CN109858674A - Monthly load forecasting method based on XGBoost algorithm - Google Patents
Monthly load forecasting method based on XGBoost algorithm Download PDFInfo
- Publication number
- CN109858674A CN109858674A CN201811612619.2A CN201811612619A CN109858674A CN 109858674 A CN109858674 A CN 109858674A CN 201811612619 A CN201811612619 A CN 201811612619A CN 109858674 A CN109858674 A CN 109858674A
- Authority
- CN
- China
- Prior art keywords
- index
- load
- model
- xgboost
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P80/00—Climate change mitigation technologies for sector-wide applications
- Y02P80/10—Efficient use of energy, e.g. using compressed air or pressurized fluid as energy carrier
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides the monthly load forecasting methods based on XGBoost algorithm, including convert to load index, carry out hot coded treatment to load factor achievement data is influenced；Using the every maximum monthly load data of user as the output of model, select with the stronger influence factor variable of maximum monthly load relevance as input variable；Sparse matrix is converted by the influence load prediction variable factors for being selected into model, forms XGboost modeling data；Maximum monthly load is defined as the output of XGboost model；Definition Model learning objective function, regression tree generate the XGboost model of the construction load prediction such as parameter；Cross validation test is carried out to each parameter of XGboost, obtains the highest parameter combination of model accuracy, load prediction is carried out based on obtained parameter combination.Verified, model result shows average relative error control 5%, and model has preferable effect for following one month peak load data of prediction individual enterprise, and to optimization of enterprises power mode is helped, reducing electric cost has certain guidance and reference.
Description
Technical field
The invention belongs to data to predict field, in particular to based on the monthly load forecasting method of XGBoost algorithm.
Background technique
Load prediction to the safe operation of electric system and the economic benefit important in inhibiting of power grid enterprises, always with
Come, important process of the load forecast of short, medium and long phase as power supply enterprise, all kinds of load forecasting methods and research layer go out
Not poor, classical prediction technique includes time series, neural network, SVM (support vector machines), grey method etc..Mainstream
Load forecast is mainly for provincialism, macroscopical load data of professional, and the load data of individual enterprise is due to being looked forward to
The uncertain factors such as industry production, macro policy influence, and lack periodic regularity, fluctuation is big, and there are biggish model prediction mistakes
Difference.
Summary of the invention
In order to solve shortcoming and defect existing in the prior art, the present invention provides based on the monthly of XGBoost algorithm
Load forecasting method can be improved predictablity rate.
In order to reach above-mentioned technical purpose, the present invention provides the monthly load forecasting method based on XGBoost algorithm, institutes
Stating prediction technique includes:
Load index is converted, carries out hot coded treatment to load factor achievement data is influenced；
Using the every maximum monthly load data of user as the output of model, selection and the stronger influence of maximum monthly load relevance
Variable factors are as input variable；
Sparse matrix is converted by the influence load prediction variable factors for being selected into model, forms XGboost modeling data；It will
Maximum monthly load is defined as the output of XGboost model；Definition Model learning objective function, regression tree generate the construction load such as parameter
The XGboost model of prediction；
Cross validation test is carried out to each parameter of XGboost, adjusting and optimizing model parameter obtains the highest ginseng of model accuracy
Array is closed, and carries out load prediction based on obtained parameter combination.
Optionally, described that load index is converted, hot coded treatment, packet are carried out to load factor achievement data is influenced
It includes:
Load data anomalous identification, replacement, including 3 σ rules are used by user, in monthly, 5 beyond the user's history
± 3 σ of use of ± 3 σ value of peak load is substituted；
Load data, relative influence index missing are filled up, including are filled up using history with time value by user；
Relative influence index is obviously filled up extremely by accidental environmental factor effect tendency, including uses history by user respectively
It is replaced with time value；
Index conversion, including index scatter plot and association analysis significance test are combined, it studies between influence index and load
Relationship, and index is accordingly converted；
Classification indicators processing, including one-hot coding processing is carried out to classified variables such as month, categorys of employment.
Optionally, the influence factor variable includes:
Associated loadings derive index: m phase peak load data A={ a before predicted month1,a2,a3···am}；
Associated weather index: the same period of predicted month temperature, humidity, wind-force, weather pattern etc., last, n weather of forecast
Index B={ b1,b2,b3···bn}；
User basic information index: predicted month user current family age, industry, contract capacity, electricity consumption classification p basic letters
Cease index C={ c1,c2,c3···cp}；
Enterprise's condition of production index: the q condition of production index D=such as predicted month user production plan, the output value, registered capital
{d1,d2,d3···dq}；
Industry overall condition index: r industry index E={ e of predicted month reflection Industrial Cycle degree1,e2,e3···
er}；
Festivals or holidays index F={ f1,f2,f3···fs}: festivals or holidays number of days contained by predicted month, whether the s section such as Spring Festival is false
Day index；
One-hot coding variable G={ g1,g2,g3···gt}: based on the derivative index of associated loadings, associated weather index, use
The classification in totally six class variables of family essential information index, enterprise's condition of production index, industry overall condition index, festivals or holidays index
The one-hot coding variable that variable generates such as the weather pattern in weather index, the category of employment in user basic information index, is used
The t one-hot coding variable index that the classification indicators such as electric classification generate；
Wherein, the value range of m, n, p, q, r, s, t are positive integer.
Technical solution provided by the invention has the benefit that
Verified, model result shows average relative error control 5%, and model is one following for prediction individual enterprise
The peak load data of the moon have preferable effect, and to optimization of enterprises power mode is helped, reducing electric cost has certain guidance
And reference.
Detailed description of the invention
It, below will be to attached drawing needed in embodiment description in order to illustrate more clearly of technical solution of the present invention
It is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow diagram of the monthly load forecasting method provided by the invention based on XGBoost algorithm；
Fig. 2 is the flow diagram of GBDT regression tree growth course provided by the invention；
Fig. 3 is the flow diagram of Xgboost regression tree growth course provided by the invention；
Fig. 4 is peak load actual value and predicted value deviation situation schematic diagram one provided by the invention；
Fig. 5 is peak load actual value and predicted value deviation situation schematic diagram two provided by the invention.
Specific embodiment
To keep structure and advantage of the invention clearer, structure of the invention is made further below in conjunction with attached drawing
Description.
Embodiment one
The present invention provides the monthly load forecasting methods based on XGBoost algorithm, as shown in Figure 1, the prediction technique
Include:
11, load index is converted, carries out hot coded treatment to load factor achievement data is influenced；
12, it is selected stronger with maximum monthly load relevance using the every maximum monthly load data of user as the output of model
Influence factor variable is as input variable；
13, sparse matrix is converted by the influence load prediction variable factors for being selected into model, forms XGboost and models number
According to；Maximum monthly load is defined as the output of XGboost model；Definition Model learning objective function, regression tree generate the structures such as parameter
Make the XGboost model of load prediction；
14, cross validation test is carried out to each parameter of XGboost, adjusting and optimizing model parameter obtains model accuracy highest
Parameter combination, load prediction is carried out based on obtained parameter combination.
In an implementation, this paper presents a kind of application XGBoost (extreme gradient rising) algorithms to carry out big industrial user month
Spend the model of load prediction.By comprehensively considering the factors such as weather, enterprise's condition of production, historical load variation, prediction enterprise month
Load variations trend is spent, provides the judgment basis of science for enterprise's Business Process System.
Xgboost algorithm principle was proposed in 2004 by Chen Tianqi, was on the basis of GBDT to boosting algorithm
It improves, solves GBDT algorithm model and be difficult to parallel computation problem, realize effective control to model overfitting problem.
GBDT is a kind of decision Tree algorithms of iteration, for convenient for solving objective function, GBDT often uses regression tree growth course
The residuals squares that mistake classification generates construct loss function by regression criterion square as loss function.With the life of tree
At loss function constantly declines；Regression tree growth course each split vertexes enumerate all characteristic values when dividing, selection is so that wrong
Misclassification is minimum, loss function declines most fast characteristic value as division points；The study of each regression tree be before all trees
Conclusion and residual error, fitting obtain a current residual error regression tree；Finally, the result for all trees of adding up is as final result.
GBDT regression tree growth course is as shown in Figure 2:
GBDT regression tree is sought objective function optimal solution and is only conveniently acquired to quadratic loss function, loses letter for others
Number becomes very complicated.The selection for determining division node is lost with least square, only accounts for each leaf node precision of prediction of regression tree,
It is pursuing high-precision while model complexity being easily caused to be promoted, the growth of regression tree is caused over-fitting occur.
Xgboost algorithm model improves the above-mentioned two deficiency of GBDT.Xgboost is increased to tree-model complexity
Measurement, consider two factors of loss and model complexity in the selection of regression tree generating process split vertexes, tradeoff mould
After the high complicated complicated high loss low with model of the low loss of type, optimal solution is sought, prevents from pursuing reduction loss function simply generating
Fitting phenomenon, and speed is fast, it is effective Ensemble Learning Algorithms that accuracy is high.Xgboost regression tree growth course such as Fig. 3 institute
Show:
By pushing over to objective function, final objective function can simplify are as follows:
Wherein T is regression tree leaf number；GjFor first derivative of all data on loss function；HjAll data are being damaged
Lose the second dervative on function.That is, objective function only depends on the first derivative and two on error function of each data point
Order derivative, by the transformation of the second Taylor series formula, solution unknown losses function in this way becomes feasible.
XGboost objective function can carry out self-defining according to the object difference of research, particularly may be divided into: be directed to continuous type
Study variable, objective function are as follows: linear regression (" reg:linear ")；Variable is studied for classifying type, objective function can are as follows: patrols
It collects and returns (" reg:logistic ")；Variable, objective function are studied for attribute are as follows: Poisson regression (" count:
poisson”)。
Optionally, described that load index is converted, hot coded treatment, packet are carried out to load factor achievement data is influenced
It includes:
Load data anomalous identification, replacement, including 3 σ rules are used by user, in monthly, 5 beyond the user's history
± 3 σ of use of ± 3 σ value of peak load is substituted；
Load data, relative influence index missing are filled up, including are filled up using history with time value by user；
Relative influence index is obviously filled up extremely by accidental environmental factor effect tendency, including uses history by user respectively
It is replaced with time value；
Index conversion, including index scatter plot and association analysis significance test are combined, it studies between influence index and load
Relationship, and index is accordingly converted；
Classification indicators processing, including one-hot coding processing is carried out to classified variables such as month, categorys of employment.
In an implementation, the every maximum monthly load of user is predicted using XGboost model herein, obtains Zhejiang districts and cities
Nearly in monthly, 5 peak load data of industrial user and influence load factor achievement data are trained modeling as sample greatly.
Since the acquisition of load data is influenced to will affect the accurate of model there may be bad data by equipment and human factor
Property, therefore need to identify and replace load abnormal data before modeling, fill up missing data；Influencing load factor index may be by accidental
The influence of environmental factor need to be identified, be replaced, being rejected, filling up operation；Relevance between relative influence index and load is carried out
There is complicated non-linear relation index between load, need to convert to index in analysis；To classifying type influence factor index into
The processing of row one-hot coding.By the pretreatment to model achievement data, fills up missing, replacement exception, finds association mining index
Between hide rule, it is ensured that integrality, availability, the validity of data, be conducive to improve model accuracy.
Optionally, the influence factor variable includes:
Associated loadings derive index: m phase peak load data A={ a before predicted month1,a2,a3···am}；
Associated weather index: the same period of predicted month temperature, humidity, wind-force, weather pattern etc., last, n weather of forecast
Index B={ b1,b2,b3···an}；
User basic information index: predicted month user current family age, industry, contract capacity, electricity consumption classification p basic letters
Cease index C={ c1,c2,c3···ap}；
Enterprise's condition of production index: the q condition of production index D=such as predicted month user production plan, the output value, registered capital
{d1,d2,d3···aq}；
Industry overall condition index: r industry index E={ e of predicted month reflection Industrial Cycle degree1,e2,e3···
ar}；
Festivals or holidays index F={ f1,f2,f3···as}: festivals or holidays number of days contained by predicted month, whether the s section such as Spring Festival is false
Day index；
One-hot coding variable G={ g1,g2,g3···at}: based on the derivative index of associated loadings, associated weather index, use
The classification in totally six class variables of family essential information index, enterprise's condition of production index, industry overall condition index, festivals or holidays index
The one-hot coding variable that variable generates such as the weather pattern in weather index, the category of employment in user basic information index, is used
The t one-hot coding variable index that the classification indicators such as electric classification generate；
Wherein, the value range of m, n, p, q, r, s, t are positive integer.
It in an implementation, is research variable with the big nearly 5 years every maximum monthly load data in industrial family in somewhere, as the defeated of model
Y out, selecting key index to carry out modeling is that model accurately guarantees.Herein in conjunction with pre-processed results, selection and maximum monthly load
The stronger influence factor variable of relevance is added in model as input variable X.Specific influence factor X, is divided into following several
Class:
(1) the derivative index of associated loadings: m phase peak load data A={ a before predicted month1,a2,a3···am}；
(2) associated weather index: it is the same period of predicted month temperature, humidity, wind-force, weather pattern etc., last, n of forecast
Weather index B={ b1,b2,b3···bn}；
(3) user basic information index: p predicted month user current family age, industry, contract capacity, electricity consumption classification base
This information index C={ c1,c2,c3···cp}；
(4) enterprise's condition of production index: q predicted month user production plan, the output value, registered capital condition of production index D
={ d1,d2,d3···dq}；
(5) industry overall condition index: r industry index E={ e of predicted month reflection Industrial Cycle degree1,e2,
e3···er}；
(6) festivals or holidays index F={ f1,f2,f3···fs}: festivals or holidays number of days contained by predicted month, whether the Spring Festival s section
Holiday index；
(7) one-hot coding variable G={ g1,g2,g3···gt}: it is generated based on the classified variable in above-mentioned several class variables
One-hot coding variable, such as the weather pattern in weather index；Category of employment, electricity consumption classification in user basic information index etc.
The t one-hot coding variable index that classification indicators generate.
Sparse matrix is converted by the influence load prediction variable factors for being selected into model, forms XGboost modeling data；It will
Maximum monthly load is defined as the output of XGboost model；Definition Model learning objective function, regression tree generate the construction load such as parameter
The XGboost model of prediction.XGboost can determine objective function according to Task.This paper Task is that prediction is monthly
Peak load, load value belong to connection attribute variable, and learning tasks are regression forecasting to be carried out to load, therefore linear model may be selected
As objective function.
The parameters such as depth capacity, learning rate, the number of iterations of XGboost regression tree will affect precision of prediction.By right
Each parameter of XGboost carries out cross validation test, and adjusting and optimizing model parameter obtains the highest parameter combination of model accuracy.
Herein using data be nearly 5 years every maximum monthly load data of the big industrial user in somewhere part and its relative influence because
Plain achievement data.Following table is the practical maximum monthly load data of in July, 2018 certain customers.
| User | Days | Maximum monthly load |
| User 1 | 2018-07 | 4250 |
| User 2 | 2018-07 | 879 |
| User 3 | 2018-07 | 2171 |
| User 4 | 2018-07 | 2534 |
| User 5 | 2018-07 | 96 |
| User 6 | 2018-07 | 945 |
| User 7 | 2018-07 | 2438 |
| User 8 | 2018-07 | 1013 |
| User 9 | 2018-07 | 4242 |
| User 10 | 2018-07 | 1063 |
| User 11 | 2018-07 | 185 |
| User 12 | 2018-07 | 449 |
The table 1 certain customers data of in maximum monthly load, 2018
The data modeling of in maximum monthly load, 5 test nearly to user, wherein extract preceding 4 years historical data samples as training
Collection, nearest 1 annual data sample are chosen as test set in conjunction with above-mentioned data prediction and variable, and maximum monthly load prediction is established
Model.
By cross validation, for data used herein in setting regression tree parameter: depth capacity 7, learning rate are
0.3, forecast result of model is best when the number of iterations is 60.Herein using average relative error come assessment models performance.Definition is such as
Under:
Wherein, yiFor history maximum monthly load actual value,For XGboost model predication value, n is test data sample
Number.When parameter is set as optimal combination, average relative error 2.5%.
By taking * * gas Co., Ltd (user 1) as an example, the user contract capacity 6060kVA, category of employment is chemical industry,
In January, 2018 to July user 1 maximum monthly load actual value and predicted value deviation situation, relative error maximum value is 2.3%, such as
Shown in Fig. 4:
By taking * * Chemical Co., Ltd. (user 9) as an example, the user contract capacity 10000kVA, category of employment is chemical industry,
Its in January, 2018 to July user 1 maximum monthly load actual value and predicted value deviation situation, relative error maximum value is 1.8%,
It is as shown in Figure 5:
Monthly load is predicted using XGboost model in terms of above-mentioned user's maximum monthly load prediction deviation rate angle
Effect is preferable, can achieve using standard.
The present invention provides the monthly load forecasting methods based on XGBoost algorithm, including convert to load index,
Hot coded treatment is carried out to load factor achievement data is influenced；Using the every maximum monthly load data of user as the output of model, choosing
It selects with the stronger influence factor variable of maximum monthly load relevance as input variable；To be selected into the influence load prediction of model because
Plain variables transformations are sparse matrix, form XGboost modeling data；Maximum monthly load is defined as the output of XGboost model；It is fixed
Adopted model learning objective function, regression tree generate the XGboost model of the construction load prediction such as parameter；To each parameter of XGboost
Cross validation test is carried out, adjusting and optimizing model parameter obtains the highest parameter combination of model accuracy, based on obtained parameter group
It closes and carries out load prediction.It is modeled and has been predicted for nearly in monthly, 5 load data of Zhejiang districts and cities chemical industry, by reality
Border verifying, model result show average relative error control within 5%, and model will be for one month future of prediction individual enterprise
Peak load data have preferable effect, and to optimization of enterprises power mode is helped, reducing electric cost has certain guidance and borrow
Mirror meaning.
Each serial number in above-described embodiment is for illustration only, the assembling for not representing each component or the elder generation in use process
Sequence afterwards.
The above description is only an embodiment of the present invention, is not intended to limit the invention, all in the spirit and principles in the present invention
Within, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (3)
1. the monthly load forecasting method based on XGBoost algorithm, which is characterized in that the prediction technique includes:
Load index is converted, carries out hot coded treatment to load factor achievement data is influenced；
Using the every maximum monthly load data of user as the output of model, selection and the stronger influence factor of maximum monthly load relevance
Variable is as input variable；
Sparse matrix is converted by the influence load prediction variable factors for being selected into model, forms XGboost modeling data；Most by the moon
Big load is defined as the output of XGboost model；Definition Model learning objective function, regression tree generate the construction load prediction such as parameter
XGboost model；
Cross validation test is carried out to each parameter of XGboost, adjusting and optimizing model parameter obtains the highest parameter group of model accuracy
It closes, load prediction is carried out based on obtained parameter combination.
2. the monthly load forecasting method according to claim 1 based on XGBoost algorithm, which is characterized in that described right
Load index is converted, and carries out hot coded treatment to load factor achievement data is influenced, comprising:
Load data anomalous identification, replacement, including 3 σ rules are used by user, for exceeding the user's history in monthly, 5 maximum
± 3 σ of use of ± 3 σ value of load is substituted；
Load data, relative influence index missing are filled up, including are filled up using history with time value by user；
Relative influence index is obviously filled up extremely by accidental environmental factor effect tendency, including uses the history same period by user respectively
Value replacement；
Index conversion, including index scatter plot and association analysis significance test are combined, relationship between influence index and load is studied,
And index is accordingly converted；
Classification indicators processing, including one-hot coding processing is carried out to classified variables such as month, categorys of employment.
3. the monthly load forecasting method according to claim 1 based on XGBoost algorithm, which is characterized in that the shadow
Ringing variable factors includes:
Associated loadings derive index: m phase peak load data A={ a before predicted month1,a2,a3…am}；
Associated weather index: the same period of predicted month temperature, humidity, wind-force, weather pattern etc., last, n weather index of forecast
B={ b1,b2,b3…bn}；
User basic information index: p predicted month user current family age, industry, contract capacity, electricity consumption classification essential information refer to
Mark C={ c1,c2,c3…cp}；
Enterprise's condition of production index: the q condition of production index D={ d such as predicted month user production plan, the output value, registered capital1,
d2,d3…dq}；
Industry overall condition index: r industry index E={ e of predicted month reflection Industrial Cycle degree1,e2,e3…er}；
Festivals or holidays index F={ f1,f2,f3…fs}: festivals or holidays number of days contained by predicted month, whether the s festivals or holidays index such as Spring Festival；
One-hot coding variable G={ g1,g2,g3…gt}: believed substantially based on the derivative index of associated loadings, associated weather index, user
The classified variable of index, enterprise's condition of production index, industry overall condition index, festivals or holidays index in totally six class variables is ceased to generate
One-hot coding variable, such as the weather pattern in weather index, the category of employment in user basic information index, electricity consumption classification
The t one-hot coding variable index that classification indicators generate；
Wherein, the value range of m, n, p, q, r, s, t are positive integer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811612619.2A CN109858674A (en) | 2018-12-27 | 2018-12-27 | Monthly load forecasting method based on XGBoost algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811612619.2A CN109858674A (en) | 2018-12-27 | 2018-12-27 | Monthly load forecasting method based on XGBoost algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109858674A true CN109858674A (en) | 2019-06-07 |
Family
ID=66892751
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811612619.2A Pending CN109858674A (en) | 2018-12-27 | 2018-12-27 | Monthly load forecasting method based on XGBoost algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858674A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062517A (en) * | 2019-11-21 | 2020-04-24 | 上海航天智慧能源技术有限公司 | GBDT-based LightGBM model cold and heat load prediction method |
| CN111178957A (en) * | 2019-12-23 | 2020-05-19 | 广西电网有限责任公司 | Method for early warning sudden increase of electric quantity of electricity consumption customer |
| CN112163714A (en) * | 2020-10-15 | 2021-01-01 | 国网冀北电力有限公司智能配电网中心 | XGboost-based campus client group load release algorithm |
| CN112614010A (en) * | 2020-12-07 | 2021-04-06 | 国网北京市电力公司 | Load prediction method and device, storage medium and electronic device |
| CN112734340A (en) * | 2021-01-21 | 2021-04-30 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for screening prediction indexes of express delivery quantity |
| CN112766569A (en) * | 2021-01-19 | 2021-05-07 | 国网能源研究院有限公司 | Integrated power consumption increase prediction method and device integrating high-frequency influence factors |
| CN113205207A (en) * | 2021-04-19 | 2021-08-03 | 深圳供电局有限公司 | XGboost algorithm-based short-term power consumption load fluctuation prediction method and system |
| CN113222245A (en) * | 2021-05-11 | 2021-08-06 | 深圳供电局有限公司 | Method and system for checking monthly electric quantity and electricity charge abnormity of residential user and storage medium |
| CN113435663A (en) * | 2021-07-15 | 2021-09-24 | 国网冀北电力有限公司唐山供电公司 | CNN-LSTM combined load prediction method considering electric vehicle charging load influence |
| CN113947201A (en) * | 2021-08-02 | 2022-01-18 | 国家电投集团电站运营技术（北京）有限公司 | Training method and device for power decomposition curve prediction model and storage medium |
| CN115994679A (en) * | 2023-03-24 | 2023-04-21 | 国网山东省电力公司青岛供电公司 | Regional power grid active planning method and system based on load prediction correction |
| CN116760033A (en) * | 2023-08-21 | 2023-09-15 | 南京博网软件科技有限公司 | Real-time power demand prediction system based on artificial intelligence |
-
2018
- 2018-12-27 CN CN201811612619.2A patent/CN109858674A/en active Pending
Non-Patent Citations (3)
| Title |
|---|
| 叶倩怡等: "基于Xgboost的商业销售预测", 《南昌大学学报(理科版)》 * |
| 张宇航 等: "一种基于LSTM神经网络的短期 用电负荷预测方法", 《远光软件》 * |
| 黄达文 等: "基于 XGBoost 算法的用电电量预测的实践应用", 《现代信息科技》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062517A (en) * | 2019-11-21 | 2020-04-24 | 上海航天智慧能源技术有限公司 | GBDT-based LightGBM model cold and heat load prediction method |
| CN111178957B (en) * | 2019-12-23 | 2023-04-14 | 广西电网有限责任公司 | Method for early warning sudden increase of electric quantity of electricity consumption customer |
| CN111178957A (en) * | 2019-12-23 | 2020-05-19 | 广西电网有限责任公司 | Method for early warning sudden increase of electric quantity of electricity consumption customer |
| CN112163714A (en) * | 2020-10-15 | 2021-01-01 | 国网冀北电力有限公司智能配电网中心 | XGboost-based campus client group load release algorithm |
| CN112614010A (en) * | 2020-12-07 | 2021-04-06 | 国网北京市电力公司 | Load prediction method and device, storage medium and electronic device |
| CN112766569A (en) * | 2021-01-19 | 2021-05-07 | 国网能源研究院有限公司 | Integrated power consumption increase prediction method and device integrating high-frequency influence factors |
| CN112734340A (en) * | 2021-01-21 | 2021-04-30 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for screening prediction indexes of express delivery quantity |
| CN112734340B (en) * | 2021-01-21 | 2023-09-01 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for screening prediction index of express delivery quantity |
| CN113205207A (en) * | 2021-04-19 | 2021-08-03 | 深圳供电局有限公司 | XGboost algorithm-based short-term power consumption load fluctuation prediction method and system |
| CN113222245A (en) * | 2021-05-11 | 2021-08-06 | 深圳供电局有限公司 | Method and system for checking monthly electric quantity and electricity charge abnormity of residential user and storage medium |
| CN113435663A (en) * | 2021-07-15 | 2021-09-24 | 国网冀北电力有限公司唐山供电公司 | CNN-LSTM combined load prediction method considering electric vehicle charging load influence |
| CN113947201A (en) * | 2021-08-02 | 2022-01-18 | 国家电投集团电站运营技术（北京）有限公司 | Training method and device for power decomposition curve prediction model and storage medium |
| CN115994679A (en) * | 2023-03-24 | 2023-04-21 | 国网山东省电力公司青岛供电公司 | Regional power grid active planning method and system based on load prediction correction |
| CN115994679B (en) * | 2023-03-24 | 2023-06-09 | 国网山东省电力公司青岛供电公司 | Regional power grid active planning method and system based on load prediction correction |
| CN116760033A (en) * | 2023-08-21 | 2023-09-15 | 南京博网软件科技有限公司 | Real-time power demand prediction system based on artificial intelligence |
| CN116760033B (en) * | 2023-08-21 | 2024-04-12 | 南京博网软件科技有限公司 | Real-time power demand prediction system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109858674A (en) | Monthly load forecasting method based on XGBoost algorithm | |
| CN108320053A (en) | A kind of region electricity demand forecasting method, apparatus and system | |
| Wang et al. | Combined probability density model for medium term load forecasting based on quantile regression and kernel density estimation | |
| Dong et al. | Applying the ensemble artificial neural network-based hybrid data-driven model to daily total load forecasting | |
| CN108171379A (en) | A kind of electro-load forecast method | |
| Shchetinin | Cluster-based energy consumption forecasting in smart grids | |
| Gupta et al. | Clustering-Classification based prediction of stock market future prediction | |
| Lolli et al. | Decision trees for supervised multi-criteria inventory classification | |
| Prasanna et al. | An analysis on stock market prediction using data mining techniques | |
| CN116186548A (en) | Power load prediction model training method and power load prediction method | |
| CN113011680A (en) | Power load prediction method and system | |
| Bajatović et al. | Application of predictive models for natural gas needs-current state and future trends review | |
| Bao-De et al. | Improved genetic algorithm-based research on optimization of least square support vector machines: an application of load forecasting | |
| Sobrino et al. | Forecasting the electricity hourly consumption of residential consumers with smart meters using machine learning algorithms | |
| Gaffar | Prediction of Regional Economic Growth in East Kalimantan using Genetic Algorithm | |
| Puiu et al. | Principled data completion of network constraints for day ahead auctions in power markets | |
| Matijaš et al. | Supplier short term load forecasting using support vector regression and exogenous input | |
| Lasota et al. | Enhancing intelligent property valuation models by merging similar cadastral regions of a municipality | |
| Kim et al. | Case study on the determination of building materials using a support vector machine | |
| Yan et al. | Water demand forecast model of least squares support vector machine based on particle swarm optimization | |
| Jovanović | The selection of optimal data mining method for small-sized hotels | |
| CN112163714B (en) | XGboost-based campus client group load release algorithm | |
| Guachimboza-Davalos et al. | Prediction of monthly electricity consumption by cantons in ecuador through neural networks: a case study | |
| Zhiyuan et al. | Research on the evaluation of enterprise competitiveness based on the wavelet neural network forecasting system | |
| Xiong | Analyzing the Influencing Factors of Economic Fluctuations in the Era of 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 |