CN113191559A - Improved neural network short-term resident load prediction method based on autoregressive selection - Google Patents
Improved neural network short-term resident load prediction method based on autoregressive selection Download PDFInfo
- Publication number
- CN113191559A CN113191559A CN202110501221.7A CN202110501221A CN113191559A CN 113191559 A CN113191559 A CN 113191559A CN 202110501221 A CN202110501221 A CN 202110501221A CN 113191559 A CN113191559 A CN 113191559A
- Authority
- CN
- China
- Prior art keywords
- data
- load
- historical
- prediction
- autoregressive
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims description 11
- 241000255789 Bombyx mori Species 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005096 rolling process 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
- 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
- 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
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an improved neural network short-term resident load prediction method based on autoregressive selection, which comprises the following steps: s1, counting the historical power load and related data of the residential area; s2, dividing historical data according to seasons; s3, selecting data and training a prediction model; s4, selecting an autoregressive mode of the model; s5, predicting the latest data and updating the prediction model; and S6, judging whether the sequence length reaches the standard. According to the method, the load is predicted according to seasons, a set of models is used for load prediction in different seasons, and meanwhile, the neural network is selected based on improved autoregressive, so that the accuracy of the load is effectively improved. The method fills the blank of residential area load prediction, can take into account the influence of external conditions and human factors, and realizes accurate load prediction.
Description
Technical Field
The invention belongs to the field of load prediction, and particularly relates to an improved neural network short-term resident load prediction method based on autoregressive selection.
Background
In recent years, with the popularization of electronic products and the change of working modes, the difference of power loads between different communities and resident families is larger and larger, and the different resident loads provide new requirements for a power grid. Meanwhile, with the further development of the distributed photovoltaic and new energy electric automobile, the load mode of the residential area is more complicated. Therefore, accurate prediction of residential load becomes an indispensable link.
In the current power system research, load prediction is an extremely important issue. However, the research on the load of residential areas is very rare, and the reason for the research is mainly two: firstly, the privacy of electricity load in residential areas is high, and especially in multi-family residences, the load data of the residential areas are difficult to collect; second, predicting the power load of a single household is generally considered challenging due to the instability of the household load data. It is therefore desirable to provide a method for efficiently and accurately predicting the short-term load in a residential area.
Aiming at the problems proposed above, an improved neural network short-term resident load prediction method based on autoregressive selection is designed.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide the improved neural network short-term residential load prediction method based on autoregressive selection, fills the blank of residential area load prediction, can take the external condition influence and the human factor influence into account, and realizes accurate load prediction.
The purpose of the invention can be realized by the following technical scheme:
an improved neural network short-term resident load prediction method based on autoregressive selection comprises the following steps:
s1, counting the historical power load and related data of the residential area;
s2, dividing historical data according to seasons;
s3, selecting data and training a prediction model;
s4, selecting an autoregressive mode of the model;
s5, predicting the latest data and updating the prediction model;
and S6, judging whether the sequence length reaches the standard.
Further, the S1 specifically includes the following steps:
s1.1, counting historical power load data of residential areas, namely counting daily average load power of buildings in one residential area and historical hourly load data of each resident;
s1.2, counting other data of the power load accidents, including historical temperature data, historical humidity data, historical time data, historical popular data and historical date data.
Further, the S2 specifically includes the following steps:
s2.1, preprocessing and data cleaning are carried out on the statistical historical data, namely abnormal data in the historical data are removed, and vacant data in the historical data are filled;
and S2.2, classifying the preprocessed data, namely classifying the preprocessed data into four groups of data in spring, summer, autumn and winter according to seasons.
Further, the S3 specifically includes the following steps:
s3.1, for data of different seasons, initializing model training, namely for one season, selecting data of the previous four weeks or other continuous four weeks;
s3.2, selecting corresponding historical characteristics for the selected data, wherein the selected characteristic data includes historical load data, humidity, temperature, time, wind speed, working days or rest days.
S3.3, using the selected data for ConvLSTM model training, wherein a calculation formula of ConvLSTM is as follows:
ft=σ(Wfx*xt+Wfh*ht-1+bf) (1)
it=σ(Wix*It+Wih*ht-1+bi) (2)
ot=σ(Wox*It+Woh*ht-1+bo) (4)
ht=tanh(Ct)·ot (6);
and S3.4, a single step prediction method and an optimal hysteresis step prediction method exist subsequently, the optimal hysteresis step model and the single step prediction model need to be trained respectively, the input characteristic quantity of the single step regression prediction model is only the load characteristic value at the previous moment, and the input characteristic quantity of the optimal hysteresis step prediction model is the temperature characteristic quantity, the humidity characteristic quantity, the time characteristic quantity, the wind speed characteristic quantity, the working day or rest day characteristic quantity and the hysteresis load characteristic quantity.
Further, in the formulas (1) to (6) of S3.3, "·" represents a hadamard product, "×" represents a convolution, and C represents a convolutiont-1Represents the output of the ConvLSTM module at t-1, ht-1Denotes the state of the ConvLSTM module at t-1, htIndicating the output of the cell at time t, CtIndicating the state of the cell at time t, Wfx、Wfh、Wix、Wih、Wcx、Wch、WoxAnd WohIs a trainable weight in which each intermediate vector appears in pairs, bf、bi、bcAnd boIs a trainable deviation.
Further, the S4 specifically includes the following steps:
s4.1, setting an autoregressive threshold theta, wherein the parameter is used for subsequent judgment and prediction;
s4.2, predicting the load condition of the next hour, inputting historical data of the past week, extracting characteristic variables, and further calculating the ACF sequence of the historical load sequence, wherein the specific calculation formula is as follows:
in formula (7), X, Y represent random variables, Cov (X, Y) represents covariance of X, Y, DX represents variance of X, and DY represents variance of Y;
for a fixed sequence, the calculation is as follows:
in the formula (8), XiItem i, X, representing a sequencei+kThe i + k-th term of the sequence is represented,represents the average of the sequence;
calculating the ACF sequence at the maximum lag of 24 hours according to the formulas (7) and (8);
s4.3, selecting the delay time when the ACF value is maximum as the optimal delay time, and comparing the ACF value with the theta, wherein if the ACF value is larger than the theta, the optimal delay step prediction method is selected, otherwise, the method is a single-step regression prediction method for avoiding overfitting the silkworm pupae;
the input of the prediction method of the optimal hysteresis step is 1 × 6 characteristic variables, which are specifically shown as follows:
Et=[Tt,Ht,t,Wst,wd&wet] (9)
It=[yt-p,Et] (10)
yt=ConvLSTM(It) (11);
the input of the single-step regression prediction method is load characteristic data at the last moment, and the specific steps are as follows:
It=[yt-1] (12)
yt=ConvLSTM(It) (13)。
further, in the S4.3 formulas (9) to (11), EtIs an exogenous feature vector, yt-pLoad at time t-p, ItAs input data to the ConvLSTM neural network, ytIs the predicted load outcome.
EtThe meanings of the variables are as follows: t istDenotes the temperature at time t, HtIndicating the humidity at time t, t being the time,WstRepresenting the wind speed at time t, wd&wetIndicating that the day is a weekday or a holiday.
Further, the S5 specifically includes the following steps:
s5.1, obtaining the latest one-hour prediction data based on the determined prediction method and data in S4;
and S5.2, updating model training data by using the predicted data, and retraining the model.
Further, the S6 specifically includes the following steps:
and judging whether the sequence length generated after the S5 is finished meets the required length, if the length of the predicted sequence meets the requirement, ending the process, and if the length of the predicted sequence does not meet the requirement, repeating the work in S4 and S5 after adding new predicted data.
The invention has the beneficial effects that:
1. according to the improved neural network short-term resident load prediction method based on autoregressive selection, disclosed by the invention, the load is predicted according to seasons, a set of models is used for load prediction in different seasons, and meanwhile, the neural network is selected based on the improved autoregressive selection, so that the accuracy of the load is effectively improved;
2. according to the improved neural network short-term resident load prediction method based on autoregressive selection, influences of historical load data, historical temperature data, historical humidity data, historical time data, historical customs data, historical date data and the like are considered in the process of predicting the resident load, and the contribution of multidimensional factors is considered, so that the influence of human behavior on the resident load can be reduced to the greatest extent, and the prediction accuracy can be effectively improved;
3. the improved neural network short-term resident load prediction method based on the autoregressive selection determines a prediction mode from the autocorrelation coefficient of a calculation sequence during load prediction, so that the maximum correlation of a prediction result and an original sequence can be ensured, and an overfitting phenomenon can be avoided;
4. the invention provides an improved neural network short-term resident load prediction method based on autoregressive selection, which adopts a rolling model method in load prediction. The method forms a new sequence by the result of the single step prediction and the original sequence, and further trains a prediction model according to the new sequence, and predicts the load result of the next time period by the new sequence and the model.
Drawings
In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present invention, the drawings used in the description of the embodiments or prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a flow chart of an overall prediction method of an embodiment of the present invention;
FIG. 2 is a flow chart of an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the 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 invention.
As shown in fig. 1 and fig. 2, an improved neural network short-term resident load prediction method based on autoregressive selection comprises the following steps:
s1, counting the historical power load and related data of the residential area;
s2, dividing historical data according to seasons;
s3, selecting data and training a prediction model;
s4, selecting an autoregressive mode of the model;
s5, predicting the latest data and updating the prediction model;
and S6, judging whether the sequence length reaches the standard.
The above steps are further explained below.
S1, counting the historical power load and related data of the residential area, and specifically comprising the following steps:
s1.1, counting historical power load data of residential areas, namely counting daily average load power of buildings in one residential area and historical hourly load data of each resident;
s1.2, counting other data of the power load accidents, including historical temperature data, historical humidity data, historical time data, historical popular data and historical date data.
S2, dividing historical data according to seasons, and specifically comprising the following steps:
s2.1, preprocessing and data cleaning are carried out on the statistical historical data, namely abnormal data in the historical data are removed, and vacant data in the historical data are filled;
and S2.2, classifying the preprocessed data, namely classifying the preprocessed data into four groups of data of spring, summer, autumn and winter according to the seasons of 1-3 months, 4-6 months, 7-9 months and 10-12 months.
S3, selecting data and training a prediction model, and specifically comprises the following steps:
s3.1, for data in different seasons, initializing model training, namely for a certain season, selecting data in the front four weeks or other continuous four weeks;
s3.2, selecting corresponding historical characteristics for the selected data, wherein the selected characteristic data includes historical load data, humidity, temperature, time, wind speed, working days or rest days.
S3.3, using the selected data for ConvLSTM (ConvLSTM is LSTM-a variant of long-short term memory artificial neural network, and changing weight calculation mainly of W into convolution operation so as to extract characteristics of images) model training-a calculation formula of ConvLSTM is as follows:
ft=σ(Wfx*xt+Wfh*ht-1+bf) (1)
it=σ(Wix*It+Wih*ht-1+bi) (2)
ot=σ(Wox*It+Woh*ht-1+bo) (4)
ht=tanh(Ct)·ot (6)
in equations (1) to (6), "·" denotes a hadamard product, "+" denotes a convolution, Ct-1Represents the output of the ConvLSTM module at t-1, ht-1Denotes the state of the ConvLSTM module at t-1, htIndicating the output of the cell at time t, CtIndicating the state of the cell at time t, Wfx、Wfh、Wix、Wih、Wcx、Wch、WoxAnd WohIs a trainable weight in which each intermediate vector appears in pairs, bf、bi、bcAnd boIs a trainable bias;
and S3.4, a single step prediction method and an optimal hysteresis step prediction method exist in the follow-up process, so that an optimal hysteresis step model and a single step prediction model need to be trained in the step respectively, wherein the input characteristic quantity of the single step regression prediction model is only the load characteristic value at the previous moment, and the input characteristic quantity of the optimal hysteresis step prediction model is temperature characteristic quantity, humidity characteristic quantity, time characteristic quantity, wind speed characteristic quantity, working day or rest day characteristic quantity and hysteresis load characteristic quantity.
S4, selecting an autoregressive mode of the model, and specifically comprising the following steps:
s4.1, setting an autoregressive threshold theta, wherein the parameter is used for subsequent judgment and prediction;
s4.2, predicting the load condition of the next hour, inputting historical data of the past week, extracting characteristic variables, and further calculating the ACF sequence of the historical load sequence, wherein the specific calculation formula is as follows:
in the formula (7), X and Y represent random variables, Cov (X, Y) represents covariance of X and Y, DX represents variance of X, and DY represents variance of Y.
For a fixed sequence, the calculation is as follows:
in the formula (8), XiItem i, X, representing a sequencei+kThe i + k-th term of the sequence is represented,mean values of the sequences are indicated.
Calculating the ACF sequence at the maximum lag of 24 hours according to the formulas (7) and (8);
s4.3, selecting the delay time when the ACF value is maximum as the optimal delay time, and comparing the ACF value with the theta, wherein if the ACF value is larger than the theta, the optimal delay step prediction method is selected, otherwise, the method is a single-step regression prediction method for avoiding overfitting the silkworm pupae;
the input of the prediction method of the optimal hysteresis step is 1 × 6 characteristic variables, which are specifically shown as follows:
Et=[Tt,Ht,t,Wst,wd&wet] (9)
It=[yt-p,Et] (10)
yt=ConvLSTM(It) (11)
in formulae (9) to (11), EtIs an exogenous feature vector, yt-pLoad at time t-p, ItAs input data to the ConvLSTM neural network, ytIs a predicted load outcome;
Etthe meanings of the variables are as follows: t istDenotes the temperature at time t, HtDenotes the humidity at time t, t denotes the time, WstThe wind speed at the time t is indicated,wd&wetindicating that the day is a weekday or a holiday;
the input of the single-step regression prediction method is load characteristic data at the last moment, and the specific steps are as follows:
It=[yt-1] (12)
yt=ConvLSTM(It) (13)。
s5, predicting the latest data and updating the prediction model, which comprises the following steps:
s5.1, obtaining the latest one-hour prediction data based on the determined prediction method and data in S4;
and S5.2, updating model training data by using the predicted data, and retraining the model.
S6, judging whether the sequence length reaches the standard, specifically comprising the following steps:
s6.1, firstly, judging whether the sequence length generated after S5 is finished meets the required length, if the length of the prediction sequence meets the requirement, ending the process, and if the length of the prediction sequence does not meet the requirement, repeating the work in S4 and S5 after adding new prediction data.
In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed.
Claims (9)
1. An improved neural network short-term resident load prediction method based on autoregressive selection is characterized by comprising the following steps:
s1, counting the historical power load and related data of the residential area;
s2, dividing historical data according to seasons;
s3, selecting data and training a prediction model;
s4, selecting an autoregressive mode of the model;
s5, predicting the latest data and updating the prediction model;
and S6, judging whether the sequence length reaches the standard.
2. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S1 specifically comprises the following steps:
s1.1, counting historical power load data of residential areas, namely counting daily average load power of buildings in one residential area and historical hourly load data of each resident;
s1.2, counting other data of the power load accidents, including historical temperature data, historical humidity data, historical time data, historical popular data and historical date data.
3. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S2 specifically comprises the following steps:
s2.1, preprocessing and data cleaning are carried out on the statistical historical data, namely abnormal data in the historical data are removed, and vacant data in the historical data are filled;
and S2.2, classifying the preprocessed data, namely classifying the preprocessed data into four groups of data in spring, summer, autumn and winter according to seasons.
4. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S3 specifically comprises the following steps:
s3.1, for data of different seasons, initializing model training, namely for one season, selecting data of the previous four weeks or other continuous four weeks;
s3.2, selecting corresponding historical characteristics for the selected data, wherein the selected characteristic data includes historical load data, humidity, temperature, time, wind speed, working days or rest days.
S3.3, using the selected data for ConvLSTM model training, wherein a calculation formula of ConvLSTM is as follows:
ft=σ(Wfx*xt+Wfh*ht-1+bf) (1)
it=σ(Wix*It+Wih*ht-1+bi) (2)
ot=σ(Wox*It+Woh*ht-1+bo) (4)
ht=tanh(Ct)·ot (6);
and S3.4, a single step prediction method and an optimal hysteresis step prediction method exist subsequently, the optimal hysteresis step model and the single step prediction model need to be trained respectively, the input characteristic quantity of the single step regression prediction model is only the load characteristic value at the previous moment, and the input characteristic quantity of the optimal hysteresis step prediction model is the temperature characteristic quantity, the humidity characteristic quantity, the time characteristic quantity, the wind speed characteristic quantity, the working day or rest day characteristic quantity and the hysteresis load characteristic quantity.
5. The method of claim 4A method for predicting short-term resident load of an improved neural network based on autoregressive selection is characterized in that in formulas (1) - (6) of S3.3, "·" represents a Hadamard product, ". indicates convolution, and C indicates convolutiont-1Represents the output of the ConvLSTM module at t-1, ht-1Denotes the state of the ConvLSTM module at t-1, htIndicating the output of the cell at time t, CtIndicating the state of the cell at time t, Wfx、Wfh、Wix、Wih、Wcx、Wch、WoxAnd WohIs a trainable weight in which each intermediate vector appears in pairs, bf、bi、bcAnd boIs a trainable deviation.
6. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S4 specifically comprises the following steps:
s4.1, setting an autoregressive threshold theta, wherein the parameter is used for subsequent judgment and prediction;
s4.2, predicting the load condition of the next hour, inputting historical data of the past week, extracting characteristic variables, and further calculating the ACF sequence of the historical load sequence, wherein the specific calculation formula is as follows:
in formula (7), X, Y represent random variables, Cov (X, Y) represents covariance of X, Y, DX represents variance of X, and DY represents variance of Y;
for a fixed sequence, the calculation is as follows:
in the formula (8), XiItem i, X, representing a sequencei+kRepresents the i + k term of the sequence, X represents the average of the sequence;
calculating the ACF sequence at the maximum lag of 24 hours according to the formulas (7) and (8);
s4.3, selecting the delay time when the ACF value is maximum as the optimal delay time, and comparing the ACF value with the theta, wherein if the ACF value is larger than the theta, the optimal delay step prediction method is selected, otherwise, the method is a single-step regression prediction method for avoiding overfitting the silkworm pupae;
the input of the prediction method of the optimal hysteresis step is 1 × 6 characteristic variables, which are specifically shown as follows:
Et=[Tt,Ht,t,Wst,wd&wet] (9)
It=[yt-p,Et] (10)
yt=ConvLSTM(It) (11);
the input of the single-step regression prediction method is load characteristic data at the last moment, and the specific steps are as follows:
It=[yt-1] (12)
yt=ConvLSTM(It) (13)。
7. the method for improving short-term resident load prediction in neural network based on autoregressive selection as claimed in claim 6, wherein in the formula (9) - (11) of S4.3, E istIs an exogenous feature vector, yt-pLoad at time t-p, ItAs input data to the ConvLSTM neural network, ytIs a predicted load outcome;
Etthe meanings of the variables are as follows: t istDenotes the temperature at time t, HtDenotes the humidity at time t, t denotes the time, WstRepresenting the wind speed at time t, wd&wetIndicating that the day is a weekday or a holiday.
8. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S5 specifically comprises the following steps:
s5.1, obtaining the latest one-hour prediction data based on the determined prediction method and data in S4;
and S5.2, updating model training data by using the predicted data, and retraining the model.
9. The method for improving neural network short-term resident load prediction based on autoregressive selection as claimed in claim 1, wherein said S6 specifically comprises the following steps:
and judging whether the sequence length generated after the S5 is finished meets the required length, if the length of the predicted sequence meets the requirement, ending the process, and if the length of the predicted sequence does not meet the requirement, repeating the work in S4 and S5 after adding new predicted data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110501221.7A CN113191559B (en) | 2021-05-08 | 2021-05-08 | Improved neural network short-term resident load prediction method based on autoregressive selection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110501221.7A CN113191559B (en) | 2021-05-08 | 2021-05-08 | Improved neural network short-term resident load prediction method based on autoregressive selection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113191559A true CN113191559A (en) | 2021-07-30 |
CN113191559B CN113191559B (en) | 2024-04-19 |
Family
ID=76984421
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110501221.7A Active CN113191559B (en) | 2021-05-08 | 2021-05-08 | Improved neural network short-term resident load prediction method based on autoregressive selection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113191559B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743072A (en) * | 2022-05-24 | 2022-07-12 | 中国科学院计算机网络信息中心 | Training method of short-term time sequence prediction model |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
US20130110756A1 (en) * | 2011-10-31 | 2013-05-02 | Siemens Corporation | Short-term Load Forecast Using Support Vector Regression and Feature Learning |
CN104636822A (en) * | 2015-01-21 | 2015-05-20 | 广州市香港科大霍英东研究院 | Residential load prediction method of elman-based neural network |
CN105976199A (en) * | 2016-04-26 | 2016-09-28 | 重庆大学 | Medicine sales prediction method and medicine sales prediction system based on hybrid model |
CN110046787A (en) * | 2019-01-15 | 2019-07-23 | 重庆邮电大学 | A kind of urban area charging demand for electric vehicles spatio-temporal prediction method |
WO2019231448A1 (en) * | 2018-05-31 | 2019-12-05 | Siemens Aktiengesellschaft | Solar irradiation prediction using deep learning with end-to- end training |
CN112152201A (en) * | 2020-08-25 | 2020-12-29 | 国网山东综合能源服务有限公司 | Electricity load prediction method and system based on convolution length time memory neural network |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107895211A (en) * | 2017-11-27 | 2018-04-10 | 上海积成能源科技有限公司 | A kind of long-medium term power load forecasting method and system based on big data |
CN110070202A (en) * | 2019-02-19 | 2019-07-30 | 浙江华云信息科技有限公司 | A method of economic output is predicted by electricity consumption data |
CN110070201A (en) * | 2019-02-19 | 2019-07-30 | 国网浙江省电力有限公司 | A method of passing through economic growth rate forecast electricity consumption growth rate |
CN110298501B (en) * | 2019-06-21 | 2022-08-16 | 河海大学常州校区 | Electrical load prediction method based on long-time and short-time memory neural network |
CN110826774B (en) * | 2019-10-18 | 2022-03-22 | 广东电网有限责任公司广州供电局 | Bus load prediction method and device, computer equipment and storage medium |
CN111651941B (en) * | 2020-04-30 | 2022-05-17 | 北京航空航天大学 | Global ionosphere electron total content prediction algorithm |
-
2021
- 2021-05-08 CN CN202110501221.7A patent/CN113191559B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
US20130110756A1 (en) * | 2011-10-31 | 2013-05-02 | Siemens Corporation | Short-term Load Forecast Using Support Vector Regression and Feature Learning |
CN104636822A (en) * | 2015-01-21 | 2015-05-20 | 广州市香港科大霍英东研究院 | Residential load prediction method of elman-based neural network |
CN105976199A (en) * | 2016-04-26 | 2016-09-28 | 重庆大学 | Medicine sales prediction method and medicine sales prediction system based on hybrid model |
WO2019231448A1 (en) * | 2018-05-31 | 2019-12-05 | Siemens Aktiengesellschaft | Solar irradiation prediction using deep learning with end-to- end training |
CN110046787A (en) * | 2019-01-15 | 2019-07-23 | 重庆邮电大学 | A kind of urban area charging demand for electric vehicles spatio-temporal prediction method |
CN112152201A (en) * | 2020-08-25 | 2020-12-29 | 国网山东综合能源服务有限公司 | Electricity load prediction method and system based on convolution length time memory neural network |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743072A (en) * | 2022-05-24 | 2022-07-12 | 中国科学院计算机网络信息中心 | Training method of short-term time sequence prediction model |
Also Published As
Publication number | Publication date |
---|---|
CN113191559B (en) | 2024-04-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
CN105069525B (en) | Round-the-clock 96 Day Load Curve Forecastings and optimization update the system | |
CN110348624B (en) | Sand storm grade prediction method based on Stacking integration strategy | |
CN111428926B (en) | Regional power load prediction method considering meteorological factors | |
CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
CN109583565A (en) | Forecasting Flood method based on the long memory network in short-term of attention model | |
CN113572157A (en) | User real-time autonomous energy management optimization method based on near-end policy optimization | |
CN112215428B (en) | Photovoltaic power generation power prediction method and system based on error correction and fuzzy logic | |
CN112508246B (en) | Photovoltaic power generation power prediction method based on similar days | |
CN111553750A (en) | Energy storage bidding strategy method considering power price uncertainty and loss cost | |
CN112418476A (en) | Ultra-short-term power load prediction method | |
CN114648176A (en) | Wind-solar power consumption optimization method based on data driving | |
CN113393339A (en) | Power distribution network short-term load prediction method based on deep neural network combination model | |
CN113191559B (en) | Improved neural network short-term resident load prediction method based on autoregressive selection | |
CN110222910B (en) | Active power distribution network situation prediction method and prediction system | |
CN108171380A (en) | One kind is based on fisher information processing multidimensional meteorologic factor short term processing methods | |
CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
Zhang et al. | A deep learning based real-time load forecasting method in electricity spot market | |
CN116167508B (en) | Short-term photovoltaic output rapid prediction method and system based on meteorological factor decomposition | |
CN110489893B (en) | Variable weight-based bus load prediction method and system | |
CN112734073A (en) | Photovoltaic power generation short-term prediction method based on long and short-term memory network | |
CN115731072A (en) | Microgrid space-time perception energy management method based on safe deep reinforcement learning | |
CN109034453A (en) | A kind of Short-Term Load Forecasting Method based on multiple labeling neural network | |
Lv et al. | Research on short-term load forecasting approach for smart grid | |
CN115392387A (en) | Low-voltage distributed photovoltaic power generation output prediction method |
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 | ||
CB03 | Change of inventor or designer information |
Inventor after: Chen Lijuan Inventor before: Chen Lijuan Inventor before: Yang Wenzhen Inventor before: Liang Shuo |
|
CB03 | Change of inventor or designer information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |