Home substation energy storage station system

Abstract

The application discloses a household substation energy storage station system. The household substation energy storage station system comprises a wind power module, a wind power collection module, a photovoltaic module, a photovoltaic collection module, a weather module, an energy storage module, a wind-solar power prediction device and a grid-connected control module. The wind power module is used for generating power by using wind power. The wind power collection module is used for collecting historical wind power data of the wind power module. The photovoltaic module is used for generating power by using solar energy. The photovoltaic collection module is used for collecting historical photovoltaic data of the photovoltaic module. The weather module is used for collecting weather information of a location. The energy storage module is used for storing electric energy. The wind-solar power prediction device is used for predicting future wind power generation capacity based on the historical wind power data and the weather information, and predicting future photovoltaic generation capacity based on the historical photovoltaic data and the weather information. The grid-connected control module is used for predicting grid-connected power transmission capacity based on the future photovoltaic generation capacity and residual energy storage capacity of the energy storage module. In the scheme, the prediction accuracy is higher, so that the household energy storage station has high regularity of input or output electric energy when being connected to the power grid, and clean energy utilization is facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation management, in particular to a household substation energy storage station system. BACKGROUND

[0002] Small household energy storage stations are an important source of energy, which can fully utilize local wind energy and electricity to alleviate power supply pressure. However, the construction cost of small household energy storage stations is low, and the capacity of energy storage batteries in household energy storage stations is low, so during the operation of household energy storage stations, only the power generation of the energy storage station in the future time period is accurately predicted, the household energy storage station can be operated with maximum benefit to avoid prediction errors, thereby making the national power grid have a large error in predicting the regional power gap, and thus making the power supply management difficult.

[0003] The current household energy storage station is a wind and solar integrated power station. In an area, the solar power generation of each user is basically the same, but it will also be affected by aging, and the wind power of each user is affected by various factors, and the wind power has greater randomness. The main reason for such a large error in wind power is that the prediction of wind power is based on the local climate conditions, but the local climate conditions are the climate conditions of an administrative division, which has no guiding significance for independent wind power. Therefore, neither wind power nor solar power can be accurately predicted. SUMMARY

[0004] The summary part of the present application is used to introduce the concept in a simple form, which will be described in detail in the specific embodiment part. The summary part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] In order to solve the technical problem that small household energy storage stations cannot accurately predict power generation, some embodiments of the present application provide a household substation energy storage station system, which comprises:

[0006] A wind power module for generating electricity by wind power;

[0007] A wind power collection module for collecting historical wind power data of the wind power module;

[0008] A solar power module for generating electricity by solar energy;

[0009] A solar power collection module for collecting historical solar power data of the solar power module;

[0010] A weather module for collecting weather information of the place;

[0011] An energy storage module for storing electricity;

[0012] The wind and light electricity prediction device predicts future wind power generation based on historical wind power data and weather information, and predicts future light power generation based on historical light power data and weather information.

[0013] The grid-connected control module predicts the grid-connected power generation based on the future light power generation and the remaining energy storage of the energy storage module.

[0014] In the technical solution provided in the present application, for each household substation energy storage system, the future power generation is predicted using the information collected by itself. Compared with the method of predicting the power generation of all energy storage stations using weather information of an area, the accuracy of the prediction in the present application is higher, so that the input or output power of the household energy storage station is regular when connected to the grid, facilitating the use of clean energy.

[0015] Further, the wind and light electricity prediction device includes a wind power prediction module and a light power prediction module, the wind power prediction module is used to predict future wind power generation, and the light power prediction module is used to predict future light power generation.

[0016] Further, the historical wind power data includes historical power output data.

[0017] The weather information includes historical wind speed, historical wind direction, historical temperature, historical humidity and historical air pressure.

[0018] In predicting wind power generation, there is great randomness, but there is also great seasonality. Therefore, if the weather information of different seasons is treated as the same type of data for wind power generation prediction, the model may be overfitted due to the lack of actual rules in the data itself. Therefore, the present application provides the following technical solution:

[0019] Further, the historical wind speed, historical wind direction, historical temperature, historical humidity and historical air pressure are data change curves for each year on the preset date, 24 hours a day.

[0020] For each item of weather information, it is arranged into an information matrix with hours horizontally and dates vertically.

[0021] In the technical solution provided in the present application, the information in the preset date of different seasons is collected as a standard for prediction, so that the influence of the different climate change trends of the remaining seasons on the prediction accuracy can be removed during prediction.

[0022] In predicting wind and solar power generation, it is necessary to maximize accuracy. Improving accuracy depends on the complexity of the model. Large neural network models have stronger learning capabilities than small neural network models; however, setting up and training a large neural network model in each energy storage power station would lead to high costs. Therefore, this application provides the following technical solution:

[0023] The wind and solar power forecasting device selects a trained neural network model for power generation prediction based on data similarity.

[0024] In the technical solution provided in this application, in order to reduce the training cost of neural network models, multiple neural network models are pre-trained with different types of data. Then, the subsequent energy storage power station selects the pre-trained neural network model based on the similarity of the information it collects. Therefore, the technical solution provided in this application can reduce the model training cost.

[0025] Furthermore, samples H collected in advance from all wind and solar power forecasting devices were pre-collected. i For sample H i The following classification method is adopted;

[0026] S1: For H i ={x, y} is normalized, where x is the predicted information and y is the actual power generation information;

[0027] S2: Randomly select a feature from x as the separator, and then apply the separator to all samples H. i The values ​​of x are divided into two subsets: the right subset is greater than the eigenvalue, and the left subset is less than the eigenvalue.

[0028] S3: Repeat S2 for samples in the right and left subsets, and generate a new split value randomly each time.

[0029] S4: Continue to segment the sample until any of the following conditions are met;

[0030] The number of segmentations has reached the preset maximum value;

[0031] There is only one sample in the subset;

[0032] All samples in the subset have the same feature values;

[0033] S5: Each time the stopping condition is reached, an isolated tree is generated. Multiple isolated trees are obtained by dividing the sample multiple times.

[0034] For each sample H i Traverse all isolated trees and record sample H i The number of edges traversed to reach a leaf node, to obtain sample H.i The height h of the sample points i ;

[0035] Calculate H for each sample i The average height E i ;

[0036] Calculate H for each sample i The density S;

[0037]

[0038] Where c(n) represents the average path length between the separator and the group, E i Indicates sample H i The average height, S i Indicates sample H i The sparsity of , where n represents the total number of samples;

[0039] S6: For all samples H i Given the sparseness density S, generate an n×n matrix R, where each element of matrix R is S. i and S j similarity distance D i,j , where i and j are the indices of the sample, i ≠ j;

[0040]

[0041] S7: Divide the matrix R into several groups using a clustering algorithm, based on the samples H in each group. i The source of the data determines the classification criteria for wind and solar power forecasting devices.

[0042] In the technical solution provided in this application, the data collected by all wind and solar power forecasting devices are first processed by isolated tree calculation to obtain sparsity. Then, a sparsity matrix D is generated based on the sparsity, and classification is performed based on the sparsity matrix D. In the whole process, the characteristics between samples can be effectively found, the data can be accurately classified, and the clustering effect between data can be clearly highlighted.

[0043] Large-scale neural network models possess sufficient model accuracy and stronger learning capabilities, continuously increasing their prediction accuracy with enough input samples. However, large-scale neural network models involve high computational costs and slow iteration speeds, making them unsuitable for residential energy storage stations. Therefore, this application provides the following technical solution:

[0044] The wind power prediction module and the photovoltaic prediction module each have a pre-trained prediction model built in.

[0045] The prediction model is trained in a training device;

[0046] The training device comprises:

[0047] a knowledge output unit, in which a large neural network model is pre-trained;

[0048] a training unit, in which the prediction model to be trained is arranged;

[0049] The training data are respectively input into the knowledge output unit and the training unit, wherein the knowledge output unit uses the information extracted from the training data to guide the training of the prediction model built in the training unit.

[0050] In the technical solution provided in the present application, the prediction model is not a large neural network model, but a small neural network model. However, in order to achieve the accuracy of a large neural network model, a large neural network model is used to guide the training of a small prediction model. Therefore, in practice, the prediction model can have sufficient accuracy, but the calculation amount of the model is small and the prediction efficiency is high when used.

[0051] Further, for the sample H i is divided into photoelectric prediction samples and wind power prediction samples, and the photoelectric prediction samples and the wind power prediction samples are respectively used for photoelectric power generation prediction and wind power generation prediction.

[0052] When predicting the wind power generation, weather information is mainly used for comprehensive prediction to capture the changes of the weather and the influence of the weather changes on the wind power generation. Therefore, the prediction information x is mainly weather information, and the wind direction, temperature and humidity in the weather information have different influences on the power generation. Therefore, the present application provides the following technical solution:

[0053] Further, the normalization method in S1 is as follows: all the wind speed, wind direction, humidity and air pressure are preliminarily normalized, then a corresponding weight coefficient is set for each wind speed, wind direction, humidity and air pressure, and x is obtained by adding them together.

[0054] x=α1A1+α2A2+α3A3+α4A4; wherein, α1, α2, α3, α4, are the set weight coefficients.

[0055] In the technical solution provided in the present application, when calculating x, the corresponding weight coefficients are pre-set, and the influence ability of each item of information is controlled according to the weight coefficients, so that the finally obtained x can represent the effective information.

[0056] When wind power generation is predicted, weather information is very important, but there is obvious information redundancy in the weather information, and the influence between the information is very complex, and it is actually impossible to reasonably give the weight coefficient of each weather information, and then accurately describe the prediction information x by using the weight coefficient. Therefore, the technical scheme is provided in the application as follows:

[0057] Further, the weight coefficient a of each information in the weather information is calculated in the following manner:

[0058] a u =C r (u,U)P;

[0059] Wherein, P is a conversion coefficient, C r (u,U) is the correlation ratio of the u-th information and the rest of the information;

[0060]

[0061]

[0062] Wherein, u represents the u-th information, U represents all the information except u, C represents the power generation, I r (r,U) is the correlation coupling rate, which measures the additional information sharing degree between u and C under the condition that all information except u is known, and the value range is 0 to 1, I(u;C|U) represents the information sharing amount between u and C under the condition that U is known, H(u) represents the entropy of u, H(C) represents the entropy of C, T r (A,B) represents the reduction degree of information sharing amount between u and C under the condition that U is known, and I(u;C) represents the mutual information of u and C.

[0063] Wherein, the entropy is calculated in the following manner:

[0064]

[0065] Wherein, H(u) represents the entropy of the u-th information in the weather information, i represents the index of the sample, n represents the total number of samples, b=2, p(u i ) represents the probability that u takes the value of u i .

[0066] In the technical scheme provided in the application, when calculating the weight coefficient a, the correlation degree between each weather information and the rest of the weather information is calculated according to the correlation ratio, so that the internal relationship between each information can be accurately measured when the weight coefficient is converted, thereby increasing the accuracy of the weight coefficient. BRIEF DESCRIPTION OF DRAWINGS

[0067] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrations, together with the description, serve to explain the application, but are not intended to limit the application.

[0068] In addition, throughout the drawings, same or similar reference numerals are used to denote same or similar elements. It should be understood that the drawings are schematic, and elements and elements are not necessarily drawn in proportion.

[0069] In the drawings:

[0070] Figure 1 The structure of the household substation energy storage station system.

[0071] Figure 2 The information matrix of temperature.

[0072] Figure 3 The network structure diagram when the prediction model is trained. DETAILED DESCRIPTION

[0073] Embodiments of the present application will be described in more detail by referring to the drawings. Although certain embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms, and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present application can be more thoroughly and completely understood. It should be understood that the drawings and embodiments of the present application are for exemplary purposes only, and are not intended to limit the scope of protection of the present application.

[0074] In addition, it should be further noted that, for ease of description, only the parts related to the invention are shown in the drawings. The embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0075] The present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0076] Reference Figure 1 The household substation energy storage station system is a household energy storage system, which is generally used in areas where wind power and photovoltaic power are abundant. After the system is installed in a household, the household can basically achieve self-sufficiency in electric energy, and then transmit the excess electric energy to the national grid. In order to ensure the quality of electric energy, the total amount of generated electric energy needs to be accurately predicted, so that when electric energy is transmitted to or received from the national grid, the required amount of electric energy can be sent in advance, and the power grid can be conveniently scheduled.

[0077] The household substation energy storage station system comprises a wind power module, a photovoltaic module, an energy storage module and a grid-connected control module. The wind power module is used for generating power by wind; the photovoltaic collection module is used for collecting historical photovoltaic data of the photovoltaic module; the energy storage module is used for storing power; and the grid-connected control module is used for grid-connected power transmission based on future photovoltaic power generation and residual energy storage of the energy storage module. The wind power module, the photovoltaic module, the energy storage module and the grid-connected control module are the main energy storage station system, and the specific power generation mode, energy storage mode and grid-connected scheduling mode are not described here. Moreover, the grid-connected control module flexibly adjusts the operation strategy to improve the energy utilization rate, preferentially ensures the stable supply of power, and gains profit by grid-connected power transmission according to the characteristics and use conditions of solar energy, wind energy and energy storage systems under the premise of ensuring the service life of each module. The specific adjustment strategy can be adjusted according to the actual situation, and the key is to accurately predict the power generation.

[0078] Specifically, the household substation energy storage station system further comprises a wind power collection module, a photovoltaic collection module, a weather module and a wind-solar power prediction device. The wind power collection module is used for collecting historical wind power data of the wind power module; the photovoltaic collection module is used for collecting historical photovoltaic data of the photovoltaic module; the weather module is used for collecting weather information of the location; and the wind-solar power prediction device is used for predicting future wind power generation based on historical wind power data and weather information, and predicting future photovoltaic power generation based on historical photovoltaic data and weather information.

[0079] The wind-solar power prediction device comprises a wind power prediction module and a photovoltaic prediction module. The wind power prediction module is used for predicting future wind power generation, and the photovoltaic prediction module is used for predicting future photovoltaic power generation.

[0080] It should be noted that the weather module in the present scheme is installed at the location of the household substation energy storage station system, and the data is directly measured, rather than public data provided by the meteorological department. The specific weather information measurement method is not described here, and the weather information measured in this way is more representative.

[0081] In the present scheme, wind power generation and photovoltaic power generation are predicted by two independent modules, but the structure of the neural network model built in the module is the same, only the training process is different. Therefore, only the neural network model of wind power and the training process are introduced here.

[0082] The historical wind power data includes historical power output data, so the historical wind power data is a power curve that changes over time. The same is true for historical photovoltaic data.

[0083] The weather information includes historical wind speed, historical wind direction, historical temperature, historical humidity and historical air pressure. In addition to the four points mentioned above, the weather information can also include other information.

[0084] For wind speed and wind direction, in practice, according to the orientation of the wind turbine in the wind power module, the wind speed and wind direction can be converted into a vector to combine the wind speed and wind direction into one information, for example, the wind direction of the wind turbine is north-south, and the wind speed is calculated according to the vector to the orientation of the wind turbine; of course, if some wind turbines have wind direction tracking function, the wind direction factor is not considered directly.

[0085] The historical wind speed, historical wind direction, historical temperature, historical humidity and historical air pressure are the data change curves of 24 hours per day in the preset date of each year;

[0086] The preset date here is actually the local season, for example, in a certain area, the north wind blows in winter and the southeast wind blows in spring, because there are two obvious differences, the data of winter and spring are selected respectively, the wind power prediction module is trained in winter, and then the wind power prediction is carried out by using the wind power prediction module, and the spring is predicted by using the spring data, and then the wind power prediction is carried out by using the spring model.

[0087] Therefore, in the present scheme, different models need to be selected for training and prediction. How to replace the model of the wind power prediction module is the prior art, and the present scheme only provides a training method of a model. The structure of the model of different seasons is the same, and under the condition of knowing a model structure and training method, the model of the remaining seasons can also be easily obtained.

[0088] For each item of weather information, it is arranged into an information matrix with hours horizontally and dates vertically;

[0089] When performing wind power prediction, it is generally predicted in units of days, so the weather information needs to be arranged. After the arrangement is completed, a plurality of information matrices can be obtained, taking the information matrix obtained by the air temperature as an example;

[0090] The average temperature of each hour in 24 hours is arranged horizontally in the information matrix, so that a string of row data can be obtained, and the row data of each day is collected and arranged downward to obtain an information matrix. As shown in Figure 2 .

[0091] In the present scheme, the weather information is processed in this way, and the historical wind power data is also processed in this way, so that the data format of the final weather information and the historical wind power data is the same, that is, the format of the information matrix. And with the progress of the date, the information matrix will be gradually improved. Of course, in order to ensure the accuracy of the prediction when just reaching the next season, the data of the previous years can be used as filling. When filling, the year of the filled data and the perfect generation can be determined according to the actual situation.

[0092] Further, the wind and light power prediction device selects a trained neural network model for power generation prediction based on data similarity.

[0093] Specifically, the household substation energy storage station is not one, but appears in a certain area in a piece, but because of the aging condition, maintenance habit, and product model of the power generation equipment of each household substation energy storage station, it is impossible to use the same power generation prediction logic for each substation energy storage station. Therefore, if the same trained neural network model is selected for prediction, the accuracy will be low. It is also difficult to directly determine which neural network model has high prediction accuracy by using multiple neural network models for prediction each time.

[0094] Therefore, in the present scheme, a trained neural network model is selected for power generation prediction according to data similarity. Here, the data refers to weather information and historical wind power data.

[0095] For example, there are a total of 10 energy storage stations, and the first 5 energy storage stations have high data similarity. It is indicated that the internal relationship between the historical wind power data and the weather information of the 5 energy storage stations is more similar, so the same neural network model can be used for prediction. The data similarity of the last 5 energy storage stations is different from that of the first 5 energy storage stations, so the last 5 energy storage stations and the first 5 energy storage stations can use the same neural network model structure, but different data is needed for training during training, so the final obtained neural network model has different weight coefficients. Therefore, the first five and the last five of the ten energy storage stations need to use two neural network models.

[0096] Therefore, all samples in the household substation energy storage station system need to be collected, and then similarity analysis and classification are performed.

[0097] The specific mode is as follows:

[0098] Collect all the samples H collected by the wind and light power prediction device i ;

[0099] Divide the sample H i into light and wind power prediction samples, and the light and wind power prediction samples are used for light and wind power generation prediction, respectively. Therefore, during classification, wind and light power samples cannot be mixed.

[0100] Classify the sample H i in the following way:

[0101] S1: Normalize H i ={x, y}, wherein x is prediction information and y is real power generation information.

[0102] Wherein, x is weather information, and y is historical wind power data. X and y are simplified representations.

[0103] When performing classification, because x includes multiple weather elements such as temperature, humidity, etc., x needs to be synthesized first, and x is combined into one piece of information, so that x and y can be matched with each other.

[0104] Specifically, the normalization method in S1 is as follows: all wind speeds, wind directions, humidities, and air pressures are preliminarily normalized, that is, the values are planned so that the wind speed, wind direction, humidity, and air pressure values are within the range of 0-1. In order to facilitate the description of the wind speed, the wind speed and the wind direction need to be combined into one value according to the orientation of the impeller in the wind power module. Here, the conversion of the wind speed according to the vector of the wind speed is involved, and the specific method is not described here.

[0105] Each wind speed, wind direction, humidity, and air pressure is set with a corresponding weight coefficient, and the sum is obtained as x;

[0106] x = a1A1 + a2A2 + a3A3 + a4A4; wherein a1, a2, a3, a4, are set weight coefficients.

[0107] Because in this scheme, there are only 4 weather information, so there are only 4 a, if more parameters are introduced, the number of a needs to be increased.

[0108] Because there is a huge information redundancy in weather information, it is actually difficult to accurately measure the weight term of each weather information. Therefore, the present application provides the following scheme:

[0109] The weight coefficient a of each information in the weather information is calculated as follows:

[0110] a u = C r (u,U)P;

[0111] Wherein, P is a conversion coefficient, C r (u,U) is the correlation ratio of the u-th information and the rest of the information;

[0112]

[0113] Wherein, u represents the u-th information, U represents all the rest of the information except u, C represents the power generation, I r (r,U) is the correlation coupling rate, which measures the additional information sharing degree between u and C under the condition that all information except u is known, the value range is 0 to 1, I(u;C|U) represents the information sharing amount between u and C under the condition that U is known, H(u) represents the entropy of u, H(C) represents the entropy of C, T r(A,B) represents the degree of reduction in the amount of information shared between u and C given U, and I(u;C) represents the mutual information between u and C;

[0114] The entropy is calculated as follows:

[0115]

[0116] Where H(u) represents the entropy of the u-th information in the weather information, i represents the index of the sample, n represents the total number of samples, b = 2, p(u i ) indicates that u takes the value u i The probability of.

[0117] In this scheme, by calculating the correlation between each piece of information and the rest of the information, the relationship between each piece of information and the rest of the information can be accurately described. The weight coefficients are set according to their inherent connections, so the redundancy relationship between information can be better considered when setting the weight information, thereby increasing the accuracy of the model.

[0118] S2: Randomly select a feature from x as the separator, and then apply the separator to all samples H. i The values ​​of x are divided into two subsets: the right subset is greater than the eigenvalue, and the left subset is less than the eigenvalue.

[0119] The separator value here is actually a randomly selected value. The value of x ranges from 0 to 1, so the separator value can be set to 5 initially.

[0120] S3: Repeat S2 for samples in the right and left subsets, and generate a new split value randomly each time.

[0121] S4: Continue to segment the sample until any of the following conditions are met;

[0122] The number of segmentations has reached the preset maximum value;

[0123] There is only one sample in the subset;

[0124] All samples in the subset have the same feature values;

[0125] S5: Each time the stopping condition is reached, an isolated tree is generated. Multiple isolated trees are obtained by dividing the sample multiple times.

[0126] For each sample H i Traverse all isolated trees and record sample H i The number of edges traversed to reach a leaf node, to obtain sample H. i The height h of the sample points i ;

[0127] Calculate H for each samplei The average height E i ;

[0128] Calculate H for each sample i The density S;

[0129]

[0130] Where c(n) represents the average path length between the separator and the group, E i Indicates sample H i The average height, S i Indicates sample H i The sparsity of , where n represents the total number of samples;

[0131] S6: For all samples H i Given the sparseness density S, generate an n×n matrix R, where each element of matrix R is S. i and S j similarity distance D i,j i and j are the indices of the sample, i ≠ j;

[0132] Similarity distance is the distance calculated based on similarity calculation formulas, such as Hamming distance and Euclidean distance.

[0133]

[0134] S7: Divide the matrix R into several groups using a clustering algorithm, based on the samples H in each group. i The source of the data determines the classification criteria for wind and solar power forecasting devices.

[0135] The clustering algorithm is an existing technology; the specific clustering analysis process will not be provided. After clustering analysis, all samples H can be... i The samples were divided into several categories. For example, all samples were divided into four categories. The first category contained most of the samples from the first and second energy storage stations. i If these two energy storage stations are classified as one category, then data from the same category will be used to predict the wind power value using a neural network model.

[0136] The above describes the data processing and classification methods. The following section introduces the training process and structure of the neural network model.

[0137] Specifically, the wind power prediction module and the photovoltaic prediction module each have a pre-trained prediction model built in; the prediction models have the same structure.

[0138] The prediction model is trained in a training device;

[0139] The training device includes a knowledge output unit and a training unit.

[0140] The knowledge output unit is internally provided with a large neural network model which is pre-trained. The large neural network here is a generative adversarial neural network. The basic structure is the same as that of the neural network of the prediction model, but the number of internal hidden layers and weight layers is larger, that is, the number of layers of the neural network is larger and the internal structure is more complex.

[0141] The knowledge output unit is pre-trained with a large amount of data. The large neural network model after training has stronger learning ability and can make more accurate predictions. However, the prediction time is long and the calculation amount is large, which is not suitable for large-scale installation and laying.

[0142] The training unit is provided with the prediction model to be trained;

[0143] The training data are respectively input into the knowledge output unit and the training unit, wherein the knowledge output unit uses the information extracted from the training data to guide the training of the prediction model built in the training unit.

[0144] Therefore, when predicting each small prediction model, it only needs to be loaded into the training unit and then the corresponding data is input and executed.

[0145] Further, the prediction model and the large neural network model are both generative adversarial networks.

[0146] Reference Figure 1 The prediction model includes a first generator, the large neural network model includes a second generator, and the prediction model and the large neural network model share a discriminator.

[0147] The first generator includes an input layer, a hidden layer and an output layer.

[0148] The input layer is used to receive prediction information.

[0149] The hidden layer includes a series of convolutional layers for extracting information therein.

[0150] The output layer is a single neuron fully connected layer using a Sigmoid activation function to output a probability value. This value represents the probability that the input sample is real data.

[0151] The second generator includes a large input layer and a large hidden layer, and the number of neurons in the large hidden layer is greater than that in the hidden layer.

[0152] The same data is input into the first generator and the second generator, and the intermediate features generated by the hidden layer of the second generator are finally input into the first generator to guide the final output of the hidden layer.

[0153] The hidden layer and the large hidden layer in the scheme are StyleGAN networks, and the hidden layer has a small number of neurons and a simple network structure, and the network can extract the change of information from the matrix information, and the prediction accuracy is higher after the predicted information is converted into an information matrix.

[0154] In order to ensure the accuracy of the trained model, the application provides the following loss function L:

[0155] L = λ G L G + λ d L d ;

[0156] Wherein, L G is the loss of adversarial training, L d is the loss of distillation training, λ G and λ d are the first loss weight and the second loss weight corresponding to them;

[0157]

[0158] Wherein, D S (x) represents the discrimination result of the discriminator on the real power generation information, D S (G S (y)) represents the discrimination result of the discriminator on the power generation information generated by the first generator;

[0159] L d = λ p L p + λ l L l ;

[0160] Wherein, L p is the loss of the difference between the predicted power generation information and the real power generation information in detail, and L l is the loss of the predicted power generation information and the real power generation information in the whole;

[0161] λ p and λ l are the third loss weight and the fourth loss weight corresponding to them;

[0162] L p = ||G s (x)-G t (x)||1;

[0163] L l = LP(G s (x),G t (x));

[0164] wherein x is the input prediction information, y is the generated power generation information, G s denotes the first generator, G t denotes the second generator D s denotes the discriminator; LP denotes a perceptual similarity function.

[0165] In the scheme, when the prediction model is trained, the prediction model is trained together with the large neural network model, and the output of the large neural network model is compared with the output of the prediction model, so that the law of the change of the parameters in the large neural network model can be learned in the process of training the prediction model, so as to accurately correct the structure parameters in the neural network model, thereby improving the training rate of the prediction model and increasing the model precision. In the scheme, the prediction model only includes the first generator. The discriminator is a neural network structure introduced during training.

[0166] After the first generator is trained, it is extracted and installed into the required wind power prediction module for prediction of wind power generation.

[0167] The above description is only some preferred embodiments of the present application and a description of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with the technical features disclosed in the embodiments of the present application (but not limited to) having similar functions to form technical solutions.

Claims

1. A household substation energy storage station system, comprising: a wind power module for generating electricity by wind power; a photovoltaic module for generating electricity by solar energy; an energy storage module for storing electrical energy; a wind power collection module for collecting historical wind power data of the wind power module; a photovoltaic collection module for collecting historical photovoltaic data of the photovoltaic module; characterized in that, a weather module for collecting weather information of the location; The wind and light electricity prediction device predicts future wind electricity generation based on historical wind electricity data and weather information, and predicts future light electricity generation based on historical light electricity data and weather information; all samples H i collected by the wind and light electricity prediction device are collected in advance i Classification is performed in the following manner; S1: to H i = {x, y} is normalized, where x is the prediction information and y is the real power generation information; S2: randomly select a feature from x as the split value, divide all samples H i The value of x is divided into two subsets, the right subset is greater than the feature value, and the left subset is less than the feature value; S3: repeating the above S2 for the samples in the right subset and the left subset, and randomly generating a new split value each time; S4: continuously splitting the samples until any of the following conditions is met; the number of splits reaches a preset maximum value; there is only one sample in the subset; all feature values of the samples in the subset are the same; S5: generating an isolated tree each time the stopping condition is met, and performing multiple divisions on the samples to obtain multiple isolated trees; For each sample H i , traverse all the isolated trees and record the number of edges that sample H i goes through when reaching the leaf nodes to get the height h i of the sample point of sample H i ; Calculate the average value E of the height of each sample H i i ;​ Compute the density S of each sample H i i ;​ where c(n) represents the average path length between the cut point and the cluster, E i represents the average value of the height of the sample H i i represents the sparsity of the sample H i n represents the total number of samples;​ S6: For all samples H i of density S, generate an n x n matrix R, where each element in the matrix R is the similarity distance D i and S j of H i,j , i and j are the indices of the samples, i≠j; S7: adopt clustering algorithm to divide the matrix R into several groups, determine the classification basis of the wind-solar power prediction device based on the source of each group of samples H i of samples H a grid-connected control module based on future photovoltaic power generation and remaining energy storage of the energy storage module, and grid-connected power transmission.

2. The home substation energy storage station system of claim 1, wherein: The wind-solar-power prediction device includes a wind power prediction module and a photovoltaic prediction module, the wind power prediction module is used for predicting future wind power generation, and the photovoltaic prediction module is used for predicting future photovoltaic power generation.

3. The household substation energy storage station system of claim 1, wherein: the historical wind power data includes historical power output data; the weather information includes historical wind speed, historical wind direction, historical temperature, historical humidity, and historical air pressure; the historical wind speed, the historical wind direction, the historical temperature, the historical humidity, and the historical air pressure are data change curves for each of the preset dates in a year, 24 hours a day; for each item of weather information, it is arranged into an information matrix with hours horizontally and dates vertically.

4. The home substation energy storage station system of claim 1, wherein: The wind-solar-power prediction device selects a trained neural network model based on data similarity for power generation prediction.

5. The home substation energy storage station system of claim 2, wherein: For sample H i The samples are divided into photoelectricity prediction samples and wind power prediction samples, which are used for photoelectricity generation prediction and wind power generation prediction respectively.

6. The home substation energy storage station system of claim 1, wherein: The normalization method in S1 is as follows: all wind speed, wind direction, humidity, and air pressure are preliminarily normalized, then a corresponding weight coefficient is set for each wind speed, wind direction, humidity, and air pressure, and x is obtained by adding them together; x = α1A1 + α2A2 + α3A3 + α4A4; wherein α1, α2, α3, α4, are set weight coefficients.

7. The home substation energy storage station system of claim 6, wherein: The weight coefficient α of each item of weather information is calculated as follows: α u = C r (u, U)P; where P is a conversion factor, C r (u, U) is the correlation ratio of the u-th item of information with the rest of the information; wherein u represents the u-th item of information, U represents all the remaining information except u, C represents the power generation amount, I r (r, U) is the correlation coupling rate, which measures the degree of additional information sharing between u and C under the condition that all information except u is known, and the value range is 0 to 1, I(u; C|U) represents the amount of information sharing between u and C under the condition that U is known, H(u) represents the entropy of u, H(C) represents the entropy of C, T r (A, B) represents the degree of reduction of the amount of information sharing between u and C under the condition that U is known, and I(u; C) represents the mutual information of u and C; The entropy is calculated as follows: where H(u) represents the entropy of the u-th information in the weather information, i represents the index of the sample, n represents the total number of samples, b=2, and p(u i ) represents the probability of u taking the value u i .

8. The home substation energy storage station system of claim 1, wherein: The wind power prediction module and the photovoltaic prediction module respectively have a pre-trained prediction model built-in; The prediction model is trained in a training device; The training device comprises: a knowledge output unit with a pre-trained large neural network model built-in; a training unit in which the prediction model to be trained is arranged; training data is input into the knowledge output unit and the training unit, wherein the knowledge output unit extracts information from the training data to guide the training of the prediction model built-in the training unit.

Images