CN110222882A - A kind of prediction technique and device of electric system Mid-long Term Load - Google Patents
A kind of prediction technique and device of electric system Mid-long Term Load Download PDFInfo
- Publication number
- CN110222882A CN110222882A CN201910424910.5A CN201910424910A CN110222882A CN 110222882 A CN110222882 A CN 110222882A CN 201910424910 A CN201910424910 A CN 201910424910A CN 110222882 A CN110222882 A CN 110222882A
- Authority
- CN
- China
- Prior art keywords
- mid
- long term
- load data
- predicted value
- load
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 88
- 230000001932 seasonal effect Effects 0.000 claims description 32
- 238000012545 processing Methods 0.000 claims description 19
- 230000007704 transition Effects 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000009499 grossing Methods 0.000 claims description 13
- 238000012937 correction Methods 0.000 claims description 9
- 238000005316 response function Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 5
- 230000001373 regressive effect Effects 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 4
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 9
- 238000013277 forecasting method Methods 0.000 abstract description 6
- 230000008569 process Effects 0.000 abstract description 6
- 230000005611 electricity Effects 0.000 description 48
- 230000007774 longterm Effects 0.000 description 22
- 238000011161 development Methods 0.000 description 15
- 238000004458 analytical method Methods 0.000 description 12
- 230000008859 change Effects 0.000 description 11
- 238000013439 planning Methods 0.000 description 7
- 238000003860 storage Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 241001269238 Data Species 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 239000013589 supplement Substances 0.000 description 5
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 239000000047 product Substances 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 240000002853 Nelumbo nucifera Species 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 2
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 238000004378 air conditioning Methods 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011549 displacement method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 244000086443 Craterellus fallax Species 0.000 description 1
- 235000007926 Craterellus fallax Nutrition 0.000 description 1
- 244000131316 Panax pseudoginseng Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000005713 exacerbation Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000005498 polishing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000005654 stationary process Effects 0.000 description 1
- 230000000638 stimulation Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
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)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides the prediction techniques and device of a kind of electric system Mid-long Term Load, which comprises collects the Mid-long Term Load data in electric system, pre-processes to the Mid-long Term Load data, obtain load data sample；The load data sample is input in gray model GM (1,1), obtains the first predicted value；Compared with first predicted value is made difference with the Mid-long Term Load data, residual sequence is obtained；Wherein, the residual sequence takes absolute value；When first predicted value does not meet preset requirement, the residual sequence is corrected using Markov Chain, obtains the second predicted value of the residual sequence；Judge whether second predicted value meets preset requirement；If meeting, using second predicted value as the prediction result of the Mid-long Term Load data.The present invention realizes a kind of Mid-long term load forecasting method that prediction effect is accurate, reasonable and stable, can overcome existing technical problem.
Description
Technical field
The present invention relates to technical field of power systems, more particularly to a kind of prediction technique of electric system Mid-long Term Load
And device.
Background technique
As China economically continues to develop, demand of the people to electric energy increasingly expands, and user is to power supply reliability
Requirement be also higher and higher.It is by non-renewable energy (such as coal, petroleum, natural gas but since electric energy belongs to secondary energy sources
Deng) convert, which results in electric energy can not mass storage therefore in order to provide power supply reliability, guarantee power supply and electricity consumption
Balance, it is necessary to understand electricity consumption situation, that is, the characteristic of load, rational prediction and the variation tendency for grasping load are
Electric energy scheduling, which provides, accurately to be suggested, to guarantee the reliability of power supply, provides the electric energy of high quality.
Guarantee power system security reliability service, meet user power utilization demand under the premise of, electric power enterprise need to also from from
Economic benefit side and the planning and operation for considering system, obtain preferable economic benefit.Therefore, for the assurance of power load,
The waste of electric power resource He other resources can be avoided as far as possible, and the supply and demand for effectively increasing power supply side and Demand-side is flat
Weighing apparatus.It follows that Load Prediction In Power Systems have important meaning to Electric Power Network Planning, rational dispatching by power grids, power supply and demand balance
Justice.The process of load forecast is extremely complex and cumbersome, and load climate environment, O&M maintenance, festivals or holidays and large user are prominent
The influence of hair event etc. shows certain fluctuation, and daily load curve is fluctuated because of the daily schedule of production and living, and week is negative
Lotus curve working day and nonworkdays load difference are obvious, and Various Seasonal load composition may be very different.Therefore, people is planned
Member can not directly predict the electric load fluctuated at any time.
Basis of the electric energy as national economic development, as the energy largely used, to the production and living of people have to
Close important role.But electric energy has a maximum defect to be not can be carried out a large amount of storage, only passes through water-storage
Mode carries out a small amount of storage.Hydroenergy storage station is to be taken out the water of lower using extra electric energy according in load low peak period
It is stored to eminence, the water stored before when peak times of power consumption generates electricity, and reconciles the electricity consumption feelings of peak period
Condition.But for the energy-accumulating power station that draws water, the size of reservoir is limited, and the construction in this power station is also to receive geographical position
The limitation of equal factors is set, so cannot largely build, so the storage of electric energy is a problem so far.
Electric system must satisfy the dynamic equilibrium of power generation and electricity consumption in the process of running, if such balance is beaten
It is broken, then it can jeopardize the operation of entire electric system.If power load increases, but generated energy does not accordingly increase, then can lead
The frequency of cause system reduces, and perhaps the relevant device of system can break down, lead to mains breakdown, will affect power supply reliability,
It is difficult to ensure power supply quality.So to maintain the balance of power generation with electricity consumption, it is necessary to the characteristic of power system load is fully understood,
Understand the fluctuation situation of load, multiple electricity when peak of power consumption, when low power consumption generates electricity less, storage energy, for load
In the fluctuation of different time, formulate corresponding generation schedule, could make in this way power grid steadily in the long term with the operation of safety.
Load Prediction In Power Systems are combined with society's electricity consumption situation, economy, weather, politics and other factors, to load
Supply and demand development trend carries out reasonable prediction.According to the correlation of the development trend of analysis of history load data influence factor associated therewith
Property, the prediction of science is made to the following electricity consumption.
According to prediction purpose and predicted time range difference, load prediction can be roughly divided into short-term load forecasting, in
Phase load prediction and long term load forecasting.Each type has important role, influences the reliability and economy of electric system.
In smart grid, electricity market is opened to electricity provider and consumer, for instance it can be possible that a company can be used as normally
The company of product treatment electric power, and weigh and consider the balance of demand response.Accurate Mid-long term load forecasting and short term
Aid decision making person is formulated in prediction, which in electricity market, suitably to be planned and to obtain maximum return most important.
Summary of the invention
The present invention provides the prediction technique and device of a kind of electric system Mid-long Term Load, to realize the middle length of electric system
Phase load is predicted.
To solve the above-mentioned problems, the invention discloses a kind of prediction technique of electric system Mid-long Term Load, the sides
Method includes:
The Mid-long Term Load data in electric system are collected, the Mid-long Term Load data are pre-processed, are born
Lotus data sample；
The load data sample is input in gray model GM (1,1), obtains the first predicted value；
Compared with first predicted value is made difference with the Mid-long Term Load data, residual sequence is obtained；Wherein, described residual
Difference sequence takes absolute value；
When first predicted value does not meet preset requirement, the residual sequence is corrected using Markov Chain, is obtained
Second predicted value of the residual sequence；
Judge whether second predicted value meets preset requirement；
If meeting, using second predicted value as the prediction result of the Mid-long Term Load data.
Further, the Mid-long Term Load data in electric system are collected, the Mid-long Term Load data are located in advance
Reason, the step of obtaining load data sample include:
Collect the Mid-long Term Load data in electric system；
First pretreatment is carried out to the Mid-long Term Load data, obtains the first Mid-long Term Load data；
Data smoothing processing is carried out to the first Mid-long Term Load data, obtains load data sample.
Further, data smoothing processing is carried out to the first Mid-long Term Load data, obtains load data sample
Sub-step further comprises:
Choose the first Mid-long Term Load data in multiple years；
The first Mid-long Term Load data are calculated in the average value of the same target time period in the multiple year, Yi Ji
The overall average in all months in the multiple year；
The average value of the same target time period is divided by with the overall average, the season for obtaining the target time period refers to
Number；
The seasonal index number of the average value and the target time period of the same target time period is divided by, when obtaining the target
The load data sample of phase.
Further, the load data sample is input in gray model GM (1,1), obtains the step of the first predicted value
Suddenly include:
The load data sample is input in GM (1,1) model；
In the GM (1,1) model, single order Accumulating generation is carried out to the load data sample, obtains single order load sequence
Column；
It is fitted the changing rule of the single order load sequence using first-order equation, obtains the time of the GM (1,1) model
Receptance function model；
The time response function model is subjected to the reduction of single order regressive, obtains the gray prediction of the load data sample
As a result；
The gray prediction result is subjected to average generation processing, obtains the first predicted value.
Further, before using the Markov Chain amendment residual sequence, which comprises
According to the residual sequence, judge whether first predicted value meets preset requirement；
If meeting, using first predicted value as the prediction result of the Mid-long Term Load data.
Further, before correcting the residual sequence using Markov Chain, the method also includes:
The residual sequence is input in the GM (1,1) model, the gray prediction sequence of the residual sequence is obtained；
The residual sequence and the gray prediction sequence are divided by, the grey degree of fitting index of the residual sequence is obtained；
According to preset state classification condition, the grey degree of fitting index is divided into several state intervals；
According to several described state intervals, the Markov state transition probability square of the gray prediction sequence is constructed
Battle array.
Further, the residual sequence is corrected using Markov Chain, obtains the second predicted value of the residual sequence
The step of include:
According to the Markov state transition probability matrix, obtains the gray prediction sequence and correspond to the state interval
Interval range；
Preset value is taken in the interval range, and the preset value is multiplied with the residual sequence, is obtained described residual
Second predicted value of difference sequence.
Further, the method also includes:
Geneva inspection is carried out to the Markov state transition probability matrix.
Further, which comprises
When judging that second predicted value does not meet preset requirement, second predicted value is added to described medium-term and long-term
In load data, and second predicted value and the Mid-long Term Load data are pre-processed.
To solve the above-mentioned problems, described the invention also discloses a kind of prediction meanss of electric system Mid-long Term Load
Device comprises the following modules:
Data preprocessing module, for collecting the Mid-long Term Load data in electric system, to the Mid-long Term Load number
According to being pre-processed, load data sample is obtained；
Gray model processing module obtains for the load data sample to be input in gray model GM (1,1)
One predicted value；
Residual sequence obtains module, compared with first predicted value is made difference with the Mid-long Term Load data, obtains
Obtain residual sequence；Wherein, the residual sequence takes absolute value；
Markov Chain correction module, for utilizing Markov when first predicted value does not meet preset requirement
Chain corrects the residual sequence, obtains the second predicted value of the residual sequence；
Preset requirement judgment module, for judging whether second predicted value meets preset requirement；
Prediction result correction verification module, it is pre- by described second for when judging that second predicted value meets preset requirement
Prediction result of the measured value as the Mid-long Term Load data.
Compared with prior art, the present invention includes the following advantages:
The present invention applies to grey forecasting model among long-medium term power load forecasting, can reduce Mid-long term load forecasting
Forecast sample is few, the biggish problem of predicted time span；Mid-long Term Load data are pre-processed first before the projection,
The fluctuation of its Mid-long Term Load data is reduced, wherein especially reduced caused by Accumulating generation using seasonal index smoothness
Fluctuation；Then state interval is divided using Markov Chain, the fluctuation tendency of load is established by probability transfer matrix, it will
Grey forecasting model is combined with Markov Chain, is again reduced the fluctuation of Mid-long Term Load data, is finally realized one
Accurate, the reasonable and stable Mid-long term load forecasting method of kind prediction effect, can overcome existing technical problem.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of the prediction technique of electric system Mid-long Term Load of the present invention；
Fig. 2 is the classification situation of Load characteristics index in the prior art；
Fig. 3 is a kind of structural schematic diagram of the prediction meanss of electric system Mid-long Term Load of the present invention；
Fig. 4 .1 is the load distribution map at any time of 2011 in monthly, 2016 electricity datas；
Fig. 4 .2 is the smoothed out load sequence of seasonal index number；
Fig. 4 .3 is that gray model match value and the smoothed out load alignment of seasonal index number scheme；
Fig. 4 .4 is residual sequence figure constructed by the present invention；
Fig. 4 .5 is that the Grey Model value after residual GM is schemed with the smoothed out load alignment of seasonal index number；
Fig. 4 .6 is the predicted value and actual comparison figure before residual GM；
Fig. 4 .7 is the predicted value and actual comparison figure after residual GM；
Fig. 4 .8 is method and other methods prediction result comparison diagram of the invention.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
In the normal operation work of electric system, there are factors that will will affect the variation of electric load, generally comprise
Controllable factor and uncontrollable factor, policy, economic season etc., which all can generate no small influence to power grid, to be influenced, and seems different
There are many connections between factor again.
Wherein major influence factors include:
(1) economic development and the influence of local policy
During national economy rapid growth, people's daily life and all trades and professions electricity consumption demand are also constantly
Soaring, corresponding change can also occur for load.The economy for determining that an area is even national of one economic policy of country
Developing direction can also generate highly important influence to electric system.
(2) influence of power grid electricity price.
Power grid electricity price also results in the variation of load, and power grid can realize the adjusting of load by adjusting electricity price, for example divide
When electricity price setting, tou power price can by reduce peak of power consumption when load, guidance user time-sharing section using electricity, play
The effect of peak load shifting.
(3) influence of meteorological condition and weather.
There are many loads more sensitive for weather and temperature change, such as refrigerator, furnace, sky for electric power demand side
Tune machine etc..These loads are the variations according to weather and change, for example, when time in summer weather is warmmer, especially in China south
A just band, the use time interval of air-conditioning greatly increases in one day, and at this moment power load aggravates；When weather is colder, heating is set
Standby investment increases, and also results in power load exacerbation.
(4) influence of time factor.
Time is also one of factor for influencing load variations, for load prediction, main time factor
Can be divided into following three kinds: two-day weekend with it is workaday intersect, replacement, the appearance of great festivals or holidays of seasonal climate transformation.For
For short term, the variation of load weekly causes short term to change, and working day is different from the load of two-day weekend
, on Monday belong to regular working day to Friday, commercial power biggish for power demand, workload demand amount is larger, in week
Six arrive Zhou Tian, and factory, which has a holiday, causes load to reduce.Great festivals or holidays are often due to the form of expression of business model increases power supply
Perhaps, demand, such as the Spring Festival, International Labour Day, the Dragon Boat Festival, National Day etc. can make electricity change.For term Load,
The replacement in season throughout the year leads to the variation of term Load, and the temperature in summer is higher, and the temperature in winter is lower, and weather is sensitive
Load using more, is become more for the demand of electricity.
(5) enchancement factor etc.
Enchancement factor, such as natural calamity, affect electric system, also result in load and generate unexpected variation.
Perhaps, some natural calamities also will affect the acquisition of load data, and load data is caused abnormal point occur, this number to load prediction
It is had an impact according to acquisition, so as to reduce the accuracy of prediction result.
Load Prediction In Power Systems are combined with society's electricity consumption situation, economy, weather, politics and other factors, to load
Supply and demand development trend carries out reasonable prediction.According to the correlation of the development trend of analysis of history load data influence factor associated therewith
Property, the prediction of science is made to the following electricity consumption.It, can be by load prediction point according to the difference of prediction purpose and predicted time range
For three classes:
Short-term load forecasting (STLF), usually 1 hour to 1 week, short-term load forecasting for generating control in real time, safety
Analysis and power exchange plan etc. operate extremely important.Compared to Mid-long term load forecasting, carrying out short-term load forecasting is in order to short
When control scheduling various regions power supply volume, on the one hand can help to formulate the fuel of generator in a short time for power supply side and supply
Plan, improves economy and reasonability；On the other hand facilitate power grid and formulate interruption maintenance plan.It can be also used for pacifying in real time
Complete analysis, Real-time Economic Dispatch etc..The characteristics of short-term load forecasting is to predict day and a few days ago with the rule of the transient change of period
Restrain it is more similar, different maturity periods it is interior have similar rule.The influence factor of short-term load forecasting mainly has week type, electricity
Valence, meteorologic factor etc..
Medium term load forecasting (MTLF), a usually Zhou Zhiyi, medium term load forecasting with from a few days to a few weeks or several
The period of the moon is related, and for meeting the load requirement in summer or winter peak period.The purpose of medium term load forecasting be for
The interruption maintenance plan of route and substation is formulated, control centre adjusts the method for operation etc. of electric system.Term Load is pre-
The load variations rule that is mainly characterized by surveyed has periodically, for example, the load variations trend in observation several years, can obtain and exist every year
The load trend in identical month has similitude.
It is more than 1 year that long term load forecasting (LTLF), which is often referred to the time,.Long term load forecasting is mainly used for Electric Power Network Planning, is
Power supply expands the macro-level policy-makings such as capacity, network structure upgrading, system scale expansion and provides strong foundation.Long term load forecasting
Feature is that data are substantially to be monotonically changed, and generally shows incremental, aperiodicity.The influence factor of long-term phase load prediction
Mainly there are national economic development situation, the variation of population, the adjustment of the industrial structure, variation of electrovalence policy etc..
Above-mentioned short-term, medium and long term load prediction, each type have important role, influence the reliable of electric system
Property and economy.In smart grid, electricity market is opened to electricity provider and consumer, for instance it can be possible that a company can
Using the company as normal product processing electric power, and weigh and consider the balance of demand response.Accurate medium term load forecasting and
Short-term load forecasting, which formulates aid decision making person in electricity market, suitably to be planned and to obtain maximum return most important.
It should be noted that medium term load forecasting is concentrated mainly on 1-5 years, the prediction generally as unit of year is main
Act on mid-term balance of electric power and ener, power supply and substation's constant volume addressing, Net Frame of Electric Network planning etc..Long term load forecasting refers generally to
Even longer-term, main function is to plan that electric system distant view develops within 10-20 years.The master of Mid-long term load forecasting
Feature is wanted to have:
(1) forecast sample is few: the time major part of Electric Power Network Planning load prediction was usually no more than 15 years in 10-15 years,
Shorter is possibly less than 10 years, this leads to have limitation in selection main reasons is that complete load sample negligible amounts.Its
His reason mainly includes that historical statistics means are insufficient, statistical error problem is larger and economic society transition problem, available
Sample size is also limited.Based on the few feature of load prediction sample, should be avoided as far as possible when carrying out the selection of load forecasting model
To the demanding model of sample size, and select efficiently excavate the model for containing effective information in small sample.
(2) predicted time span is larger: in Mid-long term load forecasting, the range of time span at least more than 1 year, this
Challenge is proposed to prediction precision.Impact factor is broadly divided into short-term factor and long term factor, and short-term factor is main
Comprising temperature change, rainfall situation, humidity condition etc., long term factor has financial crisis, industry restructuring, regional industrial to turn
Shifting, high energy-consuming industry development, income level of resident, electrovalence policy, subsidy policy stimulation etc., short-term factor to Mid-long Term Load not
Excessive influence can be generated, and long term factor is affected to Mid-long Term Load variation.It follows that economic development is in middle length
Phase load forecast has apparent influence, therefore should focus on collecting the related data information of economic development when being predicted,
The prediction model for establishing multivariable improves the precision of prediction of load prediction.
(3) demand of the prediction to section is more urgent: Mid-long term load forecasting mainly macro-plan and in long-term electric power
Obvious effect is planned, policymaker is desirable with effective information, obtains effectively as a result, providing the upper limit of predicted load under
Limit provides reasonable suggest to planning.
(4) Electric Power Network Planning, traffic control experience accurately play an important role to long term load forecasting.Mid-long term load forecasting has
The features such as time span is big, sample size is few, influence factor is more, therefore work as and possess management and running work expertise pair abundant
Its precision predicted has a preferable supplement, therefore more focuses on qualitative subjective analysis in system modelling and have with objective quantitative
Machine combines, and enhances the accuracy, reasonability and stability of forecast result of model.
Currently, experts and scholars both domestic and external propose the method for many load predictions, most of Predicting Techniques are negative
It carries prediction aspect to attempt, achieves different degrees of success.Since the time of load prediction divides the difference of classification, may lead
Cause identical method that may generate different results to different types of load prediction.These especially popular technologies in recent years
It include: to have econometrics model and statistics for long term load forecasting.There is neural network for short-term load forecasting,
Fuzzy logic, expert system, homing method, time series, similar day method and support vector machines.Due to influencing medium term load forecasting
Factor between short-term load forecasting and long term load forecasting, therefore be wherein suitable for it is long-term and short-term load forecasting perhaps
Multi-method can be applied to medium term load forecasting.
Load is divided into several types according to the length of time, for Short Term load Forecasting Technique, short term
The grasped data information of prediction is compared with horn of plenty, and the variation of load in a short time will not be particularly big, and influence factor does not have
Mid-long term load forecasting is so much, therefore the research of current short-term load forecasting method is more more mature, and artificial intelligence
With the technological innovation in terms of information processing, short-term load forecasting method has been got back further supplement and improvement.Compared to short
For phase load prediction, the span time of Mid-long term load forecasting is big, and influence factor is more and uncontrollable, and data information is collected relatively tired
It is more difficult will to carry out in this respect accurate quantitative study for difficulty.
In view of the above technical problems, referring to Fig.1, a kind of prediction technique of electric system Mid-long Term Load of the present invention is shown
Step flow chart, which comprises
Step S101: the Mid-long Term Load data in electric system are collected, the Mid-long Term Load data are located in advance
Reason, obtains load data sample；
Characteristics of Electric Load index is the quantity performance of part throttle characteristics, that is, the characteristic value of load variations.In last generation
" index explanation of power industry production statistics " file that Ministry of Energy, discipline the eighties China issues is proposed about part throttle characteristics
14 indexs, Guo Wang company has carried out addition supplement to it after 21st century, wherein most importantly increasing one
New index --- peak-valley ratio.The calculating of Load characteristics index and analysis are non-for the load variations characteristic for describing electric system
It is often necessary.By to part throttle characteristics carry out analysis obtain peak load using the development trend of hour be part throttle characteristics prediction base
This content.Load curve characteristics index is to carry out load Analysis, grasp rule and characteristic that load changes over time, is primarily determined
The method of operation in all kinds of power plant power stations and the parameter of installed capacity.
The index of Load characteristics index has very much, both at home and abroad also not no index system of unified specification, considers practicability,
Main loads characteristic index is divided into a year Load characteristics index, moon Load characteristics index and daily load characteristic index according to the time period.
The selection of Load characteristics index can largely influence the quality of load Analysis, and the method for load prediction also can be directly to pre-
Result is surveyed to have an impact.Load characteristics index can be divided into description class substantially according to content, compare class and class of a curve, referring to figure
2, show the classification situation of Load characteristics index in the prior art.As can be seen from Figure 2, description class mainly includes maximum load, most
Underload, average load, peak-valley difference.Comparing class mainly includes daily load rate, average daily load rate, ratio of minimum load to maximum load, peak-valley difference
Rate, maximum load utilize hourage etc..Class of a curve includes daily load curve, monthly load curve, yearly load curve, year continuous loading
Curve.
Mid-long Term Load data collected by the present invention refer to above-mentioned Load characteristics index.As the basis of load prediction,
Historical data will directly influence the reliability of prediction result, and load prediction is that basis carries out analysis expansion to historical data.?
It is general using SCADA system acquisition initial data (the Mid-long Term Load data i.e. in the present invention) in electric system.
In actual motion, the links of data collection system may be because of some external factors such as: communication meets with
To interference, losing occur in data in data transmission procedure, may result in the missing and exception of data acquisition.In addition, having
When some special event cutting loads have a power failure, line maintenance has a power failure, and large user, major issue impact etc. can cause load data to be dashed forward
Become.As lacked, being mutated these abnormal data, it is pre- that these exceptional values are directly applied into load without being pocessed if directly
Among survey, it will so that the accounting of load prediction technology and these abnormal datas, then, when being fitted emulation, perhaps
Normal match value can be deviateed because of these exceptional values, obtain less appropriate predicted value, cause forecasting accuracy lower.Cause
This, the present invention pre-processes initial data, to guarantee the foundation of correctly predicted model and the raising of precision of prediction.
In a preferred embodiment, the Mid-long Term Load data collected in electric system are proposed, in described
The step of long-term load data are pre-processed, and load data sample is obtained include:
Sub-step 1: the Mid-long Term Load data in electric system are collected；
Sub-step 2: the first pretreatment is carried out to the Mid-long Term Load data, obtains the first Mid-long Term Load data；
Sub-step 3: data smoothing processing is carried out to the first Mid-long Term Load data, obtains load data sample.
The embodiment of the present invention carries out the first pretreatment to collected Mid-long Term Load data first, then again to pretreatment
Data afterwards carry out data smoothing processing and remove some improper datas or special data by handling twice, reduce data
Fluctuation, to improve precision of prediction to a certain extent.
Data across comparison method, data longitudinal direction method of comparison, curve displacement method, probability statistics can be used in above-mentioned first pretreatment
One of method and the supplement of missing data or any several method, in which:
(1) data across comparison method
The across comparison method of data is using the load value of adjacent moment as reference, by the load value and ginseng of specified time point
Examine value to be compared, if the difference of the two has been more than a certain threshold value, then it is assumed that the load value at the moment be exceptional value, need to its into
Row repairing treatment.
(2) data longitudinal direction method of comparison
Electric load has certain periodicity, therefore the data variation of front and back mutually in the same time is little, they have certain
Similitude, if the difference of a certain time data and front and back load value in the same time has been more than some threshold value, then it is assumed that the data are
Bad data, using the load average value of front and back in the same time as the correction value at the moment.
(3) curve displacement method
Obvious abnormal and major break down day the load of load curve is rejected or replaced.
(4) probabilistic method
The primary election and amendment for rule of thumb carrying out Power system load data first, then further according to needing to be arranged appropriate confidence
Section reaches the identification and amendment to bad data in data acquired by this two step.
(5) supplement of missing data
Containing scarce in the data that the problems such as having a power failure as caused by information acquisition system reliability or special circumstances occurs
The case where losing data, the processing to this kind of data is that several days load datas in front and back are taken mean value to fill up.
Current data smoothing processing method includes " the simple rolling average of 2n+1 point " smothing filtering method, " weighting is mobile flat
" exponential smoothing, seasonal index smoothness etc..
In practical operation power, the production activity of people, the continuous variation of life style and meteorological condition, electric power
Supply can change with these variation, and this variation has diversity and complexity.Monthly load is come
It says, changing rule can be unfolded in terms of annual developmental sequence and monthly developmental sequence two.On the one hand, by the sequence of year development
On from the point of view of, with the development of social economy, the living standard of the people significantly improved, and continuous increasing occurs in the monthly load of electric system
Long situation；On the other hand, in each moon load in 1 year, with certain fluctuation tendency for being showed of variation in season.
Attribute annual period of load mainly has the following aspects:
The growth property of trend, two continuous time loads have same distribution character；
The fluctuation in season, general area one, February load level it is relatively low under, but to seven Augusts of summer
Its load level of part, which just has, to be obviously improved；
There is low ebb in the attribute of festivals or holidays, the load when general trend of usual regional load is grand festivals or holidays,
His period is in higher horizontal fluctuating change up and down.
The moon cyclic attributes of load are mainly:
A month the inside, load was substantially being changed using week as the period.It is workaday in a week
Load level is higher than the load level of two-day weekend.This is mainly due to the compositions of load and people's rule of life habit to be determined
's.The main constituents for the daily load that works are industrial load, but the electricity consumption proportion in two-day weekend resident living can be bright
Aobvious to rise, because of the specific gravity that network load shared by industrial load is very big, two-day weekend load should be able to be significantly lower than working day
Load.In addition to this, because industrial load on weekdays during generally lie in stable operating, workaday load
Variation has similitude.It is so analyzed from attribute annual period of load, similar load distribution performance can be to monthly load
Prediction provides accurate power load distributing trend.It is analyzed by the moon cyclic attributes of load, the prediction for the daily load that works will conduct
Influence the mostly important reason of the accuracy of the prediction modeling of monthly load.
Patentee considers the fluctuation problem of monthly load, in the present invention preferably using seasonal index smoothness to institute
It states the first Mid-long Term Load data and carries out data smoothing processing, to obtain accurate load data sample.
In a preferred embodiment, it shows according to seasonal index smoothness, to first Mid-long Term Load
Data carry out data smoothing processing, and the sub-step for obtaining load data sample further comprises:
Choose the first Mid-long Term Load data in multiple years；
The first Mid-long Term Load data are calculated in the average value of the same target time period in the multiple year, Yi Ji
The overall average in all months in the multiple year；
The average value of the same target time period is divided by with the overall average, the season for obtaining the target time period refers to
Number；
The seasonal index number of the average value and the target time period of the same target time period is divided by, when obtaining the target
The load data sample of phase.
After obtaining the load data sample, need to predict the load data sample.Current prediction technique
It is broadly divided into traditional prediction method and modern prediction technique.
Traditional prediction method is to build high-precision data model by collecting research data, and basic theories is derived from
The mathematical theories such as Probability Theory and Math Statistics.Such as: time series method, regression analysis and gray forecast approach prediction technique etc..
Modern prediction technique includes intelligent algorithm, fuzzy prediction method, wavelet analysis method and combinatorial forecast etc..Its
In, intelligent algorithm includes expert system, artificial neural network method etc..The method of load prediction is more, but each side
Method has the advantage and disadvantage and the scope of application of oneself, and inventor compares several common prediction techniques, and comparison result is such as
Shown in table 1.
Table 1
As known from Table 1, gray forecast approach prediction technique has that algorithm is simple, method is mature, calculation amount is small, speed is fast, sample
The advantages that requirement is small, and characteristic quantity need not be considered in modeling, the index prediction of nonlinear change is applied also for, and it is medium-term and long-term
The Property comparison of load data is close, so the present invention uses in gray forecast approach prediction technique as prediction side of the invention
Method, the prediction model with gray model GM (1,1) as load data sample.
Step S102: the load data sample is input in gray model GM (1,1), obtains the first predicted value；
In a preferred embodiment, it shows and the load data sample is input to gray model GM (1,1)
In, the step of obtaining the first predicted value includes:
Sub-step 4: the load data sample is input in GM (1,1) model；
Sub-step 5: in the GM (1,1) model, single order Accumulating generation is carried out to the load data sample, obtains one
Rank load sequence；
Sub-step 6: being fitted the changing rule of the single order load sequence using first-order equation, obtains the GM (1,1) mould
The time response function model of type；
Sub-step 7: the time response function model is subjected to the reduction of single order regressive, obtains the load data sample
Gray prediction result；
Sub-step 8: the gray prediction result is subjected to average generation processing, obtains the first predicted value.
It in embodiments of the present invention, can be by separation and war, irregular by carrying out single order Accumulating generation to load data sample
Over-and-over addition is generated new cumulative data sequence by initial data, after accumulated generation, so that it becomes having exponential increase rule
The ascending series of rule.
The specific implementation step of sub-step 6 includes:
The changing rule of the single order load sequence is fitted using first-order equation；
Use the undetermined parameter of single order load sequence described in least square solution；
The undetermined parameter is substituted into the first-order equation, obtains the time response function model of the GM (1,1) model.
Above-mentioned first-order equation, least square method and undetermined parameter belong to the common algorithms and data of Grey Model method, herein
Seldom repeat.
Next, by time response function model progress single order regressive reduction, at the result after Accumulating generation prediction
Reason, so that ascending series are reduced to normal ordered series of numbers.
Sub-step 8 is using average generation processing method to carry out school to the gray prediction result by the reduction of single order regressive
Just, the first predicted value of final gray model is obtained.
Above-mentioned average generation includes adjacent average generation, non-neighboring average generation.When adjacent average generation has equal ordered series of numbers
Away from the average value of adjacent two data, can subtract by the data handled in this way under using even time interval when generating new data
It is irregular under weak anomaly and emergency case, while can effective polishing defect value；With adjacent average generation on the contrary, non-neighboring average generation
Ordered series of numbers when away from be it is unequal, generally use and delete the exceptional value that occurs in ordered series of numbers, contain to be formed in entire ordered series of numbers
Vacancy so that ordered series of numbers when away from equal, after abnormal data is deleted, deleted data are often asked using the data on both sides
Average value replaces, thus the position that fills up the vacancy.
Step S103: compared with first predicted value is made difference with the Mid-long Term Load data, residual sequence is obtained；Its
In, the residual sequence takes absolute value；
Since load prediction work is the judgement to load future development rule, to predict following certain periods
Load value and load variations rule.Some uncontrollable factors are often accompanied by the prediction of load, these factors may cause to pre-
The reliability and accuracy for surveying result have an impact.If there is no selection suitable appropriate prediction model when being predicted, it is this
The accident changed at random all can make predicted value that can have a certain difference with actual value.Therefore, in step s 103, of the invention
Compared with first predicted value is made difference with the Mid-long Term Load data, school is carried out to the prediction result of above-mentioned prediction technique
It tests, to evaluate the precision of above-mentioned grey forecasting model.Since residual sequence has positive value also to have negative value, and gray model requires data
Sequence symbol is consistent, cannot directly use Grey Model, so making it meet ash herein by the absolute value of residual sequence
The requirement of color model.
Specific steps are as follows:
It is assumed that the sequence of history Mid-long Term Load data and the first predicted value is respectively X(0)(k)、Wherein, k=
At the time of 1,2 ..., n indicate record.ε (k) is represented by the residual error at k moment, i.e. actual value X(0)(k) and predicted valueIt
Between difference.The average value of residual epsilon (k)It may be expressed as:
The wherein average value of history Mid-long Term Load dataAre as follows:
History Mid-long Term Load data varianceExpression formula are as follows:
Residual varianceExpression formula are as follows:
Posteriority difference ratio C are as follows:
Small error possibility P are as follows:
Equation left side two indexes C, P value range in formula (5) and (6) are as follows: C > 0,0≤P≤1.C index value is smaller, indicates
The precision of predicted value is higher.When the residual error fitting degree that history Mid-long Term Load data are more discrete, and predicted value and actual value generate
Higher, C is smaller.When P is bigger, the difference of residual error and residual error average value is less than 0.6745S1Point it is more, thus, it can be known that with index C
On the contrary, the bigger precision of prediction of P is higher.According to above-mentioned two index parameter, establishes such as table 2 and evaluate obtained by grey forecasting model
Predicted value precision.
Table 2
| Precision of prediction grade | P | C | Precision of prediction grade | P | C |
| Good (level-one) | >0.95 | <0.35 | (three-level) reluctantly | >0.7 | <0.45 |
| Qualified (second level) | >0.8 | <0.5 | Unqualified (level Four) | ≥0.7 | ≥0.65 |
Based on precision evaluation standard described in table 2, in a preferred embodiment, showing evaluation method includes:
According to the residual sequence, judge whether first predicted value meets preset requirement；
If meeting, using first predicted value as the prediction result of the Mid-long Term Load data.
Above-mentioned preset requirement refers to using after residual sequence progress P, C metrics evaluation, whether judges the first predicted value
It is qualified.
Grey GM (1,1) model can simulate the basic trend of load sample data, and obtain predicted value, but such as
Fruit only uses grey forecasting model only come if being modeled, the prediction result that cannot be more satisfied with, trace it to its cause for,
Gray model cannot simulate the random information occurred in load sequence well.So the first predicted value be likely to occur it is qualified with not
Qualified two kinds of situations.
When the first predicted value qualification, directly using first predicted value as the prediction knot of the Mid-long Term Load data
Fruit.
Step S104: when first predicted value does not meet preset requirement, the residual error is corrected using Markov Chain
Sequence obtains the second predicted value of the residual sequence.
But when the first predicted value is unqualified, illustrate that there are certain precision problem, inventors by this gray model GM (1,1)
The residual sequence is modified using Markov Chain, the prediction that gray model GM (1,1) is obtained can be further decreased
The fluctuation of value.
Markov Chain mentioned by the present invention is the method for describing Stochastic Dynamic Process, it refers to system in each time
State in which be it is random, and unrelated with state before, that is, have the characteristics that markov property.Markov chain model is logical
It crosses and determines the current status of variable to predict the development tendency in following a period of time.If the transfer of system generating state is
Refer to the case where system variable is transferred to the value range of another state from the value range of a state.
Before formally being corrected using the Markov Chain amendment residual sequence, need gray model and Markov Chain
Be combined together, the method specifically includes the following steps:
The residual sequence is input in the GM (1,1) model, the gray prediction sequence of the residual sequence is obtained；
The residual sequence and the gray prediction sequence are divided by, the grey degree of fitting index of the residual sequence is obtained；
According to preset state classification condition, the grey degree of fitting index is divided into several state intervals；
According to several described state intervals, the Markov state transition probability square of the gray prediction sequence is constructed
Battle array.
In a preferred embodiment, before correcting the residual sequence using Markov Chain, also to the horse
Er Kefu state transition probability matrix carries out geneva inspection.
The implementation method of step S104 includes:
According to the Markov state transition probability matrix, obtains the gray prediction sequence and correspond to the state interval
Interval range；
Preset value is taken in the interval range, and the preset value is multiplied with the residual sequence, is obtained described residual
Second predicted value of difference sequence.
Above-mentioned steps are illustrated using formula below.
By obtaining a residual sequence after GM (1,1) model prediction, q can be denoted as(0)(k), it may be assumed that
Wherein x(0)(k) true Mid-long Term Load data are represented,Represent the predicted value of GM (1,1) model.
After the absolute value of residual sequence, then residual sequence ε(0)It can indicate are as follows:
ε(0)=(ε(0)(1),ε(0)(2),ε(0)(3),…,ε(0)(n)) (8)
By residual sequence ε(0)Be input in grey GM (1,1) model and predicted, establish about residual sequence GM (1,
1) model obtains the gray prediction sequence of residual sequence:
Parameter thereinWithIt can be obtained by least square method.
Calculate the grey degree of fitting index of residual sequence:
The gray model matched curve of residual sequence is rendered as the characteristic of index, and grey degree of fitting index Y (k) is mainly used
Reflect that original data surround the degree of fluctuation of matched curve, that is, reflects the degree of fluctuation of residual sequence.Due to residual
The variation of difference sequence shows a kind of the characteristics of non-stationary randomness, it is possible to be corrected by markovian application residual
The result of the prediction of GM (1,1) model of difference sequence.The basic model of Markov Model about Forecasting method are as follows:
X (k+1)=x (k) × P (11)
Wherein, x (k) indicates that quantity of state of the prediction object at the t=k moment, x (k+1) indicate prediction object in t=k+1
The quantity of state at quarter, P indicate the step transition probability matrix in Markov Chain.
1, the grey degree of fitting index of residual sequence is divided
Since grey fitting accuracy index is the non-stationary process of a random fluctuation, thus need to consider a state classification
Method, so that Y (k) is divided into multiple states.According to the size of Y (k), state interval can be divided:
Ei=[Yi1,Yi2], i=1,2 ..., k (12)
Wherein, Yi1That indicate is the lower limit of Y (k), Yi2What is indicated is the upper limit of Y (k).Under normal circumstances, the shape of division
State is more, and the precision of model is higher.
2, structural regime transition probability matrix
The prediction result of residual sequence is divided into several states:
Pij=nij(k)/ni (13)
Above formula indicates system by state EiBy k step to state EjThe number of appearance is nij(k), EiThe number of appearance is ni。
Then state transition probability matrix are as follows:
3, geneva is examined
It is (n-1) that statistic, which obeys freedom degree,2Distribution.Selected confidence level is α, can be by tabling look-up to obtain χa 2((n-1)2),
It enables:
IfThen it is with geneva.
4, the amendment of predicted value
It, can be with the corresponding section of grey degree of fitting index of prediction residual sequence by Markov state transition probability matrix
Range [Yi1,Yi2], preset value is then taken in the interval range, and the preset value is multiplied with the residual sequence, obtain
Obtain the second predicted value of the residual sequence.It is preferred that take state interval intermediate value and raw residual sequence product as residual error sequence
The correction value of column predicted value, i.e. the second predicted value.
Step S105: judge whether second predicted value meets preset requirement；
Step S106: if meeting, using second predicted value as the prediction result of the Mid-long Term Load data.
The verification method of the first predicted value of above-mentioned preset requirement reference, not described here any more.It, will when meeting preset requirement
Prediction result of second predicted value as the Mid-long Term Load data.
In addition, when second predicted value does not meet preset requirement, which comprises
When judging that second predicted value does not meet preset requirement, second predicted value is added to described medium-term and long-term
In load data, and second predicted value and the Mid-long Term Load data are pre-processed.
After second predicted value is added in the Mid-long Term Load data, subsequent processing steps are referring to centering of the present invention
The processing step of long-term load data, not described here any more.
To sum up, grey forecasting model has the characteristics that can handle " small sample ", " poor information ", and the present invention is based on its characteristics
Applied among long-medium term power load forecasting, the forecast sample to reduce Mid-long term load forecasting is few, predicted time across
Spend biggish problem.But for the load with fluctuation, the prediction of gray model it is ineffective, trace it to its cause for
Since gray model is the characteristics of input data is made accumulation process, makes it have exponential increase, once the sample data of input
When fluctuating larger, the precision of prediction will be reduced.To solve the problems, such as data fluctuations caused by Accumulating generation, centering of the present invention is long
Phase load data is pre-processed, and can be reduced to the fluctuation of Mid-long Term Load data, wherein especially utilizes seasonal index number
Exponential smoothing reduces fluctuation caused by Accumulating generation, and then by treated, sequence is predicted again, is finally gone back prediction result
Original, the prediction result which is replaced.In addition, occurring since grey forecasting model cannot simulate well in load sequence
Random information, and demand of the Mid-long term load forecasting to section is more urgent at present, the present invention using Markov Chain come
State interval is divided, the fluctuation tendency of load is established by probability transfer matrix, by grey forecasting model and Markov Chain phase
In conjunction with, the fluctuation of Mid-long Term Load data can be reduced again, finally realize a kind of prediction effect it is accurate, it is reasonable and stable in
Long term load forecasting method, to overcome existing technical problem.
Next, technical solution of the present invention is described in detail using a specific example.
Zhangye Prefecture's in January, 2011 in December, 2016 is chosen, 72 totally months Analyzing Total Electricity Consumption data are as original
Data are emulated.Specific initial data is shown in Table shown in 3.1, and the load of 2011 in monthly, 2016 electricity datas is at any time
Distribution map is as shown in Fig. 4 .1.
Table 3.1
The monthly electricity data of analysis 2011 to 2016 it can be concluded that, the song of the data of Zhangye Prefecture and Non-smooth surface
Line has trend growth property and cyclic swing.For monthly electricity, with the replacement in season, people's lives habit
Change, can all possess identical trend every year, for water power, since seasonal law has the area of dry season and wet season
Not；For people's household electricity, hot and winter the cold in summer leads to the use of the equipment of the adjusting temperature such as air-conditioning
Increase, these phenomenons are but also the characteristic of the load of every month is different.It can use gray model and solve the problems, such as trend growth,
The fluctuation pattern of load is solved using Markov Chain.
It is modeled using the monthly electricity data in Zhangye Prefecture in January, 2011 in December, 2016 as initial data,
Using the monthly electricity data in January, 2017 to December as test data, modeled using MATLAB R2016a platform.
It, can be by grey with two kinds of trend of fluctuation and trend growth property from Fig. 4 .1 it is found that for monthly
Model predicts its trend growth property, if but directly original loads data be input among gray model predict, ash
Color model is, then vulnerable to the influence of fluctuation, to cause prediction result undesirable with exponential growth.So carrying out grey
It before the prediction of model, needs to carry out initial data the smoothing processing of data, the fluctuation of initial data is reduced, then will place
Data after reason, which are input in gray model, to be predicted.
(1) 2011 to 2016 corresponding average value of each month is calculated first；
aiIndicate i-th month corresponding average value, bijRepresent the electricity data of the i month in jth year.
(2) the average electricity consumption in all months, i.e. overall average electricity consumption are calculated；
(3) electricity consumption that is averaged of each phase divided by overall average electricity consumption is obtained into the seasonal index number of each phase；
(4) it is modified monthly to be obtained into seasonal index number divided by the seasonal index number in corresponding month for monthly electricity consumption data sequence
Electricity consumption sequence.
2011 to 2016 seasonal index numbers calculated according to formula (1.1)~(1.3) are as shown in table 3.2.Table 3.2
For 2011 in monthly, the 2016 electricity data seasonal index numbers in Zhangye Prefecture.
Table 3.2
Same month sequence of average as shown in table 3.2, in data smoothing treatment process are as follows:
A=[16.458 14.225 14.667 15.378 15.339 16.328 ... 16.815 18.253
17.169 15.973 16.774 16.634]
Overall average electricity consumption is 16.168 hundred million KWh；Seasonal index number sequence are as follows:
ρ=[1.018 0.880 0.907 0.951 0.949 1.010 ... 1.040 1.129 1.062 0.988
1.038 1.029]
Through the value of revised 2011 in monthly, 2016 electricity of seasonal index number as shown in table 3.3, seasonal index number
Smoothed out load sequence is as shown in Fig. 4 .2.Table 3.3 is revised 2011 in monthly, 2016 electricity of seasonal index number
Value.
Table 3.3
From Fig. 4 .2 it is found that the smoothed out load sequence of seasonal index number compared with the load sequence before amendment for, fluctuation does not have
So obvious, only some tiny fluctuations are whole to have the trend increased upwards, can be by the smoothed out load sequence of seasonal index number
Column, which are input in gray model, to be predicted.
Establish the gray model of seasonal index number Orders Corrected:
It can be obtained by MATLAB platform emulation, parameter a and parameter u are as follows: a=-0.0027, u=14.671
A and u can be substituting in the time response function model of the GM (1,1) model, seasonal index number Orders Corrected
Gray model are as follows:
Using the gray model of foundation, in January, 2011 in December, 2016 is simulated, the value of fitting is 3.4 institute of table
Show, table 3.4 is 2011 to 2016 fitting results.The match value and the smoothed out load sequence pair of seasonal index number of gray model
Than scheming as shown in Fig. 4 .3.
Table 3.4
Utilize the monthly electricity in January, 2017 to December obtained in the smoothed out sequence inputting to gray model of seasonal index number
Predicted value is measured as shown in table 3.5.Table 3.5 is the monthly power quantity predicting value in January, 2017 to December.
Table 3.5
Analyze fitting result, table 3.5 it can be concluded that, the average relative error of gray model match value is 3.05%, effect
Preferably, but highest relative error has reached 9.20%, needs to carry out some amendments to it.
Next, calculating the error of gray model, residual sequence is constituted:
In formula:For the predicted value of gray model,For actual value.
Residual sequence figure is as shown in Fig. 4 .4 constructed by the present invention.
Can be seen that residual sequence from Fig. 4 .4 has positive value also to have negative value, and gray model requires data sequence symbol one
It causes, Grey Model cannot be directly used, so making it meet wanting for gray model the absolute value of residual sequence herein
It asks.
Will | ξ(0)(k) | then sequence uses GM (1,1) model, residual absolute value is calculated as original data sequence
Forecasting sequence
It can be obtained by MATLAB platform emulation, parameterAnd parameterAre as follows:
The match value that residual sequence can be calculated by above formula, as shown in table 3.4.Residual sequence match value calculates
Later, ash discharge precision index is calculated, several dynamical states are divided according to grey precision index, according to the point for falling into each state interval,
Markov Transition Probabilities matrix is calculated, the correction value of residual error is obtained.
Analyzing result to can be seen that the relative error after residual GM by table 3.4 is 1.72%.It is repaired compared to residual error
2.81% before just, precision of prediction has obtained better promotion.Gray model match value and seasonal index number are flat after residual GM
Match value and actual comparison figure such as Fig. 4 .6 and Fig. 4 .7 institute of the load sequence after cunning as shown in Fig. 4 .5, before and after residual GM
Show.
It is predicted using data of the gray model after residual GM in January, 2017 to December, obtained predicted value
Final predicted value should can be just reduced to multiplied by seasonal index number, final prediction result is as shown in table 3.6.Table 3.6 is
The monthly power quantity predicting value in January, 2017 to December.
Table 3.6
It can analyze and obtain from the predicted value in table 3.6, minimal error reaches 0.06% accuracy, average relative error
It is 1.86%.The result illustrates the gray model based on Markov Chain residual GM in the moon in January, 2017 to December
Degree power quantity predicting achieves preferable effect.
In order to prove the validity of this method, the present invention is by the predicted value of this method and wavelet neural network, ARIMA model
It is compared, as shown in Fig. 4 .8.
Can be seen that the method for the present invention predicted value from Fig. 4 .8 to be close with actual curve, from residual plot it can also be seen that
Compared with wavelet neural network and ARIMA model, the prediction effect of the method for the present invention is preferable.
It should be noted that for simple description, therefore, it is stated as a series of action groups for embodiment of the method
It closes, but those skilled in the art should understand that, embodiment of that present invention are not limited by the describe sequence of actions, because according to
According to the embodiment of the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art also should
Know, the embodiments described in the specification are all preferred embodiments, and the related movement not necessarily present invention is implemented
Necessary to example.
In view of the above technical problems, referring to Fig. 3, a kind of prediction meanss of electric system Mid-long Term Load of the present invention are shown
Structural schematic diagram, described device comprises the following modules:
Data preprocessing module 301, for collecting the Mid-long Term Load data in electric system, to the Mid-long Term Load
Data are pre-processed, and load data sample is obtained；
Gray model processing module 302 is obtained for the load data sample to be input in gray model GM (1,1)
First predicted value out；
Residual sequence obtains module 303, compared with first predicted value is made difference with the Mid-long Term Load data,
Obtain residual sequence；Wherein, the residual sequence takes absolute value；
Markov Chain correction module 304, for utilizing Ma Erke when first predicted value does not meet preset requirement
Husband's chain corrects the residual sequence, obtains the second predicted value of the residual sequence；
Preset requirement judgment module 305, for judging whether second predicted value meets preset requirement；
Prediction result correction verification module 306, for when judging that second predicted value meets preset requirement, by described second
Prediction result of the predicted value as the Mid-long Term Load data.
For system embodiments, since it is basically similar to the method embodiment, related so being described relatively simple
Place illustrates referring to the part of embodiment of the method.
All the embodiments in this specification are described in a progressive manner, the highlights of each of the examples are with
The difference of other embodiments, the same or similar parts between the embodiments can be referred to each other.
Above to a kind of prediction technique and device of electric system Mid-long Term Load provided by the present invention, carry out in detail
It introduces, specific examples are applied in the present invention, and principle and implementation of the present invention are described, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention；At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range, in conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.
Claims (10)
1. a kind of prediction technique of electric system Mid-long Term Load, which is characterized in that the described method includes:
The Mid-long Term Load data in electric system are collected, the Mid-long Term Load data are pre-processed, load number is obtained
According to sample；
The load data sample is input in gray model GM (1,1), obtains the first predicted value；
Compared with first predicted value is made difference with the Mid-long Term Load data, residual sequence is obtained；Wherein, the residual error sequence
Column take absolute value；
When first predicted value does not meet preset requirement, the residual sequence is corrected using Markov Chain, described in acquisition
Second predicted value of residual sequence；
Judge whether second predicted value meets preset requirement；
If meeting, using second predicted value as the prediction result of the Mid-long Term Load data.
2. the method according to claim 1, wherein the Mid-long Term Load data in electric system are collected, to institute
Stating the step of Mid-long Term Load data are pre-processed, obtain load data sample includes:
Collect the Mid-long Term Load data in electric system；
First pretreatment is carried out to the Mid-long Term Load data, obtains the first Mid-long Term Load data；
Data smoothing processing is carried out to the first Mid-long Term Load data, obtains load data sample.
3. according to the method described in claim 2, it is characterized in that, carrying out data smoothing to the first Mid-long Term Load data
Processing, the sub-step for obtaining load data sample further comprises:
Choose the first Mid-long Term Load data in multiple years；
The first Mid-long Term Load data are calculated in the average value of the same target time period in the multiple year, and described
The overall average in all months in multiple years；
The average value of the same target time period is divided by with the overall average, obtains the seasonal index number of the target time period；
The seasonal index number of the average value and the target time period of the same target time period is divided by, the target time period is obtained
Load data sample.
4. method according to claim 1 or 3, which is characterized in that the load data sample is input to gray model
In GM (1,1), the step of obtaining the first predicted value, includes:
The load data sample is input in GM (1,1) model；
In the GM (1,1) model, single order Accumulating generation is carried out to the load data sample, obtains single order load sequence；
It is fitted the changing rule of the single order load sequence using first-order equation, obtains the time response of the GM (1,1) model
Function model；
The time response function model is subjected to the reduction of single order regressive, obtains the gray prediction knot of the load data sample
Fruit；
The gray prediction result is subjected to average generation processing, obtains the first predicted value.
5. the method according to claim 1, wherein before correcting the residual sequence using Markov Chain,
The described method includes:
According to the residual sequence, judge whether first predicted value meets preset requirement；
If meeting, using first predicted value as the prediction result of the Mid-long Term Load data.
6. the method according to claim 1, wherein before correcting the residual sequence using Markov Chain,
The method also includes:
The residual sequence is input in the GM (1,1) model, the gray prediction sequence of the residual sequence is obtained；
The residual sequence and the gray prediction sequence are divided by, the grey degree of fitting index of the residual sequence is obtained；
According to preset state classification condition, the grey degree of fitting index is divided into several state intervals；
According to several described state intervals, the Markov state transition probability matrix of the gray prediction sequence is constructed.
7. according to the method described in claim 6, it is characterized in that, correcting the residual sequence, acquisition using Markov Chain
The step of second predicted value of the residual sequence includes:
According to the Markov state transition probability matrix, the area that the gray prediction sequence corresponds to the state interval is obtained
Between range；
Preset value is taken in the interval range, and the preset value is multiplied with the residual sequence, obtains the residual error sequence
Second predicted value of column.
8. the method according to the description of claim 7 is characterized in that the method also includes:
Geneva inspection is carried out to the Markov state transition probability matrix.
9. the method according to claim 1, wherein the described method includes:
When judging that second predicted value does not meet preset requirement, second predicted value is added to the Mid-long Term Load
In data, and second predicted value and the Mid-long Term Load data are pre-processed.
10. a kind of prediction meanss of electric system Mid-long Term Load, which is characterized in that described device comprises the following modules:
Data preprocessing module, for collecting the Mid-long Term Load data in electric system, to the Mid-long Term Load data into
Row pretreatment, obtains load data sample；
Gray model processing module show that first is pre- for the load data sample to be input in gray model GM (1,1)
Measured value；
Residual sequence obtains module, compared with first predicted value is made difference with the Mid-long Term Load data, obtains residual
Difference sequence；Wherein, the residual sequence takes absolute value；
Markov Chain correction module, for being repaired using Markov Chain when first predicted value does not meet preset requirement
The just described residual sequence obtains the second predicted value of the residual sequence；
Preset requirement judgment module, for judging whether second predicted value meets preset requirement；
Prediction result correction verification module, for when judging that second predicted value meets preset requirement, by second predicted value
Prediction result as the Mid-long Term Load data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910424910.5A CN110222882A (en) | 2019-05-21 | 2019-05-21 | A kind of prediction technique and device of electric system Mid-long Term Load |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910424910.5A CN110222882A (en) | 2019-05-21 | 2019-05-21 | A kind of prediction technique and device of electric system Mid-long Term Load |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110222882A true CN110222882A (en) | 2019-09-10 |
Family
ID=67821632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910424910.5A Pending CN110222882A (en) | 2019-05-21 | 2019-05-21 | A kind of prediction technique and device of electric system Mid-long Term Load |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110222882A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110727919A (en) * | 2019-09-17 | 2020-01-24 | 珠海格力电器股份有限公司 | Photovoltaic power generation capacity prediction method, device and system |
| CN111143776A (en) * | 2019-12-27 | 2020-05-12 | 新奥数能科技有限公司 | Electric quantity load prediction method and device |
| CN111582542A (en) * | 2020-03-31 | 2020-08-25 | 国网上海市电力公司 | Power load prediction method and system based on abnormal restoration |
| CN111598296A (en) * | 2019-10-16 | 2020-08-28 | 中国南方电网有限责任公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
| CN111598475A (en) * | 2020-05-22 | 2020-08-28 | 浙江工业大学 | Power grid risk prediction method based on improved gray Markov model |
| CN112184487A (en) * | 2020-09-30 | 2021-01-05 | 国网北京市电力公司 | Method and device for predicting power supply index |
| CN112446545A (en) * | 2020-12-01 | 2021-03-05 | 河北工业大学 | Load prediction method based on overlapped Markov chain |
| CN112542857A (en) * | 2020-12-07 | 2021-03-23 | 上海电气分布式能源科技有限公司 | Method, device and equipment for controlling stable output of microgrid system |
| CN112990587A (en) * | 2021-03-24 | 2021-06-18 | 北京市腾河智慧能源科技有限公司 | Method, system, equipment and medium for accurately predicting power consumption of transformer area |
| CN113191003A (en) * | 2021-05-08 | 2021-07-30 | 上海核工程研究设计院有限公司 | Nuclear power real-time data trend fitting algorithm |
| CN113408795A (en) * | 2021-06-03 | 2021-09-17 | 国网河北省电力有限公司高邑县供电分公司 | Power load prediction system and method based on grey theory |
| CN113516279A (en) * | 2021-04-27 | 2021-10-19 | 贵州电网有限责任公司 | Comprehensive energy load prediction method based on energy consumption state transfer |
| CN113792828A (en) * | 2021-11-18 | 2021-12-14 | 成都数联云算科技有限公司 | Power grid load prediction method, system, equipment and medium based on deep learning |
| CN115018209A (en) * | 2022-08-08 | 2022-09-06 | 国网湖北省电力有限公司营销服务中心(计量中心) | Long-term prediction method and equipment for operation error of digital electric energy metering system |
| CN115051416A (en) * | 2022-08-16 | 2022-09-13 | 阿里巴巴（中国）有限公司 | Data processing method, power generation method and device and cloud equipment |
| CN117216469A (en) * | 2023-09-03 | 2023-12-12 | 国网江苏省电力有限公司信息通信分公司 | Big data processing method and system for real-time monitoring and prediction of power system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016082804A (en) * | 2014-10-21 | 2016-05-16 | 国立大学法人北陸先端科学技術大学院大学 | Power price determination apparatus and power price determination method |
| US9460478B2 (en) * | 2012-12-17 | 2016-10-04 | Arizona Board Of Regents On Behalf Of Arizona State University | System and method for wind generation forecasting |
| CN109740812A (en) * | 2018-12-28 | 2019-05-10 | 广州供电局有限公司 | Methods of electric load forecasting, device, computer equipment and storage medium |
| CN109767043A (en) * | 2019-01-17 | 2019-05-17 | 中南大学 | A kind of electric load time series big data intelligent modeling and prediction technique |
-
2019
- 2019-05-21 CN CN201910424910.5A patent/CN110222882A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9460478B2 (en) * | 2012-12-17 | 2016-10-04 | Arizona Board Of Regents On Behalf Of Arizona State University | System and method for wind generation forecasting |
| JP2016082804A (en) * | 2014-10-21 | 2016-05-16 | 国立大学法人北陸先端科学技術大学院大学 | Power price determination apparatus and power price determination method |
| CN109740812A (en) * | 2018-12-28 | 2019-05-10 | 广州供电局有限公司 | Methods of electric load forecasting, device, computer equipment and storage medium |
| CN109767043A (en) * | 2019-01-17 | 2019-05-17 | 中南大学 | A kind of electric load time series big data intelligent modeling and prediction technique |
Non-Patent Citations (2)
| Title |
|---|
| 杨波: ""基于支持向量机的电力系统短期负荷预测"", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 黄银华等: ""马尔科夫理论在中长期负荷预测中的应用"", 《电力系统及其自动化学报》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110727919A (en) * | 2019-09-17 | 2020-01-24 | 珠海格力电器股份有限公司 | Photovoltaic power generation capacity prediction method, device and system |
| CN111598296A (en) * | 2019-10-16 | 2020-08-28 | 中国南方电网有限责任公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
| CN111143776A (en) * | 2019-12-27 | 2020-05-12 | 新奥数能科技有限公司 | Electric quantity load prediction method and device |
| CN111582542A (en) * | 2020-03-31 | 2020-08-25 | 国网上海市电力公司 | Power load prediction method and system based on abnormal restoration |
| CN111582542B (en) * | 2020-03-31 | 2023-10-03 | 国网上海市电力公司 | Power load prediction method and system based on anomaly repair |
| CN111598475A (en) * | 2020-05-22 | 2020-08-28 | 浙江工业大学 | Power grid risk prediction method based on improved gray Markov model |
| CN112184487A (en) * | 2020-09-30 | 2021-01-05 | 国网北京市电力公司 | Method and device for predicting power supply index |
| CN112446545B (en) * | 2020-12-01 | 2022-02-08 | 河北工业大学 | Load prediction method based on overlapped Markov chain |
| CN112446545A (en) * | 2020-12-01 | 2021-03-05 | 河北工业大学 | Load prediction method based on overlapped Markov chain |
| CN112542857A (en) * | 2020-12-07 | 2021-03-23 | 上海电气分布式能源科技有限公司 | Method, device and equipment for controlling stable output of microgrid system |
| CN112542857B (en) * | 2020-12-07 | 2023-05-16 | 上海电气分布式能源科技有限公司 | Method, device and equipment for controlling stable output of micro-grid system |
| CN112990587A (en) * | 2021-03-24 | 2021-06-18 | 北京市腾河智慧能源科技有限公司 | Method, system, equipment and medium for accurately predicting power consumption of transformer area |
| CN112990587B (en) * | 2021-03-24 | 2023-10-24 | 北京市腾河智慧能源科技有限公司 | Method, system, equipment and medium for accurately predicting power consumption of transformer area |
| CN113516279A (en) * | 2021-04-27 | 2021-10-19 | 贵州电网有限责任公司 | Comprehensive energy load prediction method based on energy consumption state transfer |
| CN113516279B (en) * | 2021-04-27 | 2022-08-30 | 贵州电网有限责任公司 | Comprehensive energy load prediction method based on energy consumption state transfer |
| CN113191003A (en) * | 2021-05-08 | 2021-07-30 | 上海核工程研究设计院有限公司 | Nuclear power real-time data trend fitting algorithm |
| CN113408795A (en) * | 2021-06-03 | 2021-09-17 | 国网河北省电力有限公司高邑县供电分公司 | Power load prediction system and method based on grey theory |
| CN113792828A (en) * | 2021-11-18 | 2021-12-14 | 成都数联云算科技有限公司 | Power grid load prediction method, system, equipment and medium based on deep learning |
| CN115018209B (en) * | 2022-08-08 | 2022-11-08 | 国网湖北省电力有限公司营销服务中心(计量中心) | Long-term prediction method and equipment for operation error of digital electric energy metering system |
| CN115018209A (en) * | 2022-08-08 | 2022-09-06 | 国网湖北省电力有限公司营销服务中心(计量中心) | Long-term prediction method and equipment for operation error of digital electric energy metering system |
| CN115051416A (en) * | 2022-08-16 | 2022-09-13 | 阿里巴巴（中国）有限公司 | Data processing method, power generation method and device and cloud equipment |
| CN117216469A (en) * | 2023-09-03 | 2023-12-12 | 国网江苏省电力有限公司信息通信分公司 | Big data processing method and system for real-time monitoring and prediction of power system |
| CN117216469B (en) * | 2023-09-03 | 2024-03-15 | 国网江苏省电力有限公司信息通信分公司 | Big data processing method and system for real-time monitoring and prediction of power system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110222882A (en) | A kind of prediction technique and device of electric system Mid-long Term Load | |
| Papavasiliou et al. | Reserve requirements for wind power integration: A scenario-based stochastic programming framework | |
| Wang et al. | Analysis of operation cost and wind curtailment using multi-objective unit commitment with battery energy storage | |
| CN107292766B (en) | Wind power consumption-oriented power system peak regulation means economical evaluation method and system | |
| CN104077651B (en) | Maintenance scheduling for power systems optimization method | |
| CN113496311A (en) | Photovoltaic power station generated power prediction method and system | |
| CN108596242A (en) | Power grid meteorology load forecasting method based on wavelet neural network and support vector machines | |
| CN107665385A (en) | A kind of short-term load forecasting method based on SVMs of micro-grid system | |
| CN116914747B (en) | Power consumer side load prediction method and system | |
| CN114372360A (en) | Method, terminal and storage medium for power load prediction | |
| Han et al. | Security-based active demand response strategy considering uncertainties in power systems | |
| CN105528668A (en) | Dynamic environment and economy scheduling method of grid-connected wind power system | |
| CN114021848A (en) | Generating capacity demand prediction method based on LSTM deep learning | |
| CN115205068A (en) | Energy storage optimal peak-valley time interval dividing method considering net load demand distribution | |
| CN103199616B (en) | Large regional grid wind-powered electricity generation group service data monitoring analysis system | |
| CN113887809A (en) | Power distribution network supply and demand balance method, system, medium and computing equipment under double-carbon target | |
| CN116722605A (en) | Power distribution network scheduling optimization method based on Internet of things | |
| Salman et al. | Optimal Power Systems Planning for IEEE-14 Bus Test System Application | |
| CN115860797A (en) | Electric quantity demand prediction method suitable for new electricity price reform situation | |
| CN116154875A (en) | Photovoltaic power station active power optimization distribution method based on TCN and error function | |
| CN115456406A (en) | Evaluation method, device, equipment and storage medium of comprehensive energy system | |
| CN108615091A (en) | Electric power meteorology load data prediction technique based on cluster screening and neural network | |
| Fu et al. | Study on wind power consumption capacity of power grid considering risk and unit commitment | |
| Alkilany et al. | Application of the k-means clustering algorithm to predict load shedding of the southern electrical grid of libya | |
| CN114372615A (en) | Short-term photovoltaic power prediction method and system considering spatial correlation |
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: Wei Mingkui Inventor after: Zhou Quan Inventor after: Cai Shaorong Inventor after: Jiang Li Inventor after: Lu Liang Inventor before: Liu Qin Inventor before: Wei Mingkui Inventor before: Zhou Quan Inventor before: Cai Shaorong Inventor before: Jiang Li Inventor before: Lu Liang |
|
| CB03 | Change of inventor or designer information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190910 |
|
| RJ01 | Rejection of invention patent application after publication |