CN114219150B - Power load interval prediction method based on self-adaptive optimization construction interval - Google Patents
Power load interval prediction method based on self-adaptive optimization construction interval Download PDFInfo
- Publication number
- CN114219150B CN114219150B CN202111535899.3A CN202111535899A CN114219150B CN 114219150 B CN114219150 B CN 114219150B CN 202111535899 A CN202111535899 A CN 202111535899A CN 114219150 B CN114219150 B CN 114219150B
- Authority
- CN
- China
- Prior art keywords
- interval
- training
- prediction
- power load
- picp
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000005457 optimization Methods 0.000 title claims abstract description 24
- 238000010276 construction Methods 0.000 title claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 70
- 230000008569 process Effects 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 108090000672 Annexin A5 Proteins 0.000 claims abstract 10
- 230000003044 adaptive effect Effects 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 8
- 238000013135 deep learning Methods 0.000 claims description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000009514 concussion Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011084 recovery 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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a power load interval prediction method based on a self-adaptive optimization construction interval, which combines PID thought, considers first-order and second-order differences of PICP and PINAW in an adjustment strategy of the self-adaptive construction interval and improves a training process. Besides the interval coverage rate PICP, the method adds the average interval width PINAW of the important index in interval prediction into the adjustment strategy of the self-adaptive construction interval. The method comprises the following steps: preprocessing a power load data sample; constructing a GRU interval prediction model; constructing intervals as training labels; training a GRU interval prediction model, and optimizing and constructing interval width based on the first-order and second-order difference of PICP and PINAW of a prediction interval obtained by each training; and predicting the power load interval based on the trained GRU interval prediction model. The method can obtain a high-quality prediction interval, namely, the PICP has a narrow interval width on the premise of meeting the confidence coefficient, and the training result has good consistency.
Description
Technical Field
The invention belongs to the field of power load interval prediction, and relates to a power load interval prediction method based on a self-adaptive optimization construction interval.
Background
Accurate power load prediction is critical in modern power system economic and safe operation. In recent years, under the large background of energy internet, load influence factors are more diversified, and load characteristics also show new characteristics and trends. Meanwhile, the requirement of efficient and economic operation of the power system on the load prediction precision is gradually increased, and the traditional point prediction method is more and more difficult to meet the actual requirement. Load interval prediction can quantify uncertainty of prediction results, more reference information can be brought to power workers, various scientific and reasonable strategies can be formulated, and therefore more and more attention is paid.
Deep learning has become a research focus in recent years, neural networks with more complex structures have stronger nonlinear mapping capability, and more intrinsic features can be extracted from data than traditional machine learning models. Representative deep learning models include convolutional neural networks, stacked autoencoders, and Long-Short Term Memory neural networks (LSTM). In recent years, a Gated Round Unit (GRU) has been developed as an extension of the LSTM with a simplified gating mechanism, with similar performance as the LSTM and with a lower computational burden. These models are rarely applied to the interval prediction problem. Based on previous research, researchers have proposed a GRU prediction model with high learning ability, which can directly generate a prediction interval and perform model training by using an efficient gradient descent algorithm, such as Root Mean Square (RMSProp) and Adaptive momentum (Adam) algorithms. Based on the nature of the gradient, these algorithms need a differentiable cost function for supervised learning, so that they cannot optimize an indistinct evaluation index such as CWC. In order to solve the problem, some scholars propose a self-adaptive optimization method based on a structural interval, and high-quality training labels are established for supervised learning of the model. However, the method does not consider the optimization of the width of the prediction interval when applied, and the prediction interval obtained by each training has larger uncertainty.
Disclosure of Invention
In order to overcome the above disadvantages of the prior art, the present invention provides an improved power load interval prediction method based on an adaptive optimization construction interval.
The purpose of the invention is realized by the following technical scheme: a power load interval prediction method based on an adaptive optimization construction interval comprises the following steps:
(1) preprocessing a power load data sample, dividing a power load time sequence into characteristics and labels, and constructing a training set;
(2) constructing a GRU (gate recovery Unit) interval prediction model, wherein the GRU interval prediction model comprises a GRU cycle input layer, a full connection layer and an output layer which are sequentially connected, and the output layer comprises two neurons which respectively output an upper bound and a lower bound of a prediction interval;
(3) the construction interval is defined as a training label as follows:
whereinAnd Y is the upper and lower bounds of the build interval, respectively; d u And d l The upper and lower boundary widths of the construction interval are respectively used as variables for adaptive optimization; y is a label marked out in the step (1);
(4) training a GRU Interval prediction model through the training set constructed in the step (1) and the training labels constructed in the step (3), and optimizing the width of a constructed Interval based on the first-order and second-order differences of PICP (prediction interaction Coverage probability) and PINAW (prediction interaction Normalized Average width) of the prediction Interval obtained by each training;
(5) and predicting the power load interval of the future time period based on the trained GRU interval prediction model.
Further, in the step (1), the power load data sample is preprocessed, including removing abnormal data, filling null values, normalizing the data, dividing the power load time sequence into features and labels by windowing, and dividing the data set into a training set, a verification set and a test set.
Further, the step (4) comprises the following substeps:
(4.1) substituting the characteristics of the training set samples into the GRU interval prediction model constructed in the step (2), and training the model by adopting an Adam deep learning optimization algorithm and the training labels constructed in the step (3);
(4.2) obtaining the upper and lower boundaries of a prediction interval after a training period, and calculating the average fitting error;
(4.3) calculating the prediction interval in the current training sampleThe PICP and the PINAW, and the width d of the constructed upper and lower bounds is updated u ,d l ;
And (4.4) if the maximum training period is reached, quitting training and saving the model, and otherwise, repeating the steps (3) - (4).
Further, in the step (4.2), the calculation formula of the average fitting error is as follows:
wherein e l ,e u Mean fitting error of lower and upper bounds, L, respectively i ,U i Respectively the ith output lower bound and the ith output upper bound of the model, n is the number of samples in the training set, i Y,respectively, the lower bound and the upper bound of the construction interval of the ith sample.
Further, in the step (4.3), the constructed upper and lower bound widths d are updated u ,d l The specific process is as follows:
(4.3.1) if the current PICP is present<PINC, updating d according to the following formula u ,d l :
Wherein the parameter k 1 The updating speed for controlling the interval width, and the updating strategy of the parameter alpha is as follows:
wherein PICP (T) is the PICP value k of the current time T i1 ,k p1 ,k d1 Respectively controlling coefficients of integral, proportional and differential terms;
(4.3.2) if the current PICP is greater than or equal to the PINC, updating d according to the following formula u ,d l :
Wherein the parameter k 3 The updating speed for controlling the interval width is the same as that of the parameter alpha updating strategy in the step (4.3.1), and the updating strategy of the parameter beta is as follows:
wherein PINAW (T) is the PINAW value, k, of the current time T i2 ,k p2 ,k d2 Respectively, are the coefficients of integral, proportional and differential term control.
Further, in the step (5), for the CWC (coverage Width criterion) index result of the verification set, a model with the smallest CWC index is selected for predicting the power load section of the future time period.
The beneficial technical effects of the invention are as follows: the method is combined with the GRU deep learning algorithm of the current mainstream, the self-adaptive optimization method based on the construction interval is improved, the average width of the interval is introduced in the optimization process, the closed-loop self-adaptive adjustment strategy based on the PID idea is adopted to improve the prediction effect, and meanwhile, the best training model is selected by applying the training indexes of the verification set, so that the prediction stability is improved.
Drawings
Fig. 1 is a diagram of a GRU interval prediction model structure according to an embodiment of the present invention;
FIG. 2 is a flow chart of a prediction method provided by an embodiment of the present invention;
FIG. 3 is a diagram of a model training process provided by an embodiment of the present invention;
fig. 4 is a diagram of an interval prediction result provided in the embodiment of the present invention.
Detailed Description
For better understanding of the technical solutions of the present application, the following detailed descriptions of the embodiments of the present application are provided with reference to the accompanying drawings.
It should be understood that the embodiments described are only a few embodiments of the present application, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The terminology used in the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the examples of this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
Fig. 1 is a diagram of a GRU interval prediction model structure, in which a GRU loop input layer is used for time-series feature extraction, and then further processes the features through a full-link layer to output upper and lower bounds of a prediction interval. The specific explanation is as follows: the first layer is a GRU cycle input layer, based on an input sequence x ═ x 1 ,…,x n Finish feature vector H t The length of the time series is n, wherein each x i The output of the last GRU unit is the extracted feature vector and is input to the following full connection layer as the load value at the ith moment. The full connection layer comprises a plurality of hidden layers and is connected with the output layer, and the output layer comprises two neurons and respectively outputs the upper bound and the lower bound of the prediction interval.
The adaptive optimization method based on the construction interval is specifically described as follows: aiming at a given Interval prediction confidence level PINC (prediction Interval Nominal Confidence), a training label is directly obtained by a method of artificially constructing an upper boundary and a lower boundary. For a series of training labels in the training set Y ═ Y 1 ,y 2 ,…,y n ]Where n is the number of samples in the training set, the construction interval is defined as follows:
whereinAnd Y is the upper and lower bounds of the construction interval, respectively;d u And d l The upper and lower bound widths of the construction interval are respectively used as the variables of the self-adaptive optimization, so that a cost function based on the mean square error can be constructed:
wherein L is i ,U i The ith output lower bound and upper bound of the model, respectively, because of f for each training sample cost The Adam algorithm can therefore use this function to calculate the gradient of each weight and bias in the neural network.
Aiming at the problem that the gradient descent algorithm cannot optimize the width like a flexible heuristic algorithm, a self-adaptive width optimization method is provided. After each training period, the prediction interval of the model output will gradually approach the constructed training interval, and a fitting error will occur between the two. The mean fit error is defined as follows:
wherein e l ,e u The average fitting error of the lower bound and the upper bound respectively, because the fitting error is gradually optimized and finally tends to be stable, the interval expected to be constructed can follow the output prediction interval. According to the above assumptions, after each training period, if e l The section width d of the structure needs to be increased l (ii) a If e u If the height of the section is increased, the section width d of the structure needs to be reduced u . Simultaneously, a parameter alpha is introduced for optimizing the PICP after each training period, and when the PICP is less than the PINC, the parameter alpha is increased so as to increase the interval width d u ,d l . The specific adjustment process is as follows:
wherein the parameter k 1 For controlling interval widthThe update speed and the update strategy of the parameter α are shown in equation (5).
Wherein PICP (T) is the PICP value, k, of the current time T i1 ,k p1 ,k d1 Respectively, are the coefficients of integral, proportional and differential term control. The integral term is used for ensuring that the PICP is converged near the PINC finally along with the iteration of the training process; the proportional term adjusts alpha according to the variation trend of v (T), if v (T) < v (T-1) and PICP (T) < PINC, it indicates that v (T) has the trend of deterioration, namely PICP shows the descending trend, alpha is further increased by reducing delta alpha, thereby increasing the interval width and preventing the PICP from descending; the derivative term adjusts alpha by further considering the variation trend of v (T) -v (T-1), improves the speed of convergence of the training process and reduces the concussion.
In order to consider the PINAW while considering the PICP in the training process, a parameter beta is added to optimize the PINAW after each training period, and the adjustment principle is similar to that of optimizing the PICP. The specific adjustment process is as follows:
wherein α and β are adjusted according to formulas (5) and (7):
wherein the parameter k 3 For controlling the update speed of the section width. Because the optimization of the PICP and the PINAW has certain contradiction in the training process, when the PICP of a training sample does not reach a given confidence coefficient, the optimized PINAW is not considered for the moment, and the constructed interval width is updated according to the formula (4). When the PICP of the training sample reaches a given confidence coefficient, the smaller the PINAW is, the better the PINAW is expected to be, so that the training index of the PINAW is preset to be 0, and the constructed interval width is updated according to the formula (6).
As shown in FIG. 2, the method of the present invention comprises the following steps:
step 1: preprocessing a power load data sample, mainly comprising removing abnormal data, filling null values, normalizing the data, dividing a power load time sequence into characteristics and labels by windowing, and dividing a data set into a training set, a verification set and a test set according to the ratio of 8:1: 1.
Step 2: initializing GRU interval prediction model parameters, including:
step 2.1: initializing GRU model parameters and presetting PID parameters k i1 ,k p1 ,k d1 ,k i2 ,k p2 ,k d2 ;
Step 2.2: initially setting the structural interval width d u ,d l And the optimization parameters alpha and beta are 0, so that the optimization parameters are subjected to self-adaptive optimization in the training process.
And step 3: training a GRU interval prediction model, comprising:
step 3.1: according to the formula (1), the current configuration d is adopted u ,d l Constructing a training interval by using the label Y, and substituting a training set sample into the GRU model;
step 3.2: calculating a gradient by adopting an Adam deep learning optimization algorithm and a cost function of the formula (2), and updating the weight and the deviation in a small batch mode;
step 3.3: calculating average fitting error according to formula (3), calculating PICP and PINAW of prediction interval in current training sample, updating alpha and beta according to formula (5) and formula (7), and if PICP is detected<The PINC updates d according to equation (4) u ,d l Otherwise, d is updated according to equation (6) u ,d l ;
Step 3.4: calculating CWC of a prediction interval in the verification set;
step 3.5: and if the maximum training period is reached, quitting training and saving the model, otherwise, repeating the steps 3.1-3.4.
And 4, step 4: and taking the model with the minimum CWC index for prediction.
As described above, the historical power load data of AEMO (2006-2010) of new south wils australia is taken as an example for verification, the set confidence level is 93%, and the change of the PICP and the PINAW in the training process is shown in fig. 3. The result of the interval prediction of the trained interval prediction model on the test set is shown in fig. 4, and it can be seen from the graph that the prediction interval has a narrow interval width while keeping the interval coverage level high. Considering the fact that the deep learning algorithm is inconsistent in the multiple training results, the deep learning algorithm is adopted for training 10 times, the PINC is 0.93, and the interval prediction result indexes are recorded as shown in the following table.
Number of times | PICP | PINAW | CWC |
1 | 0.9354 | 0.1569 | 0.1569 |
2 | 0.9370 | 0.1920 | 0.1920 |
3 | 0.9427 | 0.1838 | 0.1838 |
4 | 0.9287 | 0.1737 | 0.1737 |
5 | 0.9394 | 0.1695 | 0.1695 |
6 | 0.9121 | 0.1515 | 0.1515 |
7 | 0.9290 | 0.1588 | 0.1588 |
8 | 0.9293 | 0.1724 | 0.1724 |
9 | 0.9209 | 0.1565 | 0.1565 |
10 | 0.9277 | 0.1551 | 0.1551 |
Maximum value of | 0.9426 | 0.1920 | 0.1920 |
Minimum value | 0.9121 | 0.1515 | 0.1515 |
Mean value of | 0.9302 | 0.1670 | 0.1670 |
It can be seen that 10 model prediction results obtained after algorithm training are close, which indicates that the 10 model prediction results have strong consistency.
The above description is intended only to be exemplary of the one or more embodiments of the present disclosure, and should not be taken as limiting the one or more embodiments of the present disclosure, as any modifications, equivalents, improvements, etc. that come within the spirit and scope of the one or more embodiments of the present disclosure are intended to be included within the scope of the one or more embodiments of the present disclosure.
Claims (3)
1. A power load interval prediction method based on an adaptive optimization construction interval is characterized by comprising the following steps:
(1) preprocessing a power load data sample, dividing a power load time sequence into characteristics and labels, and constructing a training set;
(2) constructing a GRU interval prediction model, wherein the GRU interval prediction model comprises a GRU cycle input layer, a full connection layer and an output layer which are sequentially connected, and the output layer comprises two neurons and respectively outputs an upper bound and a lower bound of a prediction interval;
(3) the construction interval is defined as a training label as follows:
whereinAndYrespectively an upper bound and a lower bound of the construction interval; d u And d l The upper and lower boundary widths of the construction interval are respectively used as variables for adaptive optimization; y is a label marked out in the step (1);
(4) training a GRU interval prediction model through the training set constructed in the step (1) and the training labels constructed in the step (3), and optimizing the width of the constructed interval based on the first-order difference and the second-order difference of the PICP and the PINAW of the prediction interval obtained by each training; the method comprises the following substeps:
(4.1) substituting the characteristics of the training set samples into the GRU interval prediction model constructed in the step (2), and training the model by adopting an Adam deep learning optimization algorithm and the training labels constructed in the step (3);
(4.2) obtaining the upper and lower bounds of a prediction interval after a training period, and calculating the average fitting error, wherein the calculation formula is as follows:
wherein e l ,e u Mean fitting error of lower and upper bounds, L, respectively i ,U i Respectively the ith output lower bound and the ith output upper bound of the model, n is the number of samples in the training set, i Y,the lower boundary and the upper boundary of the construction interval of the ith sample respectively;
(4.3) calculating PICP and PINAW of a prediction interval in the current training sample, and updating the constructed upper and lower bound width d u ,d l The updating process is as follows:
(4.3.1) if the current PICP is present<PINC, updating d according to the following formula u ,d l :
Wherein the parameter k 1 The updating speed for controlling the interval width, and the updating strategy of the parameter alpha is as follows:
wherein PICP (T) is the PICP value, k, of the current time T i1 ,k p1 ,k d1 Respectively controlling coefficients of integral, proportional and differential terms;
(4.3.2) if the current PICP is greater than or equal to the PINC, updating d according to the following formula u ,d l :
Wherein the parameter k 3 The updating speed for controlling the interval width is the same as that of the parameter alpha updating strategy in the step (4.3.1), and the updating strategy of the parameter beta is as follows:
wherein PINAW (T) is the PINAW value, k, of the current time T i2 ,k p2 ,k d2 The coefficients are respectively controlled by an integral term, a proportional term and a differential term;
(4.4) if the maximum training period is reached, quitting training and saving the model, otherwise, repeating the steps (3) - (4);
(5) and predicting the power load interval of the future time period based on the trained GRU interval prediction model.
2. The method for predicting the power load interval based on the adaptive optimization structure interval as claimed in claim 1, wherein in the step (1), the power load data samples are preprocessed, the preprocessing comprises removing abnormal data, filling null values, normalizing the data, dividing a power load time sequence into features and labels by windowing, and dividing a data set into a training set, a verification set and a test set.
3. The method according to claim 1 or 2, wherein in the step (5), for the result of the CWC index in the verification set, a model with the minimum CWC index is selected for predicting the power load interval of the future time period.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111535899.3A CN114219150B (en) | 2021-12-15 | 2021-12-15 | Power load interval prediction method based on self-adaptive optimization construction interval |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111535899.3A CN114219150B (en) | 2021-12-15 | 2021-12-15 | Power load interval prediction method based on self-adaptive optimization construction interval |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114219150A CN114219150A (en) | 2022-03-22 |
CN114219150B true CN114219150B (en) | 2022-09-23 |
Family
ID=80702424
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111535899.3A Active CN114219150B (en) | 2021-12-15 | 2021-12-15 | Power load interval prediction method based on self-adaptive optimization construction interval |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114219150B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN111461404A (en) * | 2020-03-09 | 2020-07-28 | 华能四川水电有限公司 | Short-term load and hydropower prediction method based on neural network prediction interval |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110263866B (en) * | 2019-06-24 | 2023-11-10 | 苏州智睿新能信息科技有限公司 | Power consumer load interval prediction method based on deep learning |
-
2021
- 2021-12-15 CN CN202111535899.3A patent/CN114219150B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN111461404A (en) * | 2020-03-09 | 2020-07-28 | 华能四川水电有限公司 | Short-term load and hydropower prediction method based on neural network prediction interval |
Non-Patent Citations (3)
Title |
---|
《Deep Learning Method Based on Gated Recurrent Unit and Variational Mode Decomposition for Short-Term Wind Power Interval Prediction》;Ruoheng Wang等;《IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS》;20191113;第31卷(第10期);第Ⅰ、Ⅱ节C、第III节,图2、4 * |
GRU神经网络短期电力负荷预测研究;周莽等;《科技创新与应用》;20181122(第33期);全文 * |
基于改进高斯过程回归的短期负荷概率区间预测方法;刘升伟等;《电力系统保护与控制》;20200102(第01期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN114219150A (en) | 2022-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111474965B (en) | Fuzzy neural network-based method for predicting and controlling water level of series water delivery channel | |
CN112003269B (en) | Intelligent on-line control method of grid-connected shared energy storage system | |
CN111144663B (en) | Ultra-short-term wind power prediction method for offshore wind farm considering output fluctuation process | |
CN108959728A (en) | Radio-frequency devices parameter optimization method based on deep learning | |
CN106920014A (en) | A kind of short-term load forecasting method and device | |
CN109472397B (en) | Polymerization process parameter adjusting method based on viscosity change | |
CN108764539A (en) | A kind of water levels of upstream and downstream prediction technique of step hydropower station | |
CN112131206B (en) | Multi-model database OrientDB parameter configuration automatic tuning method | |
CN109409614A (en) | A kind of Methods of electric load forecasting based on BR neural network | |
CN114357852A (en) | Layered water injection optimization method based on long-short term memory neural network and particle swarm optimization algorithm | |
CN113623719B (en) | Heat exchange station prediction control method based on effective room temperature detection | |
CN115755219B (en) | STGCN-based flood forecast error real-time correction method and system | |
CN113591954A (en) | Filling method of missing time sequence data in industrial system | |
CN112330487B (en) | Photovoltaic power generation short-term power prediction method | |
CN117977568A (en) | Power load prediction method based on nested LSTM and quantile calculation | |
CN114757427A (en) | Autoregressive corrected LSTM intelligent wind power plant ultra-short term power prediction method | |
JP2681421B2 (en) | Demand forecasting method and device | |
CN110097929A (en) | A kind of blast furnace molten iron silicon content on-line prediction method | |
CN114219150B (en) | Power load interval prediction method based on self-adaptive optimization construction interval | |
CN112819224B (en) | Unit output prediction and confidence evaluation method based on deep learning fusion model | |
CN108459570B (en) | Irrigation water distribution intelligent control system and method based on generation of confrontation network architecture | |
CN110147825A (en) | A kind of strawberry greenhouse temperature intelligent detection device based on empirical mode decomposition model | |
CN105511270B (en) | A kind of PID controller parameter optimization method and system based on coevolution | |
CN116562454A (en) | Manufacturing cost prediction method applied to BIM long-short-time attention mechanism network | |
CN114386693A (en) | Reservoir water level prediction early warning method based on PSO-LSTM-Attention model |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |