Electric power quantity accumulated temperature effect duration prediction method and device and medium

Abstract

The invention relates to an electric power quantity accumulated temperature effect duration prediction method and device and a medium, and the method comprises the following steps: obtaining original data, dividing the original data into a plurality of groups of data, and carrying out the wavelet transformation of temperature-sensitive load data and temperature data in a high-temperature time period in each group, so as to obtain time-frequency distribution data; performing causal inference by using a frequency domain causal analysis method and integrating the time-frequency distribution data to obtain a causal effect of the temperature on the temperature-sensitive power and electric quantity data, and determining occurrence time of an explicit causal relationship; daily load data, daily electric quantity data and daily temperature data of multiple consecutive days are collected as to-be-analyzed data; and predicting the duration of the accumulated temperature effect according to the data to be analyzed by using a high-order Granger causality method. Compared with the prior art, the method has the advantages of high accuracy, high interpretability and the like.

Description

Technical Field

[0001] The present invention relates to the field of energy technology, and in particular to a method, device and medium for predicting the duration of accumulated temperature effect of electric power. Background Art

[0002] The existence of the accumulated temperature effect in the power system is widely recognized. Its most intuitive manifestation is that continued high temperatures will lead to further increases in load and power consumption, while when the temperature drops, the decrease in load and power consumption will have a certain hysteresis. This effect may be caused by many reasons, including the heat accumulation effect and heat storage effect under high temperature, the production and living habits of the region, and the hysteresis of various electricity consumer groups in responding to temperature changes. The accumulated temperature effect has an important impact on the power and electricity data forecasting work of the power system.

[0003] The importance of the accumulated temperature effect to the power and electricity forecast is reflected in confirming the clear causal relationship between temperature data and power and electricity data. Power and electricity data has sparse discrete characteristics and a high degree of randomness. It has both endogenous dynamic properties and is significantly affected by the external ambient temperature. If the traditional causal inference algorithm is used to explore the causal effect between the two at the coarse-grained data level, there will be great difficulties and defects. At the same time, the changing laws of temperature and power and electricity exist on three different time scales: year, week, and day. Using only the overall causal inference scheme is not suitable for causal effect scenarios with time-varying laws. For the causal analysis and mining of temperature and electricity on medium and short time scales, existing technical solutions rarely involve it.

[0004] The current prediction technology for the accumulated temperature effect can be summarized as correcting the temperature measurement index to fit the power and electricity for prediction. This type of solution is limited to improving the prediction effect, without in-depth understanding and root analysis of the occurrence and duration of the accumulated temperature effect, so as to achieve the purpose of quantitatively characterizing the accumulated temperature effect and accurately calculating the power and electricity prediction value. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method, device and medium for predicting the duration of the accumulated temperature effect of electric power.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for predicting the duration of the accumulated temperature effect of electric power comprises the following steps:

[0008] S1: Obtain the original data, divide the original data into multiple groups of data, perform wavelet transformation on the temperature-sensitive load data and temperature data in the high-temperature time period of each group, and obtain time-frequency distribution data;

[0009] S2: Use frequency domain causal analysis method to conduct causal inference based on time-frequency distribution data, obtain the causal effect of temperature on temperature-sensitive power data, and determine the time when explicit causal relationship occurs;

[0010] S3: Collect daily load data, daily power data and daily temperature data for multiple consecutive days as data to be analyzed;

[0011] S4: Using high-order Granger causality methods, the duration of the accumulated temperature effect is predicted based on the data to be analyzed, the causal effect of temperature on temperature-sensitive power data, and the time when the explicit causal relationship appears.

[0012] Further, the frequency domain causal analysis method includes the following steps:

[0013] S21: Let the time-frequency distribution data of the i-th group of load data or power data and temperature data be W i,1 (t,f) and W i,2 (t,f), where W i,1 (t,f) is the time-frequency distribution data of load or power, W i,2 (t,f) is the time-frequency distribution data of temperature;

[0014] S22: Calculate the spectrum distribution matrix S(t,f). The calculation expression of the spectrum distribution matrix is:

[0015]

[0016] S23: Calculate each component in the spectrum distribution. The calculation expression of each component is:

[0017]

[0018] S lm,i (t,f)=W i,l (t,f)W i,m (t,f) *

[0019] In the formula, S lm (t,f) is the element of the spectral distribution matrix (l,m), S lm,i (t,f) is the element of the group spectrum distribution matrix (l,m) of the i-th group of data, W i,m (t,f) * It is the companion matrix of the time-frequency distribution data of load, power or temperature.

[0020] S23: Decompose the spectrum matrix. The calculation expression of the decomposition process is:

[0021] S(t,f)=ψ(t,f)ψ * (t,f)

[0022] f=e iθ

[0023]

[0024] Where S(t,f) is the spectral distribution matrix, ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix, A k (t) is ψ(t,e iθ ) is the kth coefficient of the Fourier series expansion of the matrix.

[0025] S24: Further decomposition, we get:

[0026] S(t,f)=ψA0 -1 A0A0 T A0 -T ψ *

[0027] Where S(t,f) is the spectral distribution matrix, ψ=ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix. A0=A0(t) is ψ(t,e iθ ) is the 0th coefficient of the Fourier series expansion of the matrix.

[0028] S25: Define H(t,f)=ψ(t,f)A0 -1 (t), we can get the high-order Granger causality of temperature data on load or power in the time-frequency domain. The calculation expression of high-order Granger causality in the time-frequency domain is:

[0029]

[0030] In the formula, S lm (t,f) is the element of the spectral distribution matrix (l,m), Σ 22 (t) is the variance of the second series regression residual at time t.

[0031] Further, step S4 includes the following steps:

[0032] S41: Selecting the temperature-sensitive component corresponding to the power data of a certain day from the data to be analyzed, and extracting the temperature data of the day and the temperature data of the previous days;

[0033] S42: Establishing a learning regression of a first nonlinear function from the temperature data of the day to the temperature-sensitive component of the electric power data of the day to obtain a first error;

[0034] S43: Establishing a learning regression of a second nonlinear function from the temperature data of the current day and the temperature data of each of the previous several days to the power data of the current day to obtain a second error;

[0035] S44: Calculate the high-order Granger causal strength of the temperature data of each day in the previous several days on the power consumption forecast;

[0036] S45: According to the changing trend of the higher-order Granger causal strength, different causal strength intervals are delineated. By setting thresholds, the duration of the accumulated temperature effect is judged. Based on the changes in the higher-order Granger causal strength, the optimal number of days for the accumulated temperature effect is selected.

[0037] Furthermore, the calculation expression of the first nonlinear function is:

[0038] C1=f θ (T)+∈0,θ1=argmin(∈0(θ1))

[0039] Where C1 is the first nonlinear function, ∈0 is the first error, and θ1 is the optimization parameter of the first nonlinear function.

[0040] Furthermore, the calculation expression of the second nonlinear function is:

[0041] C2=f θ (T0,T1,…,T p )+∈ p ,θ2=argmin(∈ p (θ2)

[0042] Where C2 is the second nonlinear function, ∈ p is the second error with respect to the p-order time lag, and θ2 is the optimization parameter of the second nonlinear function.

[0043] Furthermore, the calculation expression of the high-order Granger causal strength is:

[0044]

[0045] In the formula, I p is the higher-order Granger causal strength with respect to p-order time lag, ∈ p is the second error with respect to the p-order time lag.

[0046] Furthermore, the wavelet function used in the wavelet transform is a non-orthogonal, non-compactly supported complex wavelet function with a frequency control parameter.

[0047] Furthermore, the original data are grouped by year as the time unit, and the high temperature period is the continuous period from June to September when the average temperature is greater than 22 degrees Celsius.

[0048] According to a second aspect of the present invention, a device for predicting the duration of an accumulated temperature effect of electric power includes a memory, a processor, and a program stored in the memory. When the processor executes the program, it implements any one of the above methods for predicting the duration of an accumulated temperature effect of electric power.

[0049] A third aspect of the present invention is a storage medium having a program stored thereon, which, when executed, implements any of the above methods for predicting the duration of the accumulated temperature effect of electric power.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1) The present invention obtains the time-frequency characteristics of non-stationary time series through wavelet transform and analyzes them, which can fully explore the local explicit causal laws in the time-frequency domain, and then calculate and determine the key time period and temperature range where the causal relationship mainly appears, so as to find the time when the accumulated temperature effect occurs and the corresponding temperature.

[0052] 2) The present invention adopts the mathematical methods of wavelet transform, spectral matrix decomposition and high-order Granger causality calculation to convert time series signals into time-frequency domain signals, which has strong interpretability. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0054] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0055] Example 1

[0056] The present invention is a method, device and medium for predicting the duration of the accumulated temperature effect of electric power. Figure 1 As shown, the method comprises the following steps:

[0057] S1: Obtain the original data, divide the original data into multiple groups of data, perform wavelet transformation on the temperature-sensitive load data and temperature data in the high-temperature time period of each group, and obtain time-frequency distribution data;

[0058] In this embodiment, the original data is grouped based on years as the time unit, and the high temperature time period is a continuous time period from June to September when the average temperature is greater than 22 degrees Celsius.

[0059] Preferably, the wavelet transform uses a non-orthogonal, non-compactly supported complex wavelet function with a frequency control parameter.

[0060] The original N-day data is divided into multiple groups of data by year, and the temperature-sensitive load data and daily temperature data in the high temperature period of each year are transformed using Morlet wavelet to obtain time-frequency distribution data;

[0061] S2: Use frequency domain causal analysis method to conduct causal inference based on time-frequency distribution data, obtain the causal effect of temperature on temperature-sensitive power data, and determine the time when explicit causal relationship occurs;

[0062] S21: Let the time-frequency distribution data of the i-th group of load data or power data and temperature data be W i,1 (t,f) and W i,2 (t,f), where W i,1 (t,f) is the time-frequency distribution data of load or power, W i,2 (t,f) is the time-frequency distribution data of temperature;

[0063] S22: Calculate the spectrum distribution matrix S(t,f), the calculation expression of the spectrum distribution matrix is:

[0064]

[0065] S23: Calculate each component in the spectrum distribution, and the calculation expression of each component is:

[0066]

[0067] S km,i (t,f)=W i,l (t,f)W i,m (t,f) *

[0068] In the formula, S km (t,f) is the element of the spectral distribution matrix (l,m), S lm,i (t,f) is the element of the group spectrum distribution matrix (l,m) of the i-th group of data, W i,m (t,f) * It is the companion matrix of the time-frequency distribution data of load, power or temperature.

[0069] S23: Decompose the spectrum matrix. The calculation expression of the decomposition process is:

[0070] S(t,f)=ψ(t,f)ψ * (t,f)

[0071] f=e iθ

[0072]

[0073] Where S(t,f) is the spectral distribution matrix, ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix, A k (t) is ψ(t,e iθ ) is the kth coefficient of the Fourier series expansion of the matrix.

[0074] S24: Further decomposition, we get:

[0075] S(t,f)=ψA0 -1 A0A0 T A0 -T ψ *

[0076] Where S(t,f) is the spectral distribution matrix, ψ=ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix, A0=A0(t) is ψ(t,e iθ ) is the 0th coefficient of the Fourier series expansion of the matrix.

[0077] S25: Define H(t,f)=ψ(t,f)A0 -1 (t), and obtain the time-frequency domain high-order Granger causality of temperature data on load or power. The calculation expression of the time-frequency domain high-order Granger causality is:

[0078]

[0079] In the formula, S lm (t,f) is the element of the spectral distribution matrix (l,m), Σ 22 (t) is the variance of the second series regression residual at time t.

[0080] S3: Collect daily load data, daily power data and daily temperature data for multiple consecutive days as data to be analyzed;

[0081] S4: Using high-order Granger causality methods, the duration of the accumulated temperature effect is predicted based on the data to be analyzed, the causal effect of temperature on temperature-sensitive power data, and the time when the explicit causal relationship appears.

[0082] S41: Selecting the temperature-sensitive component corresponding to the power data of a certain day from the data to be analyzed, and extracting the temperature data of the day and the temperature data of the previous days;

[0083] S42: Establishing a learning regression of a first nonlinear function from the temperature data of the day to the temperature-sensitive component of the electric power data of the day to obtain a first error;

[0084] The calculation expression of the first nonlinear function is:

[0085] C1=f θ (T)+∈0,θ1=argmin(∈0(θ1))

[0086] Where C1 is the first nonlinear function, ∈0 is the first error, and θ1 is the optimization parameter of the first nonlinear function.

[0087] S43: Establishing a learning regression of a second nonlinear function from the temperature data of the current day and the temperature data of each of the previous several days to the power data of the current day to obtain a second error;

[0088] The calculation expression of the second nonlinear function is:

[0089] C2=f θ (T0,T1,…,T p )+∈ p ,θ2=argmin(∈ p (θ2)

[0090] Where C2 is the second nonlinear function, ∈ p is the second error with respect to the p-order time lag, and θ2 is the optimization parameter of the second nonlinear function.

[0091] S44: Calculate the high-order Granger causal strength of the temperature data of each day in the previous several days on the power forecast; the calculation expression of the high-order Granger causal strength is:

[0092]

[0093] In the formula, I p is the higher-order Granger causal strength with respect to p-order time lag, ∈ p is the second error with respect to the p-order time lag.

[0094] S45: According to the changing trend of the higher-order Granger causal strength, different causal strength intervals are delineated. By setting thresholds, the duration of the accumulated temperature effect is judged. Based on the changes in the higher-order Granger causal strength, the optimal number of days for the accumulated temperature effect is selected.

[0095] In the above embodiment, causal strength > 0.2 is marked as significant, 0.1-0.2 is marked as obvious, 0.05-0.2 is marked as slight, and less than 0.05 is marked as insignificant. If the p-order lag causal strength is not obvious, it means that the p-order lag has no obvious accumulated temperature effect, and the best number of days is p-1 day / order.

[0096] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0097] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0098] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for predicting the duration of the accumulated temperature effect of electric power, characterized in that: The following steps are involved: S1: Obtain the original data, divide the original data into multiple groups of data, perform wavelet transformation on the temperature-sensitive load data and temperature data in the high-temperature time period of each group, and obtain time-frequency distribution data; S2: Use frequency domain causal analysis method to conduct causal inference based on time-frequency distribution data, obtain the causal effect of temperature on temperature-sensitive power data, and determine the time when explicit causal relationship occurs; S3: Collect daily load data, daily power data and daily temperature data for multiple consecutive days as data to be analyzed; S4: Using high-order Granger causality methods, the duration of the accumulated temperature effect is predicted based on the data to be analyzed, the causal effect of temperature on temperature-sensitive power data, and the time when the explicit causal relationship appears.

2. A method for predicting the duration of the accumulated temperature effect of electric power according to claim 1, characterized in that: The frequency domain causal analysis method comprises the following steps: S21: Let the time-frequency distribution data of the i-th group of load data or power data and temperature data be W i,1 (t,f) and W i,2 (t,f), where W i,1 (t,f) is the time-frequency distribution data of load or power, W i,2 (t,f) is the time-frequency distribution data of temperature; S22: Calculate the spectrum distribution matrix S(t,f), the calculation expression of the spectrum distribution matrix is: S23: Calculate each component in the spectrum distribution, and the calculation expression of each component is: S lm,i (t,f)=W i,l (t,f)W i,m (t,f) * In the formula, S lm (t,f) is the element of the spectral distribution matrix (l,m), S lm,i (t,f) is the element of the group spectrum distribution matrix (l,m) of the i-th group of data, W i,m (t,f) * It is the companion matrix of the time-frequency distribution data of load, power or temperature. S23: Decompose the spectrum matrix. The calculation expression of the decomposition process is: S(t,f)=ψ(t,f)ψ * (t,f) f=e iθ Where S(t,f) is the spectral distribution matrix, ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix, A k (t) is ψ(t,e iθ ) is the kth coefficient of the Fourier series expansion of the matrix. S24: Further decomposition, we get: S(t,f)=ψA0 -1 A0A0 T A0 -T ψ * Where S(t,f) is the spectral distribution matrix, ψ=ψ(t,e iθ ) is the symmetric decomposition matrix of the spectral distribution matrix, ψ * is the corresponding adjoint matrix. A0=A0(t) is ψ(t,e iθ ) is the 0th coefficient of the Fourier series expansion of the matrix. S25: Define H(t,f)=ψ(t,f)A0 -1 (t), and obtain the time-frequency domain high-order Granger causality of temperature data on load or power. The calculation expression of the time-frequency domain high-order Granger causality is: In the formula, A lm (t,f) is the element of the spectral distribution matrix (l,m), Σ 22 (t) is the variance of the second series regression residual at time t.

3. The method for predicting the duration of the accumulated temperature effect of electric power according to claim 1 is characterized in that: Step S4 includes the following steps: S41: Selecting the temperature-sensitive component corresponding to the power data of a certain day from the data to be analyzed, and extracting the temperature data of the day and the temperature data of the previous days; S42: Establishing a learning regression of a first nonlinear function from the temperature data of the day to the temperature-sensitive component of the electric power data of the day to obtain a first error; S43: Establishing a learning regression of a second nonlinear function from the temperature data of the current day and the temperature data of each of the previous several days to the power data of the current day to obtain a second error; S44: Calculate the high-order Granger causal strength of the temperature data of each day in the previous several days on the power consumption forecast; S45: According to the changing trend of the higher-order Granger causal strength, different causal strength intervals are delineated. By setting thresholds, the duration of the accumulated temperature effect is judged. Based on the changes in the higher-order Granger causal strength, the optimal number of days for the accumulated temperature effect is selected.

4. The method for predicting the duration of the accumulated temperature effect of electric power according to claim 1 is characterized in that: The calculation expression of the first nonlinear function is: C1=f θ (T)+∈0,θ1=argmin(∈0(θ1)) Where C1 is the first nonlinear function, ∈0 is the first error, and θ1 is the optimization parameter of the first nonlinear function.

5. The method for predicting the duration of the accumulated temperature effect of electric power according to claim 1 is characterized in that: The calculation expression of the second nonlinear function is: C2=f θ (T0,T1,…,T p )+∈ p ,θ2=argmin(∈ p (θ2)) Where C2 is the second nonlinear function, ∈ p is the second error with respect to the p-order time lag, and θ2 is the optimization parameter of the second nonlinear function.

6. A method for predicting the duration of the accumulated temperature effect of electric power according to claim 1, characterized in that: The calculation expression of the high-order Granger causal strength is: In the formula, I p is the higher-order Granger causal strength with respect to p-order time lag, ∈ p is the second error with respect to the p-order time lag.

7. The method for predicting the duration of the accumulated temperature effect of electric power according to claim 1, characterized in that: The wavelet function used in the wavelet transform is a complex wavelet function that is non-orthogonal, non-compactly supported and has a frequency control parameter.

8. The method for predicting the duration of the accumulated temperature effect of electric power according to claim 1, characterized in that: The original data is grouped with years as the time unit, and the high temperature time period is a continuous time period from June to September when the average temperature is greater than 22 degrees Celsius.

9. A device for predicting the duration of the accumulated temperature effect of electric power, comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, a method for predicting the duration of the accumulated temperature effect of electric power as described in any one of claims 1-8 is implemented.

10. A storage medium having a program stored thereon, characterized in that: When the program is executed, a method for predicting the duration of the accumulated temperature effect of electric power as described in any one of claims 1-8 is implemented.

Images