Power distribution system simulation model library construction method based on data model hybrid drive

Abstract

The invention particularly relates to a power distribution system simulation model library construction method based on data model hybrid driving. The method comprises the following steps: performing physical characteristic analysis on each power electronic device in a power distribution system to obtain a physical driving model of each power electronic device; establishing a data driving model of each power electronic device by using the physical driving model of each power electronic device; establishing an electromagnetic transient model of each piece of power electronic equipment by utilizing the physical driving model of each piece of power electronic equipment and the data driving model of each piece of power electronic equipment, and constructing a power distribution system simulation model library by utilizing the electromagnetic transient model of each piece of power electronic equipment; and establishing an object state transition model of the power distribution system simulation model library, and performing model switching of the power distribution system simulation model library by using the object state transition model so as to perform deduction simulation of the power distribution system. According to the method, the safety level and the operation efficiency of the power distribution network are improved, the multi-level modeling speed and accuracy of the power distribution system can be improved, and the modeling requirements under different conditions are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution system simulation, and in particular relates to a method for constructing a power distribution system simulation model library based on data model hybrid drive. Background Art

[0002] With the introduction of a large number of new components or systems such as distributed power sources, microgrids, and electric vehicle charging piles, the distribution system has shown new operating characteristics such as extensive access to power electronic devices, extensive interconnection of AC and DC, and deep integration of information and physics. The complexity of its power grid has grown by leaps and bounds, further increasing the difficulty of simulation analysis. Compared with traditional power grids, the power electronic distribution system uses a large number of power electronic devices, control and protection equipment, etc., and its system structure is more complex; on the other hand, with the access of a large number of distributed power sources, the operating state of the distribution system will continue to change with the changes in external conditions, topology, load demand, and the occurrence of faults or disturbances, and its transient characteristics are more complex. The time and space scales corresponding to different dynamic processes of the power electronic distribution system are different. In the dynamic processes of different time and space scales, the resolution of modeling of distributed power sources, energy storage, power electronic devices and other equipment components is different from the perspective of system resources and simulation accuracy, thus forming the modeling requirements at the corresponding time and space scales. For example, using the detailed model of the distribution network components for steady-state calculations will seriously occupy system resources and affect the calculation speed; while the simplified model of the components for transient calculations will produce erroneous simulation results, which are obviously deviated from the actual production.

[0003] In the 1960s, Professor Dommel proposed the theory of electromagnetic transient simulation of power systems and constructed the basic framework of the electromagnetic transient program (EMTP), marking the beginning of this field. Some literature points out that the simulation of inverters is a bottleneck for accelerating EMT analysis. A novel average value modeling method is proposed for grid-connected inverters, which greatly reduces the calculation time. The research team proposed a parameterized constant admittance model of the converter based on cross-initialization, which has the same calculation efficiency and higher accuracy as the traditional constant admittance model, and its operating characteristics are closer to the ideal switch model. For high-voltage and large-capacity equipment, it is usually composed of a large number of circuit modules with the same structure cascaded, and the circuit elements are often tens of thousands. Therefore, for these cascaded devices, an internal circuit equivalent modeling method can be used to eliminate internal circuit nodes and reduce the overall model order. The dynamic phasor model uses phasor description. Its essence is to transform the high-frequency changing instantaneous value model represented by the switching function in the electromagnetic transient simulation into a dynamic phasor model described by a slow-changing analytical signal through a certain transformation method.

[0004] However, most of the existing distribution system simulation models are single-resolution offline models. On the one hand, there is a lack of simulation models suitable for different typical events, and on the other hand, there is a lack of online correction technology based on actual operating data, resulting in low model accuracy. In addition, the emergence of photovoltaic power generation, charging piles, energy storage systems, etc. in the distribution system, their output volatility has aggravated the time-varying characteristics of the distribution system, and their operating characteristics have increased the nonlinear characteristics of the distribution system, resulting in a lack of new component models in the existing simulation model library. New component models such as distributed power sources, power electronic devices, charging piles and fuel cells, and some simplified models need to be verified and optimized. Summary of the invention

[0005] In order to overcome the problems existing in the above-mentioned related technologies, the present invention provides a method for constructing a distribution system simulation model library based on data model hybrid drive.

[0006] According to a first aspect of an embodiment of the present invention, a method for constructing a power distribution system simulation model library based on data model hybrid drive is provided, comprising:

[0007] Analyzing the physical characteristics of each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device;

[0008] Using the physical drive model of each power electronic device, a data drive model of each power electronic device is established;

[0009] Using the physical drive model of each power electronic device and the data drive model of each power electronic device, an electromagnetic transient model of each power electronic device is established, and using the electromagnetic transient model of each power electronic device to construct a distribution system simulation model library;

[0010] An object state transfer model of the power distribution system simulation model library is established, and the object state transfer model is used to switch the models of the power distribution system simulation model library to perform deduction simulation of the power distribution system.

[0011] Preferably, the performing of physical characteristic analysis on each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device includes:

[0012] Performing physical characteristic analysis on each power electronic device in the power distribution system to obtain relevant parameters of each power electronic device;

[0013] Based on the laws of physics, a functional relationship between the relevant parameters is established by formula derivation to obtain equations of the power electronic devices;

[0014] The equations of each power electronic device are the physical driving model.

[0015] Preferably, the step of using the physical drive model of each power electronic device to establish the data drive model of each power electronic device includes:

[0016] Simulate multiple preset power distribution scenarios to obtain simulation data;

[0017] Extracting relevant data of the differential equation group in the physical driving model from the simulation data as training data;

[0018] The improved neural network model is trained using the training data to obtain a trained improved neural network model, and the trained improved neural network model is the data-driven model.

[0019] Preferably, the relevant data of the differential equation group includes: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

[0020] Preferably, the using the training data to train the improved neural network model comprises:

[0021] The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain the trained improved neural network model.

[0022] Preferably, the process of acquiring the improved neural network model includes:

[0023] The weights and thresholds of the neural network model are optimized using a particle swarm optimization algorithm to obtain the improved neural network model.

[0024] Preferably, the use of the physical drive model of each power electronic device and the data drive model of each power electronic device to establish the electromagnetic transient model of each power electronic device includes:

[0025] After replacing the differential equations in the physical driving model with the data driving model, hybrid driving is performed with the physical driving model to obtain a hybrid driving model;

[0026] Parameter correction is performed on the hybrid drive model to obtain a corrected hybrid drive model, where the corrected hybrid drive model is the electromagnetic transient model.

[0027] Preferably, the performing parameter correction on the hybrid drive model includes:

[0028] collecting input data of the hybrid drive model;

[0029] Inputting the input data into the power distribution system, and outputting the actual value of active power and the actual value of reactive power;

[0030] Inputting the input data into the hybrid drive model to obtain an active power simulation value and a reactive power simulation value;

[0031] Using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function;

[0032] Taking the minimum function value of the target loss function as the goal, the target loss function is optimized using an extended Kalman filter method to correct the model parameters of the hybrid drive model to obtain a corrected hybrid drive model.

[0033] Preferably, the object state transition model of establishing the power distribution system simulation model library includes:

[0034] The electromagnetic transient model in the power distribution system simulation model library is taken as an object, and the object state transfer model of the power distribution system simulation model library is established by using the Markov model.

[0035] According to a second aspect of an embodiment of the present invention, there is provided a device for constructing a power distribution system simulation model library based on data model hybrid drive, comprising:

[0036] An acquisition unit, configured to analyze the physical characteristics of each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device;

[0037] A first establishing unit, configured to establish a data driven model of each power electronic device by using a physical driven model of each power electronic device;

[0038] A second establishing unit is used to establish an electromagnetic transient model of each power electronic device by using the physical driving model of each power electronic device and the data driving model of each power electronic device, and to construct a distribution system simulation model library by using the electromagnetic transient model of each power electronic device;

[0039] The model switching unit is used to establish an object state transfer model of the power distribution system simulation model library, and use the object state transfer model to switch the model of the power distribution system simulation model library to perform deduction simulation of the power distribution system.

[0040] Preferably, the acquisition unit is specifically used for:

[0041] Performing physical characteristic analysis on each power electronic device in the power distribution system to obtain relevant parameters of each power electronic device;

[0042] Based on the laws of physics, a functional relationship between the relevant parameters is established by formula derivation to obtain equations of the power electronic devices;

[0043] The equations of each power electronic device are the physical driving model.

[0044] Preferably, the first establishing unit includes:

[0045] A simulation module is used to simulate multiple preset power distribution scenarios to obtain simulation data;

[0046] An extraction module, used to extract relevant data of the differential equation group in the physical drive model from the simulation data as training data;

[0047] The training module is used to train the improved neural network model using the training data to obtain the trained improved neural network model, and the trained improved neural network model is the data-driven model.

[0048] Preferably, the relevant data of the differential equation group includes: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

[0049] Preferably, the training module is specifically used for:

[0050] The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain the trained improved neural network model.

[0051] Preferably, the first establishing unit further includes: a first acquisition module, used to acquire the improved neural network model; the first acquisition module is specifically used to:

[0052] The weights and thresholds of the neural network model are optimized using a particle swarm optimization algorithm to obtain the improved neural network model.

[0053] Preferably, the second establishing unit comprises:

[0054] A second acquisition module is used to replace the differential equations in the physical drive model with the data drive model, and then perform hybrid drive with the physical drive model to obtain a hybrid drive model;

[0055] The correction module is used to perform parameter correction on the hybrid drive model to obtain a corrected hybrid drive model, where the corrected hybrid drive model is the electromagnetic transient model.

[0056] Preferably, the correction module is specifically used for:

[0057] collecting input data of the hybrid drive model;

[0058] Inputting the input data into the power distribution system, and outputting the actual value of active power and the actual value of reactive power;

[0059] Inputting the input data into the hybrid drive model to obtain an active power simulation value and a reactive power simulation value;

[0060] Using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function;

[0061] Taking the minimum function value of the target loss function as the goal, the target loss function is optimized using an extended Kalman filter method to correct the model parameters of the hybrid drive model to obtain a corrected hybrid drive model.

[0062] Preferably, the model switching unit comprises:

[0063] The module is used to establish the object state transfer model of the power distribution system simulation model library by using the electromagnetic transient model in the power distribution system simulation model library as the object and utilizing the Markov model.

[0064] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0065] The memory is used to store one or more programs;

[0066] When the one or more programs are executed by the at least one processor, the method for constructing a power distribution system simulation model library based on data model hybrid drive is implemented.

[0067] According to a fourth aspect of an embodiment of the present invention, a readable storage medium is provided, on which an execution program is stored. When the execution program is executed, the method for constructing a distribution system simulation model library based on data model hybrid drive is implemented.

[0068] The technical solution provided by the present invention has the following beneficial effects:

[0069] The method for constructing a distribution system simulation model library based on data-model hybrid drive provided by the present invention analyzes the physical characteristics of each power electronic device in the distribution system to obtain the physical drive model of each power electronic device, establishes the data drive model of each power electronic device by using the physical drive model of each power electronic device, establishes the electromagnetic transient model of each power electronic device by using the physical drive model of each power electronic device and the data drive model of each power electronic device, and constructs the distribution system simulation model library by using the electromagnetic transient model of each power electronic device, establishes the object state transfer model of the distribution system simulation model library, and uses the object state transfer model to switch the model of the distribution system simulation model library to perform deduction simulation of the distribution system, which not only realizes the simulation modeling based on data-model hybrid drive for the equipment elements in the distribution network at different time scales, supports the deduction simulation and effect verification of various working conditions such as fault recovery, optimization strategy and comprehensive management of the power electronic distribution system, improves the safety level and operation efficiency of the distribution network, but also can improve the multi-level modeling speed and accuracy of the distribution system, meet the modeling needs under different situations, and provide an innovative solution for the optimization of distribution network modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0071] Figure 1 It is a flow chart of a method for constructing a power distribution system simulation model library based on data model hybrid drive provided by an embodiment of the present invention;

[0072] Figure 2 is a schematic diagram of constructing an electromagnetic transient model provided by an embodiment of the present invention;

[0073] Figure 3 It is a schematic diagram of the model parameter correction principle provided by an embodiment of the present invention;

[0074] Figure 4 It is a structural block diagram of a device for constructing a power distribution system simulation model library based on data model hybrid drive provided by an embodiment of the present invention;

[0075] Figure 5 It is a structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the following embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] Embodiment 1

[0078] The present invention provides a method for constructing a power distribution system simulation model library based on data model hybrid drive, such as Figure 1 As shown, the following steps are included:

[0079] Step 11: Analyze the physical characteristics of each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device;

[0080] Step 12: Using the physical driving model of each power electronic device, establish a data driving model of each power electronic device;

[0081] Step 13: using the physical driving model of each power electronic device and the data driving model of each power electronic device, establishing the electromagnetic transient model of each power electronic device, and using the electromagnetic transient model of each power electronic device to construct a distribution system simulation model library;

[0082] Step 14: Establish an object state transfer model of the distribution system simulation model library, and use the object state transfer model to switch the model of the distribution system simulation model library to perform deduction simulation of the distribution system.

[0083] Further, step 11 includes:

[0084] Step 111: Analyze the physical characteristics of each power electronic device in the power distribution system to obtain relevant parameters of each power electronic device;

[0085] Step 112: Based on the laws of physics, a functional relationship between relevant parameters is established by formula derivation to obtain equations for each power electronic device;

[0086] Step 113: The equations of each power electronic device are physically driven models.

[0087] Further, step 12 includes:

[0088] Step 121: simulating multiple preset power distribution scenarios to obtain simulation data;

[0089] Step 122: extracting relevant data of the differential equation group in the physical driving model from the simulation data as training data;

[0090] Step 123: Use the training data to train the improved neural network model to obtain a trained improved neural network model, where the trained improved neural network model is a data-driven model.

[0091] Furthermore, the relevant data of the differential equation group includes: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

[0092] Furthermore, in step 123, the improved neural network model is trained using the training data, including:

[0093] The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain an improved neural network model after training.

[0094] Furthermore, step 12 also includes: step 120: obtaining an improved neural network model; step 120, including:

[0095] The particle swarm optimization algorithm is used to optimize the weights and thresholds of the neural network model to obtain an improved neural network model.

[0096] Further, step 13 includes:

[0097] Step 131: After replacing the differential equations in the physical driving model with the data driving model, hybrid driving is performed with the physical driving model to obtain a hybrid driving model;

[0098] Step 132: calibrate the parameters of the hybrid drive model to obtain a calibrated hybrid drive model, where the calibrated hybrid drive model is an electromagnetic transient model.

[0099] Further, step 132 includes:

[0100] Step 1321: Collect input data of the hybrid drive model;

[0101] Step 1322: Input the input data to the power distribution system, and output the actual value of active power and the actual value of reactive power;

[0102] Step 1323: inputting the input data into the hybrid drive model to obtain active power simulation value and reactive power simulation value;

[0103] Step 1324: using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function;

[0104] Step 1325: Taking the minimum function value of the target loss function as the goal, the target loss function is optimized using the extended Kalman filter method to correct the model parameters of the hybrid drive model to obtain a corrected hybrid drive model.

[0105] It should be noted that the embodiment of the present invention does not limit the “input data of the hybrid drive model”. The input data of different power electronic devices are different, and those skilled in the art can set them according to engineering needs, experimental data or expert experience.

[0106] The "optimization of the target loss function by using the extended Kalman filter method to correct the model parameters of the hybrid drive model" involved in the embodiment of the present invention is well known to those skilled in the art, and therefore, its specific implementation method will not be described in detail.

[0107] Further, step 14 includes:

[0108] Step 141: Taking the electromagnetic transient model in the power distribution system simulation model library as an object, using the Markov model to establish an object state transition model of the power distribution system simulation model library.

[0109] It should be noted that the "Markov model" involved in the embodiments of the present invention is well known to those skilled in the art, and therefore, its specific implementation will not be described in detail. In some embodiments, step 141 includes:

[0110] Step 1411: taking the electromagnetic transient model in the power distribution system simulation model library as an object and establishing the state space of the object;

[0111] Step 1412: Determine that when there is no external event input, after a time segment Δt, the state transition probability matrix of the object in the natural state is P(Δt);

[0112] Step 1413: Determine that the state transfer matrix of the object under the action of the random external event x is P(x);

[0113] Step 1414: Define the row vector N(t) as the actual state distribution of an object at time t, and determine t based on the row vector N(t). i The state distribution probability of the object at the moment;

[0114] Step 1415: Determine the state of the object using a direct sampling method to obtain a state transfer function;

[0115] Step 1416: Determine the transfer result of the object state in the object state transfer modeling process according to the state transfer function, and use the transfer result to switch the model of the distribution system simulation model library to perform deduction simulation of the distribution system.

[0116] The present invention provides a method for constructing a distribution system simulation model library based on data-model hybrid drive. As a research on the construction technology of the distribution system simulation model library based on data-model hybrid drive, the method combines the transient model parameter correction method and the new theory of nonlinear dynamic modeling in the distribution system research to construct an electromagnetic transient simulation model library for the power electronic distribution system, meet the modeling and simulation needs of the distribution network, support the intelligent and efficient analysis of the distribution system, and effectively support the deduction, simulation and effect verification of various working conditions such as fault recovery, optimization strategy, and comprehensive management of the power electronic distribution system, so as to achieve the purpose of improving the safety level and operation efficiency of the distribution network.

[0117] In order to further illustrate the above-mentioned method for constructing a power distribution system simulation model library based on data model hybrid drive, the present invention also provides a specific example, including the following steps:

[0118] Step S1: Modeling of power electronic equipment of the distribution system. According to the physical characteristics of distributed photovoltaic, charging piles, energy storage and other power electronic equipment, the basic model power electronic equations are constructed and the physical driving model is established, including:

[0119] (1) Photovoltaic cell model

[0120] Common mathematical models of photovoltaic cells mainly include ideal models, single diode models and double diode models. Among them, the ideal model ignores all internal losses of the cell and consists of a photocurrent source Iph and a diode D in parallel. The diode expresses the continuity and nonlinear relationship between voltage and current; the single diode model further simulates the internal loss through two resistors; the double diode model needs to add a parallel diode to simulate the diffusion effect of space charge. Because the ideal model is too simple and cannot reflect the internal loss and operating efficiency of photovoltaic cells, and the double diode model is too complex, the single diode model is often used for modeling in practical applications.

[0121] The output volt-ampere characteristic of the photovoltaic cell given by the single diode model is:

[0122]

[0123] In the above formula, I is the output current of the photovoltaic cell, I ph is the photocurrent source current, I s is the diode saturation current; q is the electron charge constant, which can be but is not limited to 1.602e-19C; V is the photovoltaic cell output voltage, R s is the series resistance of the photovoltaic cell; A is the diode characteristic fitting coefficient, which is a variable in the single diode model and can be, but not limited to, 2 in the dual diode model; k is the Boltzmann constant, which can be, but not limited to, 1.831e-23 J / K; T is the absolute operating temperature of the photovoltaic cell, R shis the parallel resistance of the photovoltaic cell.

[0124] Photocurrent source current I ph It is a function of light intensity and battery temperature, and its typical formula is:

[0125]

[0126] In the above formula, S is the actual irradiance (unit: W / m 2 ), S ref is the irradiance under standard conditions, I ph,ref is the photocurrent value under standard conditions, C T is the temperature coefficient (in A / K), T ref It is the absolute operating temperature of photovoltaic cells under standard conditions.

[0127] The diode saturation current changes with the battery temperature and satisfies the relationship:

[0128]

[0129] In the above formula, I s,ref is the diode saturation current under standard conditions (in A); E g is the bandgap width (in eV), which is related to the photovoltaic cell material.

[0130] Since the output power of a single photovoltaic cell is small, photovoltaic cells are usually connected in series or parallel to form a photovoltaic array to obtain a larger output power. When photovoltaic modules are connected in series or parallel to form a photovoltaic array, it is usually believed that the photovoltaic modules connected in series and parallel have ideal consistency.

[0131] The relationship between the output voltage and current of this equivalent circuit is shown in the following equation, where N s and N p are the number of photovoltaic cells connected in series and in parallel, respectively.

[0132]

[0133] Photovoltaic cells are usually connected to the grid through inverters, mainly including single-stage grid-connected structures and bipolar grid-connected structures. Among them, the single-stage grid-connected structure is easier to implement and has a simpler control system, and is widely used in practice. Here, taking the unipolar grid-connected structure as an example, the photovoltaic inverter adopts a dual-loop control strategy. The outer loop includes two parts: DC voltage control and reactive power control based on the MPPT algorithm, and the inner loop adopts current control. The MPPT algorithm is used to track the maximum output power of photovoltaics to achieve the maximum efficiency of the photovoltaic power generation system.

[0134] (2) Battery model

[0135] The battery is an electrochemical cell that directly converts the energy released by the chemical redox reaction into DC power. Depending on the chemical substances used, the battery can be divided into lead-acid batteries, nickel-cadmium batteries, nickel-metal hydride batteries, lithium-ion batteries, etc. At present, the most widely used in distributed battery energy storage systems is lead-acid batteries, which have the advantages of relatively low cost, simple use, convenient maintenance, abundant raw materials, and can be produced.

[0136] Battery models can usually be divided into three types: experimental models, electrochemical models, and equivalent circuit models. The equivalent circuit model is most suitable for dynamic characteristics simulation. The equivalent circuit model can be further divided into many types, such as the ideal battery model and its improved model, Thevenin model, fourth-order dynamic model, general equivalent circuit model, etc.

[0137] Among these models, the general equivalent circuit model has a simple circuit structure and takes into account the nonlinear characteristics of the battery. The calculation method of the circuit parameters is simple and is applicable to lead-acid, nickel-cadmium, nickel-metal hydride batteries and lithium batteries. It has a high degree of fit in the short-term dynamic simulation process. Among the other models, the ideal model has a simple circuit but has limitations and it is difficult to fully consider the nonlinear characteristics of the battery. In order to fully consider the model fit, the Thevenin model and the fourth-order dynamic model have too complex circuit structures, which is not conducive to simulation implementation. Therefore, the modeling method of the general battery model is mainly adopted.

[0138] The general battery model consists of a voltage source and a constant internal resistance in series, with the battery's charge and discharge state SOC as the only state variable. The model has the following assumptions:

[0139] 1) Assume that the internal resistance of the battery remains constant during the battery charging and discharging process;

[0140] 2) The parameters of the battery model are obtained through the discharge characteristic curve and are assumed to be completely applicable to the charging characteristics;

[0141] 3) Assume that the capacity of the battery does not change with the change of current;

[0142] 4) Assume that temperature has no effect on the battery model;

[0143] 5) The self-discharge and memory characteristics of the battery are not considered.

[0144] The controlled voltage source voltage, that is, the no-load voltage E of the battery b The expression is:

[0145]

[0146] In the above formula, E bis the no-load voltage of the battery, E0 is the constant voltage of the battery, K is the polarization voltage, Q is the capacity of the battery (in Ah), A is the amplitude in the exponential region, B is the inverse of the time constant in the exponential region, t is time, and i is current.

[0147] (3) Line

[0148] For the line, the calculation formula is shown in formula (6). The differential equation group of the three-phase asymmetric distribution line can be expressed as:

[0149]

[0150] In the above formula, a is phase a, b is phase b, c is phase c, p = {a, b, c}; R and L are both three-dimensional vectors, including self / mutual resistance and self / mutual inductance respectively; v p1 is the primary terminal voltage of phase p, v p2 is the secondary terminal voltage of phase p, i p is the p-phase current, i j To measure current, R pj To measure resistance, L pj To measure inductance.

[0151] Through the trapezoidal integration method, the Norton form dynamic phasor of the line is shown as follows:

[0152]

[0153]

[0154] in:

[0155]

[0156] In the above formula, a is phase a, b is phase b, c is phase c, k is line constant, t is time, Δt is time interval, <v p1 > k is the voltage v p1 The average value of <v p2 > k is the voltage v p2 The average value of is the positive sequence impedance, p > k (t) is the current i at time t p The average value during the period, I p (t-Δt) is the current at time t-Δt, is the negative sequence impedance, p > k (t-Δt) is the current i at the time point t-Δt p ​​The average value within the time period, j is the imaginary unit, and ω is the angular frequency.

[0157] Step S2: Extract actual equipment operation data through simulation, perform data processing and data training, establish a data-driven model, and perform hybrid driving with the physical-driven model to realize the construction of the electromagnetic transient model of the distribution system, which specifically includes:

[0158] (1) Data training

[0159] An improved particle swarm optimization method is used to train the data-driven model, specifically training the data-driven model corresponding to the differential equation group in photovoltaic, wind power, energy storage and other models. The output of the data-driven model is the calculation result of the differential equation group.

[0160] To build a data-driven model to replace the differential equations of power electronic components, it is necessary to extract the required training data from the simulation results of multiple scenarios. By setting the duration and output of different working conditions, multiple scenarios are simulated to obtain a large amount of sample data for data extraction and training of data-driven models.

[0161] Knowledge enhancement is performed when the data-driven model is input to calculate the state variables and algebraic variables required for the differential equations, and to supplement the equation-related parameters calculated by the physical-driven model. These parameters are introduced into data learning as knowledge-enhanced feature quantities, which will guide the data-driven model to better learn scenarios with extremely large parameter changes under different working conditions.

[0162] The improved particle swarm optimization algorithm is as follows:

[0163] 1) Determine the structure of each layer of the neural network;

[0164] 2) Initialize parameters, including particle speed and position;

[0165] 3) Input training samples and calculate particle fitness values;

[0166] 4) Record the global optimal value and individual historical optimal value;

[0167] 5) Update the speed and position of the particle, recalculate the particle fitness value, determine the global optimal value and the individual historical optimal value, and repeat the optimization until the global optimal fitness value meets the accuracy requirements;

[0168] 6) Mapping the global optimal particle position to the network weights and thresholds;

[0169] 7) Input sample data and use the neural network for training.

[0170] (2) Data-model hybrid driven modeling

[0171] like Figure 2As shown in the figure, by constructing a data-driven model and a physical model and then performing hybrid driving, an electromagnetic transient model of the distribution system is obtained, which can be integrated through a weighted average strategy to improve the accuracy of the prediction. Physical laws and constraints are considered in the data-driven model, for example, differential equations, algebraic equations, and boundary conditions are used to improve the reliability of the model. Experimental data or simulation data are used to optimize the physical model, and model predictions are used to fill in data gaps, and vice versa. The feature-enhanced data set calculated by the physical model enables the data-driven model to capture more complex dynamic behaviors. Combining the physical model with the data-driven model can improve the degree of fit between the model and the actual system, reduce the computational cost, and improve the performance of the model.

[0172] Step S3: Perform data correction and iterative optimization to achieve hybrid drive model adaptation and parameter correction of the power distribution system under dynamic time-varying conditions, specifically including:

[0173] In the modeling of transient processes of distribution networks, after the distribution network model structure is determined, the most critical thing is to calibrate the model parameters. By definition, parameter correction is the optimization of data fitting, and its essence is to find a set of optimal parameter vectors to minimize the model error and get closer to the real system. Parameter correction includes three elements: input-output data pairs, mathematical models, and correction criteria. Model parameter correction starts from the definition and seeks a set of optimal model parameters by selecting a suitable optimization algorithm to minimize the value of the predetermined objective function. It can be seen that it is essentially a parameter optimization problem.

[0174] like Figure 3 As shown in the figure, U(0)…U(k) is used as the input signal (i.e., the input data of the hybrid drive model). When it is sent to the actual distribution component system, the actual output value can be obtained: the actual value of active power P(k) and the actual value of reactive power Q(k). At the same time, when the input signal U(k) is sent to the distribution component simulation model, the calculated value can be obtained: the active power simulation value P m (k) and reactive power simulation value Q m (k), and then the error between the output value of the actual distribution component system and the calculated value of the distribution component simulation model can be obtained, that is, the target loss function J(e) can be obtained. Finally, the target loss function J(e) is optimized by extending the Kalman filter method through the optimization algorithm, and the model parameters are continuously corrected to make the distribution component simulation model continuously approach the actual distribution component system. Finally, the parameters that minimize the target function value can be obtained, that is, the distribution component simulation model can well express the actual distribution component system.

[0175] The distribution network system model has the characteristics of high order and severe nonlinearity. Its objective function and solution space are relatively complex. There are not only multiple extreme points in the solution space, but also the differences between some extreme points are very small. The use of traditional algorithm optimization methods to perform parameter correction often leads to large deviations in the correction results. Therefore, new algorithm optimization methods are needed for parameter correction.

[0176] In this project, the extended Kalman filter is mainly used to optimize the model parameters of photovoltaic power supply, the model parameters of electromagnetic transient of electrical components, the model parameters of lithium battery energy storage unit and the model parameters of power electronic converter.

[0177] The extended Kalman filter is based on the linear Kalman filter and is a nonlinear parameter estimation method. It also has many advantages, such as being able to obtain unbiased optimal estimates, recursive parameters with sufficient number of steps and variance matrix estimates that do not depend on their initial values, and having good stability. Nonlinear discrete systems and time-varying systems can be represented by the following nonlinear difference equations.

[0178] X(k+1)=F[X(k),W(k),k] (11)

[0179] Z(k)=h[X(k),k]+V(k) (12)

[0180] In the above formula, k is the kth sampling moment, X(k+1) is the vector sequence of the k+1th sampling moment, X(k) is the n-dimensional random state vector sequence, Z(k) is the n-dimensional random measurement vector sequence, F(·) is the n-dimensional function, h(·) is the m-dimensional vector function, and W(k) and V(k) are both noise sequences. For nonlinear discrete systems, an approximate linearization method can be used. Analogous to the basic Kalman filter equation, after mathematical transformation, the linearized system Kalman filter equation can be obtained as follows:

[0181]

[0182]

[0183]

[0184]

[0185]

[0186] The extended Kalman filter method can be used to correct system parameters. The dynamic equation of the discrete system is:

[0187] X(k+1)=F[X(K),k,E]+B k U(k)+W(k) (18)

[0188] Z(k+1)=h[X(k+1),k+1,H]+V(k+1) (19)

[0189] In the above formula, k is the kth sampling time, k+1 is the k+1th sampling time, is the estimated value at time k+1, is the estimated derivative value at time k+1, K k+1 is the intermediate variable at time k+1, Z(k+1) is the random measurement vector sequence at time k+1, is the estimated value at time k, p k+1 is the time k+1, δ is the partial derivative, h k+1 is the vector function at time k+1, Find the inverse of R at time k+1, T is the transpose, F k is the n-dimensional function at time k, p k is the middle value at time k, Q k is the covariance matrix of the noise at time k, Find the inverse of the intermediate value at time k+1, p′ k+1 is the derivative of the intermediate value at time k+1, X(K) is the target vector sequence, V(k+1) is the noise sequence at time k+1, Z(k+1) is the vector sequence at time k+1, U(k) is a known non-random vector, B k To control the coefficient matrix, let E and H be unknown parameter vectors. E and H can be regarded as newly added state vectors, and a new augmented state vector X″(k)=[X(k)E(k)H(k)] is defined. Then the dynamic equation after the augmented vector is X″, which contains X and parameter space E and H. When X″ is estimated, parameters E and H are naturally estimated as well. In this way, the problem of parameter correction is transformed into a filtering problem, and the calculation formula is the same as the filtering equation.

[0190] Step S4: Dynamically switch the model level, dynamically switch the model level according to the system state to complete the model adaptation, and realize the switching of the model library level, that is, take the electromagnetic transient model in the distribution system simulation model library as the object, and use the Markov model to establish the object state transfer model of the distribution system simulation model library, which specifically includes:

[0191] First, determine the state space of the object S = {s1, s2, ..., s n}, s1 is the first state, s2 is the second state, s n is the nth state, n is the total number of states, and the state of the object may be transferred between these n states.

[0192] Assuming that there is no external event input, after a time segment Δt, the state transition probability matrix of the object in the natural state is P(Δt), i={1,2,…,n},j={1,2,…,n},n is the total number of states, P ij (Δt) is the probability that the object will transition from the second state to the first state after a time segment Δt.

[0193]

[0194] In the above formula, P(Δt) is the state transition probability matrix of the object under natural state, n is the total number of states, P 11 (Δt) is the probability that the object moves from state 1 to state 1 after a time segment Δt, P 12 (Δt) is the probability that the object moves from the first state to the second state after the time segment Δt, P 1n (Δt) is the probability that the object transfers from the first state to the nth state after the time segment Δt, P 21 (Δt) is the probability that the object transfers from the second state to the first state after the time segment Δt, P 22 (Δt) is the probability that the object moves from state 2 to state 2 after a time segment Δt, P 2n (Δt) is the probability that the object transfers from the second state to the nth state after the time segment Δt, P n1 (Δt) is the probability that the object transfers from the nth state to the first state after the time segment Δt, P n2 (Δt) is the probability that the object moves from the nth state to the second state after the time segment Δt, P nn (Δt) is the probability that the object moves from the nth state to the nth state after the time segment Δt.

[0195] Under the influence of random external events, the state transfer matrix of the object is P(x), P i1 (x)+P i2 (x)+…+P in (x) = 1, P i1 (x) is the probability that the object transfers from the i-th state to the first state under the influence of the random external event x, P i2 (x) is the probability that the object transfers from the i-th state to the second state under the influence of random external event x, P in (x) is the probability that the object transfers from the i-th state to the n-th state under the influence of the random external event x.

[0196]

[0197] In the above formula, P(x) is the state transition probability matrix of the object under the action of random external event x; n is the total number of states, P 11(x) is the probability that the object transfers from state 1 to state 1 under the action of random external event x, P 12 (x) is the probability that the object transfers from the first state to the second state under the influence of random external event x, P 1n (x) is the probability that the object transfers from the first state to the nth state under the influence of random external event x, P 21 (x) is the probability that the object transfers from the second state to the first state under the influence of random external event x, P 22 (x) is the probability that the object transfers from state 2 to state 2 under the influence of random external event x, P 2n (x) is the probability that the object transfers from the second state to the nth state under the influence of random external event x, P n1 (x) is the probability that the object transfers from the nth state to the first state under the influence of random external event x, P n2 (x) is the probability that the object transfers from the nth state to the second state under the influence of random external event x, P nn (x) is the probability that the object will transfer from the nth state to the nth state under the influence of the random external event x.

[0198] Define the row vector N(t) as the actual state distribution of an object at time t:

[0199]

[0200] In the above formula, N(t) is a row vector, P[S(t) = s1] is the probability that S(t) = s1, P[S(t) = s2] is the probability that S(t) = s2, and P[S(t) = s n ] is S(t)=s n The probability of , T is the transpose;

[0201] In t i At this moment, the state distribution probability of the object is as follows:

[0202]

[0203] In the above formula, N(t i ) is t i The state distribution probability vector of the object at the moment, N(t i-1 ) is t i-1 The state distribution probability vector of the object at that moment.

[0204] The state distribution of the object in the simulation model belongs to discrete distribution, and the possible value of the random variable S is s k (k=1,2,…,n), n is the total number of states; among them, s kis the kth state, which is also the possible value of the random variable S; S is the probability of taking each possible value:

[0205] P{S=s k =p k , k=1,2,…,n (24)

[0206] In the above formula, p k For sampling k The probability of P[S = s k ] is S = s k The probability of

[0207] For discrete distribution, direct sampling method can be used to determine the state of the object. The direct sampling method is as follows:

[0208]

[0209] In the above formula, S F is the final sampling result, δ is the random number of direct sampling, and D is the sampling method; s I is the I-th state, which is also the possible value of the random variable S; i∈[1,I], I≤n, n is the total number of states, p i For sampling i The probability of i is the i-th state.

[0210] Then the state transfer function F in the object state transfer model is:

[0211] F:s(t i )=D[N(t i ),δ] (26)

[0212] In the above formula, s(t i ) is t i The object state is finally determined by direct sampling method at the moment, D is the sampling method, N(t i ) is t i The state distribution probability vector of the object at the moment, δ is the random number sampled directly.

[0213] Based on the above analysis, the transfer result of the object state in the object state transfer modeling process can be expressed by formula (27), which describes the state transfer law and state transfer output of the object under working condition driving.

[0214]

[0215] In the above formula, s(t i) can be used as a prediction model, that is, given the state S at the current moment, in the absence of external event input, the object state sequence of T time segments in the future can be obtained at the current moment, so as to switch the model; F(·) is an n-dimensional function, s(t i-1 ) is the tth i-1 The state at the moment, y is the state function, y is the output state, and x is a random external event.

[0216] In view of the characteristics of the power electronic distribution system with numerous grid equipment and dense lines and loads, according to the modeling method of power electronic equipment such as distributed photovoltaics, charging piles, and energy storage, and according to the calculation process of the time domain simulation method, the physical equation relationship in the power electronic component model is partially replaced by the data-driven model, giving full play to the advantages of the hybrid drive of the data model and the physical model. During the calculation process, the physical-driven model will correct the calculation results of the data-driven model, thereby avoiding the error superposition problem caused by the entire power electronic component being completely driven by data.

[0217] In view of the high dimensionality and complex structure of power electronic converters, the temporal and spatial uncertainty of distributed resources is analyzed, and the model adaptation and parameter correction methods of distribution systems under dynamic time-varying conditions are studied. Model adaptation includes hierarchical modeling and model switching, and dynamic switching of models to cope with changes in system status and demand. Parameter correction is the parameter optimization of data fitting. By selecting a suitable optimization algorithm, a set of optimal parameter vectors is found to minimize the predetermined objective function value and minimize the model error.

[0218] By integrating various component modeling methods and correction methods, a distribution system electromagnetic transient simulation model library is constructed to meet the needs of different business scenarios. In various fault recovery, optimization strategies, and comprehensive management conditions of the power electronic distribution network, measured data correction, iterative optimization, and sensitivity analysis are performed. Measured data correction adjusts model parameters through real-time data, iterative optimization uses algorithms to adjust parameters to reduce prediction errors, and sensitivity analysis determines key parameters and performs key corrections. At the same time, model adaptation requires dynamic switching of model levels according to system status, and time-varying adjustments to adapt to dynamic load and power generation conditions to improve the accuracy and real-time performance of the model.

[0219] The present invention can analyze the spatiotemporal uncertainty of distributed resources and study the model adaptation and parameter correction methods of distribution systems under dynamic time-varying conditions in view of the high dimensionality and complex structure of power electronic converters. Model adaptation includes hierarchical modeling and model switching. The appropriate model level is selected according to the analysis requirements and goals. The upper layer uses a simplified model for system evaluation, and the lower layer uses a detailed model for detailed analysis. At the same time, different models are dynamically switched to cope with changes in system status and demand. The present invention meets the needs of different business scenarios by constructing a distribution system data-model hybrid-driven electromagnetic transient simulation model library, and by constructing a multi-layer model library, it ensures that the model can cover different analysis requirements from macro to micro.

[0220] Through the above method and steps, the present invention finally constructs a comprehensive and accurate electromagnetic transient simulation model library for power electronic distribution network, which can support the simulation and effect verification of various fault recovery, optimization strategies, comprehensive management, etc. of power electronic distribution network, thereby providing support for the safe and reliable operation of power electronic distribution network.

[0221] Embodiment 2

[0222] The present invention also provides a device for constructing a power distribution system simulation model library based on a data model hybrid drive, such as Figure 4 As shown, including:

[0223] An acquisition unit, used to analyze the physical characteristics of each power electronic device in the power distribution system and obtain a physical driving model of each power electronic device;

[0224] A first establishing unit is used to establish a data driven model of each power electronic device by using a physical driving model of each power electronic device;

[0225] The second establishing unit is used to establish an electromagnetic transient model of each power electronic device by using the physical driving model of each power electronic device and the data driving model of each power electronic device, and to construct a distribution system simulation model library by using the electromagnetic transient model of each power electronic device;

[0226] The model switching unit is used to establish an object state transfer model of the distribution system simulation model library, and use the object state transfer model to switch the model of the distribution system simulation model library to perform deduction simulation of the distribution system.

[0227] Furthermore, the acquisition unit is specifically used for:

[0228] Analyze the physical characteristics of each power electronic device in the power distribution system and obtain the relevant parameters of each power electronic device;

[0229] Based on the laws of physics, the functional relationship between relevant parameters is established by formula derivation to obtain the equations of each power electronic device;

[0230] The equations for each power electronic device are the physics-driven model.

[0231] Furthermore, the first establishing unit includes:

[0232] A simulation module is used to simulate multiple preset power distribution scenarios to obtain simulation data;

[0233] An extraction module, used for extracting relevant data of a differential equation group in a physical driving model from simulation data as training data;

[0234] The training module is used to train the improved neural network model using the training data to obtain the trained improved neural network model, and the trained improved neural network model is a data-driven model.

[0235] Furthermore, the relevant data of the differential equation group includes: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

[0236] Furthermore, the training module is specifically used to:

[0237] The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain an improved neural network model after training.

[0238] Furthermore, the first establishing unit further includes: a first acquisition module, used to acquire an improved neural network model; the first acquisition module is specifically used to:

[0239] The particle swarm optimization algorithm is used to optimize the weights and thresholds of the neural network model to obtain an improved neural network model.

[0240] Furthermore, the second establishing unit includes:

[0241] A second acquisition module is used to replace the differential equations in the physical drive model with the data drive model, and then perform hybrid drive with the physical drive model to obtain a hybrid drive model;

[0242] The correction module is used to perform parameter correction on the hybrid drive model to obtain a corrected hybrid drive model, where the corrected hybrid drive model is an electromagnetic transient model.

[0243] Furthermore, the correction module is specifically used for:

[0244] Collect input data for the hybrid drive model;

[0245] Input data to the power distribution system and output the actual value of active power and the actual value of reactive power;

[0246] Inputting the input data into the hybrid drive model to obtain active power simulation values ​​and reactive power simulation values;

[0247] Using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function;

[0248] Taking the minimum function value of the target loss function as the goal, the extended Kalman filter method is used to optimize the target loss function to correct the model parameters of the hybrid drive model and obtain the corrected hybrid drive model.

[0249] Furthermore, the model switching unit comprises:

[0250] A module is established, which is used to take the electromagnetic transient model in the distribution system simulation model library as an object and use the Markov model to establish the object state transfer model of the distribution system simulation model library.

[0251] It can be understood that the device embodiment provided above corresponds to the method embodiment described above, and the corresponding specific contents can be referenced to each other and will not be repeated here.

[0252] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0253] Embodiment 3

[0254] like Figure 5 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected via a bus; the memory may be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory may also be used to store data, which may be called and / or modified when the instructions are executed.

[0255] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in a storage medium to implement corresponding method flows or corresponding functions, so as to implement the steps of a method for constructing a distribution system simulation model library based on a data model hybrid drive in the above-mentioned embodiment.

[0256] Embodiment 4

[0257] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in electronic devices and, of course, extended storage media supported by electronic devices. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage. The processor loads and executes one or more instructions stored in the storage medium, which can implement the steps of a method for constructing a distribution system simulation model library based on a data model hybrid drive in the above embodiment.

[0258] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0259] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0260] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0261] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0262] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a power distribution system simulation model library based on data model hybrid drive, characterized in that: include: Analyzing the physical characteristics of each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device; Using the physical drive model of each power electronic device, a data drive model of each power electronic device is established; Using the physical drive model of each power electronic device and the data drive model of each power electronic device, an electromagnetic transient model of each power electronic device is established, and a distribution system simulation model library is constructed using the electromagnetic transient model of each power electronic device; An object state transfer model of the power distribution system simulation model library is established, and the object state transfer model is used to switch the models of the power distribution system simulation model library to perform deduction simulation of the power distribution system.

2. The method according to claim 1, characterized in that The physical characteristics of each power electronic device in the power distribution system are analyzed to obtain a physical driving model of each power electronic device, including: Performing physical characteristic analysis on each power electronic device in the power distribution system to obtain relevant parameters of each power electronic device; Based on the laws of physics, a functional relationship between the relevant parameters is established by formula derivation to obtain equations of the power electronic devices; The equations of each power electronic device are the physical driving model.

3. The method according to claim 1, characterized in that The method of using the physical drive model of each power electronic device to establish a data drive model of each power electronic device includes: Simulate multiple preset power distribution scenarios to obtain simulation data; Extracting relevant data of the differential equation group in the physical driving model from the simulation data as training data; The improved neural network model is trained using the training data to obtain a trained improved neural network model, and the trained improved neural network model is the data-driven model.

4. The method according to claim 3, characterized in that The relevant data of the differential equation group include: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

5. The method according to claim 4, characterized in that The step of training the improved neural network model using the training data comprises: The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain the trained improved neural network model.

6. The method according to claim 3, characterized in that The process of acquiring the improved neural network model includes: The weights and thresholds of the neural network model are optimized using a particle swarm optimization algorithm to obtain the improved neural network model.

7. The method according to claim 2, characterized in that The method of using the physical drive model of each power electronic device and the data drive model of each power electronic device to establish the electromagnetic transient model of each power electronic device includes: After replacing the differential equations in the physical driving model with the data driving model, hybrid driving is performed with the physical driving model to obtain a hybrid driving model; Parameter correction is performed on the hybrid drive model to obtain a corrected hybrid drive model, where the corrected hybrid drive model is the electromagnetic transient model.

8. The method according to claim 7, characterized in that The performing parameter correction on the hybrid drive model includes: collecting input data of the hybrid drive model; Inputting the input data into the power distribution system, and outputting the actual value of active power and the actual value of reactive power; Inputting the input data into the hybrid drive model to obtain an active power simulation value and a reactive power simulation value; Using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function; Taking the minimum function value of the target loss function as the goal, the target loss function is optimized using an extended Kalman filter method to correct the model parameters of the hybrid drive model to obtain a corrected hybrid drive model.

9. The method according to claim 1, characterized in that: The object state transfer model of establishing the power distribution system simulation model library comprises: The electromagnetic transient model in the power distribution system simulation model library is taken as an object, and the object state transfer model of the power distribution system simulation model library is established by using the Markov model.

10. A device for constructing a power distribution system simulation model library based on data model hybrid drive, characterized in that: include: An acquisition unit, configured to analyze the physical characteristics of each power electronic device in the power distribution system to obtain a physical driving model of each power electronic device; A first establishing unit, configured to establish a data driven model of each power electronic device by using a physical driven model of each power electronic device; A second establishing unit is used to establish an electromagnetic transient model of each power electronic device by using the physical driving model of each power electronic device and the data driving model of each power electronic device, and to construct a distribution system simulation model library by using the electromagnetic transient model of each power electronic device; The model switching unit is used to establish an object state transfer model of the power distribution system simulation model library, and use the object state transfer model to switch the model of the power distribution system simulation model library to perform deduction simulation of the power distribution system.

11. The device according to claim 10, characterized in that The acquisition unit is specifically used for: Performing physical characteristic analysis on each power electronic device in the power distribution system to obtain relevant parameters of each power electronic device; Based on the laws of physics, a functional relationship between the relevant parameters is established by formula derivation to obtain equations of the power electronic devices; The equations of each power electronic device are the physical driving model.

12. The device according to claim 10, characterized in that The first establishing unit comprises: A simulation module is used to simulate multiple preset power distribution scenarios to obtain simulation data; An extraction module, used to extract relevant data of the differential equation group in the physical drive model from the simulation data as training data; The training module is used to train the improved neural network model using the training data to obtain the trained improved neural network model, and the trained improved neural network model is the data-driven model.

13. The device according to claim 12, characterized in that The relevant data of the differential equation group include: the calculation results of the differential equation group, and the state variables and algebraic variables required for calculating the differential equation group.

14. The device according to claim 13, characterized in that The training module is specifically used for: The improved neural network model is trained using the state variables and algebraic variables in the training data as input layer training samples of the improved neural network model, and the calculation results of the differential equation group in the training data as output layer training samples of the improved neural network model to obtain the trained improved neural network model.

15. The device according to claim 12, characterized in that The first establishing unit further includes: a first acquisition module, used to acquire the improved neural network model; the first acquisition module is specifically used to: The weights and thresholds of the neural network model are optimized using a particle swarm optimization algorithm to obtain the improved neural network model.

16. The device according to claim 11, characterized in that The second establishing unit comprises: A second acquisition module is used to replace the differential equations in the physical drive model with the data drive model, and then perform hybrid drive with the physical drive model to obtain a hybrid drive model; The correction module is used to perform parameter correction on the hybrid drive model to obtain a corrected hybrid drive model, where the corrected hybrid drive model is the electromagnetic transient model.

17. The device according to claim 16, characterized in that The correction module is specifically used for: collecting input data of the hybrid drive model; Inputting the input data into the power distribution system, and outputting the actual value of active power and the actual value of reactive power; Inputting the input data into the hybrid drive model to obtain an active power simulation value and a reactive power simulation value; Using the error between the actual value of active power and the simulated value of active power and the error between the actual value of reactive power and the simulated value of reactive power as the target loss function; Taking the minimum function value of the target loss function as the goal, the target loss function is optimized using an extended Kalman filter method to correct the model parameters of the hybrid drive model to obtain a corrected hybrid drive model.

18. The device according to claim 10, characterized in that The model switching unit comprises: The module is used to establish the object state transfer model of the power distribution system simulation model library by using the electromagnetic transient model in the power distribution system simulation model library as an object and utilizing the Markov model.

19. An electronic device, characterized in that: include: at least one processor and memory; The memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method for constructing a distribution system simulation model library based on data model hybrid drive as described in any one of claims 1 to 9 is implemented.

20. A readable storage medium, characterized in that: An execution program is stored thereon, and when the execution program is executed, the method for constructing a distribution system simulation model library based on data model hybrid drive as described in any one of claims 1 to 9 is implemented.

Images