CN112488396A - Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model - Google Patents
Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model Download PDFInfo
- Publication number
- CN112488396A CN112488396A CN202011387066.2A CN202011387066A CN112488396A CN 112488396 A CN112488396 A CN 112488396A CN 202011387066 A CN202011387066 A CN 202011387066A CN 112488396 A CN112488396 A CN 112488396A
- Authority
- CN
- China
- Prior art keywords
- model
- holt
- winters
- formula
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 9
- 230000015654 memory Effects 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000005265 energy consumption Methods 0.000 claims abstract description 5
- 238000013135 deep learning Methods 0.000 claims abstract description 4
- 230000007787 long-term memory Effects 0.000 claims abstract description 4
- 230000006403 short-term memory Effects 0.000 claims abstract description 4
- 230000001932 seasonal effect Effects 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 7
- 238000013016 damping Methods 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 150000001875 compounds Chemical class 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 230000005611 electricity Effects 0.000 description 9
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/148—Wavelet transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Tourism & Hospitality (AREA)
- Databases & Information Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Operations Research (AREA)
- Algebra (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Development Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Water Supply & Treatment (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
Abstract
The invention relates to a wavelet transform-based power load prediction method of a Holt-Winters and LSTM combined model, which comprises the following steps of: step S1, acquiring the actual energy consumption data of the power consumer load and preprocessing the data; step S2, removing potential noise in the load data by wavelet denoising with a preset threshold according to the preprocessed load data, and performing discrete wavelet decomposition; step S3, constructing and training a Holt-Winters model according to the decomposed wavelet coefficients of each layer as training samples; step S4, according to the preprocessed load data, a deep learning framework is adopted to build a deep long-term and short-term memory network model; step S5: combining the Holt-Winters model and the depth long-time memory network model, and calculating the weight of each model in the combined model; and step S6, carrying out weighted average on outputs of the Holt-Winters model and the depth long-time memory network model according to the obtained weights to obtain a final prediction curve. The invention effectively improves the reliability of power load prediction.
Description
Technical Field
The invention relates to the field of power load prediction, in particular to a power load prediction method based on a Holt-Winters and LSTM combined model of wavelet transformation.
Background
The load prediction plays a very important role in the economic and reliable operation of a power grid system and also plays an important role in the electric power trading market. For power generation enterprises, accurate load prediction is beneficial to reasonably arranging the operation mode of a power grid and formulating reasonable power supply construction planning, and for power selling companies, the accurate load prediction is beneficial to realizing maximized profit in electric quantity transaction of a spot market, and meanwhile, value-added services such as energy conservation and the like can be made for enterprise users. At the present stage, the number of electricity selling companies in China increases year by year, at the moment, the strength of the load forecasting capability determines the competitive power of the electricity selling companies in the market, and because the electricity selling companies cannot give reasonable and effective quotations in the spot market without accurate load forecasting results, high deviation cost is easily caused, so the load forecasting capability is very important for the current electricity selling companies.
Disclosure of Invention
In view of this, the present invention provides a power load prediction method based on a wavelet transform Holt-Winters and LSTM combined model, which combines Holt-Winters and LSTM to effectively improve reliability of power load prediction.
In order to achieve the purpose, the invention adopts the following technical scheme:
a power load prediction method based on a Holt-Winters and LSTM combined model of wavelet transformation comprises the following steps:
step S1, acquiring the actual energy consumption data of the power consumer load and preprocessing the data;
step S2, removing potential noise in the load data by wavelet denoising with a preset threshold according to the preprocessed load data, and performing discrete wavelet decomposition;
step S3, constructing and training a Holt-Winters model according to the decomposed wavelet coefficients of each layer as training samples;
step S4, according to the preprocessed load data, a deep learning framework is adopted to build a deep long-term and short-term memory network model;
step S5: combining the Holt-Winters model and the depth long-time memory network model, and calculating the weight of each model in the combined model;
and step S6, carrying out weighted average on outputs of the Holt-Winters model and the depth long-time memory network model according to the obtained weights to obtain a final prediction curve.
Further, the preprocessing comprises elimination processing and normalization processing of negative values, null values, zero values and abnormal values.
Further, the normalization processing specifically includes:
in the formula, PjIs the load, P, at the j-th point of a certain user on a certain dayj.minAnd Pj.maxThe load is the minimum value and the maximum value of the load of the user at the jth point of the day respectively; pj' is a value normalized by the load at the j point of a certain day of a certain user.
Further, the step S3 is specifically:
and step S31, decomposing the discrete signal to obtain a corresponding wavelet coefficient, wherein the decomposition formula is as follows:
constructing a new discrete signal according to the wavelet coefficient, wherein a reconstruction formula is shown as follows:
the formula can be abbreviated as:
in the formula (I), the compound is shown in the specification,is the wavelet coefficient of the j-th layer, k is the position index,for wavelet functions, x (t) is the original signal, Aj(t) For low frequency components of the original signal at layer j, and Dj(t) is a high frequency component of the j-th layer;
step S32, the Holt-Winters model is composed of three exponential smoothing equations and a prediction equation, and the specific formula is as follows:
si=α(xi-pi-l)+(1+α)(si-1+ti-1)
ti=β(si-si-1)+(1-β)ti-1
pi=γ(xi-si)+(1-γ)pi-l
xi+h=si+hti+pi-k+h
in the formula, si、ti、piRespectively representing the horizontal characteristic, the trend characteristic and the seasonal characteristic of the original data at the ith time point, respectively corresponding to damping factors of three components, wherein the damping factors are respectively alpha, beta and gamma and range from 0 to 1, and xiThe actual value of the data at the ith time point, h is the step size from the observation point to the prediction point, and l is the period of the seasonal component;
further, the Holt-Winters model optimizes the parameters of the model by using a PSO algorithm in the training process, specifically, the PSO algorithm is used for searching the optimal value of the seasonal parameter l in the Holt-Winters model, the fitness function of the algorithm is the average absolute error between the predicted value of the Holt-Winters model and the true value of the test sample, and the calculation formula is as follows:
in the formula, yiRepresents the actual value of the sample sequence, andit is the predicted value of the sequence.
Further, the step S5 is specifically:
step S51, determining two prediction modes according to the fitting effect of the two modelsWeight of type in combinatorial model ωiThen the formula of its combined prediction model can be recorded as:
in the formula: f. oftCombining the predicted values of the prediction models for the time t; f. ofitThe predicted value of the ith prediction model at the moment t;
and step S52, according to the optimization theory, defining the loss function as the minimum sum of squares of the prediction error, wherein the formula is as follows:
and satisfies the condition:
in the formula, etCombining predicted errors for time t; y istIs an observed value; e is the sum of squares of errors;
step S53, respectively setting the weights of the Holt-Winters model and the LSTM model as omega1、ω2Predicted value is Y1、Y2The final predicted value of the two models is YcThen the formula can be written as:
Yc=ω1Y1+ω2Y2
obtaining the variance and covariance of the errors of the two models through the prediction errors of the two models, and setting the variance and covariance as sigma11,σ12,σ22And finally calculating the weight coefficients of the two prediction models as follows:
compared with the prior art, the invention has the following beneficial effects:
the method combines Holt-Winters and LSTM, and effectively improves the reliability of power load prediction.
Drawings
FIG. 1 is a functional block diagram of the method of the present invention;
FIG. 2 is a diagram of a seasonal Holt-Winters principle based on wavelet transforms in an embodiment of the present invention;
FIG. 3 is a diagram of the structure of an LSTM cell unit in accordance with an embodiment of the present invention.
Detailed Description
The invention is further explained below with reference to the drawings and the embodiments.
Referring to fig. 1, the present invention provides a power load prediction method based on a wavelet transform Holt-Winters and LSTM combined model, comprising the following steps:
step S1, acquiring the actual energy consumption data of the power consumer load and preprocessing the data;
in this embodiment, the actual energy consumption data of the power consumer load with a period of one year and a time resolution of 15min is acquired, and since the load data in the database is the result of multiple transmissions of the metering meter, unexpected factors may occur in each transmission process to affect the accuracy of the final data, and in order to reduce the influence of these data on the model predictive performance, the preprocessing operation needs to be performed on the original data before the experiment is performed. Including but not limited to negative, null, zero, outlier processing.
In order to eliminate the difference of dimension and magnitude between the variables, the preprocessed load data is normalized. The normalization method adopts a maximum and minimum normalization method, normalizes all the user electricity load data to a [0,1] interval, and only reserves the electricity utilization habits and characteristics of a user electricity load curve so as to find out the similarity between the user electricity utilization modes. The normalization companies are as follows:
in the formula, PjIs the load, P, at the j-th point of a certain user on a certain dayj.minAnd Pj.maxRespectively, the minimum value and the maximum value of the load of the user at 96 points of the day. Pj' is a value normalized by the load at the j point of a certain day of a certain user.
Step S2, removing potential noise in the load data by wavelet denoising with a preset threshold according to the preprocessed load data, and performing discrete wavelet decomposition;
in this embodiment, the normalized load data is drawn as a daily electricity load curve, and because the daily load curve obtained from the raw data is relatively tortuous and has more burrs, in order to enable the model to better fit the input load sequence during training to find the optimal parameters, the wavelet denoising with a proper threshold value is adopted to eliminate the potential noise in the load data, and the discrete wavelet decomposition is performed on the load data subjected to the wavelet denoising. As shown in fig. 2, Daubechies8 wavelet (db 8 for short) is taken as a mother wavelet, the number of layers of decomposition is selected to be 3, and the curve is smoothed to some extent, so that the characteristics of the curve are retained while noise data are filtered.
Step S3, constructing and training a Holt-Winters model according to the decomposed wavelet coefficients of each layer as training samples; the method specifically comprises the following steps:
and step S31, decomposing the discrete signal to obtain a corresponding wavelet coefficient, wherein the decomposition formula is as follows:
constructing a new discrete signal according to the wavelet coefficient, wherein a reconstruction formula is shown as follows:
the formula can be abbreviated as:
in the formula (I), the compound is shown in the specification,is the wavelet coefficient of the j-th layer, k is the position index,for wavelet functions, x (t) is the original signal, Aj(t) is the low frequency component of the original signal at layer j, and Dj(t) is a high frequency component of the j-th layer;
and step S32, the power load is obviously influenced by meteorological factors such as ambient temperature, humidity, sunshine, rainfall and the like in different seasons, so that the load data has a time sequence with obvious seasonal characteristics. The Holt-Winters model is a method which is based on an exponential smoothing method and specially predicts a time series with seasonal characteristics. The core idea of the exponential smoothing method is to predict the data size of an observation point by weighted summation of historical data, and the weight size of each historical data point is related to the distance of the point from the predicted point, and the longer the distance is, the smaller the weight is. The Holt-Winters model adds a seasonal component to an original smoothing formula, namely the Holt-Winters model not only performs weighted summation on historical data, but also performs weighted summation on trends and seasonal characteristics of the data. The Holt-Winters model consists of three exponential smoothing equations and a prediction equation, and the specific formula is as follows:
si=α(xi-pi-l)+(1+α)(si-1+ti-1)
ti=β(si-si-1)+(1-β)ti-1
pi=γ(xi-si)+(1-γ)pi-l
xi+h=si+hti+pi-k+h
in the formula, si、ti、piRespectively representing the horizontal characteristic, the trend characteristic and the seasonal characteristic of the original data at the ith time point, respectively corresponding to damping factors of three components, wherein the damping factors are respectively alpha, beta and gamma and range from 0 to 1, and xiIs the actual value of the data at the ith time point, h is the step size from the observation point to the prediction point, and l is the period of the seasonal component.
Preferably, the Holt-Winters model optimizes the parameters of the model by using a PSO algorithm in the training process, specifically, the PSO algorithm is used for searching the optimal value of the seasonal parameter l in the Holt-Winters model, the fitness function of the algorithm is the average absolute error between the predicted value of the Holt-Winters model and the true value of the test sample, and the calculation formula is as follows:
in the formula, yiRepresents the actual value of the sample sequence, andit is the predicted value of the sequence.
Step S4, according to the preprocessed load data, a deep learning framework is adopted to build a deep long-term and short-term memory network model;
in this example, the cell unit structure of LSTM is as shown in fig. 3, and the cell unit of LSTM has three gates, i.e., Input Gate, Forget Gate, and Output Gate. The input gate is used for controlling information input, the forgetting gate is used for controlling the retention of cell historical state information, and the output gate is used for controlling information output. Activating a function to enable the output value of the forgetting gate to be between [0 and 1], and when the output of the forgetting gate is 0, the function shows that all information in the last state is discarded; when 1, all the information indicating the previous state is retained. The calculation process is as follows:
ft=σ(Wf·[ht-1,xt]+bf)
it=σ(Wi·[ht-1,xt]+bi)
ot=σ(Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
in the formula, xtAs an input vector of the model, htFor each LSTM cell output vector, itTo the output of the input gate, ftFor forgetting the output of the gate, CtCell state at the present time t, otIs the output of the output gate, where W and b are parameter matrices.
In the model building process, the output mode of the model is designed to enable the load prediction model to be direct multi-step prediction, namely, the power user load curves of a plurality of time periods in the future can be directly predicted.
Step S5: combining the Holt-Winters model and the depth long-time memory network model, and calculating the weight of each model in the combined model; the method specifically comprises the following steps:
step S51, determining the weight omega of the two prediction models in the combined model according to the fitting effect of the two modelsiThen the formula of its combined prediction model can be recorded as:
in the formula: f. oftCombining the predicted values of the prediction models for the time t; f. ofitThe predicted value of the ith prediction model at the moment t;
and step S52, according to the optimization theory, defining the loss function as the minimum sum of squares of the prediction error, wherein the formula is as follows:
and satisfies the condition:
in the formula, etCombining predicted errors for time t; y istIs an observed value; e is the sum of squares of errors;
step S53, respectively setting the weights of the Holt-Winters model and the LSTM model as omega1、ω2Predicted value is Y1、Y2The final predicted value of the two models is YcThen the formula can be written as:
Yc=ω1Y1+ω2Y2
obtaining the variance and covariance of the errors of the two models through the prediction errors of the two models, and setting the variance and covariance as sigma11,σ12,σ22And finally calculating the weight coefficients of the two prediction models as follows:
and step S6, carrying out weighted average on outputs of the Holt-Winters model and the depth long-time memory network model according to the obtained weights to obtain a final prediction curve.
In this embodiment, the raw data is divided into training data and test data to determine the accuracy of the algorithm's predictions. 80% were randomly selected as training samples and 20% as test samples. And (5) carrying out performance test by using the trained model to obtain the performance evaluation index of the trained model. The performance indexes of the model are phase-to-fraction error MAPE and root mean square error RMSE values, the smaller the index value is, the higher the prediction precision is, and the calculation formula of the amphoteric performance index is as follows:
the above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.
Claims (6)
1. A power load prediction method based on a Holt-Winters and LSTM combined model of wavelet transformation is characterized by comprising the following steps:
step S1, acquiring the actual energy consumption data of the power consumer load and preprocessing the data;
step S2, removing potential noise in the load data by wavelet denoising with a preset threshold according to the preprocessed load data, and performing discrete wavelet decomposition;
step S3, constructing and training a Holt-Winters model according to the decomposed wavelet coefficients of each layer as training samples;
step S4, according to the preprocessed load data, a deep learning framework is adopted to build a deep long-term and short-term memory network model;
step S5: combining the Holt-Winters model and the depth long-time memory network model, and calculating the weight of each model in the combined model;
and step S6, carrying out weighted average on outputs of the Holt-Winters model and the depth long-time memory network model according to the obtained weights to obtain a final prediction curve.
2. The wavelet transform-based electric power load prediction method of the Holt-Winters and LSTM combined model, according to claim 1, characterized in that the preprocessing includes elimination processing and normalization processing for negative values, null values, zero values, outliers.
3. The wavelet transform-based electric power load prediction method of the Holt-Winters and LSTM combined model according to claim 2, wherein the normalization process specifically comprises:
in the formula, PjIs the load, P, at the j-th point of a certain user on a certain dayj.minAnd Pj.maxThe load is the minimum value and the maximum value of the load of the user at the jth point of the day respectively; pj' is a value normalized by the load at the j point of a certain day of a certain user.
4. The method for predicting power load based on the wavelet transform Holt-Winters and LSTM combined model as claimed in claim 1, wherein said step S3 is specifically:
and step S31, decomposing the discrete signal to obtain a corresponding wavelet coefficient, wherein the decomposition formula is as follows:
constructing a new discrete signal according to the wavelet coefficient, wherein a reconstruction formula is shown as follows:
the formula can be abbreviated as:
in the formula (I), the compound is shown in the specification,is the wavelet coefficient of the j-th layer, k is the position index,for wavelet functions, x (t) is the original signal, Aj(t) is the low frequency component of the original signal at layer j, and Dj(t) is a high frequency component of the j-th layer;
step S32, the Holt-Winters model is composed of three exponential smoothing equations and a prediction equation, and the specific formula is as follows:
si=α(xi-pi-l)+(1+α)(si-1+ti-1)
ti=β(si-si-1)+(1-β)ti-1
pi=γ(xi-si)+(1-γ)pi-l
xi+h=si+hti+pi-k+h
in the formula, si、ti、piRespectively representing the horizontal characteristic, the trend characteristic and the seasonal characteristic of the original data at the ith time point, respectively corresponding to damping factors of three components, wherein the damping factors are respectively alpha, beta and gamma and range from 0 to 1, and xiIs the actual value of the data at the ith time point, h is the step size from the observation point to the prediction point, and l is the period of the seasonal component.
5. The wavelet transform-based electric power load prediction method for the Holt-Winters and LSTM combined model is characterized in that the Holt-Winters model uses a PSO algorithm to optimize parameters of the model in a training process, specifically, the PSO algorithm is used for finding an optimal value of a seasonal parameter l in the Holt-Winters model, a fitness function of the algorithm is an average absolute error between a predicted value of the Holt-Winters model and a true value of a test sample, and a calculation formula is as follows:
6. The method for predicting power load based on the wavelet transform Holt-Winters and LSTM combined model as claimed in claim 1, wherein said step S5 is specifically:
step S51, determining the weight omega of the two prediction models in the combined model according to the fitting effect of the two modelsiThen the formula of its combined prediction model can be recorded as:
in the formula: f. oftCombining the predicted values of the prediction models for the time t; f. ofitThe predicted value of the ith prediction model at the moment t;
and step S52, according to the optimization theory, defining the loss function as the minimum sum of squares of the prediction error, wherein the formula is as follows:
and satisfies the condition:
in the formula, etCombining predicted errors for time t; y istIs an observed value; e is the sum of squares of errors;
step S53, respectively setting the weights of the Holt-Winters model and the LSTM model as omega1、ω2Predicted value is Y1、Y2The final predicted value of the two models is YcThen the formula can be written as:
Yc=ω1Y1+ω2Y2
obtaining the variance and covariance of the errors of the two models through the prediction errors of the two models, and setting the variance and covariance as sigma11,σ12,σ22And finally calculating the weight coefficients of the two prediction models as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011387066.2A CN112488396A (en) | 2020-12-01 | 2020-12-01 | Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011387066.2A CN112488396A (en) | 2020-12-01 | 2020-12-01 | Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112488396A true CN112488396A (en) | 2021-03-12 |
Family
ID=74938718
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011387066.2A Pending CN112488396A (en) | 2020-12-01 | 2020-12-01 | Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112488396A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111508598A (en) * | 2020-05-06 | 2020-08-07 | 万达信息股份有限公司 | Method for predicting outpatient quantity of respiratory system diseases |
CN113131476A (en) * | 2021-04-28 | 2021-07-16 | 南方电网科学研究院有限责任公司 | Power load prediction method |
CN113283155A (en) * | 2021-03-22 | 2021-08-20 | 中国农业科学院农业资源与农业区划研究所 | Near-surface air temperature estimation method, system, storage medium and equipment |
CN114064203A (en) * | 2021-10-28 | 2022-02-18 | 西安理工大学 | Cloud virtual machine load prediction method based on multi-scale analysis and deep network model |
CN114113773A (en) * | 2021-11-11 | 2022-03-01 | 国网辽宁省电力有限公司经济技术研究院 | Non-invasive load monitoring method based on zero sample learning |
CN115081702A (en) * | 2022-06-14 | 2022-09-20 | 国网信息通信产业集团有限公司 | Power load prediction method with interpretable characteristic, system and terminal |
CN116341189A (en) * | 2023-01-12 | 2023-06-27 | 山西大学 | Prophet and Holt-windows combined model thermal comfort prediction method based on wavelet transformation |
CN117390895A (en) * | 2023-12-08 | 2024-01-12 | 华科五洲(天津)海洋工程有限公司 | Semi-submersible ship ballast system simulation method, device and storage medium |
CN117977579A (en) * | 2024-03-28 | 2024-05-03 | 西华大学 | DWT-LSTM neural network-based adjustable load prediction method |
CN117996816A (en) * | 2024-03-29 | 2024-05-07 | 江苏谷峰电力科技股份有限公司 | Intelligent control method and system for wind, light and firewood storage and team-level energy storage |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110160927A1 (en) * | 2009-12-30 | 2011-06-30 | Wilson Kevin W | Method for Prediction for Nonlinear Seasonal Time Series |
CN109214586A (en) * | 2018-09-27 | 2019-01-15 | 国网河南省电力公司濮阳供电公司 | Area power grid electricity sales amount prediction technique based on Holt-Winters model |
CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
US20190265768A1 (en) * | 2018-02-24 | 2019-08-29 | Hefei University Of Technology | Method, system and storage medium for predicting power load probability density based on deep learning |
-
2020
- 2020-12-01 CN CN202011387066.2A patent/CN112488396A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110160927A1 (en) * | 2009-12-30 | 2011-06-30 | Wilson Kevin W | Method for Prediction for Nonlinear Seasonal Time Series |
US20190265768A1 (en) * | 2018-02-24 | 2019-08-29 | Hefei University Of Technology | Method, system and storage medium for predicting power load probability density based on deep learning |
CN109214586A (en) * | 2018-09-27 | 2019-01-15 | 国网河南省电力公司濮阳供电公司 | Area power grid electricity sales amount prediction technique based on Holt-Winters model |
CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
Non-Patent Citations (3)
Title |
---|
刘月灿: ""基于大数据负荷预测的需求响应博弈策略的研究与实现"", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
周先国: ""提升小波变换的应用研究及FPGA实现"", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
程诗尧 等: ""基于Holt-Winters 和LSTM 的组合模型在电能表需求预测中的应用"", 《中国设备工程》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111508598A (en) * | 2020-05-06 | 2020-08-07 | 万达信息股份有限公司 | Method for predicting outpatient quantity of respiratory system diseases |
CN111508598B (en) * | 2020-05-06 | 2023-12-08 | 万达信息股份有限公司 | Respiratory disease outpatient service quantity prediction method |
CN113283155A (en) * | 2021-03-22 | 2021-08-20 | 中国农业科学院农业资源与农业区划研究所 | Near-surface air temperature estimation method, system, storage medium and equipment |
CN113283155B (en) * | 2021-03-22 | 2024-04-09 | 中国农业科学院农业资源与农业区划研究所 | Near-surface air temperature estimation method, system, storage medium and equipment |
CN113131476A (en) * | 2021-04-28 | 2021-07-16 | 南方电网科学研究院有限责任公司 | Power load prediction method |
CN114064203A (en) * | 2021-10-28 | 2022-02-18 | 西安理工大学 | Cloud virtual machine load prediction method based on multi-scale analysis and deep network model |
CN114113773B (en) * | 2021-11-11 | 2024-03-19 | 国网辽宁省电力有限公司经济技术研究院 | Non-invasive load monitoring method based on zero sample learning |
CN114113773A (en) * | 2021-11-11 | 2022-03-01 | 国网辽宁省电力有限公司经济技术研究院 | Non-invasive load monitoring method based on zero sample learning |
CN115081702A (en) * | 2022-06-14 | 2022-09-20 | 国网信息通信产业集团有限公司 | Power load prediction method with interpretable characteristic, system and terminal |
WO2023241439A1 (en) * | 2022-06-14 | 2023-12-21 | 国网信息通信产业集团有限公司 | Power load forecasting method and system having interpretable characteristics, and terminal |
CN116341189A (en) * | 2023-01-12 | 2023-06-27 | 山西大学 | Prophet and Holt-windows combined model thermal comfort prediction method based on wavelet transformation |
CN117390895B (en) * | 2023-12-08 | 2024-02-09 | 华科五洲(天津)海洋工程有限公司 | Semi-submersible ship ballast system simulation method, device and storage medium |
CN117390895A (en) * | 2023-12-08 | 2024-01-12 | 华科五洲(天津)海洋工程有限公司 | Semi-submersible ship ballast system simulation method, device and storage medium |
CN117977579A (en) * | 2024-03-28 | 2024-05-03 | 西华大学 | DWT-LSTM neural network-based adjustable load prediction method |
CN117996816A (en) * | 2024-03-29 | 2024-05-07 | 江苏谷峰电力科技股份有限公司 | Intelligent control method and system for wind, light and firewood storage and team-level energy storage |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112488396A (en) | Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model | |
Dong et al. | Hourly energy consumption prediction of an office building based on ensemble learning and energy consumption pattern classification | |
Deihimi et al. | Application of echo state networks in short-term electric load forecasting | |
Rai et al. | Analysis of classical and machine learning based short-term and mid-term load forecasting for smart grid | |
CN107506868B (en) | Method and device for predicting short-time power load | |
CN111260136A (en) | Building short-term load prediction method based on ARIMA-LSTM combined model | |
Wang et al. | Wind speed forecasting based on multi-objective grey wolf optimisation algorithm, weighted information criterion, and wind energy conversion system: A case study in Eastern China | |
CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
CN110705743A (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
Zhu et al. | Short‐Term Electricity Consumption Forecasting Based on the EMD‐Fbprophet‐LSTM Method | |
CN115759371A (en) | GCN-LSTM-based short-term load prediction method for power system | |
CN116845874A (en) | Short-term prediction method and device for power load | |
Zhang et al. | Load Prediction Based on Hybrid Model of VMD‐mRMR‐BPNN‐LSSVM | |
Dong et al. | Short-term building cooling load prediction model based on DwdAdam-ILSTM algorithm: A case study of a commercial building | |
CN114595861A (en) | MSTL (modeling, transformation, simulation and maintenance) and LSTM (least Square TM) model-based medium-and-long-term power load prediction method | |
CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
CN116169670A (en) | Short-term non-resident load prediction method and system based on improved neural network | |
Arpogaus et al. | Short-term density forecasting of low-voltage load using Bernstein-polynomial normalizing flows | |
CN116091118A (en) | Electricity price prediction method, device, equipment, medium and product | |
Chen et al. | Short term load forecasting based on SBiGRU and CEEMDAN-SBiGRU combined model | |
Wu et al. | Automatic fuzzy model identification for short-term load forecast | |
CN114266602A (en) | Deep learning electricity price prediction method and device for multi-source data fusion of power internet of things | |
Li et al. | Short-term probabilistic load forecasting method based on uncertainty estimation and deep learning model considering meteorological factors | |
CN113570414A (en) | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm | |
Xu et al. | Federated learning for interpretable short-term residential load forecasting in edge computing network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210312 |