition module 310 is used to obtain the working state data of the synchronous generator at the current time point.
  • the working state data includes d-axis voltage, q-axis voltage, excitation voltage and mechanical power input to the synchronous generator;
  • the prediction module 320 is used to input the working status data into a preset hybrid drive model, wherein the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model; process the working status data through the synchronous generator physical model to obtain an initial predicted d-axis current and an initial predicted q-axis current; and correct the initial predicted d-axis current and the initial predicted q-axis current through the neural network model to obtain a predicted d-axis current and a predicted q-axis current of the synchronous generator.
  • the prediction module 320 is further configured to:
  • the rotor operation equation group and the generator winding flux differential equation group are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the prediction module 320 is further configured to:
  • the discretized rotor operation equations and the generator winding flux differential equations are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the neural network model includes a cascaded first long short-term memory network model, a first fully connected layer, a Dropout layer, and a second fully connected layer.
  • a device for predicting output current of a synchronous generator comprises:
  • a sample acquisition module (not shown), used to acquire a training sample set and an initial hybrid drive model, wherein the initial hybrid drive model includes a cascaded initial synchronous generator physical model and an initial neural network model;
  • a training module (not shown), configured to perform first-stage model training and second-stage model training on the initial hybrid drive model based on the training sample set to obtain the preset hybrid drive model;
  • the first stage model training includes: fixing the parameters of the initial synchronous generator physical model unchanged, training the initial neural network model using the training sample set until the error of the initial neural network model is less than a first threshold, so as to obtain a candidate neural network model;
  • the second stage training process includes: using the training sample set to jointly train the initial synchronization The generator physical model and the candidate neural network model until the error of the initial hybrid drive model is less than a second threshold, wherein the second threshold is less than the first threshold.
  • the training module (not shown) is further used to: use the training sample set to train the initial synchronous generator physical model at a first learning rate, and train the candidate neural network model at a second learning rate.
  • the sample acquisition module (not shown) is further used to collect the working data of the generator when it is disturbed at a preset frequency as the training sample.
  • Each module in the above-mentioned synchronous generator output current prediction device can be implemented in whole or in part by software, hardware and a combination thereof.
  • Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.
  • a computer device which may be a server, and its internal structure diagram may be shown in FIG4.
  • the computer device includes a processor, a memory, and a network interface connected via a system bus.
  • the processor of the computer device is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, a computer program, and a database.
  • the internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium.
  • the database of the computer device is used to store data such as training samples and parameters of synchronous generators.
  • the network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for predicting the output current of a synchronous generator is implemented.
  • FIG. 4 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.
  • a computer device including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:
  • the working state data includes the d-axis voltage, the q-axis voltage, the excitation voltage and the mechanical power input to the synchronous generator;
  • the preset hybrid drive model includes a cascaded synchronous generator physical model and a neural network model
  • the initial predicted d-axis current and the initial predicted q-axis current are corrected by the neural network model to obtain the predicted d-axis current and the predicted q-axis current of the synchronous generator.
  • the rotor operation equation group and the generator winding flux differential equation group are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • solving the rotor operation equations and the generator winding flux differential equations to obtain the initial predicted d-axis current and the initial predicted q-axis current includes:
  • the discretized rotor operation equations and the generator winding flux differential equations are solved to obtain the initial predicted d-axis current and the initial predicted q-axis current.
  • the neural network model includes a cascaded first long short-term memory network model, a first fully connected layer, a Dropout layer, and a second fully connected layer.
  • the initial hybrid drive model includes a cascaded initial synchronous generator physical model and an initial neural network model
  • the initial hybrid driving model is subjected to first-stage model training and second-stage model training to obtain the preset hybrid driving model;
  • the first stage model training includes: fixing the parameters of the initial synchronous generator physical model unchanged, training the initial neural network model using the training sample set until the error of the initial neural network model is less than a first threshold, so as to obtain a candidate neural network model;
  • the second stage training process includes: using the training sample set to jointly train the initial synchronous generator physical model and the candidate neural network model until the error of the initial hybrid drive model is is smaller than a second threshold, wherein the second threshold is smaller than the first threshold.
  • the initial synchronous generator physical model is trained using the training sample set at a first learning rate, and the candidate neural network model is trained at a second learning rate.
  • the working data of the generator when it is disturbed is collected at a preset frequency as the training sample.
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.
  • a computer program product comprising a computer program, wherein when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented, which will not be described in detail here.
  • any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory.
  • Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc.
  • Volatile memory can include random access memory (RAM) or external cache memory, etc.
  • RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • the database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database.
  • Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this.
  • the processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Operations Research (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Control Of Eletrric Generators (AREA)

Abstract

一种同步发电机输出电流的预测方法、装置、设备和存储介质。方法包括：获取同步发电机当前时刻点的工作状态数据，工作状态数据包括d轴电压、q轴电压、励磁电压和输入同步发电机的机械功率；将工作状态数据输入预设混合驱动模型，预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过同步发电机物理模型对工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过神经网络模型对初始预测d轴电流和初始预测q轴电流进行修正处理，获得同步发电机的预测d轴电流和预测q轴电流。采用本方法能够提高预测准确率，并且提升模型在不同场景下的泛化能力。

Description

同步发电机输出电流的预测方法、装置、设备和存储介质 技术领域
本申请涉及电网技术领域，特别是涉及一种同步发电机输出电流的预测方法、装置、设备和存储介质。
背景技术
随着电力技术的发展，新型输电装置和多样柔性负荷不断接入，电网的结构日益复杂。如何对日益复杂的电网进行建模和分析成为电力领域重要的研究方向。
传统技术中，对电网中发电机的分析一般基于神经网络算法建立分析模型实现。然而基于神经网络算法的准确度受限于样本量，若样本量不足容易导致准确率不高。
发明内容
基于此，有必要针对上述技术问题，提供一种能够提高预测准确率的同步发电机输出电流的预测方法、装置、设备和存储介质。
第一方面，本申请提供了一种同步发电机输出电流的预测方法。所述方法包括：
获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
在其中一个实施例中，所述通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流，包括：
将所述d轴电压、所述q轴电压、所述励磁电压和所述机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组；
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在其中一个实施例中，所述对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流，包括：
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行离散化处理；
对离散化后的所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在其中一个实施例中，所述神经网络模型包括级联的第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
在其中一个实施例中，所述将所述工作状态数据输入预设混合驱动模型之前，所述方法还包括：
获取训练样本集以及初始混合驱动模型，所述初始混合驱动模型包括级联的初始同步发电机物理模型和初始神经网络模型；
基于所述训练样本集对所述初始混合驱动模型进行第一阶段模型训练和第二阶段模型训练，以得到所述预设混合驱动模型；
其中，所述第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用所述训练样本集训练所述初始神经网络模型，直至所述初始神经网络模型的误差小于第一阈值为止，以得到候选神经网络模型；
所述第二阶段训练过程包括：利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差小于第二阈值，其中，所述第二阈值小于所述第一阈值。
在其中一个实施例中，所述利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，包括：
利用所述训练样本集，以第一学习率训练所述初始同步发电机物理模型，并以第二学习率训练所述候选神经网络模型。
在其中一个实施例中，所述获取训练样本集，包括：
以预设频率采集所述发电机在受到扰动时的工作数据，作为所述训练样本。
第二方面，本申请还提供了一种同步发电机输出电流的预测装置。所述装置包括：
工作状态数据获取模块，用于获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
预测模块，用于将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
第三方面，本申请还提供了一种计算机设备。所述计算机设备包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：
获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
第四方面，本申请还提供了一种计算机可读存储介质。所述计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
获取同步发电机当前时刻点的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始 预测d轴电流和初始预测q轴电流；
通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
第五方面，本申请还提供了一种计算机程序产品。所述计算机程序产品，包括计算机程序，该计算机程序被处理器执行时实现以下步骤：
获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
上述同步发电机输出电流的预测方法、装置、计算机设备、存储介质和计算机程序产品，通过获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。通过上述方式，本申请通过先将发电机的工作状态数据通过同步发电机物理模型进行预测，获得初始预测d轴电流和初始预测q轴电流，然后再经过神经网络模型修正，从而获得准确率高的预测结果，提高了预测结果的准信率。
此外，由于同步发电机物理模型是基于发电机的物理性质建立，因此获得的初始预测d轴电流和初始预测q轴电流与真实的d轴电流和q轴电流相差较小，导致在训练神经网络模型的过程中，样本数量的需求也大幅度减少。
附图说明
图1为一个实施例中同步发电机输出电流的预测方法的应用环境图；
图2为一个实施例中同步发电机输出电流的预测方法的流程示意图；
图3为一个实施例中同步发电机输出电流的预测装置的结构框图；
图4为一个实施例中计算机设备的内部结构图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请实施例提供的同步发电机输出电流的预测方法，可以应用于如图1所示的应用环境中。其中，同步发电机102通过网络与计算机设备104进行通信。数据存储系统可以存储计算机设备104需要处理的数据。数据存储系统可以集成在计算机设备104上，也可以放在云上或其他网络设备上。计算机设备104能够获取到同步发电机102的工作状态数据以及同步发电机102的铭牌值，该铭牌值包括同步发电机102的各种参数。具体的，计算机设备104可以通过扫描同步发电机102的铭牌或者用户通过输入装置输入，从而使得计算机设备104获得同步发电机102的铭牌值。其中，计算机设备104可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑、物联网设备和便携式可穿戴设备，物联网设备可为智能音箱、智能电视、智能空调、智能车载设备等。便携式可穿戴设备可为智能手表、智能手环、头戴设备等。
在一个实施例中，如图2所示，提供了一种同步发电机输出电流的预测方法，以该方法应用于图1中的计算机设备为例进行说明，包括以下步骤：
步骤210，获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
在使用过过程中，计算机设备可以获取同步发电机当前(时刻点)的工作状态数据，该工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率。具体的，计算机设备可以直接或者间接通过各种测量仪器获取同步发电机当前(时刻点)的工作状态数据。
步骤220，将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动 模型包括级联的同步发电机物理模型和神经网络模型；
步骤230，通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
步骤240，通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
在获得同步发电机当前时刻点的工作状态数据后，将该工作状态数据输入到预设的混合驱动模型中，该预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型，具体的，将d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率输入同步发电机物理模型，通过同步发电机物理模型对输入的d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率进行处理，得到初始预测d轴电流和初始预测q轴电流。
具体的，同步发电机物理模型包括转子运动方程组和同步发电机绕组磁链微分方程组，其中，转子运动方程组如下：
其中，δ为发电机功角，ω表示电机转速，ωs为电力系统同步转速，Tj为发电机的惯性常数，Pe表征发电机组的电磁功率，Pm代表发电机组的机械功率。
同步发电机绕组磁链微分方程组如下：
其中，T′表征同步发电机各绕组的开路暂态时间常数，T″征各绕组的开路次暂态时间常数，E′为暂态电动势，E″为次暂态电动势，Vf表示同步发电机机端励磁电压，I代表同步发电机定子电流，X′与X″分别为同步电机暂态与次暂态下 的电抗，Xl表示与饱和相关的电机漏抗，Std表示d轴同步发电机的磁链饱和程度，Stq表示q轴同步发电机的磁链饱和程度。
然后对上述公式进行求解，从而获得初始预测d轴电流和初始预测q轴电流。由于上述同步发电机物理模型属于物理理论构成，在建模时便进行了高度的理论简化以及线性化，并且在实际应用中同步发电机实际输出的d轴电流和q轴电流，会受到外界的干扰，这些都导致同步发电机物理模型与真实结果存在一定的误差。因此再将获得的初始预测d轴电流和初始预测q轴电流通过神经网络进行修正。该神经网络为预先训练好的神经网络模型。通过神经网络修正初始预测d轴电流和初始预测q轴电流，或者预测的d轴电流和q轴电流。
作为一种实施例，神经网络模型包括级联的第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
具体的，在修正过程中，初始预测d轴电流和初始预测q轴电流经过依次经过第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层，即初始预测d轴电流和初始预测q轴电流经过第一长短记忆网络模型，第一长短记忆网络模型的输出结果输入第一全连接层，第一全连接层的输出结果输入至Dropout层，Dropout层的输出结果输入第二全连接层，最后得到修正后的预测的d轴电流和q轴电流。
作另一种实施例，神经网络模型包括级联的多层第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
具体的，在修正过程中，初始预测d轴电流和初始预测q轴电流经过依次经过多层第一长短记忆网络模型(示例性的，第一长短记忆网络模型可以为3层)、第一全连接层、Dropout层和第二全连接层，即初始预测d轴电流和初始预测q轴电流经过每层第一长短记忆网络模型，最后一层第一长短记忆网络模型的输出结果输入第一全连接层，第一全连接层的输出结果输入至Dropout层，Dropout层的输出结果输入第二全连接层，最后得到修正后的预测的d轴电流和q轴电流。
需要说明的是，计算机设备可以实时或者定时获取同步发电机的工作状态数据，将每次获得的工作状态数据输入该预设的混合驱动模型，从而持续获得该 同步发电机的预测d轴电流和q轴电流。
上述同步发电机输出电流的预测方法，通过获取同步发电机当前时刻点的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。通过上述方式，本申请通过先将发电机的工作状态数据通过同步发电机物理模型进行预测，获得初始预测d轴电流和初始预测q轴电流，然后再经过神经网络模型修正，从而获得准确率高的预测结果，提高了预测结果的准信率。
在一个实施例中，通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流的步骤，包括：
将所述d轴电压、所述q轴电压、所述励磁电压和所述机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组；
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
具体的，通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流的过程可以包括：
将d轴电压、q轴电压、励磁电压和机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组，转子运行方程组和发电机绕组磁链微分方程组如上一实施例所示。根据当前采集的d轴电压、q轴电压、励磁电压和机械功率即可进行求解，获得初始预测d轴电流和初始预测q轴电流。
作为一种实施例，对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流，包括：
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行离散化处理；
对离散化后的所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
具体的，采用隐式梯形积分法进行离散化，以确保同步相量量测装置(PMU)采样频率下模型的稳定性。
Δt表示时间步长，k表示第k个时间点，u为输入变量，如励磁电压等，p为同步发电机的动态参数的集合，f表征同步电机电磁暂态的微分方程组，h则表示所有输出代数方程。该离散化处理，可以参考现有技术。
离散化处理后，利用如下求解方程进行求解：

其中，Fj为l个离散方程中第j个方程，xi为m个状态量中第i个状态量，x0是t0时刻的初始值。作为另一实施例，也可以直接利用上述求解方程对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解。
在一个实施例中，所述将所述工作状态数据输入预设混合驱动模型之前，所述方法还包括：
获取训练样本集以及初始混合驱动模型，所述初始混合驱动模型包括级联的初始同步发电机物理模型和初始神经网络模型；
基于所述训练样本集对所述初始混合驱动模型进行第一阶段模型训练和第 二阶段模型训练，以得到所述预设混合驱动模型；
其中，所述第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用所述训练样本集训练所述初始神经网络模型，直至所述初始神经网络模型的误差小于第一阈值为止，以得到候选神经网络模型；
所述第二阶段训练过程包括：利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差小于第二阈值，其中，所述第二阈值小于所述第一阈值。
具体的，计算机设备可以预设频率采集所述发电机在受到扰动时的工作数据，作为所述训练样本。示例性的，预设频率可以为50至100Hz中任一频率。当电网收到扰动时，采集同步发电机的工作状态数据(即变量)，包括d轴电压、q轴电压、励磁电压(或励磁电流)以及发电机输入机械功率，以及同步发电机实际输出的d轴电流、q轴电流。从而获得多个样本数据，每个样本数据都包括：d轴电压、q轴电压、励磁电压(或励磁电流)以及发电机输入机械功率。
通过获取发电机的铭牌值获得同步发电机物理模型的参数集。
根据同步发电机物理模型的参数集初始化同步发电机物理模型，并初始化神经网络模型，然后将初始化同步发电机物理模型和初始化神经网络模型进行拼接，获得初始混合驱动模型。
然后基于训练样本集对初始混合驱动模型进行第一阶段模型训练和第二阶段模型训练，以得到所述预设混合驱动模型。具体的，由于同步发电机物理模型是基于物理性质构建，其准确度较高，因此第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用训练样本集训练初始神经网络模型，即将训练样本输入初始同步发电机物理模型，获得初始预测的d轴电流、初始预测的q轴电流，然后将初始预测的d轴电流、初始预测的q轴电流作为初始神经网络模型的输入，实现对初始神经网络模型的训练。直至所述初始神经网络模型的误差小于第一阈值为止，将最终获得的神经网络模型作为候选神经网络模型。
作为一种实施例，初始神经网络模型包括级联的第一初始长短记忆网络模型、第一初始全连接层、初始Dropout层和第二初始全连接层。
具体的，在修正过程中，初始预测d轴电流和初始预测q轴电流经过依次经过第一初始长短记忆网络模型、第一初始全连接层、初始Dropout层和第二初始全连接层，即初始预测d轴电流和初始预测q轴电流经过第一初始长短记忆网络模型，第一初始长短记忆网络模型的输出结果输入第一初始全连接层，第一初始全连接层的输出结果输入至初始Dropout层，初始Dropout层的输出结果输入第二初始全连接层，最后得到修正后的预测的d轴电流和q轴电流。
作另一种实施例，神经网络模型包括级联的多层第一初始长短记忆网络模型、第一初始全连接层、初始Dropout层和第二初始全连接层。
具体的，在修正过程中，初始预测d轴电流和初始预测q轴电流经过依次经过多层第一初始长短记忆网络模型(示例性的，第一初始长短记忆网络模型可以为3层)、第一初始全连接层、初始Dropout层和第二初始全连接层，即初始预测d轴电流和初始预测q轴电流经过每层第一初始长短记忆网络模型，最后一层第一初始长短记忆网络模型的输出结果输入第一初始全连接层，第一初始全连接层的输出结果输入至初始Dropout层，初始Dropout层的输出结果输入第二初始全连接层，最后得到修正后的预测的d轴电流和q轴电流。
得到修正后的预测的d轴电流和q轴电流与训练样本中的同步发电机实际输出的d轴电流、q轴电流计算误差，若该误差大于或者等于第一误差，则更新初始神经网络模型内的参数，重新将初始预测的d轴电流、初始预测的q轴电流作为初始神经网络模型的输入。依次类推，直到初始神经网络模型的误差小于第一误差，此时结束第一阶段训练，将误差小于第一误差的初始神经网络模型作为候选神经网络模型。
然后进行第二阶段的训练，第二阶段的训练包括：利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差小于第二阈值，其中，所述第二阈值小于所述第一阈值。即同时更新初始同步发电机物理模型和候选神经网络模型中的参数，重复上述序列过程，并计算输出结果与同步发电机实际输出的d轴电流、q轴电流计算误差，直到该误差小于第二阈值。如此获得训练好的混合驱动模型。
具体的，利用所述训练样本集共同训练所述初始同步发电机物理模型和所 述候选神经网络模型，包括：
利用所述训练样本集，以第一学习率训练所述初始同步发电机物理模型，并以第二学习率训练所述候选神经网络模型。
其中，第一学习率和第二学习率可以相同，也可以不同，通过第一学习率更新初始同步发电机物理模型的参数，从而训练初始同步发电机物理模型，通过第二学习率更新候选神经网络模型的参数，从而训练候选神经网络模型。
通过上述方式训练的混合驱动模型，能够保持物理上一致的结果(即准确度很高，接近100％)，同时缩小了参数的搜索范围(在第一阶段，缩小了初始神经网络模型的参数搜索范围，在第二阶段缩小了初始同步发电机物理模型和候选神经网络模型的参数搜索范围)，能够提高训练效率，由于同步发电机物理模型的输出结果比较高，在整个训练过程中对数据的依赖比较小，即只需要少量训练样本即可完成训练。
应该理解的是，虽然如上所述的各实施例所涉及的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，如上所述的各实施例所涉及的流程图中的至少一部分步骤可以包括多个步骤或者多个阶段，这些步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤中的步骤或者阶段的至少一部分轮流或者交替地执行。
基于同样的发明构思，本申请实施例还提供了一种用于实现上述所涉及的同步发电机输出电流的预测方法的同步发电机输出电流的预测装置。该装置所提供的解决问题的实现方案与上述方法中所记载的实现方案相似，故下面所提供的一个或多个同步发电机输出电流的预测装置实施例中的具体限定可以参见上文中对于同步发电机输出电流的预测方法的限定，在此不再赘述。
在一个实施例中，如图3所示，提供了一种同步发电机输出电流的预测装置，包括：
工作状态数据获取模块310，用于获取同步发电机当前时刻点的工作状态数 据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
预测模块320，用于将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
在一个实施例中，预测模块320，还用于：
将所述d轴电压、所述q轴电压、所述励磁电压和所述机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组；
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在一个实施例中，预测模块320，还用于：
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行离散化处理；
对离散化后的所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在一个实施例中，所述神经网络模型包括级联的第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
在一个实施例中，同步发电机输出电流的预测装置，包括：
样本获取模块(图未示)，用于获取训练样本集以及初始混合驱动模型，所述初始混合驱动模型包括级联的初始同步发电机物理模型和初始神经网络模型；
训练模块(图未示)，用于基于所述训练样本集对所述初始混合驱动模型进行第一阶段模型训练和第二阶段模型训练，以得到所述预设混合驱动模型；
其中，所述第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用所述训练样本集训练所述初始神经网络模型，直至所述初始神经网络模型的误差小于第一阈值为止，以得到候选神经网络模型；
所述第二阶段训练过程包括：利用所述训练样本集共同训练所述初始同步 发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差小于第二阈值，其中，所述第二阈值小于所述第一阈值。
在一个实施例中，训练模块(图未示)，还用于：利用所述训练样本集，以第一学习率训练所述初始同步发电机物理模型，并以第二学习率训练所述候选神经网络模型。
在一个实施例中，样本获取模块(图未示)，还用于：以预设频率采集所述发电机在受到扰动时的工作数据，作为所述训练样本。
上述同步发电机输出电流的预测装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图4所示。该计算机设备包括通过系统总线连接的处理器、存储器和网络接口。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质和内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储训练样本、同步发电机的参数等数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种同步发电机输出电流的预测方法。
本领域技术人员可以理解，图4中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
在一个实施例中，提供了一种计算机设备，包括存储器和处理器，存储器中存储有计算机程序，该处理器执行计算机程序时实现以下步骤：
获取同步发电机当前时刻点的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
将所述d轴电压、所述q轴电压、所述励磁电压和所述机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组；
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在一个实施例中，所述对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流，包括：
对所述转子运行方程组和所述发电机绕组磁链微分方程组进行离散化处理；
对离散化后的所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
在一个实施例中，所述神经网络模型包括级联的第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
获取训练样本集以及初始混合驱动模型，所述初始混合驱动模型包括级联的初始同步发电机物理模型和初始神经网络模型；
基于所述训练样本集对所述初始混合驱动模型进行第一阶段模型训练和第二阶段模型训练，以得到所述预设混合驱动模型；
其中，所述第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用所述训练样本集训练所述初始神经网络模型，直至所述初始神经网络模型的误差小于第一阈值为止，以得到候选神经网络模型；
所述第二阶段训练过程包括：利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差 小于第二阈值，其中，所述第二阈值小于所述第一阈值。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
利用所述训练样本集，以第一学习率训练所述初始同步发电机物理模型，并以第二学习率训练所述候选神经网络模型。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
以预设频率采集所述发电机在受到扰动时的工作数据，作为所述训练样本。
在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现第一方面所述方法的步骤。此处不再赘述。
在一个实施例中，提供了一种计算机程序产品，包括计算机程序，该计算机程序被处理器执行时实现第一方面所述方法的步骤。此处不再赘述。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、数据库或其它介质的任何引用，均可包括非易失性和易失性存储器中的至少一种。非易失性存储器可包括只读存储器(Read-Only Memory，ROM)、磁带、软盘、闪存、光存储器、高密度嵌入式非易失性存储器、阻变存储器(ReRAM)、磁变存储器(Magnetoresistive Random Access Memory，MRAM)、铁电存储器(Ferroelectric Random Access Memory，FRAM)、相变存储器(Phase Change Memory，PCM)、石墨烯存储器等。易失性存储器可包括随机存取存储器(Random Access Memory，RAM)或外部高速缓冲存储器等。作为说明而非局限，RAM可以是多种形式，比如静态随机存取存储器(Static Random Access Memory，SRAM)或动态随机存取存储器(Dynamic Random Access Memory，DRAM)等。本申请所提供的各实施例中所涉及的数据库可包括关系型数据库和非关系型数据库中至少一种。非关系型数据库可包括基于区块链的分布式数据库等，不限于此。本申请所提供的各实施例中所涉及的处理器可为通用处理器、中央处理器、图形处理器、数字信号处理器、可编程逻辑器、基于量子计算的数据处理逻辑器等，不限于此。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对本申请专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请的保护范围应以所附权利要求为准。

Claims (11)

  1. 一种同步发电机输出电流的预测方法，其特征在于，所述方法包括：
    获取同步发电机当前的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
    将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；
    通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；
    通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
  2. 根据权利要求1所述的方法，其特征在于，所述通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流，包括：
    将所述d轴电压、所述q轴电压、所述励磁电压和所述机械功率对应输入所述同步发电机物理模型中的转子运行方程组和发电机绕组磁链微分方程组；
    对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
  3. 根据权利要求2所述的方法，其特征在于，所述对所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流，包括：
    对所述转子运行方程组和所述发电机绕组磁链微分方程组进行离散化处理；
    对离散化后的所述转子运行方程组和所述发电机绕组磁链微分方程组进行求解，获得所述初始预测d轴电流和所述初始预测q轴电流。
  4. 根据权利要求1所述的方法，其特征在于，所述神经网络模型包括级联的第一长短记忆网络模型、第一全连接层、Dropout层和第二全连接层。
  5. 根据权利要求1至4任一项所述的方法，其特征在于，所述将所述工作状态数据输入预设混合驱动模型之前，所述方法还包括：
    获取训练样本集以及初始混合驱动模型，所述初始混合驱动模型包括级联的初始同步发电机物理模型和初始神经网络模型；
    基于所述训练样本集对所述初始混合驱动模型进行第一阶段模型训练和第二阶段模型训练，以得到所述预设混合驱动模型；
    其中，所述第一阶段模型训练包括：固定所述初始同步发电机物理模型的参数不变，利用所述训练样本集训练所述初始神经网络模型，直至所述初始神经网络模型的误差小于第一阈值为止，以得到候选神经网络模型；
    所述第二阶段训练过程包括：利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，直至所述初始混合驱动模型的误差小于第二阈值，其中，所述第二阈值小于所述第一阈值。
  6. 根据权利要求5所述的方法，其特征在于，所述利用所述训练样本集共同训练所述初始同步发电机物理模型和所述候选神经网络模型，包括：
    利用所述训练样本集，以第一学习率训练所述初始同步发电机物理模型，并以第二学习率训练所述候选神经网络模型。
  7. 根据权利要求5所述的方法，其特征在于，所述获取训练样本集，包括：
    以预设频率采集所述发电机在受到扰动时的工作数据，作为所述训练样本。
  8. 一种同步发电机输出电流的预测装置，其特征在于，所述装置包括：
    工作状态数据获取模块，用于获取同步发电机当前时刻点的工作状态数据，所述工作状态数据包括d轴电压、q轴电压、励磁电压和输入所述同步发电机的机械功率；
    预测模块，用于将所述工作状态数据输入预设混合驱动模型，所述预设混合驱动模型包括级联的同步发电机物理模型和神经网络模型；通过所述同步发电机物理模型对所述工作状态数据进行处理，以得到初始预测d轴电流和初始预测q轴电流；通过所述神经网络模型对所述初始预测d轴电流和所述初始预测q轴电流进行修正处理，获得所述同步发电机的预测d轴电流和预测q轴电流。
  9. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至7中任一项所述的方法的步骤。
  10. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至7中任一项所述的方法的步骤。
  11. 一种计算机程序产品，包括计算机程序，其特征在于，该计算机程序被处理器执行时实现权利要求1至7中任一项所述的方法的步骤。
PCT/CN2023/081309 2022-12-07 2023-03-14 同步发电机输出电流的预测方法、装置、设备和存储介质 WO2024119654A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211564166.7 2022-12-07
CN202211564166.7A CN115577573B (zh) 2022-12-07 2022-12-07 同步发电机输出电流的预测方法、装置、设备和存储介质

Publications (1)

Publication Number Publication Date
WO2024119654A1 true WO2024119654A1 (zh) 2024-06-13

Family

ID=84590200

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/081309 WO2024119654A1 (zh) 2022-12-07 2023-03-14 同步发电机输出电流的预测方法、装置、设备和存储介质

Country Status (2)

Country Link
CN (1) CN115577573B (zh)
WO (1) WO2024119654A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115577573B (zh) * 2022-12-07 2023-03-10 南方电网数字电网研究院有限公司 同步发电机输出电流的预测方法、装置、设备和存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150039545A1 (en) * 2013-08-05 2015-02-05 City University of London Systems, methods and devices for vector control of permanent magnet synchronous machines using artificial neural networks
CN112214899A (zh) * 2020-10-16 2021-01-12 哈尔滨理工大学 双轴励磁同步发电机的2s-dirk电磁暂态建模方法
CN115392110A (zh) * 2022-07-27 2022-11-25 广东电网有限责任公司广州供电局 基于ppo算法的数据模型混合驱动风电场建模方法
CN115577573A (zh) * 2022-12-07 2023-01-06 南方电网数字电网研究院有限公司 同步发电机输出电流的预测方法、装置、设备和存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2476316B (en) * 2009-12-21 2014-07-16 Vestas Wind Sys As A wind turbine having a control method and controller for predictive control of a wind turbine generator
RU2498494C1 (ru) * 2012-07-20 2013-11-10 Федеральное государственное бюджетное образовательное учреждение высшего профессионального образования "Московский государственный технический университет имени Н.Э. Баумана" (МГТУ им. Н.Э. Баумана) Способ управления возбуждением синхронного электрического генератора
EP3907663B1 (en) * 2020-05-06 2024-02-21 Robert Bosch GmbH Predicting a state of a computer-controlled entity
DE102020206575A1 (de) * 2020-05-26 2021-12-02 Siemens Aktiengesellschaft Prognose eines Systemzustands eines technischen Systems
CN115313930A (zh) * 2022-06-22 2022-11-08 中国大唐集团科学技术研究总院有限公司华东电力试验研究院 一种大型汽轮发电机励磁电流实时预测方法
CN114997065B (zh) * 2022-06-22 2024-07-26 沈阳工学院 基于改进随机配置网络的同步发电机励磁电流预测方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150039545A1 (en) * 2013-08-05 2015-02-05 City University of London Systems, methods and devices for vector control of permanent magnet synchronous machines using artificial neural networks
CN112214899A (zh) * 2020-10-16 2021-01-12 哈尔滨理工大学 双轴励磁同步发电机的2s-dirk电磁暂态建模方法
CN115392110A (zh) * 2022-07-27 2022-11-25 广东电网有限责任公司广州供电局 基于ppo算法的数据模型混合驱动风电场建模方法
CN115577573A (zh) * 2022-12-07 2023-01-06 南方电网数字电网研究院有限公司 同步发电机输出电流的预测方法、装置、设备和存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LI YUNLU, MA GUIQING, YANG JUNYOU, XU YAN: "Dynamic modeling for VSG cluster by using data-physical driven method", ENERGY REPORTS, ELSEVIER, vol. 8, 1 August 2022 (2022-08-01), pages 227 - 234, XP093180746, ISSN: 2352-4847, DOI: 10.1016/j.egyr.2022.02.106 *

Also Published As

Publication number Publication date
CN115577573B (zh) 2023-03-10
CN115577573A (zh) 2023-01-06

Similar Documents

Publication Publication Date Title
Yao et al. Adaptive variational quantum dynamics simulations
Taylor et al. Conic AC transmission system planning
Dou et al. Universal approach to quantum thermodynamics in the strong coupling regime
Shen et al. Adaptive L2 disturbance attenuation of Hamiltonian systems with parametric perturbation and application to power systems
JP7103341B2 (ja) 関係性探索システム、情報処理装置、方法およびプログラム
Zeng et al. An online transfer learning-based remaining useful life prediction method of ball bearings
JP5431261B2 (ja) 情報管理システム、方法及びプログラム
Liu et al. Real-time particle swarm optimization based parameter identification applied to permanent magnet synchronous machine
WO2024119654A1 (zh) 同步发电机输出电流的预测方法、装置、设备和存储介质
Chen et al. A lifelong learning method for gearbox diagnosis with incremental fault types
Morrone et al. Charging a quantum battery in a non-Markovian environment: a collisional model approach
Kamel et al. Data-driven and model-based hybrid reinforcement learning to reduce stress on power systems branches
Foggo et al. pmuBAGE: The benchmarking assortment of generated PMU data for power system events
CN116168770A (zh) 分子数据的处理方法、装置电子设备以及存储介质
Long et al. Multilayer deep deterministic policy gradient for static safety and stability analysis of novel power systems
CN114679098A (zh) 永磁同步电机前馈补偿方法、装置、计算机设备和介质
Xiang et al. An Intelligent Parameter Identification Method of DFIG Systems Using Hybrid Particle Swarm Optimization and Reinforcement Learning
Song Memory-based control of nonlinear dynamic systems part I-Design and analysis
De Tuglie et al. Identification of dynamic voltage-current power system equivalents through artificial neural networks
Alanis et al. Discrete-time inverse optimal neural control for synchronous generators
Chen et al. New insights into the extended Malkus-Robbins dynamo
GB2572975A (en) Quantum bundle adjustment
Otkun Newton–Raphson based scalar speed control and optimization of IM
Liao et al. Explainable Neural Dynamics Models for Electric Motor Temperature Estimation
Nazemi et al. A novel collaborate neural dynamic system model for solving a class of min–max optimization problems with an application in portfolio management

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23899252

Country of ref document: EP

Kind code of ref document: A1