Key section identification method and system for factors hindering new energy consumption

Abstract

The present invention discloses a method for identifying key sections of factors hindering new energy consumption, comprising: obtaining a collection of historical data on factors hindering new energy consumption; establishing a neural network model of hindering factors, and training the neural network model of hindering factors according to the collection of historical data; finding out the hindering factors for new energy consumption through correlation analysis based on the trained neural network model of hindering factors. Compared with the current offline section selection method of power grids that relies on manual experience, the technology of the present invention eliminates the process of regular manual section selection, and only needs to import data from the power grid historical database, and intelligently identifies the blocked sections and hindering factors for new energy through the model, which greatly saves manpower and improves efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid monitoring, and in particular relates to a method and system for identifying key sections of factors hindering new energy consumption. Background Art

[0002] In recent years, with the development of social economy, my country's new energy such as wind power and photovoltaics has developed rapidly. New energy has become one of the important engines of my country's energy transformation. In the critical period of my country's energy transformation, with the continuous increase in the scale of new energy access and the continuous construction and operation of ultra-high voltage AC and DC transmission channels, the operating characteristics of long-distance, chain-type relay transmission channels are becoming more and more complex, and the time and space correlation between channel capacity and new energy output is becoming closer. In particular, the gradual increase in the penetration rate of new energy power generation, the volatility, low controllability and incomplete predictability of new energy power generation due to the influence of natural factors have brought obvious uncertainty factors to long-distance and large-scale new energy transmission. Long-distance cascade transmission channels have problems such as explosion of uncertain scenario combinations, complex factors restricting the export of new energy, and difficulty in time and space coordination of cascade transmission channels. The problem of new energy consumption is becoming more and more severe. The rapid development of new energy and the continuous increase in installed capacity have further aggravated the uncertainty and diversity of power grid operation.

[0003] In the current grid operation and management mode, key sections are generally manually selected by grid operation experts based on long-term grid operation experience and offline analysis of the grid. On this basis, operators calculate the section limit transmission capacity TTC (Total Transfer Capability) based on some typical and extreme grid operation modes through offline simulation methods. Dispatchers monitor the overload and over-limit conditions in the above sections online, and adjust the grid operation mode according to operation experience to ensure that the section flow is less than the section TTC. The sections and their control rules determined by manual experience are organized and stored as knowledge of safe grid operation to form a knowledge base for safe grid operation, which further guides the safe, stable and economical operation of the grid.

[0004] This traditional method of determining critical sections using manual experience usually only reflects the weak links of the power grid under extreme operating modes, and is not very adaptable to changes in online operating modes. With the rapid development of the economy, the scale of the power system has gradually expanded, and the penetration rate of new energy has gradually increased, resulting in increasingly variable operating modes of the power system. As an important safety feature of the power system, critical sections may also change frequently, and the key factors that affect the obstruction of energy consumption also change frequently. Therefore, the traditional method of determining critical sections of the power grid using manual experience can no longer adapt to the rapid changes in the operation mode of the power grid. First, the workload of offline section selection relying on manual experience is large. For example, in Guangdong Central Dispatching, there are nearly 200 long-term sections and more than 1,000 temporary sections per year. It is time-consuming and laborious to find these sections manually. Second, the section control rules formulated based on manual experience are conservative and simple, lacking quantitative and lean management and control, and cannot maximize the transmission capacity of the transmission channel, affecting the utilization rate of network resources. The calculation of section limits is subject to a variety of stability checks. Traditional solutions are difficult to determine key influencing factors, and the calculation method cannot meet actual needs in terms of efficiency, making it very difficult to flexibly release the section transmission potential and improve the level of new energy consumption.

[0005] Dispatching and operating personnel are unable to accurately tap the power transmission potential of the renewable energy transmission section and can only adopt overly conservative operating restrictions, resulting in a large waste of section transmission capacity and unnecessary disconnection or restriction of renewable energy. Summary of the invention

[0006] In order to solve the problems existing in the prior art, the present invention provides a method for identifying key sections of factors hindering the consumption of new energy, which can automatically identify the key sections of factors hindering the consumption of new energy.

[0007] The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0008] First, a method for identifying key sections of factors hindering the consumption of new energy is provided, including:

[0009] Obtain a collection of historical data on factors hindering the consumption of new energy;

[0010] Establish a neural network model of blocking factors, and train the neural network model of blocking factors based on the historical data collection;

[0011] Based on the trained neural network model of obstruction factors, the obstruction factors for new energy consumption are found out through correlation analysis.

[0012] In combination with the first aspect, further, the training of the obstruction factor neural network model according to the historical data collection includes:

[0013] Determine the training samples of the neural network model of the blocking factors based on the historical data collection;

[0014] The training samples are used to train the neural network model of the obstruction factors;

[0015] The obstruction factor neural network model adopts a back-propagation neural network model, wherein the optimizer adopts an adam optimizer and the optimization algorithm adopts a stochastic gradient descent algorithm.

[0016] In combination with the first aspect, further, the number of hidden layers of the back propagation neural network model is set to 1, and the number of neurons in the hidden layer satisfies one or more of the following conditions:

[0017] 1) The number of hidden neurons should not be greater than twice the number of input neurons;

[0018] 2) The number of hidden neurons is between the number of input neurons and the number of output neurons;

[0019] 3) The number of hidden neurons can be two-thirds of the total number of input and output neurons.

[0020] In combination with the first aspect, further, obtaining training samples of the obstruction factor neural network model according to the historical data collection includes:

[0021] Extract data for a certain period of time from the historical data collection;

[0022] The subdivided data of the same factors in the extracted data of a certain time period are merged according to the factors to obtain the training sample set of the neural network model of the blocked factors.

[0023] In combination with the first aspect, further, the factors hindering the consumption of new energy sources found out through correlation analysis include:

[0024] Construct the influencing factor matrix X of new energy consumption and the blocked power Y of new energy;

[0025] X and Y are used as the input matrix and output vector of the obstruction factor neural network model respectively;

[0026]

[0027] Y=[y 1 y 2 …y m ] T

[0028] Among them, n is the number of factors affecting the consumption of new energy, m is the total number of samples, and x is mn represents the mth sample value of the nth factor affecting the consumption of new energy, y m Indicates the amount of new energy obstruction corresponding to the mth sample value;

[0029] Each influencing factor in the influencing factor matrix X of new energy consumption is increased or decreased by δ on its own basis to generate two new sets of input quantities, namely X + i and X - i ;

[0030]

[0031]

[0032] X + i and X - i Input the obstruction factor neural network model to obtain two sets of new output vectors Y + i and Y - i ;

[0033]

[0034]

[0035] Among them, δ is the regulation rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor increases according to the adjustment rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor is reduced according to the adjustment rate;

[0036] The sensitivity of the influencing factors is calculated by the following formula

[0037]

[0038] I MIV，i represents the sensitivity of the i-th influencing factor;

[0039] The contribution of the influencing factors is calculated by the following formula:

[0040]

[0041] C i Indicates the contribution of the i-th influencing factor;

[0042] Sort all influencing factors from high to low, take the first k influencing factors as key influencing factors, and determine the corresponding sections based on the key influencing factors.

[0043] Secondly, a key section identification system for factors hindering the consumption of new energy is provided, including:

[0044] A data acquisition module is used to obtain a collection of historical data on factors hindering the consumption of new energy;

[0045] A modeling module, used to establish a neural network model of obstruction factors and train the neural network model of obstruction factors according to a collection of historical data;

[0046] The section identification module is used to find out the obstacles to the consumption of new energy through correlation analysis based on the trained neural network model of obstacles.

[0047] In combination with the second aspect, further, the modeling module includes:

[0048] A sample determination module, used to determine the training samples of the obstruction factor neural network model based on the historical data collection;

[0049] Model training module, using training samples to train the neural network model of the obstruction factor;

[0050] The obstruction factor neural network model adopts a back-propagation neural network model, wherein the optimizer adopts an adam optimizer optimization algorithm and adopts a stochastic gradient descent algorithm.

[0051] In combination with the second aspect, further, the number of hidden layers of the back propagation neural network model in the modeling module is set to 1, and the number of neurons in the hidden layer satisfies one or more of the following conditions:

[0052] 1) The number of hidden neurons should not be greater than twice the number of input neurons;

[0053] 2) The number of hidden neurons is between the number of input neurons and the number of output neurons;

[0054] 3) The number of hidden neurons can be two-thirds of the total number of input and output neurons.

[0055] In combination with the second aspect, further, the section recognition module includes:

[0056] Vector construction module, used to construct the influencing factor matrix X of new energy consumption and the blocked power Y of new energy;

[0057] X and Y are used as the input matrix and output vector of the obstruction factor neural network model respectively;

[0058]

[0059] Y=[y 1 y 2 … y m ] T

[0060] Among them, n is the number of factors affecting the consumption of new energy, m is the total number of samples, and x is mnrepresents the mth sample value of the nth factor affecting the consumption of new energy, y m Indicates the amount of new energy obstruction corresponding to the mth sample value;

[0061] Each influencing factor in the influencing factor matrix X of new energy consumption is increased or decreased by δ on its own basis to generate two new sets of input quantities, namely X + i and X - i ;

[0062]

[0063]

[0064] X + i and X - i Input the obstruction factor neural network model to obtain two sets of new output vectors Y + i and Y - i ;

[0065]

[0066]

[0067] Among them, δ is the regulation rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor increases according to the adjustment rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor is reduced according to the adjustment rate;

[0068] The sensitivity calculation module is used to calculate the sensitivity of the influencing factors through the following formula

[0069]

[0070] I MIV，i represents the sensitivity of the i-th influencing factor;

[0071] The contribution calculation module is used to calculate the contribution of the influencing factors through the following formula

[0072]

[0073] C i Indicates the contribution of the i-th influencing factor;

[0074] The section confirmation module is used to sort all the influencing factors from high to low, take the first k influencing factors as key influencing factors, and determine the corresponding section according to the key influencing factors.

[0075] The beneficial effects of the present invention are mainly as follows:

[0076] Compared with the current offline section selection method of power grids that relies on manual experience, the technology of the present invention eliminates the process of regular manual section selection. It only needs to import data from the power grid historical database and intelligently identify the blocked sections and blocking factors of new energy through the model, which greatly saves manpower and improves efficiency.

[0077] More reasonable section control rules. The previous rules adopted relatively conservative section control rules, which in many cases would overly limit the transmission potential of new energy sections. At the same time, in certain special operating modes, overly optimistic results may appear, posing safety hazards. The technology of the present invention adopts BP neural network, and the results are based on the big data of the power grid. Through multiple training of artificial intelligence algorithms, the key influencing factors and key sections that hinder new energy are analyzed, providing a basis for the targeted formulation of section control rules, which is more reasonable than the previous method.

[0078] A computer storage medium, wherein a computer program is stored in the computer-readable storage medium, and the computer program is suitable for being loaded and executed by a processor, so that a computer device having the processor executes the method according to any one of claims 1 to 5.

[0079] The neural network model has high accuracy. When the number of samples is sufficient, the data can be mined through deep learning by establishing a neural network. Compared with some traditional algorithms or some machine learning algorithms, the obtained model is more accurate. The model has good adaptability, self-organization, strong learning and fault tolerance and anti-interference capabilities, and can achieve highly nonlinear mapping from input to output, thereby improving the accuracy of correlation analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is a flow chart of the present invention;

[0081] Figure 2 A flow chart of section identification in the present invention;

[0082] Figure 3 It is a schematic diagram of the correlation coefficient results of the factors hindering energy consumption in the present invention. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solution and advantages of the embodiments 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 described embodiments are part of the embodiments of the present invention, not 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.

[0084] In order to better understand the present invention, the relevant technologies in the technical solution of the present invention are explained below.

[0085] Example 1

[0086] like Figure 1-3 As shown, a key section identification method of a new energy consumption obstruction factor of the present invention comprises the following steps:

[0087] Step 1: Analyze the cascade relationship of the blocked surfaces, sort out the main blocked surfaces, and combine the blocked surface grid diagram of the region to obtain a collection S1 of historical data on possible factors of new energy obstruction. S1 includes the real-time power of AC lines related to new energy obstruction, the real-time power of substation main transformers and the corresponding limits, as well as the real-time power of affected DC lines, the overall real-time load of the region, the overall real-time power of regional power generation, the overall real-time power of regional thermal power and hydropower generation, etc.

[0088] Step 2: Select data within a certain time period from the collection S1 of historical data (this article uses the power grid data of a certain region throughout 2020, with a sampling interval of 15 minutes).

[0089] The data in the selected time period are processed for missing and anomalies, and then the processed data are merged according to factors to generate training data. For example, "Real-time power on the high-voltage side of No. 1 main transformer of the 750 kV Yellow River substation" and "Real-time power on the high-voltage side of No. 2 main transformer of the 750 kV Yellow River substation" are both data (factors) of the 750 kV main transformer of the Yellow River. Here, the above data are summed up and merged into the same data to generate the training sample set required for the neural network model of factors hindering the consumption of new energy.

[0090] Step three, train the neural network model of new energy consumption obstruction factors through the obtained training sample set (including influencing factors as input and new energy consumption obstruction as output), and the neural network model adopts the back propagation neural network model (Backpropagation neural network), in which the optimizer adopts the adam optimizer and the optimization algorithm adopts the stochastic gradient descent algorithm.

[0091] In order to further improve the accuracy of the neural network, the number of hidden layers is set to 1, and the number of neurons in the hidden layer meets one or more of the following conditions:

[0092] 1) The number of hidden neurons should not be greater than twice the number of input neurons;

[0093] 2) The number of hidden neurons is between the number of input neurons and the number of output neurons;

[0094] 3) The number of hidden neurons can be two-thirds of the total number of input and output neurons.

[0095] Regularization can be gradually added during the training process to prevent overfitting, and training is stopped when the accuracy of the model reaches 97%.

[0096] Step 4: Based on the trained neural network model of obstruction factors, find out the obstruction factors for new energy consumption through correlation analysis.

[0097] Construct the influencing factor matrix X (input neuron) of new energy consumption and the blocked power Y (output neuron) of new energy;

[0098] X and Y are used as the input matrix and output vector of the obstruction factor neural network model respectively;

[0099]

[0100] Y=[y 1 y 2 … y m ] T

[0101] Among them, n is the number of factors affecting the consumption of new energy, m is the total number of samples, and x is mn represents the mth sample value of the nth factor affecting the consumption of new energy, y m Indicates the amount of new energy obstruction corresponding to the mth sample value;

[0102] Each influencing factor in the influencing factor matrix X of new energy consumption is increased or decreased by δ on its own basis to generate two new sets of input quantities, namely X + i and X - i ;

[0103]

[0104]

[0105] X + i and X - i Input the obstruction factor neural network model to obtain two sets of new output vectors Y + i and Y- i ;

[0106]

[0107]

[0108] Among them, δ is the regulation rate, which is 10% here. It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor increases according to the adjustment rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor is reduced according to the adjustment rate;

[0109] The sensitivity of the influencing factor is obtained by taking the average of the differences between the results after each factor changes.

[0110]

[0111] I MIV，i represents the sensitivity of the i-th influencing factor;

[0112] The contribution of the influencing factors is calculated by the following formula:

[0113]

[0114] C i Indicates the contribution of the i-th influencing factor;

[0115] All influencing factors are sorted from high to low, and the first k (k=3) influencing factors are taken as key influencing factors. The corresponding sections are determined according to the key influencing factors, and each key influencing factor corresponds to a section.

[0116] Example 2

[0117] The present invention also provides a key section identification system for factors hindering the consumption of new energy, including:

[0118] A data acquisition module is used to obtain a collection of historical data on factors hindering the consumption of new energy;

[0119] A modeling module, used to establish a neural network model of obstruction factors and train the neural network model of obstruction factors according to a collection of historical data;

[0120] The section identification module is used to find out the obstacles to the consumption of new energy through correlation analysis based on the trained neural network model of obstacles.

[0121] The aforementioned modeling modules include:

[0122] A sample determination module, used to determine the training samples of the obstruction factor neural network model based on the historical data collection;

[0123] Model training module, using training samples to train the neural network model of the obstruction factor;

[0124] The obstruction factor neural network model adopts a back-propagation neural network model, wherein the optimizer adopts an adam optimizer and the optimization algorithm adopts a stochastic gradient descent algorithm.

[0125] The present invention has been verified in the laboratory. After preprocessing the power grid data of a certain region throughout 2020, it was sorted into a collection of 16 factors that may affect the amount of regional renewable energy obstruction. After calculating these 16 input factors using this method, the results are as follows: Figure 3 As shown in the figure, the horizontal axis ap is the 16 possible factors that may affect the regional new energy obstruction amount, and the vertical axis is the correlation coefficient of each factor on the overall regional new energy obstruction amount, thus concluding that the three factors m, h, and i are the three factors with the greatest impact on the regional new energy obstruction amount.

[0126] A computer storage medium, wherein a computer program is stored in the computer-readable storage medium, and the computer program is suitable for being loaded and executed by a processor, so that a computer device having the processor executes the method according to any one of claims 1 to 5.

[0127] 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.

[0128] 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.

[0129] 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 comprising 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.

[0130] 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. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A method for identifying key sections of factors hindering the consumption of new energy. It is characterized in that include: Obtain a collection of historical data on factors hindering the consumption of new energy; Establish a neural network model of blocking factors, and train the neural network model of blocking factors based on the historical data collection; Based on the trained neural network model of obstruction factors, the obstruction factors of new energy consumption are found through correlation analysis; The training of the obstruction factor neural network model according to the historical data collection includes: Determine the training samples of the neural network model of the blocking factors based on the historical data collection; The training samples are used to train the neural network model of the obstruction factors; The obstruction factor neural network model adopts a back propagation neural network model, wherein the optimizer adopts an adam optimizer and the optimization algorithm adopts a stochastic gradient descent algorithm; The factors hindering the consumption of new energy sources found through correlation analysis include: Construct the influencing factor matrix X of new energy consumption and the blocked power Y of new energy; X and Y are used as the input matrix and output vector of the obstruction factor neural network model respectively; And=[and 1 and 2 …and m ] T Among them, n is the number of factors affecting the consumption of new energy, m is the total number of samples, and x is mn represents the mth sample value of the nth factor affecting the consumption of new energy, y m Indicates the amount of new energy obstruction corresponding to the mth sample value; Each influencing factor in the influencing factor matrix X of new energy consumption is increased or decreased by δ on its own basis to generate two new sets of input quantities, namely X + i and X - i ; X + i and X - i Input the obstruction factor neural network model to obtain two sets of new output vectors Y + i and Y - i ; Among them, δ is the regulation rate, It indicates the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor increases according to the adjustment rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor is reduced according to the adjustment rate; The sensitivity of the influencing factors is calculated by the following formula I MIV,i represents the sensitivity of the i-th influencing factor; The contribution of the influencing factors is calculated by the following formula: C i Indicates the contribution of the i-th influencing factor; Sort all influencing factors from high to low, take the first k influencing factors as key influencing factors, and determine the corresponding sections based on the key influencing factors.

2. According to claim 1, a method for identifying key sections of factors hindering the consumption of new energy, It is characterized in that The number of hidden layers of the back propagation neural network model is set to 1, and the number of neurons in the hidden layer meets one or more of the following conditions: 1) The number of hidden neurons should not be greater than twice the number of input neurons; 2) The number of hidden neurons is between the number of input neurons and the number of output neurons; 3) The number of hidden neurons is two-thirds of the total number of input and output neurons.

3. According to claim 1, a critical section identification method for factors hindering the consumption of new energy, It is characterized in that The training samples of the neural network model of the obstruction factor obtained according to the historical data collection include: Extract data for a certain period of time from the historical data collection; The subdivided data of the same factors in the extracted data of a certain time period are merged according to the factors to obtain the training sample set of the neural network model of the blocked factors.

4. A critical section identification system for factors hindering the consumption of new energy. It is characterized in that include: A data acquisition module is used to obtain a collection of historical data on factors hindering the consumption of new energy; A modeling module, used to establish a neural network model of obstruction factors and train the neural network model of obstruction factors according to a collection of historical data; The cross-section identification module is used to find out the obstacles to the consumption of new energy through correlation analysis based on the trained neural network model of obstacles; The modeling module includes: A sample determination module, used to determine the training samples of the obstruction factor neural network model based on the historical data collection; Model training module, using training samples to train the neural network model of the obstruction factor; The obstruction factor neural network model adopts a back propagation neural network model, wherein the optimizer adopts an adam optimizer and the optimization algorithm adopts a stochastic gradient descent algorithm; The cross-section recognition module comprises: Vector construction module, used to construct the influencing factor matrix X of new energy consumption and the blocked power Y of new energy; X and Y are used as the input matrix and output vector of the obstruction factor neural network model respectively; And=[and 1 and 2 …and m ] T Among them, n is the number of factors affecting the consumption of new energy, m is the total number of samples, and x is mn represents the mth sample value of the nth factor affecting the consumption of new energy, y m Indicates the amount of new energy obstruction corresponding to the mth sample value; Each influencing factor in the influencing factor matrix X of new energy consumption is increased or decreased by δ on its own basis to generate two new sets of input quantities, namely X + i and X - i ; X + i and X - i Input the obstruction factor neural network model to obtain two sets of new output vectors Y + i and Y - i ; Among them, δ is the regulation rate, It indicates the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor increases according to the adjustment rate, It represents the amount of new energy obstruction corresponding to the mth sample after the i-th influencing factor is reduced according to the adjustment rate; The sensitivity calculation module is used to calculate the sensitivity of the influencing factors through the following formula I MIV,i represents the sensitivity of the i-th influencing factor; The contribution calculation module is used to calculate the contribution of the influencing factors through the following formula C i Indicates the contribution of the i-th influencing factor; The section confirmation module is used to sort all the influencing factors from high to low, take the first k influencing factors as key influencing factors, and determine the corresponding section according to the key influencing factors.

5. According to claim 4, a critical section identification system for factors hindering the consumption of new energy, It is characterized in that The number of hidden layers of the back propagation neural network model in the modeling module is set to 1, and the number of neurons in the hidden layer meets one or more of the following conditions: 1) The number of hidden neurons should not be greater than twice the number of input neurons; 2) The number of hidden neurons is between the number of input neurons and the number of output neurons; 3) The number of hidden neurons is two-thirds of the total number of input and output neurons.

6. A computer storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded and executed by a processor, so that a computer device having the processor executes the method according to any one of claims 1 to 3.

Images