Distributed new energy output prediction method and device based on sequence decomposition

Abstract

The invention relates to a distributed new energy output prediction method and device based on sequence decomposition. The method comprises the following steps: generating an influence factor time sequence corresponding to each influence factor according to state information of each influence factor influencing the new energy output within a prediction time period, and performing reconstruction and decomposition processing on the influence factor time sequence to obtain candidate subsequences; constructing an influence factor sequence set according to the candidate subsequences, and screening out a target subsequence from the candidate subsequences according to fitness information corresponding to the influence factor sequence set; and inputting the target sub-sequence into the new energy output prediction model to obtain a new energy output prediction result. By adopting the method, the influence factor time sequence can be reconstructed and decomposed, the key information related to the influence factors can be extracted, the target sub-sequence can be screened out by combining the fitness information obtained by the genetic algorithm, the target sub-sequence can be processed through the prediction model, the new energy output prediction result can be obtained, and the accuracy of the new energy output prediction result can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a distributed renewable energy output prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on sequence decomposition. Background Art

[0002] With the increase in the proportion of renewable energy in distribution networks, the imbalance between distributed renewable energy output and load demand has become increasingly prominent. Accurate output forecasting is of great significance for the rational dispatch and stable operation of distribution networks. Accurate forecasting of renewable energy output can improve the forecasting accuracy of wind and solar power generation and optimize the dispatch and operation efficiency of power systems.

[0003] Traditional technology mainly divides meteorological data that affects power generation output into multiple frequency bands, establishes corresponding prediction models based on the fluctuation characteristics of each frequency band, and realizes power generation output prediction by combining the output results of multiple prediction models. However, traditional technology cannot fully capture the relationship between the influence of multiple influencing factors, and simply grouping the data is prone to losing key information, affecting the prediction accuracy, which is not conducive to improving the accuracy of new energy output prediction results. Summary of the invention

[0004] Based on this, it is necessary to provide a distributed renewable energy output prediction method, device, computer equipment, computer-readable storage medium and computer program product based on sequence decomposition, which can improve the accuracy of renewable energy output prediction results in response to the above technical problems.

[0005] In a first aspect, the present application provides a distributed renewable energy output prediction method based on sequence decomposition, comprising:

[0006] Generate an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy within the forecast time period, reconstruct and decompose the influencing factor time series to obtain a candidate subsequence;

[0007] constructing an influencing factor sequence set according to the candidate subsequences, and screening out a target subsequence from the candidate subsequences according to fitness information corresponding to the influencing factor sequence set;

[0008] The target subsequence is input into a new energy output prediction model to obtain a new energy output prediction result.

[0009] In one embodiment, the reconstructing and decomposing the influencing factor time series to obtain a candidate subsequence includes:

[0010] Obtaining time delay information and embedding dimension information corresponding to the time series of the influencing factors;

[0011] Decomposing the influencing factor time series into a phase space according to the time delay information and the embedding dimension information;

[0012] Superimposing the phase space to obtain a reconstructed time series;

[0013] The reconstructed time series is decomposed into the candidate subsequences according to the mean value information and the residual information corresponding to the reconstructed time series.

[0014] In one embodiment, the method further comprises:

[0015] Determine the average value information corresponding to the reconstructed time series according to the maximum value information and the minimum value information corresponding to the reconstructed time series;

[0016] Determining intrinsic mode function information according to a difference between the reconstructed time series and the average value information corresponding to the reconstructed time series;

[0017] Residual information corresponding to the reconstructed time series is determined according to a difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series.

[0018] In one embodiment, the step of selecting a target subsequence from the candidate subsequences according to the fitness information corresponding to the influencing factor sequence set includes:

[0019] Filtering a candidate influencing factor sequence set from the influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set;

[0020] exchanging a candidate subsequence in the candidate influencing factor sequence set with a candidate subsequence in the influencing factor sequence set to obtain an exchanged influencing factor sequence set;

[0021] The candidate subsequence in the exchanged influencing factor sequence set is used as the target subsequence.

[0022] In one embodiment, inputting the target subsequence into a new energy output prediction model to obtain a new energy output prediction result includes:

[0023] Obtaining correlation coefficient information corresponding to the target subsequence;

[0024] When the correlation coefficient information is greater than a preset correlation coefficient threshold, normalizing the target subsequence to obtain a processed subsequence;

[0025] The processed subsequence is input into the new energy output prediction model to obtain the new energy output prediction result.

[0026] In one embodiment, the method further comprises:

[0027] Constructing a sample subsequence according to the state information corresponding to each influencing factor within the sample time period, and inputting the sample subsequence into the new energy output prediction model to obtain a sample prediction result;

[0028] Acquire the new energy output information under the influence of the various influencing factors in the test time period corresponding to the sample time period as test data;

[0029] The new energy output prediction model to be trained is trained according to the test data and the sample prediction results.

[0030] In a second aspect, the present application also provides a distributed new energy output prediction device based on sequence decomposition, comprising:

[0031] A decomposition module, used to generate an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy within the forecast time period, and to reconstruct and decompose the influencing factor time series to obtain a candidate subsequence;

[0032] A screening module, used to construct an influencing factor sequence set according to the candidate subsequences, and screen out a target subsequence from the candidate subsequences according to fitness information corresponding to the influencing factor sequence set;

[0033] The prediction module is used to input the target subsequence into the new energy output prediction model to obtain the new energy output prediction result.

[0034] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0036] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0037] The above-mentioned distributed renewable energy output prediction method, device, computer equipment, computer-readable storage medium and computer program product based on sequence decomposition generate a time series of influencing factors corresponding to each influencing factor according to the state information of each influencing factor affecting the renewable energy output within the prediction time period, reconstruct and decompose the time series of influencing factors to obtain candidate subsequences, thereby constructing an accurate time series of influencing factors based on the state of the influencing factors within the prediction time period, reconstructing and decomposing the time series of influencing factors to remove noise in the time series, and facilitating subsequent prediction of renewable energy output based on the reconstructed and decomposed time series; constructing a set of influencing factor sequences according to the candidate subsequences, screening out target subsequences from the candidate subsequences according to the fitness information corresponding to the set of influencing factor sequences, thereby calculating each influencing factor based on a genetic algorithm. The fitness information of the influencing factor sequence set is obtained, and an accurate target subsequence is screened out based on the fitness information; the target subsequence is input into the new energy output prediction model to obtain the new energy output prediction result, and then the new energy output prediction model is used to process the target subsequence to obtain an accurate new energy output prediction result. The state information of each influencing factor within the prediction time period can be used to construct an accurate influencing factor time series, and the key information related to the influencing factor can be accurately extracted by reconstructing and decomposing the influencing factor time series. At the same time, the genetic algorithm is combined to obtain the fitness information corresponding to each influencing factor sequence set, and the fitness information is used to screen out the accurate target subsequence. Finally, the target subsequence is processed by the new energy output prediction model to obtain an accurate new energy output prediction result, thereby improving the accuracy of the new energy output prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a flow chart of a distributed renewable energy output prediction method based on sequence decomposition in one embodiment;

[0040] Figure 2 is a schematic diagram of an initial population of a genetic algorithm in one embodiment;

[0041] Figure 3 A schematic diagram of a flow chart of a distributed renewable energy output prediction method based on sequence decomposition in one embodiment;

[0042] Figure 4 is a schematic diagram of a gene mutation in one embodiment;

[0043] Figure 5 A schematic diagram of a flow chart of a genetic algorithm in one embodiment;

[0044] Figure 6 It is a structural block diagram of a distributed new energy output prediction device based on sequence decomposition in one embodiment;

[0045] Figure 7 The figure is a diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] In one embodiment, Figure 1 As shown, a distributed renewable energy output prediction method based on sequence decomposition is provided. This embodiment uses the method applied to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0048] Step S102, generating an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy within the prediction time period, reconstructing and decomposing the influencing factor time series to obtain a candidate subsequence.

[0049] The prediction time period may refer to a time period during which new energy output analysis and prediction are required. In practical applications, the prediction time period may be any time period, such as the next day, the next week, etc.

[0050] Among them, new energy output can refer to the electric energy output by new energy power generation equipment. In actual applications, new energy output can include power generation and power generation.

[0051] Among them, the influencing factors may refer to factors that affect the amount of electric energy output by the new energy power generation equipment. In practical applications, the influencing factors may include but are not limited to wind force, temperature, humidity, and light intensity.

[0052] Among them, the status information may refer to information that characterizes the actual status of the influencing factors at a specific time. For example, the wind status information may include wind level information, the temperature status information may include temperature information, the humidity status information may include humidity information, and the light intensity status information may include light intensity information.

[0053] Among them, the influencing factor time series can refer to a series obtained by arranging the state information of influencing factors at several time points in chronological order. Taking the wind force in the influencing factors as an example, the wind force level at time point T1 in area A is X1, the wind force level at time point T2 is X2, and the wind force level at time point T3 is X3, where T1 < T2 < T3. The influencing factor time series corresponding to the wind force can be expressed as X1X2X3. The influencing factor time series can contain both time information and state information.

[0054] Among them, the candidate subsequence can refer to a sequence obtained by reconstructing and decomposing the influencing factor time series.

[0055] As an example, in order to analyze and predict the new energy output situation in area A, the server can first obtain the state information of each influencing factor that affects the new energy output within the prediction time period. Then, the server can sort the state information of each influencing factor that affects the new energy output within the prediction time period to generate the influencing factor time series corresponding to each influencing factor. The server can perform phase space reconstruction and empirical mode decomposition processing on the influencing factor time series to obtain candidate subsequences.

[0056] Step S104, construct an influencing factor sequence set according to the candidate subsequences, and screen out the target subsequence from the candidate subsequences according to the fitness information corresponding to the influencing factor sequence set.

[0057] Among them, the influencing factor sequence set can refer to the individuals that make up the initial population of the genetic algorithm. In practical applications, as Figure 2 shown, a schematic diagram of the initial population of the genetic algorithm is provided. The influencing factor sequence set can be used as the individuals that make up the initial population of the genetic algorithm. Each influencing factor sequence set can include several chromosomes. Each chromosome has a corresponding influencing factor, and the genes on each chromosome can correspond to the candidate subsequences obtained by performing phase space reconstruction and empirical mode decomposition processing on the influencing factor time series.

[0058] Among them, the fitness information can refer to the information that characterizes the performance degree of the individuals represented by the influencing factor sequence set for the genetic algorithm target. In practical applications, the genetic algorithm target can include minimizing the mean absolute error of the neural network, and the fitness information can be calculated based on a preset fitness function.

[0059] Among them, the target subsequence can refer to a sequence whose fitness information meets the preset screening requirements.

[0060] As an example, the server can construct an influencing factor sequence set based on the candidate subsequences, and construct an initial population of the genetic algorithm based on the influencing factor sequence set as individuals. The server can then use the genetic algorithm to obtain the fitness information corresponding to the individuals represented by the influencing factor sequence set, and filter out the target subsequence from the candidate subsequences based on the fitness information.

[0061] Step S106, inputting the target subsequence into the new energy output prediction model to obtain the new energy output prediction result.

[0062] Among them, the new energy output prediction model may refer to a model used to output a target subsequence and output a new energy output prediction result. In practical applications, the new energy output prediction model may include but is not limited to a long short-term memory neural network and a BP (back propagation) neural network.

[0063] The new energy output prediction result may refer to information representing the predicted value of the electric energy output by the new energy power generation equipment.

[0064] As an example, the server can input the target subsequence into the new energy output prediction model, analyze the target subsequence through the new energy output prediction model, determine the working status and output of each new energy power generation equipment under the influence of the influencing factors corresponding to the target subsequence, and obtain the new energy output prediction result.

[0065] In the above-mentioned distributed renewable energy output prediction method based on sequence decomposition, the influence factor time series corresponding to each influence factor is generated according to the state information of each influence factor affecting the renewable energy output within the prediction time period, and the influence factor time series is reconstructed and decomposed to obtain a candidate subsequence, thereby constructing an accurate influence factor time series based on the state of the influence factor within the prediction time period, and reconstructing and decomposing the influence factor time series to remove the noise in the time series, so as to facilitate the subsequent new energy output prediction based on the reconstructed and decomposed time series; construct an influence factor sequence set according to the candidate subsequences, and select the target subsequence from the candidate subsequences according to the fitness information corresponding to the influence factor sequence set, so as to calculate the fitness information of each influence factor sequence set based on the genetic algorithm. information, and screen out accurate target subsequences based on fitness information; input the target subsequence into the new energy output prediction model to obtain the new energy output prediction result, and then use the new energy output prediction model to process the target subsequence to obtain accurate new energy output prediction results, and can use the state information of each influencing factor in the prediction time period to construct an accurate influencing factor time series, and reconstruct and decompose the influencing factor time series to accurately extract key information related to the influencing factors, and at the same time combine with the genetic algorithm to obtain the fitness information corresponding to each influencing factor sequence set, and use the fitness information to screen out accurate target subsequences, and finally use the new energy output prediction model to process the target subsequence to obtain accurate new energy output prediction results, thereby improving the accuracy of the new energy output prediction results.

[0066] In an exemplary embodiment, the time series of influencing factors is reconstructed and decomposed to obtain candidate subsequences, including: obtaining time delay information and embedding dimension information corresponding to the time series of influencing factors; decomposing the time series of influencing factors into phase space according to the time delay information and the embedding dimension information; superimposing the phase space to obtain a reconstructed time series; and decomposing the reconstructed time series into candidate subsequences according to the average value information and residual information corresponding to the reconstructed time series.

[0067] The time delay information may refer to information used to construct a multi-dimensional phase space vector from an original time series. In practical applications, the time delay information may be represented as t.

[0068] The embedding dimension information may refer to the information of the number of dimensions required to characterize the coordinates of a time series data point in the phase space. In practical applications, the embedding dimension information may be expressed as m.

[0069] The phase space may refer to a phase space vector obtained by decomposing the time series of influencing factors based on time delay information and embedded dimension information.

[0070] The reconstructed time series may refer to a sequence obtained by superimposing the phase space. In practical applications, the reconstructed time series may be used as the reconstructed phase space.

[0071] The average value information may refer to information representing a smooth curve obtained by averaging the upper envelope and the lower envelope of the time series. In practical applications, the average value information may be expressed as m(t).

[0072] The residual information may refer to information representing the residual component of the time series.

[0073] As an example, the server can obtain the time delay information and embedding dimension information corresponding to the influencing factor time series, and then the server can decompose the influencing factor time series into phase space according to the time delay information and the embedding dimension information, and superimpose the phase space to obtain the reconstructed time series. The server can obtain the upper envelope and lower envelope corresponding to the reconstructed time series, and determine the mean value information and residual information corresponding to the reconstructed time series based on the upper envelope and the lower envelope. Then the server can decompose the reconstructed time series into candidate subsequences according to the mean value information and residual information corresponding to the reconstructed time series.

[0074] In this embodiment, by obtaining the time delay information and the embedded dimension information corresponding to the influencing factor time series; decomposing the influencing factor time series into a phase space according to the time delay information and the embedded dimension information; superimposing the phase space to obtain a reconstructed time series; decomposing the reconstructed time series into candidate subsequences according to the average value information and the residual information corresponding to the reconstructed time series, the influencing factor time series can be reconstructed based on the time delay information and the embedded dimension information corresponding to the influencing factor time series, and the influencing factor time series can be accurately decomposed into candidate subsequences in combination with the average value information and the residual information corresponding to the reconstructed time series, and the influencing factor time series can be denoised, key information can be extracted, and the accuracy of the candidate subsequences can be improved, so as to provide a data basis for the subsequent generation of new energy output prediction results based on the target subsequences obtained by screening the candidate subsequences, thereby improving the accuracy of the new energy output prediction results.

[0075] In some embodiments, the above method also includes: determining the average value information corresponding to the reconstructed time series based on the maximum value information and minimum value information corresponding to the reconstructed time series; determining the intrinsic mode function information based on the difference between the reconstructed time series and the average value information corresponding to the reconstructed time series; determining the residual information corresponding to the reconstructed time series based on the difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series.

[0076] Among them, the maximum value information can refer to the information representing the value of the state information of the point corresponding to the maximum value in the reconstructed time series. In practical applications, each state information in the reconstructed time series can be regarded as a point, and the value of the state information of the point corresponding to the maximum value is greater than the value of the state information of the points on both sides of the point. In specific implementation, the maximum value information can be expressed as e max (t).

[0077] Among them, the minimum value information can refer to the information that represents the value of the state information of the point corresponding to the minimum value in the reconstructed time series. In practical applications, each state information in the reconstructed time series can be regarded as a point, and the value of the state information of the point corresponding to the minimum value is smaller than the value of the state information of the points on both sides of the point. In specific implementation, the maximum value information can be expressed as e min (t).

[0078] Among them, the intrinsic mode function information can refer to the information used to determine the residual information. In practical applications, the intrinsic mode function information can be expressed as Imf i (t).

[0079] As an example, the server can first analyze the data distribution in the reconstructed time series, determine the maximum value information and the minimum value information corresponding to the reconstructed time series, and average the maximum value information and the minimum value information corresponding to the reconstructed time series to obtain the average value information corresponding to the reconstructed time series. Then, the server can use the difference between the reconstructed time series and the average value information corresponding to the reconstructed time series as a new function, replace the reconstructed time series with the new function, determine the average value information of the new function in the above manner, and calculate the difference between the new function and the average value information corresponding to the new function until the new function meets the intrinsic mode function (IMF) condition. The server can use the new function at this time as an intrinsic mode function information Imf i (t), the terminal can determine the residual information corresponding to the reconstructed time series according to the difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series. The residual information can be used to replace the reconstructed time series, and the server can obtain the residual information corresponding to the residual information in the above manner until the residual information corresponding to the residual information is a monotonic function or is less than a preset value, and the server determines that it cannot be further decomposed. The server can combine each residual information and the intrinsic mode function information corresponding to each residual information to generate a candidate subsequence.

[0080] In this embodiment, the average value information corresponding to the reconstructed time series is determined according to the maximum value information and the minimum value information corresponding to the reconstructed time series; the intrinsic mode function information is determined according to the difference between the reconstructed time series and the average value information corresponding to the reconstructed time series; the residual information corresponding to the reconstructed time series is determined according to the difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series. The average value and the intrinsic mode function can be determined based on the maximum and minimum values ​​corresponding to the reconstructed time series, and the residual information can be determined in combination with the reconstructed time series, the average value and the intrinsic mode function, so as to improve the accuracy of the residual information, thereby improving the accuracy of the candidate subsequence, and facilitating the subsequent generation of a new energy output prediction result based on the target subsequence obtained by screening the candidate subsequences, thereby improving the accuracy of the new energy output prediction result.

[0081] In some embodiments, a target subsequence is selected from a candidate subsequence according to fitness information corresponding to an influencing factor sequence set, including: selecting a candidate influencing factor sequence set from an influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set; exchanging candidate subsequences in the candidate influencing factor sequence set with candidate subsequences in the influencing factor sequence set to obtain an exchanged influencing factor sequence set; and using the candidate subsequences in the exchanged influencing factor sequence set as the target subsequence.

[0082] The candidate influencing factor sequence set may refer to an influencing factor sequence set whose fitness information meets a preset fitness requirement.

[0083] The exchanged influencing factor sequence set may refer to a set obtained by exchanging a candidate subsequence in the candidate influencing factor sequence set with a candidate subsequence in the influencing factor sequence set.

[0084] As an example, the server may first obtain the fitness information corresponding to the influencing factor sequence set, and then screen out a candidate influencing factor sequence set from the influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set. The server may screen out a candidate subsequence from the candidate influencing factor sequence set based on the roulette algorithm and exchange it with the candidate subsequence in the influencing factor sequence set to obtain the influencing factor sequence set after the exchange, thereby realizing gene exchange. The server may use the candidate subsequence in the influencing factor sequence set after the exchange as the target subsequence. It is understandable that the server may also perform gene mutation on the candidate subsequence in the influencing factor sequence set after the exchange. The server may update the population of the genetic algorithm through gene exchange and gene mutation. When the number of population updates reaches a preset number, the server may use the candidate subsequence in the population at this time as the target subsequence.

[0085] In this embodiment, a candidate influencing factor sequence set is screened out from the influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set; the candidate subsequences in the candidate influencing factor sequence set are exchanged with the candidate subsequences in the influencing factor sequence set to obtain the exchanged influencing factor sequence set; the candidate subsequences in the exchanged influencing factor sequence set are used as target subsequences, and the candidate influencing factor sequence set can be screened based on the fitness information, and the candidate subsequences in the candidate influencing factor sequence set are exchanged with the candidate subsequences in the influencing factor sequence set, thereby realizing gene exchange, which can improve the accuracy of the target subsequence, provide a data basis for the subsequent new energy output prediction model to process the target subsequence, and thus improve the accuracy of the new energy output prediction result.

[0086] In some embodiments, the target subsequence is input into a new energy output prediction model to obtain a new energy output prediction result, including: obtaining correlation coefficient information corresponding to the target subsequence; when the correlation coefficient information is greater than a preset correlation coefficient threshold, normalizing the target subsequence to obtain a processed subsequence; and inputting the processed subsequence into the new energy output prediction model to obtain a new energy output prediction result.

[0087] Among them, the correlation coefficient information may refer to information characterizing the degree of influence of the target subsequence on the power generation of the new energy power generation equipment. In practical applications, the larger the value of the correlation coefficient information, the greater the influence of the target subsequence on the power generation of the new energy power generation equipment, and vice versa.

[0088] The preset correlation coefficient threshold may refer to information used to determine whether the correlation coefficient information corresponding to the target subsequence meets the preset correlation degree requirement.

[0089] The processed subsequence may refer to a sequence obtained after normalizing the target subsequence.

[0090] As an example, the server can calculate the correlation coefficient information corresponding to the target subsequence according to a preset correlation coefficient calculation expression. When the correlation coefficient information is less than or equal to a preset correlation coefficient threshold, the server determines that the target subsequence at this time has no obvious impact on the power generation of the new energy power generation equipment. The server generates a prompt message to prompt the user to check whether the initial influencing factor time series and / or status information is incorrect. When the correlation coefficient information is greater than the preset correlation coefficient threshold, the server can normalize the target subsequence to obtain a processed subsequence, and then the server can input the processed subsequence into the new energy output prediction model to obtain a new energy output prediction result.

[0091] In this embodiment, by obtaining the correlation coefficient information corresponding to the target subsequence; when the correlation coefficient information is greater than a preset correlation coefficient threshold, the target subsequence is normalized to obtain a processed subsequence; the processed subsequence is input into the new energy output prediction model to obtain a new energy output prediction result, and the influence of the target subsequence on the power generation of the new energy power generation equipment can be analyzed in combination with the correlation coefficient, and the target subsequence whose influence meets the preset correlation degree requirement is screened, and the target subsequence is normalized to facilitate the new energy output prediction model to process the target subsequence, so that the new energy output prediction model can output accurate new energy output prediction results, thereby improving the accuracy of the new energy output prediction results.

[0092] In some embodiments, the above method also includes: constructing a sample subsequence based on the state information corresponding to each influencing factor within the sample time period, and inputting the sample subsequence into the new energy output prediction model to obtain a sample prediction result; obtaining the new energy output information under the influence of each influencing factor in the test time period corresponding to the sample time period as test data; and training the new energy output prediction model to be trained based on the test data and the sample prediction results.

[0093] The sample time period may refer to any time period in which the power generation of the new energy power generation equipment is known. In practical applications, the sample time period may include any time period before the current moment.

[0094] The sample subsequence may refer to information used to provide output to the new energy output prediction model and obtain sample prediction results.

[0095] The sample prediction result may refer to information representing the predicted value of the power generation or working state of the new energy power generation equipment within the test time period when the state of the influencing factor within the sample time period is the influencing factor corresponding to the sample subsequence.

[0096] The test time period may refer to the time period after the sample time period. In practical applications, the sample prediction result may represent the predicted value of the power generation or working status of the new energy power generation equipment during the test time period.

[0097] The new energy output information may refer to information representing the actual power generation or actual working status of the new energy power generation equipment during the test period.

[0098] As an example, the server can construct a sample subsequence based on the status information corresponding to each influencing factor within the sample time period, and input the sample subsequence into the new energy output prediction model to obtain the sample prediction result. The server can then obtain the actual new energy output information within the test time period corresponding to the sample time period as test data. The server can train the new energy output prediction model to be trained based on the test data and the sample prediction results. When the similarity between the test data and the sample prediction results meets the preset similarity requirements, the server determines that the training of the new energy output prediction model is completed.

[0099] In this embodiment, a sample subsequence is constructed according to the state information corresponding to each influencing factor in the sample time period, and the sample subsequence is input into the new energy output prediction model to obtain a sample prediction result; the new energy output information under the influence of each influencing factor in the test time period corresponding to the sample time period is obtained as test data; according to the test data and the sample prediction result, the new energy output prediction model to be trained is trained, and the new energy output prediction model can be trained based on the state information of the influencing factors in the sample time period and the actual new energy output information in the test time period, and the model performance of the new energy output prediction model is optimized, thereby improving the accuracy of the new energy output prediction result output by the new energy output prediction model.

[0100] In some embodiments, the analysis of the output characteristics of distributed renewable energy shows that wind and solar power output has strong randomness, volatility and anti-peak characteristics. It is necessary to improve the prediction accuracy of renewable energy output through technical means, and take corresponding dispatching strategies in the active distribution network based on this to reduce the impact of large-scale grid connection of renewable energy on the active distribution network. In order to improve the prediction accuracy of renewable energy output, a neural network prediction model based on sequence decomposition and selection is established, such as Figure 3 As shown, a flow chart of a distributed renewable energy output prediction method based on sequence decomposition is provided. The server first reconstructs the time series of each influencing factor through phase space and empirical mode decomposition to generate a large number of subsequences, and then selects the subsequences by combining an improved three-dimensional genetic algorithm with a neural network. Finally, the optimized subsequences of related influencing factors are input into the BP neural network and the long short-term memory neural network (Long Short-Term Memory, LSTM) for prediction to obtain the prediction results.

[0101] (2) In the specific implementation, the server can first perform data preprocessing on the status information corresponding to the influencing factors: (1) Data cleaning. In simulation experiments, experimental data often have data missing, data duplication, and data anomaly problems due to storage device damage, repeated storage, sensor failure, etc. Missing data will affect the normal operation of the program, and the automatically filled data will differ too much from the actual value. Data or outliers that deviate greatly from the actual will affect the results of phase space reconstruction and empirical mode decomposition, and increase the training error of the neural network. Data duplication will also affect the normal operation of the program and reduce the effectiveness of decomposition and prediction. Therefore, it is necessary to clean the data before conducting the experiment. Process the data of new energy output power and related influencing factors, and treat points outside the reasonable range as outlier points. After deleting the data of outlier points, treat them as missing data for processing. For missing data, use the average value of the data at the previous moment and the next moment of the missing value to supplement the data at that moment. For duplicate data, the deletion method is used for processing; (2) The maximum offset rate of node voltage, correlation analysis is an important means to determine whether the relevant influencing factor sequence (such as the influencing factor time series) is closely related to the output of renewable energy. Common correlation coefficients include Pearson, Spearman and Kendall. The Pearson correlation coefficient is the covariance of the two sequences divided by the product of their respective standard deviations. The calculation expression of the Pearson correlation coefficient can be expressed as:

[0102] .

[0103] Among them, X and Y can be two time series of influencing factors.

[0104] Covariance can reflect the strength of the correlation between two variables, but if the variable itself has a large degree of dispersion, it will affect the covariance, so the covariance is divided by the standard deviation as an analysis indicator. The Pearson correlation coefficient is distributed between -1 and 1. The closer it is to 1, the greater the positive correlation is, the closer it is to -1, the greater the negative correlation is, and the closer it is to 0, the lower the correlation is. The Spearman correlation coefficient and the Kendall correlation coefficient pay more attention to the size order in the sequence, reducing the impact of extreme numbers on the correlation, but the sensitivity to numbers is reduced.

[0105] After data preprocessing, the server can reconstruct the phase space of the time series of influencing factors. Specifically, multivariate phase space reconstruction (MPSR) is superimposed on the single variable phase space reconstruction. The time series matrix obtained by single variable phase space reconstruction can be expressed as:

[0106] .

[0107] That is, a chaotic time series I is decomposed into m M-dimensional subsequences as the chaotic attractors of the system. i It can be a phase space state point (i = 1, 2, …M), m is the embedding dimension, t is the time delay, and M is the subsequence length.

[0108] The acquisition of time delay t and embedding dimension m is the key to whether the phase space reconstruction is reasonable. The server can select time delay t and embedding dimension m based on the CC (Convergence-Criterion algorithm). The calculation expression of time delay t and embedding dimension m can be expressed as:

[0109] .

[0110] The server can remain embedded in the window frame W Under the premise of unchanged, adjust the time delay t and embedding dimension m. First, the server can construct the correlation integral C s , the calculation expression of the correlation integral can be expressed as:

[0111] .

[0112] Among them, the correlation integral C s It can represent the probability that the spatial distance between any two points in the sth sequence is less than r. θ(x) can be a unit function. When x is greater than zero, θ is 1, otherwise it is 0. i 、x j are two points in the time series, d ij is the Euclidean distance between two points.

[0113] The server can also construct an objective function, the calculation expression of which can be expressed as:

[0114]

[0115] .

[0116] The reconstructed time delay t can be obtained by , and The three statistics are calculated and the optimal delay time is The first zero point or The time corresponding to the first minimum value of The time corresponding to the minimum value.

[0117] Based on the above single variable phase space reconstruction, the server can decompose the d influencing factor time series into d phase spaces, and then superimpose them. The reconstructed new space can be expressed as:

[0118] .

[0119] Among them, I all can be the space after superposition, I i It can be a decomposition sequence for each phase space (I = 1, 2, …d).

[0120] Embedding dimension m all It can be expressed as:

[0121] .

[0122] Among them, the embedding dimension m all It can be the sum of all univariate dimensions.

[0123] Distributed renewable energy output forecast is greatly affected by meteorological factors and shows strong nonlinearity and non-stationarity. Empirical mode decomposition is an adaptive decomposition method that uses extreme point information to decompose the waveform into several intrinsic mode functions. It has no restrictions on the decomposed waveform and has greater flexibility. Therefore, the server can perform empirical mode decomposition. Specifically, the server can find the maximum and minimum points for a given time series u(t) (such as the reconstructed time series), connect the extreme points, obtain the maximum and minimum envelopes, and calculate the average line (such as the average information). The calculation expression of the average information can be expressed as:

[0124] .

[0125] Among them, e min (t) and e max (t) are the maximum and minimum envelopes respectively, and m(t) is the mean line.

[0126] The server uses the reconstructed time series to subtract the average value information to obtain a new function. The calculation expression of the new function can be expressed as:

[0127] .

[0128] The server can use the new function h 1 (t) replaces u(t) until h k (t) satisfies the IMF conditions, denoted by Imf 1 (t) = h k (t), repeat the above steps of calculating the average value and obtaining the new function. nWhen (t) is a monotonic function or is less than a given value, it cannot be further decomposed and n intrinsic mode functions Imf are obtained. i (t), the original data (time series of influencing factors / candidate subsequences) can be expressed as:

[0129] .

[0130] Since each influencing factor sequence obtains many subsequences after phase space reconstruction and empirical mode decomposition, there is useless information and redundant information with poor correlation. If all the information is used as the input of the neural network, the neural network will have a large amount of calculation and it is easy to cause overfitting. Therefore, the server needs to group and select the subsequences of each influencing factor. Specifically, the grouping requirements may include independent selection between influencing factors to ensure that each subsequence has the opportunity to participate in the prediction; evolution within the influencing factors to reasonably select each influencing factor subsequence; the sequences at the same gene position of different influencing factors may have similar effects on the prediction effect, and the effects between the influencing factors must also be considered. The server can construct an initial population based on the candidate subsequences, and the initial population can be used as an input for an improved three-dimensional genetic algorithm. Figure 2 As shown, a schematic diagram of the initial population of a genetic algorithm is provided. The server can select the most suitable influencing factors for the input and combine them to achieve group selection of particle internal parameters.

[0131] The steps of the above three-dimensional genetic algorithm may include: (1) defining a structural framework as shown in the figure above, where the population contains n individuals, the number of chromosomes in the individual is the number of influencing factors (d), the number of genes on the chromosome is the number of subsequences of the influencing factors, and they are initialized; (2) bringing all individuals (such as the influencing factor sequence set) into the fitness function and calculating the fitness; (3) selection, the server replicates the chromosome based on the roulette algorithm according to the fitness; (4) crossover, the server can regard the same chromosome of different individuals as different chromosomes of the traditional genetic algorithm, and exchange genes for different chromosomes according to a certain probability; (5) the server can perform gene mutation according to a certain probability based on the gene distribution of other chromosomes of the individual, such as Figure 4 As shown in Figure 1, a schematic diagram of gene mutation is provided. First, the sliding window size is set to n / 3 to divide the genes on the chromosome; second, the gene probability S is calculated. i , gene probability S i The calculation expression can be expressed as:

[0132] .

[0133] Among them, S iis the gene probability of the window, i is the window number, n is the number of genes on the chromosome, and n(j) is the value of the jth gene, where the value of n is 0 or 1.

[0134] Gene probability S i It can represent the gene distribution. If S i Less than 0.2, indicating that there are more 0s in this sequence, S i Between 0.2 and 0.8, the distribution is relatively uniform, S i is greater than 0.8, then 1 is more; finally, if S i At (0, 0.2) or (0.8, 1), the genes in the corresponding parts of other chromosomes have a 0.05 chance of mutating in that direction.

[0135] (6) The updated population recalculates the fitness and repeats operations (2) to (5) until the specified number of cycles / updates is reached. Since the optimization process of the genetic algorithm is random and the correlation between most sequences and the predicted sequence is weak, the result obtained by inputting such sequence combinations into the neural network is poor and has no reference value. Therefore, in order to shorten the optimization time of the optimization algorithm, when calculating the fitness, the server can first perform a correlation analysis on the sequence to determine the correlation coefficient information F. The calculation expression of the correlation coefficient information F can be expressed as:

[0136] .

[0137] Among them, F is the fitness / correlation coefficient information of the current subsequence combination, N is the total number of selected subsequences, and ρ i is the correlation between the ith sequence and the predicted sequence, FA is the prediction accuracy after neural network prediction, m is the correlation threshold, and fa is the prediction accuracy threshold. After experiments, m is taken as 0.2 and fa is taken as 50%.

[0138] When the average value of the correlation coefficient is less than the preset correlation coefficient threshold, no prediction is performed and the fitness is the average value of the correlation coefficient; when it is greater than the threshold, neural network prediction is performed and the prediction accuracy is normalized to an appropriate range as the fitness. Considering that a large number of input sequences will affect the training speed of the neural network, a penalty function is designed. The calculation expression of the penalty function can be expressed as:

[0139] .

[0140] Among them, N r is the number of chromosomes, N j is the number of genes on the chromosome, n(ij) is the value of the gene, the denominator represents the total number of selected subsequences, M is the sequence cardinality, which is 1000. C is the penalty function.

[0141] The penalty function is the total number of genes with a value of 1 in the individual. The fitness is subtracted from the penalty function to obtain the new fitness / correlation coefficient information F. The new correlation coefficient information F new The calculation expression can be expressed as:

[0142] .

[0143] like Figure 5 As shown, a flow chart of a genetic algorithm is provided. The server can obtain the pre-set hyperparameters such as population size and crossover rate for the three-dimensional genetic algorithm and randomly generate an initial population using candidate subsequences, thereby completing the initialization of the genetic algorithm. Then the server can convert the chromosomes and genes of the individuals in the population into sequence combinations. The server can calculate the correlation coefficient between each sequence and the distributed new energy output. When the correlation coefficient reaches the threshold, it is input into the neural network for training. The server can obtain the fitness of the group of sequence combinations based on the correlation and prediction accuracy. After calculating the fitness of all individuals in the population, the population is updated (such as selection, crossover and mutation) until the number of iterations is reached. The server can output the target subsequence, which can be used as the optimal sequence combination.

[0144] The new energy output prediction model can include a long short-term memory neural network. The long short-term memory neural network is a type of recurrent neural network and is a commonly used prediction model in time series prediction problems. The long short-term memory neural network also retains the influence of the previous input characteristics on the output results, avoiding the gradient disappearance and gradient explosion problems, and has a strong nonlinear fitting ability. In the specific implementation, the iterative process of the long short-term memory neural network includes three parts: memory gate, input gate and output gate. Among them, the working process of the memory gate can be expressed as:

[0145] .

[0146] Among them, x t and h t is the input information and output information of the tth round, σ is the sigmoid function, W f and b f are the weights and thresholds of the memory gate, which are trainable parameters.

[0147] The working process of the input gate can be expressed as:

[0148] .

[0149] Among them, C t is the memory information of the tth round, Wi, bi, W C 、b C is the weight and threshold of the input gate.

[0150] The working process of the output gate can be expressed as:

[0151] .

[0152] Among them, W o and b o is the weight and threshold of the input gate.

[0153] In order to optimize the performance of the new energy output prediction model, the server can obtain the state information corresponding to each influencing factor within the sample time period, and then the server can construct a sample subsequence based on the state information corresponding to each influencing factor within the sample time period, and perform sample normalization on the sample subsequence. Specifically, since the data set composed of the state information includes distributed new energy output, temperature, wind speed, light intensity and other data that are not in the same order of magnitude, and the activation function in the neural network is very sensitive to the data scale, when the value reaches about 3, the neurons will be inactivated, causing the network to lose its learning ability. Therefore, the data needs to be normalized before being input into the neural network. The expression of the normalization formula can be expressed as:

[0154] .

[0155] Among them, x min and x max are the maximum and minimum values ​​in the sample data, x is the data before normalization, and x in is the normalized data.

[0156] Based on the different sample time periods, the sample prediction results can be divided into two types: day-ahead prediction results and rolling prediction results. Among them, the prediction task corresponding to the day-ahead prediction result can be the day-ahead prediction. The day-ahead prediction is to predict the output of the day on the previous day. The time interval is 1 hour. The new energy output prediction model outputs 24 prediction values; the prediction task corresponding to the rolling prediction result can be the rolling prediction. The rolling prediction is to predict the next 4 hours with an interval of 15 minutes. A rolling prediction is performed every 1 hour to cover the previous prediction result. The new energy output prediction model outputs 16 prediction values. Due to the shift of the influencing factor data during phase space reconstruction, the maximum shift is 12 hours, so the data on January 1 is deleted. The training set can be the new energy output information (such as wind and solar output) and the decomposed influencing factor data (such as the target subsequence) from January 2 to the 8 days before the prediction date (such as the prediction time period), the validation set is the data one week before the prediction date, and the test set is the data on the prediction date. The server can input the sample subsequence in the training set into the new energy output prediction model to obtain the sample prediction results. The server can also obtain the new energy output information under the influence of various influencing factors in the test time period corresponding to the sample time period as the test data. The server can train the new energy output prediction model to be trained based on the test data and the sample prediction results. In the specific implementation, the evaluation indicators of the new energy output prediction model may include but are not limited to the root mean square error (RMSE), mean absolute percentage error (MAPE) and forecast accuracy (FA). Among them, the calculation expression of the root mean square error can be expressed as:

[0157] .

[0158] Among them, y act is the true value, y pre is the predicted value, and m is the sample size of the predicted result.

[0159] The calculation expression of mean absolute percentage error can be expressed as:

[0160] .

[0161] Among them, y act is the true value, y pre is the predicted value, and m is the sample size of the predicted result.

[0162] The calculation expression of prediction accuracy can be expressed as:

[0163] .

[0164] Among them, yact is the true value, y pre is the predicted value, and m is the sample size of the predicted result.

[0165] In this embodiment, by introducing phase space reconstruction and empirical mode decomposition algorithms, complex and unstable time series data can be more accurately decomposed into multiple stable subsequences, which not only improves the accuracy of data processing, but also effectively retains high-frequency data and reduces the loss of key information. Secondly, in the selection part, based on the improved genetic algorithm, the concept of individuals is introduced to achieve the optimal selection of sequences, making the sequence input to the neural network more representative and effective. In the prediction part, a more efficient LSTM neural network is used, and the parameters are optimized through the improved three-dimensional genetic algorithm, which further improves the accuracy and computational efficiency of the prediction model. In addition, in view of the imbalance between the output of new energy and the load demand, it can better cope with the challenges brought by the large-scale grid connection of new energy and ensure the stable operation of the power grid. Not only does it significantly improve the accuracy and real-time performance of the prediction, but it also greatly reduces the computational complexity, providing strong technical support for the efficient use of new energy and the smooth operation of the power system.

[0166] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0167] Based on the same inventive concept, the embodiment of the present application also provides a distributed new energy output prediction device based on sequence decomposition for implementing the above-mentioned distributed new energy output prediction method based on sequence decomposition. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the distributed new energy output prediction device based on sequence decomposition provided below can refer to the limitations of the distributed new energy output prediction method based on sequence decomposition above, and will not be repeated here.

[0168] In an exemplary embodiment, Figure 6 As shown, a distributed new energy output prediction device based on sequence decomposition is provided, including: a decomposition module 602, a screening module 604 and a prediction module 606, wherein:

[0169] The decomposition module 602 is used to generate an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy in the prediction time period, and reconstruct and decompose the influencing factor time series to obtain a candidate subsequence.

[0170] The screening module 604 is used to construct an influencing factor sequence set according to the candidate subsequences, and screen out a target subsequence from the candidate subsequences according to the fitness information corresponding to the influencing factor sequence set.

[0171] The prediction module 606 is used to input the target subsequence into the new energy output prediction model to obtain the new energy output prediction result.

[0172] In one of the exemplary embodiments, the decomposition module 602 is further specifically used to obtain the time delay information and the embedding dimension information corresponding to the influencing factor time series; decompose the influencing factor time series into a phase space according to the time delay information and the embedding dimension information; superimpose the phase space to obtain a reconstructed time series; and decompose the reconstructed time series into the candidate subsequences according to the average value information and the residual information corresponding to the reconstructed time series.

[0173] In one of the exemplary embodiments, the decomposition module 602 is further specifically used to determine the average value information corresponding to the reconstructed time series based on the maximum value information and the minimum value information corresponding to the reconstructed time series; determine the intrinsic mode function information based on the difference between the reconstructed time series and the average value information corresponding to the reconstructed time series; determine the residual information corresponding to the reconstructed time series based on the difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series.

[0174] In one of the exemplary embodiments, the above-mentioned screening module 604 is specifically used to screen out a candidate influencing factor sequence set from the influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set; exchange the candidate subsequences in the candidate influencing factor sequence set with the candidate subsequences in the influencing factor sequence set to obtain an exchanged influencing factor sequence set; and use the candidate subsequences in the exchanged influencing factor sequence set as the target subsequence.

[0175] In one of the exemplary embodiments, the prediction module 606 is further specifically used to obtain the correlation coefficient information corresponding to the target subsequence; when the correlation coefficient information is greater than a preset correlation coefficient threshold, the target subsequence is normalized to obtain a processed subsequence; and the processed subsequence is input into the new energy output prediction model to obtain the new energy output prediction result.

[0176] In one of the exemplary embodiments, the above-mentioned device also includes a training module, which is specifically used to construct a sample subsequence based on the state information corresponding to the various influencing factors within the sample time period, and input the sample subsequence into the new energy output prediction model to obtain a sample prediction result; obtain the new energy output information under the influence of the various influencing factors in the test time period corresponding to the sample time period as test data; and train the new energy output prediction model to be trained based on the test data and the sample prediction result.

[0177] Each module in the above-mentioned distributed renewable energy output prediction device based on sequence decomposition can be implemented in whole or in part by software, hardware and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0178] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a distributed new energy output prediction method based on sequence decomposition is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0179] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0180] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0181] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0182] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0183] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0184] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0185] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0186] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A distributed renewable energy output prediction method based on sequence decomposition, characterized in that: The method comprises: Generate an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy within the forecast time period, reconstruct and decompose the influencing factor time series to obtain a candidate subsequence; constructing an influencing factor sequence set according to the candidate subsequences, and screening out a target subsequence from the candidate subsequences according to fitness information corresponding to the influencing factor sequence set; The target subsequence is input into a new energy output prediction model to obtain a new energy output prediction result.

2. The method according to claim 1, characterized in that The reconstructing and decomposing the influencing factor time series to obtain a candidate subsequence includes: Obtaining time delay information and embedding dimension information corresponding to the time series of the influencing factors; Decomposing the influencing factor time series into a phase space according to the time delay information and the embedding dimension information; Superimposing the phase space to obtain a reconstructed time series; The reconstructed time series is decomposed into the candidate subsequences according to the mean value information and the residual information corresponding to the reconstructed time series.

3. The method according to claim 2, characterized in that The method further comprises: Determine the average value information corresponding to the reconstructed time series according to the maximum value information and the minimum value information corresponding to the reconstructed time series; Determining intrinsic mode function information according to a difference between the reconstructed time series and the average value information corresponding to the reconstructed time series; Residual information corresponding to the reconstructed time series is determined according to a difference between the reconstructed time series and the intrinsic mode function information corresponding to the reconstructed time series.

4. The method according to claim 1, characterized in that: The step of selecting a target subsequence from the candidate subsequences according to the fitness information corresponding to the influencing factor sequence set includes: Filtering a candidate influencing factor sequence set from the influencing factor sequence set according to the fitness information corresponding to the influencing factor sequence set; exchanging a candidate subsequence in the candidate influencing factor sequence set with a candidate subsequence in the influencing factor sequence set to obtain an exchanged influencing factor sequence set; The candidate subsequence in the exchanged influencing factor sequence set is used as the target subsequence.

5. The method according to claim 1, characterized in that The step of inputting the target subsequence into a new energy output prediction model to obtain a new energy output prediction result includes: Obtaining correlation coefficient information corresponding to the target subsequence; When the correlation coefficient information is greater than a preset correlation coefficient threshold, normalizing the target subsequence to obtain a processed subsequence; The processed subsequence is input into the new energy output prediction model to obtain the new energy output prediction result.

6. The method according to claim 1, characterized in that The method further comprises: Constructing a sample subsequence according to the state information corresponding to each influencing factor within the sample time period, and inputting the sample subsequence into the new energy output prediction model to obtain a sample prediction result; Acquire the new energy output information under the influence of each influencing factor in the test time period corresponding to the sample time period as test data; The new energy output prediction model to be trained is trained according to the test data and the sample prediction results.

7. A distributed new energy output prediction device based on sequence decomposition, characterized in that: The device comprises: A decomposition module, used to generate an influencing factor time series corresponding to each influencing factor according to the state information of each influencing factor affecting the output of new energy within the forecast time period, and to reconstruct and decompose the influencing factor time series to obtain a candidate subsequence; A screening module, used to construct an influencing factor sequence set according to the candidate subsequences, and screen out a target subsequence from the candidate subsequences according to fitness information corresponding to the influencing factor sequence set; The prediction module is used to input the target subsequence into the new energy output prediction model to obtain the new energy output prediction result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images