Reduction current and voltage monitoring data missing value interpolation method and system

Abstract

The invention relates to the technical field of data processing, in particular to a reduction current and voltage monitoring data missing value interpolation method and system, and the method comprises the steps: determining an isolated missing value of historical load data in a microgrid; estimating a target missing value based on an adjacent value of the isolated missing value; performing data randomness restoration on the historical load data based on the target missing value by adopting a missing value interpolation network to obtain restored load data; wherein the missing value interpolation network is obtained by combining a particle swarm algorithm and a back propagation neural network; training the missing value interpolation network based on the repaired load data to obtain a trained network; and performing load prediction on the micro-grid based on the trained network to obtain load data of the micro-grid. According to the method, the data randomness of the historical load data can be accurately restored, and the accuracy of load data prediction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an interpolation method and system for restoring missing values of current and voltage monitoring data in the technical field of data processing. Background Art

[0002] When monitoring current and voltage data online, missing values are very common in condition monitoring data, which can be divided into natural factors and human factors: natural factors such as sensor failures, data transmission signal interruptions, data loss caused by data compression and decompression, etc.; human factors include deleting invalid abnormal data and error values, etc. Missing values destroy the continuity and integrity of time-series data. Since the vast majority of fault identification methods cannot accept input data containing missing values, it is of great significance to carry out the repair work of missing values in condition monitoring data.

[0003] In the research of different fields on missing data and its causes, generally due to some known or unknown reasons, data is missing, which increases the difficulty of data analysis work and there will also be inaccuracies in the final analysis results. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent predictive control method with extreme stability, and the specific technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present invention provides an interpolation method for restoring missing values of current and voltage monitoring data, and the method includes:

[0006] In a microgrid, determine the isolated missing values of historical load data;

[0007] Estimate the target missing value based on the neighboring values of the isolated missing value;

[0008] Use a missing value interpolation network to perform data randomness repair on the historical load data based on the target missing value to obtain repaired load data; wherein, the missing value interpolation network is obtained by combining a particle swarm algorithm and a backpropagation neural network;

[0009] Train the missing value interpolation network based on the repaired load data to obtain a trained network;

[0010] Perform load prediction on the microgrid based on the trained network to obtain the load data of the microgrid.

[0011] In a second aspect, an interpolation system for restoring missing values of current and voltage monitoring data is provided, and the system includes:

[0012] A determination module, configured to determine the isolated missing values of historical load data in a microgrid;

[0013] An estimation module for estimating a target missing value based on neighboring values of the isolated missing value;

[0014] A repair module for performing data randomness repair on the historical load data based on the target missing value by using a missing value imputation network to obtain repaired load data; wherein, the missing value imputation network is obtained by combining a particle swarm algorithm and a backpropagation neural network;

[0015] A training module for training the missing value imputation network based on the repaired load data to obtain a trained network;

[0016] A prediction module for performing load prediction on the microgrid based on the trained network to obtain the load data of the microgrid.

[0017] In a third aspect, a computer program product is provided, which includes: computer program code that, when running on a computer, causes the computer to execute the method in the first aspect or any one of the possible implementation manners described in the first aspect.

[0018] In a fourth aspect, a computer-readable storage medium is provided, which stores computer program code that, when running on a computer, causes the computer to execute the method in the first aspect or any one of the possible implementation manners described in the first aspect.

[0019] The present invention has the following beneficial effects: In a microgrid, an isolated missing value of historical load data is determined; and a target missing value is estimated based on neighboring values of the isolated missing value; in this way, the target missing value in the historical load data can be accurately estimated through the neighboring values of the isolated missing value. Then, data randomness repair is performed on the historical load data based on the target missing value by using a missing value imputation network to obtain repaired load data; since the missing value imputation network is obtained by combining a particle swarm algorithm and a backpropagation neural network, the random characteristics in the historical load data can be accurately repaired through the missing value imputation network, so as to complete the missing value imputation work in the dataset to obtain more complete repaired load data. Finally, the missing value imputation network is trained based on the repaired load data to obtain a trained network; since the repaired load data is obtained after repairing the data random characteristics, training the missing value imputation network with the repaired load data can make the performance of the obtained trained network better; thus, using the trained network obtained by optimizing the backpropagation neural network through the particle swarm algorithm to perform load prediction on the microgrid can make the predicted load data more accurate and improve the prediction accuracy. Description of the Drawings

[0020] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0021] Figure 1 is a schematic diagram of an application scenario of an interpolation method for missing values in reduction current and voltage monitoring data provided in the related art;

[0022] Figure 2 is a schematic diagram of the implementation process of an interpolation method for missing values in reduction current and voltage monitoring data provided in an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of an application scenario of an interpolation method for missing values in reduction current and voltage monitoring data provided in an embodiment of the present invention;

[0024] Figure 4 is another schematic diagram of an application scenario of an interpolation method for missing values in reduction current and voltage monitoring data provided in an embodiment of the present invention;

[0025] Figure 5 is a schematic diagram of the composition structure of an interpolation system for missing values in reduction current and voltage monitoring data provided in an embodiment of the present invention;

[0026] Figure 6 is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Embodiment

[0027] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, will detail the specific implementation manner, structure, features and effects of an interpolation method for missing values in reduction current and voltage monitoring data proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0028] Among them, in the description of the embodiments of the present invention, unless otherwise specified, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present invention, "a plurality" means two or more than two.

[0029] Hereinafter, the terms "first" and "second" are for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features.

[0030] Unless otherwise defined, all technical and scientific terms used in the embodiments of the present invention have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0031] In the related art, the reasons for data missing are mainly as follows: information cannot be obtained temporarily, information is omitted, one or some attributes are unavailable, intentionally concealed or limited by history, doomed to abandon obtaining information, strong real-time performance of the system resulting in data missing, etc.

[0032] If the continuous missing length of a certain state quantity is not greater than the time threshold T ht , then this state quantity belongs to the missing value.

[0033] Define the missing sample rate (MSR) and the missing variable rate (MVR) of the sample to evaluate the integrity of the data set. The smaller the missing sample rate and the missing variable rate, the higher the integrity of the data.

[0034] Let the data segment X ∈ R m where X is the data segment to be analyzed. Then the calculation method of the missing sample rate MSR of X is: MSR(k) = 1 - m com (k) / mk; where m com (k) represents the number of samples, and m represents the total number of data segments.

[0035] Let the measured value in the k-th data segment X(k) be m com，j (i.e., the number of samples), then the calculation method of the missing variable rate MVR of the state quantity is: MVR(k) = 1 - m com，j / m. The integrity of the sample requires that all state quantities in the sample are known. Therefore, for the entire complete sample, the missing variable rate is less than the missing sample rate.

[0036] The filling process only fills the unknown values with subjective estimated values, which may not exactly conform to the objective facts. While processing the incomplete information, the original information system is more or less changed, especially the loss of randomness. For example, Figure 1 as shown, the volatility of the predicted value (i.e., the prediction curve 11) is much smaller than that of the true value (i.e., the prediction curve 12), and the interpolation cannot reflect the random volatility. Based on this, the embodiments of the present invention provide a method for interpolating missing values of on-line monitoring data of current and voltage that can reflect high-fidelity random volatility.

[0037] Based on this, in the embodiments of the present invention, a fluctuation value sampling interpolation function is constructed using the complete measurement values adjacent to the missing value before and after to estimate the missing value; the daily complete variables in the data segment are used as input data, and a repair model based on the combined algorithm of the particle swarm optimization (PSO) and BP neural network is established to repair the random characteristics between the data segments, thereby completing the missing value interpolation work in the dataset.

[0038] The following specifically describes the specific solution of a method for interpolating missing values in restored current and voltage monitoring data provided by the present invention with reference to the accompanying drawings. Please refer to Figure 2 , which shows a schematic implementation flow diagram of a method for interpolating missing values in restored current and voltage monitoring data provided by an embodiment of the present invention. The method includes:

[0039] 201. In the microgrid, determine the isolated missing values in the historical load data.

[0040] Here, a microgrid refers to a small power generation and distribution system composed of distributed power sources, energy storage devices, energy conversion devices, loads, monitoring and protection devices, etc. A microgrid is an autonomous system that can achieve self-control, protection, and management, and can operate either in parallel with the external power grid or independently. A microgrid consists of the following main parts: Distributed power sources include: photovoltaic power generation, wind power generation, diesel generators, etc.; Energy storage devices, used to store energy to ensure that the microgrid can continue to supply power during power outages; Energy conversion devices, used to convert one form of energy into another form, for example, converting direct current into alternating current; Loads include: various general loads and important loads; Monitoring and protection devices, used to ensure the safe operation and fault handling of the microgrid.

[0041] The historical load data is the historical current and historical voltage data of the microgrid. An isolated missing value refers to the value of a certain data point in the historical load data being missing in a specific dimension, and the values of this data point in other dimensions are irrelevant to the missing value. Isolated missing values usually appear in certain specific dimensions of the historical load data. For example, in time series data, the data at a certain time point may be missing while the data at other time points is complete.

[0042] In some possible implementation manners, by segmenting the historical load data of the microgrid, multiple data segments are obtained; then, through regression analysis of the multiple data segments, the isolated missing values can be accurately obtained. For example, according to the position of the missing samples, the historical load data is segmented, and the historical load data is divided into k segments, where k is an integer greater than 1, that is, the historical load data X = [X(1), X inc (1), …, X inc (k - 1), X(k)], and then, by normalizing each data segment and traversing and analyzing one by one, the isolated missing values in each data segment are determined.

[0043] 202. Estimate the target missing value based on the neighboring values of the isolated missing value.

[0044] Here, by searching for multiple neighboring values of the isolated missing value in the historical load data, a fluctuation value sampling interpolation function is constructed to estimate the target missing value through this fluctuation value sampling interpolation function.

[0045] In some possible implementation manners, first, in the multiple data segments, determine multiple neighboring values of the isolated missing value that meet the preset conditions; where the preset conditions include: being adjacent to the isolated missing value and meeting the preset quantity and preset integrity. In some possible implementation manners, in the multiple data segments, determine multiple candidate data values that are adjacent to the isolated missing value and meet the preset quantity and preset integrity; and determine the multiple candidate data values as the multiple neighboring values of the isolated missing value that meet the preset conditions. Among them, the preset quantity can be a custom quantity. For example, the preset quantity is at least greater than or equal to 6, and the preset integrity can be a custom data integrity. For example, the preset integrity is at least greater than or equal to 90% data integrity, for example, the data integrity is 100% complete data. In a specific example, in the multiple data segments, at least six or more complete neighboring values adjacent to the isolated missing value, for example, neighboring values Among them, X i-3 , X i-2 , X i-1 , X i+1 , X i+2 , X i+3 represent 6 adjacent values. is the i-th variable, i is the serial number, and f spline () represents the neighboring value estimation function.

[0046] Secondly, based on the multiple neighboring values, construct a fluctuation value sampling interpolation function.

[0047] Here, constructing a fluctuation value sampling interpolation function requires at least six or more complete neighboring values. In some possible implementation manners, in the multiple neighboring values, determine the difference between each two neighboring values corresponding to the symmetric positions determined with the isolated missing value as the center point; for example, the difference between x i+2 , x i-1 , or the difference between x i+3 , x i-1 etc. And fuse the differences to obtain the fluctuation value sampling interpolation function. For example, neighboring values When X i , X i+1 are both missing at the same time, that is, the time threshold T ht= 2, at this time, the fluctuation value sampling interpolation function can be obtained through the following process:

[0048] The first step: Construct the fluctuation characteristics, as shown in formulas (1) and (2):

[0049]

[0050] The second step: Construct the continuous characteristics, as shown in formula (3):

[0051]

[0052] Finally, estimate the target missing value based on the fluctuation value sampling interpolation function.

[0053] Here, by estimating the missing value through the fluctuation value sampling interpolation function shown in the above formulas (1) to (3), the target missing value can be obtained. Even when three values are missing simultaneously, a similar method can be used for interpolation.

[0054] 203. Use the missing value imputation network to perform data randomness repair on the historical load data based on the target missing value to obtain the repaired load data.

[0055] Among them, the missing value imputation network is obtained by combining the Particle Swarm Optimization (PSO) and the BackPropagation (BP) neural network.

[0056] In the related art, due to defects such as slow convergence speed and poor global search ability of the BP neural network algorithm, the BP neural network has a low accuracy when predicting data, resulting in a poor prediction effect. However, for the particle swarm optimization algorithm, its characteristics of fast convergence speed and strong global search ability enable the particle swarm optimization algorithm to exactly make up for the defects of the traditional BP neural network itself. Therefore, the combined algorithm of the particle swarm optimization algorithm and the BP neural network has become a feasible solution to improve the load prediction accuracy. In the embodiment of the present invention, the short-term load of the microgrid is predicted by using the particle swarm optimization BP neural network algorithm prediction model. The advantage of this model is that it can perform excellent global search and non-linear generalization on data, enabling it to have a high prediction accuracy ability for a lot of data. Moreover, due to the fast convergence speed of the particle swarm optimization algorithm, the prediction speed of this model is faster than that of other models to a certain extent. Since the advantages and disadvantages of these two algorithms complement each other, this combined algorithm model can be applied to various data in most cases. In the embodiment of the present invention, based on the combined method of the particle swarm optimization algorithm and the BP neural network algorithm, the BP neural network algorithm can be started with. By establishing a neural network model, its input layer, hidden layer, and output layer are determined. After setting the BP neural network parameters, the particle swarm optimization algorithm is used to optimize the BP neural network. First, the particle swarm parameters are set, and then by setting the initial population and performing iterative optimization on the initial population, that is, a series of operations of the particle swarm optimization algorithm are run. The optimal initial threshold and weight searched by the particle swarm are given to the neural network prediction, and finally the network training is carried out to obtain its prediction value.

[0057] In some possible implementation manners, first, in the historical load data, the target data segment to which the target missing value belongs is determined; for example, after determining the target missing value, in the historical load data, the target data segment to which the target missing value belongs is analyzed. Then, the missing value imputation network is used to repair the data randomness of the target data segment in the historical load data to obtain the repaired load data. For example, one-dimensional interpolation and multiple regression analysis models are used to repair all the isolated missing values in the data segment, and finally [X(1), X inc (1), …, X inc (k - 1), X(k)] is used as input data, and the particle swarm optimization algorithm and the BP neural network algorithm are used to repair the random missing samples in [X(1), X inc (1), …, X inc (k - 1), X(k)] to obtain the repaired load data.

[0058] 204, based on the repaired load data, the missing value imputation network is trained to obtain the trained network.

[0059] Here, the historical load data can be obtained by collecting at certain time intervals and then performing normalization processing. In some possible implementation manners, first, initial load data of the microgrid is collected at a preset time interval; wherein, the preset time interval can be a custom time length. For example, the preset time interval is one hour, so that the microgrid is sampled once every 1 hour to obtain the initial load data. Second, the maximum load data and the minimum load data are determined from the initial load data; finally, based on the maximum load data and the minimum load data, the initial load data is normalized to obtain the historical load data. For example, a training set and a test set are set up and data normalization is performed. According to the normalization formula (4), the normalized value y of the historical load data is restricted within the range of 0 to 1.

[0060] y = (x - min) / (max - min) (4);

[0061] Wherein, x represents the initial load data, min represents the minimum load data, and max represents the maximum load data. In this way, the initial load data is normalized through the normalization formula, so that the obtained historical load data is convenient for subsequent calculations.

[0062] In some possible implementation manners, the training process of the missing value imputation network can be implemented through the following process:

[0063] In the first step, based on the data type of the repaired load data, the number of the first nodes of the input layer in the missing value imputation network is determined.

[0064] Here, the setting of the input layer is determined according to the type of the input samples.

[0065] In the second step, based on the number of nodes of the input layer, the number of the second nodes of the hidden layer and the number of the third nodes of the output layer in the missing value imputation network are set.

[0066] Here, the number of nodes of the hidden layer can be indefinite. When setting the number of the hidden layer, it is necessary to continuously adjust to achieve the best effect. The number of the second nodes of the hidden layer can be equal to the number of nodes of the input layer multiplied by two plus one. The number of the third nodes of the output layer can be set to 1.

[0067] In the third step, based on the number of the first nodes, the number of the second nodes and the number of the third nodes, the backpropagation neural network is constructed.

[0068] Here, a BP neural network is constructed through an input layer with the first number of nodes, a hidden layer with the second number of nodes, and an output layer with the third number of nodes, and the network parameters are set. The network parameters include the learning rate of the neural network, the maximum number of training times, and the setting of the minimum error value, etc. Then, by setting the key parameters of the particle swarm optimization algorithm, including: setting the initial particles and velocities, setting the individual extreme values and the global extreme values, etc. Finally, an iterative optimization operation is performed. The iterative optimization operation includes inertia weight update, velocity update, position update, as well as individual optimal update and global optimal update, etc. The optimal initial thresholds and weights are assigned to the network prediction. The optimal solutions searched by the particle swarm optimization algorithm, that is, the optimal initial thresholds and weights, are substituted into the neural network to obtain the trained network.

[0069] Step 4: Determine the key parameters of the particle swarm optimization algorithm.

[0070] Here, the key parameters include: the first acceleration constant, the second acceleration constant, the population size, the inertia weight, the number of iterations, and the error index, etc.

[0071] In some possible implementation manners, the key parameters of the particle swarm optimization algorithm can be determined through the following process:

[0072] First, in the particle swarm optimization algorithm, determine the first acceleration constant and the second acceleration constant with equal values. For example, the first acceleration constant = the second acceleration constant = 1.5.

[0073] Then, based on the accuracy requirement of the microgrid for the load data, determine the population size, the inertia weight, and the number of iterations of the particle swarm optimization algorithm.

[0074] Here, by analyzing the accuracy requirement of the microgrid for the predicted load data, the population size, the inertia weight, and the number of iterations of the particle swarm optimization algorithm are set, which can not only meet the requirements of the microgrid but also reduce the computational amount.

[0075] Step 5: Based on the particle swarm optimization algorithm carrying the key parameters, perform iterative training on the backpropagation neural network to obtain the trained network.

[0076] Here, the backpropagation neural network is iteratively trained according to the number of iterations until the network parameters (such as weights, learning rate, etc.) of the trained network converge and the accuracy of the obtained prediction results meets the accuracy requirement of the microgrid, and the trained network is obtained. In this way, the BP neural network model is optimized by the particle swarm optimization algorithm, so that the prediction results of the obtained trained network are more accurate.

[0077] In some embodiments, to verify the authenticity of the trained network, the embodiments of the present application select the load data of a certain city's power system for prediction. The data sampling time is about one month, a total of 32 days. The sampling interval is one hour for one point, and there are 10 categories in total, with 768 samples in each category. Before training the data, since the data is relatively scattered and has different dimensional units, directly training will cause the neural network to converge slowly, resulting in a longer training time. Therefore, for the convenience of data training, when inputting the data, it is necessary to normalize the data, and the normalization formula is implemented according to Formula (4). The embodiments of the present application select the first 648 samples as the training set, and in the last five days, according to the method of sampling once per hour, 120 samples are used as the test set.

[0078] After inputting the sample data, according to the model steps, it is necessary to determine the number of nodes in the neural network. Here, since there are 10 categories of data samples, the input layer has 10 nodes. Next is the design of the number of nodes in the hidden layer. The number of nodes in the hidden layer can be indefinite, and any number of nodes can make the BP neural network model operate. However, part of the prediction accuracy depends on the number of nodes in the hidden layer. If the number of nodes is too small, the network algorithm will be too simple, resulting in a significant reduction in its information processing ability, making it unable to learn effectively and systematically, thus reducing the prediction accuracy; if the number of nodes is too large, it will increase the complexity of the neural network, making the neural network have a slow learning rate and easily fall into local minimum points, making the predicted value of the neural network have great randomness, just like using two arrays to solve a five-variable equation system, there are many optimal solutions. Therefore, the number of nodes in the hidden layer of the neural network needs to be continuously debugged and optimized by the experimenter to achieve the optimal prediction accuracy. However, through the search for the hidden layer data in the embodiments of the present invention, it is known that the number of nodes in the hidden layer of the neural network can be approximately determined by the following formula to obtain the optimal number of neural network nodes: the number of nodes in the hidden layer = the number of nodes in the input layer * 2 + 1. Therefore, in the embodiments of the present invention, 10 nodes in the input layer are used here. According to the formula calculation, 21 nodes in the hidden layer should be taken as the optimal number. Finally, the number of nodes in the output layer is set to 1.

[0079] After determining the number of nodes in the neural network, set the key parameters of the particle swarm algorithm, including:

[0080] First, determine the first acceleration constant c1 and the second acceleration constant c2. c1 represents the flight acceleration of an individual particle, and c2 represents the flight acceleration of the entire particle group. When only c1 is present, each particle searches independently. Due to the lack of communication, the probability of each particle finding the optimal solution decreases. When only c2 is present, the particles act together as a group. Although the speed of finding the optimal solution increases, the search for the optimal value is concentrated in a certain area, causing the particle group to fall into a local optimum. If neither c1 nor c2 exists, the search range cannot be further expanded, and the particle group is restricted to a limited search range. Only when c1 and c2 are equal can the optimal solution be searched more effectively. Therefore, in the embodiment of the present invention, when c1 = c2 = 1.5, the particle group achieves better search results.

[0081] Secondly, determine the size of the particle swarm: If the swarm size is to be determined, it should be determined based on the specific situation. For general problems, the set swarm size will most likely be around 20. However, for some complex or special problems, the particle swarm size can be very large, sometimes 100, or even 400 for some problems. Of course, the more particle swarms are set, the larger the search range, making the prediction accuracy higher. However, for problems that do not require particularly high prediction accuracy, it is recommended to set the swarm size as small as possible. After all, the larger the swarm size, the longer the algorithm runs. Therefore, the swarm size set in the embodiment of the present invention is 20.

[0082] Next, determine the inertia weight: When setting the range of the inertia weight, the range cannot be too large or too small. Therefore, the inertia weight interval range set in this embodiment of the present invention is [0.4-0.9].

[0083] Next, determine the number of iterations: This number also requires continuous debugging to find the optimal value. Setting it too low will result in an overly simple algorithm, with particles updating their speed and position a few times before stopping. This will narrow the search range and minimize the probability of finding the optimal value. Setting it too high will result in excessive computational time, reducing the algorithm's efficiency. Therefore, through continuous debugging, the present embodiment determined that 30 iterations of the particle swarm algorithm are appropriate.

[0084] Finally, set the error evaluation index: There are two types of indicators, mean absolute percentage error and root mean square error, as shown in formulas (5) and (6):

[0085]

[0086] Where: n represents the number of predicted load data, represents the mean value, yi Represents the i-th load data.

[0087] 205. Based on the trained network, load forecasting is performed on the microgrid to obtain the load data of the microgrid.

[0088] Here, short-term load forecasting is performed on the microgrid through the trained network to obtain the forecasting data. In a specific example, the microgrid is forecasted through a particle swarm optimization BP neural network forecasting model, and the forecasting results are as Figure 3 shown. Among them, the forecasting curve 31 represents the PSO-BP error value, that is, the error of the particle swarm optimization BP neural network model (i.e., the trained network), and the forecasting curve 32 represents the BP forecasting error value, that is, the error of the BP neural network model. By comparing the errors of the two models, it can be seen that the error range of the traditional BP neural network is between (-130) and 110, while the error range of the particle swarm optimization BP neural network is between (-70) and 90, indicating that the forecasting error value of the optimized algorithm model is significantly smaller than that of the BP neural network algorithm model. Since the error movement range of the optimized algorithm model is smaller, it indicates that its error value is more concentrated, and the forecasting accuracy will be higher than that of the traditional BP neural network.

[0089] In the embodiment of the present invention, by calculating formulas (5) and (6), the root mean square error values and mean absolute percentage error values of the two algorithm models are obtained, and the results are shown in Table 1.

[0090] Table 1 Comparison of error results between the particle swarm optimization BP neural network model and the BP neural network model

[0091]

[0092] From the comparison of the error indexes of the two models in Table 1, it can be seen that the two error indexes of the particle swarm optimization BP neural network model are both smaller than those of the traditional BP neural network model. It shows that the optimized algorithm is superior to the traditional BP neural network in terms of forecasting accuracy.

[0093] Among them, the comparison of the forecasting results of the two forecasting models is as Figure 4 shown, including: the PSO-BP forecasting load curve 41, that is, the forecasting load curve of the particle swarm optimization BP neural network model (i.e., the trained network), the traditional BP forecasting load curve 42, that is, the forecasting load curve of the BP neural network model, and the expected value 43.

[0094] According to Figure 4From the data in , it can be seen that the prediction model of the particle swarm optimization BP neural network is closer to the true value than the traditional BP neural network in its predicted load value. Therefore, for this optimized algorithm model, it has higher prediction accuracy and better prediction effect compared to the BP neural network model.

[0095] An embodiment of the present invention provides an interpolation system for missing values in restored current and voltage monitoring data. Please refer to Figure 5 , which shows a schematic structural diagram of an interpolation system for missing values in restored current and voltage monitoring data provided by an embodiment of the present invention. The system 500 includes:

[0096] A determination module 501, configured to determine isolated missing values in historical load data in a microgrid;

[0097] An estimation module 502, configured to estimate a target missing value based on adjacent values of the isolated missing value;

[0098] A repair module 503, configured to perform data randomness repair on the historical load data based on the target missing value by using a missing value interpolation network, and obtain repaired load data; wherein, the missing value interpolation network is obtained by combining a particle swarm algorithm and a backpropagation neural network;

[0099] A training module 504, configured to train the missing value interpolation network based on the repaired load data to obtain a trained network;

[0100] A prediction module 505, configured to perform load prediction on the microgrid based on the trained network to obtain load data of the microgrid.

[0101] In some possible implementation manners, the determination module 501 is further configured to segment the historical load data of the microgrid to obtain a plurality of data segments; perform regression analysis on the plurality of data segments to obtain the isolated missing values.

[0102] In some possible implementation manners, the determination module 501 is further configured to determine a plurality of adjacent values of the isolated missing value that meet a preset condition in the plurality of data segments; construct a fluctuation value sampling interpolation function based on the plurality of adjacent values; estimate the target missing value based on the fluctuation value sampling interpolation function.

[0103] In some possible implementation manners, the determination module 501 is further configured to determine a plurality of candidate data values that are adjacent to the isolated missing value and meet a preset quantity and preset integrity in the plurality of data segments; determine the plurality of candidate data values as the plurality of adjacent values of the isolated missing value that meet the preset condition.

[0104] In some possible implementation manners, the determining module 501 is further configured to determine, among the multiple adjacent values, the difference between each two adjacent values corresponding to the symmetric positions determined with the isolated missing value as the center point; fuse the differences to obtain the fluctuation value sampling interpolation function.

[0105] In some possible implementation manners, the repairing module 503 is further configured to determine, in the historical load data, the target data segment to which the target missing value belongs; use the missing value interpolation network to repair the data randomness of the target data segment in the historical load data to obtain the repaired load data.

[0106] In some possible implementation manners, the determining module 501 is further configured to collect the initial load data of the microgrid at preset time intervals; determine the maximum load data and the minimum load data in the initial load data; perform normalization processing on the initial load data based on the maximum load data and the minimum load data to obtain the historical load data.

[0107] In some possible implementation manners, the training module 504 is further configured to determine, based on the data type of the repaired load data, the number of the first nodes in the input layer of the missing value interpolation network; set the number of the second nodes in the hidden layer and the number of the third nodes in the output layer of the missing value interpolation network based on the number of nodes in the input layer; construct the backpropagation neural network based on the number of the first nodes, the number of the second nodes, and the number of the third nodes; determine the key parameters of the particle swarm algorithm; perform iterative training on the backpropagation neural network based on the particle swarm algorithm carrying the key parameters to obtain the trained network.

[0108] In some possible implementation manners, the training module 504 is further configured to determine, in the particle swarm algorithm, the first acceleration constant and the second acceleration constant that are equal in value; determine the population size, the inertia weight, and the number of iterations of the particle swarm algorithm based on the accuracy requirement of the microgrid for the load data; where the key parameters at least include: the first acceleration constant, the second acceleration constant, the population size, the inertia weight, and the number of iterations.

[0109] Optionally, the transmission medium may be a wired link (such as, but not limited to, coaxial cable, optical fiber, and Digital Subscriber Line (DSL), etc.) or a wireless link (such as, but not limited to, Wireless Fidelity (WIFI), Bluetooth, and mobile device network, etc.). It should be noted that: for the system provided in the above embodiments, only the division of the above functional modules is used for illustration. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the method embodiments provided in the above embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiments and will not be elaborated here.

[0110] Figure 6 is a schematic structural diagram of a computer device provided by an embodiment of the present invention. Exemplarily, as Figure 6 shown, the computer device 600 includes: a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and running on the processor 602. When the processor 602 executes the computer program 603, the computer device can execute any one of the above-described methods for interpolating missing values of the reduction current voltage monitoring data.

[0111] In addition, an embodiment of the present invention also protects a system, which may include a memory and a processor. The memory stores executable program code, and the processor is used to call and execute the executable program code to execute a method for interpolating missing values of the reduction current voltage monitoring data provided by an embodiment of the present invention. In this embodiment, the system can be divided into functional modules according to the above method examples. For example, it can correspond to each functional module, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is illustrative, only a logical function division, and there may be other division methods in actual implementation. It should be noted that all relevant content of each step involved in the above method embodiments can be cited in the function description of the corresponding functional modules and will not be elaborated here.

[0112] It should be understood that the system provided in this embodiment is used to execute the above method for interpolating missing values in the monitored current and voltage data, so the same effect as the above implementation method can be achieved. In the case of adopting an integrated unit, the system may include a processing module and a storage module. Among them, when the system is applied to a device, the processing module can be used to control and manage the actions of the device. The storage module can be used to support the device to execute mutual program codes, etc. Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logic blocks, modules and circuits described in combination with the disclosure of the present invention. The processor can also be a combination that realizes computing functions, such as including a combination of one or more microprocessors, a combination of a digital signal processing (DSP) and a microprocessor, etc. The storage module can be a memory.

[0113] In addition, the system provided in the embodiment of the present invention can specifically be a chip, a component or a module. The chip may include a connected processor and a memory. Among them, the memory is used to store instructions. When the processor calls and executes the instructions, the chip can execute the method for interpolating missing values in the monitored current and voltage data provided in the above embodiment. This embodiment also provides a computer-readable storage medium. Computer program code is stored in the computer-readable storage medium. When the computer program code runs on a computer, the computer is enabled to execute the above-related method steps to implement the method for interpolating missing values in the monitored current and voltage data provided in the above embodiment.

[0114] This embodiment also provides a computer program product. When the computer program product runs on a computer, it causes the computer to execute the above-related steps to implement an interpolation method for missing values in restored current and voltage monitoring data provided by the above embodiment. Among them, the system, computer-readable storage medium, computer program product or chip provided in this embodiment are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be elaborated here. Through the description of the above embodiments, those skilled in the art can understand that for the convenience and brevity of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the system is divided into different functional modules to complete all or part of the functions described above. In the embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the system or unit can be in an electrical, mechanical or other form.

[0115] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. The above content is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention.

Claims

1. A method for interpolating missing values in reduction current voltage monitoring data, characterized in that, The method includes: In a microgrid, determining isolated missing values in historical load data; Estimating target missing values based on neighboring values of the isolated missing values; Using a missing value imputation network to perform data randomness repair on the historical load data based on the target missing values to obtain repaired load data; wherein, the missing value imputation network is obtained by combining a particle swarm algorithm and a backpropagation neural network; Training the missing value imputation network based on the repaired load data to obtain a trained network; Performing load forecasting on the microgrid based on the trained network to obtain the load data of the microgrid.

2. The interpolation method for missing values of the reduction current voltage monitoring data according to claim 1, wherein The determining, in the microgrid, of isolated missing values in historical load data includes: Segmenting the historical load data of the microgrid to obtain a plurality of data segments; Performing regression analysis on the plurality of data segments to obtain the isolated missing values.

3. The interpolation method for missing values in the reduction current voltage monitoring data according to claim 2, characterized in that, The estimating of target missing values based on neighboring values of the isolated missing values includes: In the plurality of data segments, determining a plurality of neighboring values of the isolated missing values that meet a preset condition; Constructing a fluctuation value sampling interpolation function based on the plurality of neighboring values; Estimating the target missing values based on the fluctuation value sampling interpolation function.

4. A method for interpolating missing values in restoring current and voltage monitoring data according to claim 3, characterized in that: The determining, in the plurality of data segments, of a plurality of neighboring values of the isolated missing values that meet a preset condition includes: In the plurality of data segments, determining a plurality of candidate data values that are adjacent to the isolated missing value and meet a preset quantity and preset integrity; Determining the plurality of candidate data values as the plurality of neighboring values of the isolated missing values that meet the preset condition.

5. A method for interpolating missing values in reduction current voltage monitoring data according to claim 3, characterized in that The constructing of a fluctuation value sampling interpolation function based on the plurality of neighboring values includes: In the plurality of neighboring values, determining the difference between each two neighboring values corresponding to symmetric positions determined with the isolated missing value as the center point; Fusing the differences to obtain the fluctuation value sampling interpolation function.

6. The method for interpolating missing values in the reduction current voltage monitoring data according to claim 1, characterized in that The using of a missing value imputation network to perform data randomness repair on the historical load data based on the target missing values to obtain repaired load data includes: In the historical load data, determining a target data segment to which the target missing value belongs; Using the missing value imputation network to repair the data randomness of the target data segment in the historical load data to obtain the repaired load data.

7. A method for interpolating missing values in restoring current and voltage monitoring data according to claim 1, characterized in that: The method further includes: Collecting initial load data of the microgrid at preset time intervals; In the initial load data, determining the maximum load data and the minimum load data; Performing normalization processing on the initial load data based on the maximum load data and the minimum load data to obtain the historical load data.

8. A method for interpolating missing values in reduction current voltage monitoring data according to claim 1, characterized in that, The training of the missing value imputation network based on the repaired load data to obtain a trained network includes: Based on the data type of the repaired load data, determining the number of first nodes in the input layer of the missing value imputation network; Based on the number of nodes in the input layer, setting the number of second nodes in the hidden layer and the number of third nodes in the output layer of the missing value imputation network; Construct the backpropagation neural network based on the number of the first nodes, the number of the second nodes, and the number of the third nodes; Determine the key parameters of the particle swarm optimization algorithm; Based on the particle swarm optimization algorithm carrying the key parameters, iteratively train the backpropagation neural network to obtain the trained network.

9. A method for interpolating missing values in reduction current voltage monitoring data according to claim 8, characterized in that, The determining the key parameters of the particle swarm optimization algorithm includes: In the particle swarm optimization algorithm, determine a first acceleration constant and a second acceleration constant with equal values; Based on the accuracy requirement of the microgrid for load data, determine the population size, inertia weight, and number of iterations of the particle swarm optimization algorithm; wherein, the key parameters at least include: the first acceleration constant, the second acceleration constant, the population size, the inertia weight, and the number of iterations.

10. A system for interpolating missing values in reduction current voltage monitoring data, characterized in that, The system includes: A determination module, configured to determine isolated missing values of historical load data in a microgrid; An estimation module, configured to estimate target missing values based on neighboring values of the isolated missing values; A repair module, configured to perform data randomness repair on the historical load data based on the target missing values by using a missing value imputation network to obtain repaired load data; wherein, the missing value imputation network is obtained by combining a particle swarm optimization algorithm and a backpropagation neural network; A training module, configured to train the missing value imputation network based on the repaired load data to obtain a trained network; A prediction module, configured to perform load prediction on the microgrid based on the trained network to obtain load data of the microgrid.

Images