Multi-energy consumption data completion method based on wavelet decomposition and Fourier transform

Abstract

The present invention discloses a multi-energy consumption data completion method based on wavelet decomposition and Fourier transform, which includes the following steps: 1. performing wavelet decomposition on the collected carbon-containing characteristic energy consumption sequence to obtain an energy consumption cycle characteristic sequence and an energy consumption trend characteristic sequence; 2. obtaining a predicted energy consumption trend characteristic sequence based on curve fitting; 3. obtaining a predicted energy consumption cycle characteristic sequence based on Fourier series fitting; and 4. completing missing data through wavelet reconstruction based on steps 2 and 3. The present invention constructs a data completion model based on wavelet decomposition and Fourier transform, thereby enabling the completion of massive multi-energy consumption data for key emission-controlled enterprises.

Description

Technical Field

[0001] The invention relates to a multi-energy usage data completion method and belongs to the field of data analysis. Background Art

[0002] Since 2010, mankind has entered the era of big data. The advent of the big data era has brought many opportunities and challenges to the data mining technology of various types of carbon-containing energy consumption. Today, our research on various types of carbon-containing energy consumption data no longer uses the sampling survey method, but a comprehensive analysis of all types of carbon-containing energy consumption data. The significant characteristics of various types of carbon-containing energy consumption data are obvious periodicity and trend, many types, fast flow rate and large data volume. When collecting various types of carbon-containing energy consumption data on a daily basis, the phenomenon of missing various types of carbon-containing energy consumption data often occurs. There are many reasons for the loss of various types of carbon-containing energy consumption data, such as accidental omission of information, inability to obtain information, etc. The missing of various types of carbon-containing energy consumption data will affect the establishment of carbon-related real-time monitoring models for key emission-controlled enterprises. The commonly used methods for handling missing values ​​of various types of carbon-containing energy consumption data are as follows:

[0003] The first approach involves simply deleting the usage data for each carbon-containing energy source for a single cycle. This approach is simple and effective when the deleted usage data for a single cycle accounts for a small proportion of the overall data. However, this approach can degrade the quality of the data mining algorithm when the proportion of missing values ​​fluctuates significantly. Furthermore, the deleted usage data for a single cycle may contain important information, distorting the data and even leading to erroneous conclusions.

[0004] The second type of method infers and fills in missing values ​​for various carbon-containing energy consumption data. This method is generally based on statistical principles and uses different algorithms to fill in missing values. Common data filling algorithms include: mean (or median) filling, special value filling, manual filling, k-nearest neighbor method, etc. However, these algorithms are only suitable for the overall trend of various carbon-containing energy consumption data and cannot assess the cyclical fluctuations of various carbon-containing energy consumption data.

[0005] The third approach uses no processing and instead utilizes deep mining methods, such as Bayesian networks and artificial neural networks. This approach involves directly mining data series containing missing data on various carbon-containing energy consumption data. However, due to the massive amount of data available, deep mining can reduce the efficiency of data completion and lead to excessive processing time. Summary of the Invention

[0006] In view of the above deficiencies in the prior art, the present invention provides a multi-energy consumption data completion method based on wavelet decomposition and Fourier transform, aiming to minimize the processing time of massive data of various carbon-containing energy consumptions on the basis of meeting data completion, so as to improve the efficiency of massive data completion of various carbon-containing energy consumptions.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A multi-energy consumption data completion method based on wavelet decomposition and Fourier transform of the present invention is characterized by including the following steps:

[0009] Step 1: Collect data on various carbon-containing energy consumptions of key emission control enterprises according to the sampling period to obtain a characteristic carbon-containing energy consumption data set represents the carbon-containing characteristic energy consumption data of the jth type collected at the mth sampling moment; 1 ≤ m ≤ i, and the carbon-containing characteristic energy consumption data at the mth sampling moment has characteristics including periodic characteristics and trend characteristics; i represents the total sampling moments;

[0010] Assume that the carbon-containing energy consumption data of the jth type collected at the mth sampling moment is missing data, then select the carbon-containing energy consumption data at the n sampling moments before the missing data to form a carbon-containing energy consumption sequence and n is an even number divisible by 8, 1 ≤ n < m; where, represents the carbon-containing energy consumption data of the jth type collected at the (m - n + 1)th sampling moment;

[0011] Perform 3-layer wavelet decomposition on the carbon-containing energy consumption sequence at n sampling moments through equations (1-1)-(1-2) to obtain the energy consumption period characteristic sequence of the p-layer wavelet decomposition of the energy consumption sequence and the energy consumption trend characteristic sequence of the p-layer wavelet decomposition

[0012]

[0013]

[0014] In equations (1-1)-(1-3), n , , p , , ,

[0013] , ,

[0012] , p , ,

[0014] , is the sampling moment of the p-layer wavelet decomposition; s is the abscissa of the wavelet space, s ≤ 2n p ; is the sequence of the p-layer wavelet decomposition; Z is the set of all integers, s ∈ Z; is the jth type of energy consumption period characteristic obtained by the p-layer wavelet decomposition; The energy consumption trend feature of the j-th type obtained by the p-th layer wavelet decomposition; G(s - 2n p ) is the high-pass filter function; H(s - 2n p ) is the low-pass filter function;

[0015] Step 2: Obtain the fitting prediction function of the energy consumption trend feature sequence and use Equation (2-1) to determine the four curve fitting coefficients ω0, ω1, ω2, ω3 of the energy consumption trend feature prediction function In Equations (2-1) - (2-2), ε is a constant;

[0016]

[0017]

[0018]

[0019] Step 3: Use Equation (3-1) to obtain the k-th main frequency of the energy consumption cycle feature sequence 1 ≤ k < N - 1, and then use Equation (3-2) to obtain the energy consumption cycle feature prediction function of the energy consumption cycle feature sequence

[0020] <q

[0021]

[0022] <000008A><000008B><000008C><000008D>In Equations (3-1) - (3-2), ω k is the angular frequency of the Fourier series of the k-th main frequency;

[0023]

[0024] Step 4: Use Steps 4.1 - 4.5 to complete the missing data F m ; Step 4.1: Based on the energy consumption cycle feature prediction function Obtain the predicted values at (n1 + 1), (n1 + 2), (n1 + 3) and (n1 + 4) moments as and respectively, so as to obtain the energy consumption cycle feature prediction sequence at (n2 + 4) sampling moments

[0025]

[0026] <00000A0> <00000A1> <00000A2> <00000A3> <00000A4> <00000A5> <00000A6> <00000A7> <00000A8> <00000A9> <00000AA> <00000AB> <00000AC> <00000AD> <00000AE> <00000AF> <00000B0> <00000B1> <00000B2> <00000B3> <00000B4> <00000B5> <00000B6> <00000B7> <00000B8> <00000B9> <00000BA> <00000BB> <00000BC> <00000BD> <00000BE> <00000BF> <00000C0> <00000C1> <00000C2> <00000C3> <00000C4> <00000C5> <00000C6> <00000C7> <00000C8> <00000C9> <00000CA> <00000CB> <00000CC> <00000CD> <00000CE> <00000CF> <00000D0> <00000D1> <00000D2> <00000D3> <00000D4> <00000D5> <00000D6> <00000D7> <00000D8> <00000D9> <00000DA> <00000DB> <00000DC> <00000DD> <00000DE> <00000DF> <00000E0> <00000E1> <00000E2P> <00000E3> <00000E4> <00000E5> <00000E6> <00000E'7> <00000E8> <00000E9> <00000EA> <00000EB> <00000EC> <00000ED> <00000EE> <00000EF> <00000F0> <00000F1> <00000F2> <00000F3> <00000F4> <00000F5> <00000F6> <00000F7> <00000F8> <00000F9> <00000FA> <00000FB> <00000FC> <00000FD> <00000FE> <00000FF>

[0027] [[ID=]165]

[0028] [[ID=)169]

[0029] [[ID=]173]

[0030]

[0031]

[0032]

[0033]

[0034] Figure 1

[0035] Figure 1

[0036] <X

[0037] [[ID=1m4]]<X

[0038] [[END

[0027] In formula (4-1), It is the energy consumption trend feature prediction sequence obtained through two-layer wavelet reconstruction; They are the j-th carbon-containing energy consumption data from the m-th sampling moment to the m+7-th sampling moment. The predicted value of

[0028] Step 4.3, keep only As the carbon-containing energy consumption data at the mth sampling moment The predicted value of , in order to achieve the missing data of the j-th carbon-containing energy consumption collected at the m-th sampling moment The completion of , thus obtaining the complete j-th carbon-containing energy consumption characteristic data set collected at the i sampling moment

[0029] The present invention provides an electronic device, comprising a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the multi-energy usage data completion method, and the processor is configured to execute the program stored in the memory.

[0030] The present invention provides a computer-readable storage medium having a computer program stored thereon. The computer-readable storage medium is characterized in that when the computer program is run by a processor, the steps of the multi-energy usage data completion method are executed.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The present invention analyzes two unique characteristics shared by various types of carbon-containing energy usage data, namely trend characteristics and periodic characteristics, and uses wavelet decomposition to decompose various types of carbon-containing energy usage data sequences into sequences with trend characteristics and sequences with periodic characteristics, namely energy consumption trend characteristic sequences and energy consumption periodic characteristic sequences, thereby laying a data foundation for completing missing data in subsequent various types of carbon-containing energy usage data sequences.

[0033] 2. The present invention analyzes the characteristics of the energy consumption trend characteristic sequence and the characteristics of the energy consumption cycle characteristic sequence, and respectively uses curve fitting to establish a fitting function model that adapts to the energy consumption trend characteristic sequence, and obtains the predicted energy consumption trend characteristic sequence through the fitting function model, and uses Fourier series to establish a fitting function model that adapts to the energy consumption cycle characteristic sequence, and obtains the predicted cycle trend characteristic sequence through the fitting function model. Finally, the missing data is obtained by reconstructing the predicted energy consumption trend characteristic sequence and the predicted cycle trend characteristic sequence, thereby realizing data completion. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The figure is a flow chart of the steps of the method of the present invention. Specific implementation mode

[0035] In this embodiment, a multi-energy consumption data completion method based on wavelet decomposition and Fourier transform is as follows Figure 1 shown, and includes the following steps:

[0036] Step 1: Collect various carbon-containing energy consumption data of key emission control enterprises according to the sampling period to obtain a carbon-containing characteristic energy consumption data set represents the carbon-containing energy consumption data of the jth type collected at the mth sampling moment; 1 ≤ m ≤ i, the carbon-containing characteristic energy consumption data at the mth sampling moment has characteristics including periodic characteristics and trend characteristics; i represents the total sampling moment, and the carbon-containing energy consumption data sets of various types have two unique characteristics, namely trend characteristics and cycle characteristics;

[0037] Assume that the carbon-containing energy consumption data of the jth type collected at the mth sampling moment is missing data, then select the carbon-containing energy consumption data of the previous n sampling moments of the missing data to form a carbon-containing energy consumption sequence and n is an even number divisible by 8, and 1 ≤ n < m; where represents the carbon-containing energy consumption data of the jth type collected at the (m - n + 1)th sampling moment; Perform three-layer wavelet decomposition on the carbon-containing energy consumption sequence of n sampling moments through equations (1-1) - (1-2)

[0038] Use wavelet decomposition to decompose the carbon-containing energy consumption sequence into a sequence with trend characteristics and a sequence with cycle characteristics, and obtain the energy consumption cycle characteristic sequence of the p-layer wavelet decomposition of the energy consumption sequence and the energy consumption trend characteristic sequence of the p-layer wavelet decomposition

[0039]

[0040]

[0041]

[0041] In equations (1-1) - (1-3), n p is the sampling moment of the p-layer wavelet decomposition; s is the abscissa of the wavelet space, s ≤ 2n p , is the sequence of the p-layer wavelet decomposition; Z is the set of all integers, s ∈ Z; ]>is the energy consumption cycle characteristic sequence obtained by the p-layer wavelet decomposition; is the energy consumption trend characteristic sequence obtained by the p-layer wavelet decomposition; G(s - 2n p ) is the high-pass filter function; H(s - 2n p) is a low-pass filter function;

[0042] Step 2: Use Equation (2-1) to obtain the energy consumption trend feature sequence through curve fitting of the fitting prediction function and use Equation (2-2) to minimize the predicted value and the energy consumption data to determine the four curve fitting coefficients ω0, ω1, ω2, ω3 of the energy consumption trend feature prediction function ;

[0043]

[0044]

[0045] In Equations (2-1) - (2-2), ε is a constant;

[0046] Step 3: Use Equation (3-1) to obtain the energy consumption period feature sequence through Fourier series fitting of the kth main frequency 1 ≤ k < N - 1, and thus use Equation (3-2) to obtain the energy consumption period feature prediction function of the energy consumption period feature sequence

[0047]

[0048]

[0049] In Equations (3-1) - (3-2), ω k is the angular frequency of the Fourier series of the kth main frequency;

[0050] Step 4: Use Steps 4.1 - 4.5 to complete the missing data F m ;

[0051] Step 4.1: Based on the energy consumption trend feature prediction function obtain the predicted value at the (n3 + 1)th moment and thus obtain the energy consumption trend feature prediction sequence at (n3 + 1) sampling moments

[0052] Based on the energy consumption period feature prediction function obtain the predicted values at (n3 + 1) moments and thus obtain the energy consumption period feature prediction sequence at (n3 + 1) sampling moments

[0053] Based on the energy consumption period feature prediction function The predicted values ​​at (n2+1) and (n2+2) moments are and Thus, the energy consumption cycle characteristic prediction sequence of (n2+2) sampling moments is obtained

[0054] Based on the prediction function of energy consumption cycle characteristics The predicted values ​​at (n1+1), (n1+2), (n1+3) and (n1+4) moments are and Thus, the energy consumption cycle characteristic prediction sequence of (n2+4) sampling moments is obtained

[0055] Through formula (4-1)-formula (4-2), the energy consumption trend feature prediction sequence is obtained through two-layer wavelet reconstruction

[0056]

[0057]

[0058] Step 4.2: Obtain the carbon-containing energy consumption prediction sequence after three-layer wavelet reconstruction through formula (4-3) in, The carbon-containing energy consumption data from the mth sampling moment to the m+7th sampling moment The predicted value of:

[0059]

[0060] Step 4.3, The predicted value of Discard, only keep As the carbon-containing energy consumption data at the mth sampling moment The predicted value of the j-th type of carbon-containing energy consumption missing data collected at the m-th sampling moment The completion of the above data finally results in the complete j-th carbon-containing characteristic energy consumption dataset collected at the i sampling moment.

[0061] In this embodiment, an electronic device includes a memory and a processor. The memory is used to store a program that supports the processor to execute the above-mentioned multi-energy usage data completion method. The processor is configured to execute the program stored in the memory.

[0062] In this embodiment, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the multi-energy usage data completion method are executed.

Claims

1. A multi-energy usage data completion method based on wavelet decomposition and Fourier transform, characterized in that: The following steps are involved: Step 1: Collect various carbon-containing energy consumption data of key emission-controlled enterprises according to the sampling period to obtain a characteristic carbon-containing energy consumption data set represents the j-th carbon-containing characteristic energy consumption data collected at the m-th sampling moment; 1≤m≤i, the carbon-containing characteristic energy consumption data at the m-th sampling moment The characteristics include periodicity and trend characteristics; i represents the total sampling time; Assume that the carbon-containing energy consumption data of the j-th category collected at the m-th sampling moment is missing data, and then select the missing data The carbon-containing energy consumption data of the previous n sampling moments constitutes a carbon-containing energy consumption sequence and n is an even number divisible by 8, 1 ≤ n < m; where represents the carbon-containing energy consumption data of the j-th category collected at the (m - n + 1)-th sampling moment; The carbon-containing energy consumption sequence at n sampling moments is calculated by formula (1-1)-formula (1-2) Perform 3-layer wavelet decomposition to obtain the energy consumption period characteristic sequence of the p-layer wavelet decomposition of the energy consumption sequence and energy consumption trend characteristic sequence of p-layer wavelet decomposition In formula (1-1) to formula (1-3), n p is the sampling time of the p-th layer wavelet decomposition; s is the horizontal coordinate of the wavelet space, s≤2n p ; is the sequence of the p-th layer wavelet decomposition; Z is the set of all integer numbers, s∈Z; is the j-th energy consumption cycle characteristic obtained by the p-th layer wavelet decomposition; is the energy consumption trend characteristic of the jth type obtained by the pth layer wavelet decomposition; G(s-2n p ) is the high-pass filter function; H(s-2n p ) is a low-pass filter function; Step 2: Use formula (2-1) to obtain the energy consumption trend characteristic sequence The fitted prediction function And use formula (2-2) to determine the energy consumption trend characteristic prediction function The four curve fitting coefficients ω0, ω1, ω2, ω3; In formula (2-1)-formula (2-2), ε is a constant; Step 3: Use formula (3-1) to obtain the energy consumption cycle characteristic sequence The kth main frequency Thus, the energy consumption cycle characteristic sequence is obtained using formula (3-2) Energy consumption cycle characteristic prediction function In formula (3-1)-formula (3-2), ω k is the angular frequency of the Fourier series of the kth principal frequency; Step 4: Use steps 4.1 to 4.5 to implement missing data F m Complete the task; Step 4.1: Prediction function based on energy consumption cycle characteristics The predicted values ​​at (n1+1), (n1+2), (n1+3) and (n1+4) moments are and Thus, the energy consumption cycle characteristic prediction sequence of (n2+4) sampling moments is obtained Step 4.2: Obtain the carbon-containing energy consumption prediction sequence after three-layer wavelet reconstruction through formula (4-1) In formula (4-1), It is the energy consumption trend feature prediction sequence obtained through two-layer wavelet reconstruction; They are the j-th carbon-containing energy consumption data from the m-th sampling moment to the m+7-th sampling moment. The predicted value of Step 4.3, keep only As the carbon-containing energy consumption data at the mth sampling moment The predicted value of , in order to achieve the missing data of the j-th carbon-containing energy consumption collected at the m-th sampling moment The completion of , thus obtaining the complete j-th carbon-containing energy consumption characteristic data set collected at the i sampling moment 2. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store a program for supporting a processor to execute the method according to claim 1 , and the processor is configured to execute the program stored in the memory.

3. 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 claim 1 are performed.

Images