CN108846528B - Long-term load prediction method for large-scale industrial power consumer - Google Patents
Long-term load prediction method for large-scale industrial power consumer Download PDFInfo
- Publication number
- CN108846528B CN108846528B CN201811097409.4A CN201811097409A CN108846528B CN 108846528 B CN108846528 B CN 108846528B CN 201811097409 A CN201811097409 A CN 201811097409A CN 108846528 B CN108846528 B CN 108846528B
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- model
- load
- training
- 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 39
- 230000007774 longterm Effects 0.000 title claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 230000005611 electricity Effects 0.000 claims abstract description 15
- 238000012731 temporal analysis Methods 0.000 claims abstract description 12
- 238000000700 time series analysis Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 28
- 238000005096 rolling process Methods 0.000 claims description 25
- 238000009826 distribution Methods 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 10
- 230000001932 seasonal effect Effects 0.000 claims description 7
- 238000010998 test method Methods 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 230000001788 irregular Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000007476 Maximum Likelihood Methods 0.000 claims description 2
- 238000012300 Sequence Analysis Methods 0.000 claims 1
- 238000005457 optimization Methods 0.000 claims 1
- 238000003066 decision tree Methods 0.000 abstract description 14
- 230000001174 ascending effect Effects 0.000 abstract description 10
- 238000004140 cleaning Methods 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 8
- 230000006870 function Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000004088 simulation 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
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)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A long-term load prediction method for a large-scale industrial power consumer comprises the following steps: step S10, selecting the object to be predicted, extracting the electricity consumption data, and cleaning the data; step S20, performing time series analysis on the cleaned electric quantity data; step S30, determining the power usage in the past several days using time series analysis; step S40, inputting the data after completing the feature engineering into a gradient ascending decision tree algorithm model; and step S50, adding some service rules for predicting special time intervals, and finely adjusting and optimizing the last trained model to obtain a final prediction result. The method solves the problem of strong dependence of load prediction on meteorological data, greatly improves the prediction time, can perform long-term prediction, and has the advantages of clear thought, high economic value and suitability for popularization and use.
Description
Technical Field
The invention belongs to the field of analysis and calculation of an electric power system, and particularly relates to the technical field of power consumption prediction of different large-scale industrial users of the electric power system.
Background
At present, load prediction of large industrial users is mainly completed manually, the load prediction depends on the reported power consumption of the large users, and the load prediction needs to invest a lot of human resources for different users to respectively predict, and the effect is not ideal. In recent years, many enterprises and universities try to predict the load of large industrial users by using a machine learning algorithm, but data characteristics related to the load of the large industrial users are too complex and cannot be collected mostly, for example, user product inventory, large industrial user production orders, product market price fluctuation and the like, weather data are only dynamic data which have large influence on the load and can be collected at present, so the methods all use the weather prediction data when selecting the dynamic prediction data characteristics, the weather prediction can only reach high accuracy within a short period of one week, the weather of a future month cannot be predicted accurately, the current electricity fee payment or power generation planning takes a month as a unit, and the short-term load prediction cannot have practical value even if the short-term load prediction can reach a high level. The load prediction based on the time series analysis algorithm needs to have obvious periodicity and stability of an analysis object to ensure high accuracy, and unstable residual errors cannot be too large, but the power utilization behaviors of most of large industrial users depend on comprehensive factors such as product sales volume, product orders and the like, and are not periodic and extremely unstable. The method provides a long-term load prediction method for large industrial users without using meteorological data, can prolong the prediction unit to the month, and achieves a very high level on the accuracy rate.
Disclosure of Invention
In summary, it is necessary to provide a new intelligent large-scale user load prediction method, which can accurately predict the large-scale user load for a long time in a month unit and improve the stability of the power grid operation.
A rolling window type large industrial user long-term load prediction method based on a gradient ascending decision tree algorithm and an autocorrelation sequence mainly comprises the following steps: step S10, selecting a large user needing prediction to extract power consumption data, fusing original 15-minute power data into a daily power structure in order to ensure sufficient data quantity and not too much consumption in the calculation process, and screening abnormal data by using a normal distribution test method in statistical probability distribution, wherein the probability of occurrence of one data under the condition of normal distribution is
Mu is a data expectation value, sigma is a data variance value, a probability threshold value is set according to the difference of each group of data objects, when the probability of the data is smaller than the threshold value, the data is regarded as abnormal data and is replaced by using a Newton interpolation method, and in order to prevent a few extremely large abnormal values from having great influence on the data expectation value and the variance, the data which is more than 10 times of a median is replaced by using the median; step S20, time series correlation analysis is carried out on the modified electric quantity data, firstly, the electric quantity data sequence is changed into a stable sequence by using a difference method according to the precondition of the time series analysis, usually, a time series is composed of a long-term trend T, a seasonal change S, a cyclic change C and an irregular change I, and due to the property of a time series analysis algorithm, T, C and I lose part of data at the front end and the tail end of the data in the extraction process, only a historical seasonal change coefficient S of a prediction object is extracted as a data characteristic of a later training model for the purpose of completely fusing data information into metadata; step S30, performing characteristic engineering on the historical electricity consumption data of the object, taking the electricity consumption of a certain day and the difference of the electricity consumption of the day as new data characteristics based on the theoretical idea of time series prediction because the meteorological data is not used, selecting days according to the autocorrelation coefficient of the object in the time series analysis result, wherein the autocorrelation coefficient describes the degree of mutual influence of the sequence between two different time periods, and the calculation formula is that
Wherein musFor data expectation over s time period, σsIs the variance of the data over a time period of s, μtFor data expectation over time period t, σtSelecting the number of days with the maximum mutual influence coefficient of the historical electricity consumption characteristics preferentially according to different conditions of each month for the variance of the data in the time period t; step S40, constructing a gradient ascent algorithm model, inputting the data after completing the characteristic engineering into a gradient ascent decision tree algorithm model, optimizing the model parameters, and generating a weak classifier through multiple iterations, wherein each iteration generates a weak classifierThe classifier is trained on the residual error of the last round of classifier, and the parameter maximum simulation method of the loss function of each weak classifier is
Fm-1(x) For the current model, T (x)i,θm) Linearly overlapping the classifier of each iteration cycle and the newly generated weak classifier for the new weak classifier generated by each iteration cycle, and finally training a load prediction model; step S50, changing the model into a rolling window prediction model, theoretically predicting the load of any number of days by the model trained in the previous step, but the prediction days are still too many and inaccurate, adding a rolling prediction window in order to accurately predict the load for one day, only using the previous model to predict the load, then adding new prediction data into a training data set, removing the data of the training data which is farthest away from the current training set, wherein the length of the training set is the length of the rolling window, so that the length of the training data set can be kept unchanged, the influence of the data which is too long on the current prediction can be eliminated, then retraining the model, and predicting the next day, so that the prediction data are combined after one month of rolling prediction to predict the load of the month; and step S60, adding some service rules for predicting special time intervals according to the characteristics of different users, and finely adjusting and optimizing the trained model, such as the fixed annual inspection time of a large user, employee vacation arrangement and the like, to finally obtain a monthly load prediction result.
The method is based on a gradient ascending decision tree algorithm and a rolling window type power enterprise aiming at an autocorrelation sequence, does not need external data characteristics, and only depends on the characteristics of curve data of the method to predict aperiodic non-stability data. The invention is mainly characterized in that:
1. external meteorological data is not needed, and only historical load data of a large user is relied on;
2. determining a data feature selection length through an autocorrelation coefficient of an autocorrelation sequence;
3. and predicting the daily load data one by using a rolling window, and finally accumulating to obtain the monthly load prediction.
The method has the advantages that under the condition of departing from meteorological data limitation, a new data characteristic is constructed by using the autocorrelation coefficient, long-term load prediction with high accuracy is realized through a rolling window, the data limitation that only short-term load prediction can be carried out is broken through, and the long-term load prediction method has universal applicability. The invention has the advantages of clear thought, better universality, high economic value and suitability for popularization and use.
Drawings
FIG. 1 is a flowchart of a method for predicting long-term load of a rolling window type large industrial user based on a gradient ascending decision tree algorithm and an autocorrelation sequence, which is provided by the invention;
FIG. 2 is a flowchart of an abnormal data cleaning process based on probability statistics normal distribution according to the present invention;
FIG. 3 is a code flow diagram of a gradient ascending decision tree algorithm provided by the present invention;
FIG. 4 is a flowchart of a rolling window model training process provided by the present invention;
fig. 5 is a diagram of a prediction result of the monthly load prediction method provided by the present invention for the first half of 2018 of a certain industrial user.
Detailed Description
The technical scheme of the invention is further detailed in the following description and the accompanying drawings in combination with specific embodiments.
Referring to fig. 1, fig. 1 is a flowchart of a method for predicting a long-term load of a rolling window type large industrial user based on a gradient ascending decision tree algorithm and an autocorrelation sequence, which mainly includes the following steps:
and step S10, selecting an object to be predicted, extracting power consumption data, fusing original 15-minute power data into a daily power structure in order to ensure that the data volume is sufficient and the consumption in the calculation process is not too large, and identifying abnormal data by using a normal distribution test method in statistical probability distribution.
The probability of occurrence of a data in the case of a normal distribution is
Wherein μ is a data expected value, σ is a data variance value, a probability threshold is set according to the difference of each group of data objects, and when the probability of the data occurrence is smaller than the threshold, the data is regarded as abnormal data, see fig. 2, the data is replaced by using an equidistant newton interpolation method, and the computation method of the newton interpolation method is as follows:
before calculating the data expectation and the variance, in order to prevent a few extremely large abnormal values from having great influence on the data expectation and the variance, replacing data more than 10 times of a median by using the median;
step S20, performing time series correlation analysis on the modified electric quantity data, and firstly changing the electric quantity data sequence into a stable sequence by using a difference method according to a precondition of the time series analysis, where a time series generally consists of a long-term trend T, a seasonal variation S, a cyclic variation C, and an irregular variation I, and can be generally expressed as the following form:
yt＝Tt+St+Ct+It
t is a time subscript, and due to the property of a time series analysis algorithm, T, C and I will lose part of data at the front end and the tail end of the data in the extraction process, so that only the historical seasonal change coefficient S of a prediction object is extracted as a data feature of a later training model for the complete integration of data information into metadata;
step S30, performing characteristic engineering on the historical electricity consumption data of the object, wherein only the historical load data is left by the unique dynamic data because the meteorological data is not used, screening the historical load data through time series analysis to create new data characteristics, taking the electricity consumption of a certain number of historical days and the difference of the electricity consumption of the days as new data characteristics based on the theoretical idea of time series prediction, selecting days according to the object autocorrelation coefficient in the time series analysis result, wherein the autocorrelation coefficient describes the degree of mutual influence of the sequence between two different time periods, namely the data of the current time period is influenced most by the time periods of recent history, and the correlation calculation formula is as follows:
wherein musFor data expectation over s time period, σsIs the variance of the data over a time period of s, μtFor data expectation over time period t, σtPreferentially selecting days with the largest mutual influence coefficient of the historical electricity consumption characteristics for the variance of the data in the time period t according to different conditions of each time period, carrying out differential processing on load data of the days before the predicted target date, and then taking the load data together as new data characteristics;
and step S40, constructing a gradient ascending decision tree algorithm model, inputting the data after completing the characteristic engineering into the algorithm model, adjusting and optimizing the parameters of the model, generating a weak classifier by each iteration through multiple iterations, and training each classifier on the basis of the residual error generated by the previous iteration.
The gradient boosting decision tree can be represented in the form: fm(x) For the current model, T (x)i,θm) The new weak classifiers, i.e., the m-th decision tree, θ, generated for each iteration roundmParameters representing a decision tree; strong classifier Fm(x) The linear sum of a plurality of weak classifiers.
The strong classifier iterates m times: fm(x)＝Fm-1(x)+T(x,θm)
Then a loss function is obtained: l (F)m(x),y)＝L(Fm-1(x)+T(x,θm),y)
The purpose of the iteration is to find the parameter thetamConstruction of T (x)i,θm) So that the current round loses L (F)m(x) Y) is minimal and the huber loss function is chosen to be optimal experimentally.
huber loss function:
the parameter maximum likelihood method of the loss function of each weak classifier is as follows:
Fm-1(x) For the current model, T (x)i,θm) And finally training a prediction model which can be used for the electricity consumption of one month in the future for the new weak classifier generated by each iteration. Please refer to fig. 3 for the flow of this step;
and step S50, adding a rolling prediction window to change the model into a rolling window prediction model, performing load prediction for one day by using the previous model, then adding new prediction data into a training data set, removing the data of the training data which is farthest away from the current training set, wherein the length of the training set is the length of the rolling window, then retraining the model by using the new training data set, performing prediction for the next day, repeating the step, rolling the training set window forwards along with the change of the training set, and combining the prediction data after rolling prediction for one month to obtain the load prediction for the month. Please refer to fig. 4 for the flow of this step;
and step S60, adding some service rules for predicting special time intervals according to the characteristics of different users, and finely adjusting and optimizing the trained model, such as the fixed annual inspection time of a large user, employee vacation arrangement and the like, to finally obtain a monthly load prediction result.
The method does not need external data characteristics, and only depends on the characteristics of curve data per se to predict the aperiodic non-stability data. The invention is mainly characterized in that:
1. external meteorological data is not needed, and only historical load data of a large user is relied on;
2. determining a data feature selection length through an autocorrelation coefficient of an autocorrelation sequence;
3. and predicting the daily load data one by using a rolling window, and finally accumulating to obtain the monthly load prediction.
The method has the advantages that under the condition of departing from meteorological data limitation, a new data characteristic is constructed by using the autocorrelation coefficient, long-term load prediction with high accuracy is realized through a rolling window, the data limitation that only short-term load prediction can be carried out is broken through, and the long-term load prediction method has universal applicability. The invention has the advantages of clear thought, better universality, high economic value and suitability for popularization and use.
Example one
The invention relates to a rolling window type large industrial user long-term load forecasting method based on a gradient ascending decision tree algorithm and an autocorrelation sequence, which is used for forecasting a monthly load from 1 month to 6 months in 2018 based on load data of a certain large industrial user under a certain power grid and comparing the monthly load with actual power consumption, and comprises the following specific steps of:
(1) the method comprises the steps that original load data of a large industrial user from 2010 to 2018 in 6 months are taken in a metering automation system, 15-minute data are fused into load data with the unit of day at the acquisition frequency of 15 minutes, most of the data are 0 before 2013 in 10 months, and an acquisition terminal is not erected yet, so that the data are used as effective data from 2013 in 10 months.
(2) Detecting abnormal values by normal distribution test method, and respectively introducing data into
Setting the threshold value to be 0.0001, and judging that the data is an abnormal value if the probability of the data appearing in the current positive distribution is less than 1 ten-thousandth, and replacing the data after calculation according to the previous data and the next data by using a Newton interpolation method.
(3) And carrying out time series analysis on the modified load data through a statmodel framework in a python language library, and adding the extracted seasonal change component S into a data table to become a new data characteristic.
(4) Based on data to be predicted that is 1 month before 2018, by
And calculating the autocorrelation coefficient of 0 th order to 10 th order of lag of the load sequence, wherein the maximum autocorrelation coefficient of 10 th order is 0.673, so that the load data of 10 days before 1 month and 1 day in 2018 is added into the data table as a new data characteristic and differential processing is carried out.
(5) And (3) constructing a gradient ascending decision tree model, inputting the processed data from 8/month 1 in 2013 to 12/month 31 in 2017 as training data into the model, and performing parameter adjustment and model training.
(6) And predicting by using the trained model, starting predicting from 1/2018, adding a rolling window, adding a prediction result into an original training data set to form a new training data set, predicting the load of 1/2/2018, and repeating the steps until 31 times are predicted, and adding the 31 values to obtain the predicted value of the load of 1/2018.
(7) And analogizing in turn to obtain the predicted load value of 2018 in 1-6 months, and according to analysis, the user can leave no stop production during the spring festival, adding a corresponding rule and carrying out error calculation with the actual power consumption, wherein the error calculation formula is as follows:
the final result is shown in FIG. 5.
From the calculation process, the rolling window type large industrial user long-term load prediction method based on the gradient ascending decision tree algorithm and the autocorrelation sequence can accurately provide long-term monthly load prediction of the large industrial user for a power supply bureau, the average accuracy rate is over 90%, and a foundation is provided for stable operation of a power system. 1. In addition, other modifications within the spirit of the invention will occur to those skilled in the art, and it is understood that such modifications are included within the scope of the invention as claimed.
Claims (1)
1. A long-term load prediction method for a large-scale industrial power consumer is characterized by comprising the following steps:
step S10, selecting the predictionThe method comprises the steps of extracting power consumption data from large users, fusing original 15-minute power data into a daily power structure, screening abnormal data by using a normal distribution test method in statistical probability distribution, wherein the probability of one data appearing in the normal distribution is
Mu is a data expected value, sigma is a data variance value, a probability threshold value is set according to the difference of each group of data objects, when the probability of the data is smaller than the threshold value, the data is regarded as abnormal data and is replaced by a Newton interpolation method, and data which is more than 10 times of the median is replaced by the median;
step S20, performing time series correlation analysis on the modified electric quantity data, firstly changing the electric quantity data sequence into a stable sequence by using a difference method according to the precondition of the time series analysis, wherein one time series is composed of a long-term trend T, a seasonal variation S, a cyclic variation C and an irregular variation I, and only extracting a historical seasonal variation coefficient S of a prediction object as a data feature of a later training model;
step S30, performing characteristic engineering on the historical electricity consumption data of the object, taking the electricity consumption of a certain historical day and the difference of the electricity consumption of the day as new data characteristics, selecting the day according to the object autocorrelation coefficient in the time sequence analysis result, wherein the autocorrelation coefficient describes the degree of mutual influence of the sequence between two different time periods, and the calculation formula is
Wherein musFor data expectation over s time period, σsIs the variance of the data over a time period of s, μtFor data expectation over time period t, σtSelecting the number of days with the maximum mutual influence coefficient of the historical electricity consumption characteristics preferentially according to different conditions of each month for the variance of the data in the time period t;
step S40, constructing a gradient ascent algorithm model, inputting the data after completing the characteristic engineering into the gradient ascent algorithm model, and inputting the model parametersLine optimization, through multiple iterations, each iteration generates a weak classifier, each classifier is trained on the basis of the residual error of the last classifier, and the parameter maximum likelihood solution of the loss function of each weak classifier is
Fm-1(xi) For the current model, T (x)i,θm) Linearly overlapping the classifier of each iteration cycle and the newly generated weak classifier for the new weak classifier generated by each iteration cycle, and finally training a load prediction model;
step S50, changing the model into a rolling window prediction model, theoretically predicting the load of any number of days by the model trained in the previous step, adding the rolling prediction window, only using the previous model to predict the load of one day, then adding the new prediction data into a training data set, removing the data of the training data farthest from the current time from the training set, wherein the length of the training set is the length of the rolling window, retraining the model, and predicting the next day, so that after one month of rolling prediction, the prediction data are combined to predict the load of the month;
and step S60, adding a service rule for predicting a special time interval according to the self characteristics of different users, and finely adjusting and optimizing the trained model to finally obtain a monthly load prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811097409.4A CN108846528B (en) | 2018-09-19 | 2018-09-19 | Long-term load prediction method for large-scale industrial power consumer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811097409.4A CN108846528B (en) | 2018-09-19 | 2018-09-19 | Long-term load prediction method for large-scale industrial power consumer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108846528A CN108846528A (en) | 2018-11-20 |
| CN108846528B true CN108846528B (en) | 2021-07-09 |
Family
ID=64187947
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811097409.4A Active CN108846528B (en) | 2018-09-19 | 2018-09-19 | Long-term load prediction method for large-scale industrial power consumer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108846528B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378541A (en) * | 2019-08-08 | 2019-10-25 | 上海交通大学 | A kind of short-term multistep forecasting method of wind power and device |
| CN111582551B (en) * | 2020-04-15 | 2023-12-08 | 中南大学 | Wind power plant short-term wind speed prediction method and system and electronic equipment |
| CN111553516A (en) * | 2020-04-17 | 2020-08-18 | 贵州电网有限责任公司 | Short-term electric quantity load accurate prediction method |
| CN111598302B (en) * | 2020-04-19 | 2023-09-05 | 华电郑州机械设计研究院有限公司 | AP-TS-SVR model-based thermal power plant short-term industrial heat load prediction method |
| CN111950805B (en) * | 2020-08-25 | 2022-09-20 | 华润数字科技有限公司 | Medium-and-long-term power load prediction method and device, computer equipment and storage medium |
| CN112598181A (en) * | 2020-12-23 | 2021-04-02 | 国网河北省电力有限公司经济技术研究院 | Load prediction method, device, equipment and storage medium |
| CN112668765A (en) * | 2020-12-23 | 2021-04-16 | 韶关东阳光科技研发有限公司 | Method and device for load prediction and energy management of power consumers of aluminum processing enterprises |
| CN114399098A (en) * | 2021-12-30 | 2022-04-26 | 昆明能讯科技有限责任公司 | High-applicability and high-precision industrial power utilization user classification peak-shifting power utilization method |
| CN114638425B (en) * | 2022-03-24 | 2023-05-23 | 国网江苏省电力有限公司南通供电分公司 | Historical data-based large-user monthly electricity consumption prediction method and system |
| CN115759236B (en) * | 2022-12-30 | 2024-01-12 | 北京德风新征程科技股份有限公司 | Model training method, information sending method, device, equipment and medium |
| CN117592770B (en) * | 2024-01-19 | 2024-04-12 | 成都秦川物联网科技股份有限公司 | Intelligent gas industry and commercial tenant flowmeter regulation and control method, internet of things system and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
| CN107563539A (en) * | 2017-07-24 | 2018-01-09 | 佛山市顺德区中山大学研究院 | Short-term and long-medium term power load forecasting method based on machine learning model |
| CN108510113A (en) * | 2018-03-21 | 2018-09-07 | 中南大学 | A kind of application of XGBoost in short-term load forecasting |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9620959B2 (en) * | 2013-03-15 | 2017-04-11 | Accenture Global Services Limited | Enhanced grid reliability through predictive analysis and dynamic action for stable power distribution |
-
2018
- 2018-09-19 CN CN201811097409.4A patent/CN108846528B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
| CN107563539A (en) * | 2017-07-24 | 2018-01-09 | 佛山市顺德区中山大学研究院 | Short-term and long-medium term power load forecasting method based on machine learning model |
| CN108510113A (en) * | 2018-03-21 | 2018-09-07 | 中南大学 | A kind of application of XGBoost in short-term load forecasting |
Non-Patent Citations (1)
| Title |
|---|
| 基于梯度提升决策树的电力短期负荷预测模型;毕云帆等;《青岛大学学报》;20180831;第70-75页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108846528A (en) | 2018-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108846528B (en) | Long-term load prediction method for large-scale industrial power consumer | |
| Madhukumar et al. | Regression model-based short-term load forecasting for university campus load | |
| CN111489036A (en) | Resident load prediction method and device based on electrical appliance load characteristics and deep learning | |
| CN111045907A (en) | System capacity prediction method based on traffic | |
| CN103853939A (en) | Combined forecasting method for monthly load of power system based on social economic factor influence | |
| Mi et al. | Low-carbon generation expansion planning considering uncertainty of renewable energy at multi-time scales | |
| CN114154716B (en) | Enterprise energy consumption prediction method and device based on graph neural network | |
| CN115860797A (en) | Electric quantity demand prediction method suitable for new electricity price reform situation | |
| CN111311026A (en) | Runoff nonlinear prediction method considering data characteristics, model and correction | |
| CN114519470A (en) | Method and device for predicting medium-and-long-term power generation amount of wind power | |
| CN105389637A (en) | Effective workday construction method for monthly electricity consumption prediction model | |
| CN114548493A (en) | Method and system for predicting current overload of electric energy meter | |
| CN117236800B (en) | Multi-scene self-adaptive electricity load prediction method and system | |
| Gochhait et al. | Regression model-based short-term load forecasting for load despatch centre | |
| CN114676931B (en) | Electric quantity prediction system based on data center technology | |
| CN116979503A (en) | Commercial building load prediction method and system based on Informir network | |
| CN113449933B (en) | Regional medium-term load prediction method and device based on clustering electric quantity curve decomposition | |
| Parreno | Forecasting the total non-coincidental monthly system peak demand in the Philippines: A comparison of seasonal autoregressive integrated moving average models and artificial neural networks | |
| CN115619447A (en) | Monthly electricity sales combined prediction method, equipment and medium | |
| CN113743994A (en) | Provider's season-busy prediction method, system, equipment and storage medium | |
| Pang et al. | Short-term power load forecasting based on gray relational analysis and support vector machine optimized by artificial bee colony algorithm | |
| Tao et al. | Load forecasting based on short-term correlation clustering | |
| Hou | Short Term Load Forecasting Based on Bayesian Forecasting Model | |
| Davarzani et al. | Study of missing meter data impact on domestic load profiles clustering and characterization | |
| Wang et al. | Research on Electricity Forecasting Method Based on Big Data |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220214 Address after: 650034 No. 73, Tuodong Road, Kunming, Yunnan Patentee after: INFORMATION CENTER OF YUNNAN GRID POWER Co.,Ltd. Address before: 650217 No. 105 Yunda West Road, Kunming Economic and Technological Development Zone, Yunnan Province Patentee before: INFORMATION CENTER OF YUNNAN GRID POWER Co.,Ltd. Patentee before: Kunming Nengxun Technology Co., Ltd |