CN108448572A - A kind of short-term micro-grid load interval probability prediction technique - Google Patents
A kind of short-term micro-grid load interval probability prediction technique Download PDFInfo
- Publication number
- CN108448572A CN108448572A CN201810227968.6A CN201810227968A CN108448572A CN 108448572 A CN108448572 A CN 108448572A CN 201810227968 A CN201810227968 A CN 201810227968A CN 108448572 A CN108448572 A CN 108448572A
- Authority
- CN
- China
- Prior art keywords
- coral
- formula
- interval
- micro
- grid 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
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Farming Of Fish And Shellfish (AREA)
Abstract
The invention discloses a kind of short-term micro-grid load interval probability prediction technique, this method includes：Several historical load datas of micro-capacitance sensor are obtained as sample set；Optimality Criteria is built in conjunction with forecast interval coverage rate, forecast interval bandwidth root mean square；The short-term micro-grid load interval probability prediction model based on Recognition with Recurrent Neural Network is established, carrying out optimizing to Optimality Criteria by coral reef algorithm updates neural network weight threshold value；Basic coral reef algorithm is improved using eliminative mechanism optimisation strategy and accelerates convergence rate to improve algorithm performance.The present invention preferably overcomes the not high disadvantage of the basic slow precision of coral reef algorithm the convergence speed by the improvement of coral reef algorithm, effectively increases micro-grid load prediction level.
Description
Technical field
The invention belongs to technical field of power systems, and in particular to a kind of short-term micro-grid load interval probability prediction side
Method.
Background technology
Micro-capacitance sensor be by distributed generation resource, energy storage device, energy conversion device and monitoring, protective device etc. be tied to
The small-sized decentralized system of customer power supply.The accurate prediction of micro-grid load is the important foundation of micro-capacitance sensor operation and energy management,
Micro-capacitance sensor operation reserve will be directly affected.In load prediction field, it is broadly divided at present for the method for micro-grid load prediction
Traditional prediction method and modern intelligent Forecasting.Traditional prediction method mainly has extra curvature pushing manipulation, grey method, returns and divide
Analysis method, time series method and Load Derivation etc..Such approach application probability theory or mathematical statistics, are obtained by statistical analysis
The fitting function of the historical data gone out carries out load prediction, and algorithmic procedure is concise, it is easy to accomplish.Modern intelligent Forecasting master
There are artificial neural network and SVM prediction method.Such approach application artificial intelligence technology, by empirical learning and
Sample training carries out optimal fitting to load variations rule, and can fully consider shadow of the extraneous factor to load variations
It rings, the precision of prediction result is higher.
But micro-grid load prediction technique is single point prediction at present, only provides a determining numerical value, it can not
Determine the possible fluctuation range of prediction result.And many uncertain factors are contained in micro-grid system so that decision-making work is deposited
In risk, the uncertainty of electricity needs is must take into consideration in decision, therefore how to realize that interval prediction is more in line with objective need
It asks.
Invention content
In view of the deficiencies of the prior art, the present invention proposes a kind of short-term micro-grid load interval probability prediction technique.
Technical scheme is as follows：
A kind of short-term micro-grid load interval probability prediction technique, including：
Step 1 obtains micro-grid load data；
Step 2, in conjunction with forecast interval coverage rate and forecast interval bandwidth root mean square, and introduce mean deviation index, construct
Width coverage criterion is as Optimization goal function；
Step 3 is improved basic coral reef algorithm using eliminative mechanism optimisation strategy；
Step 4 establishes the short-term micro-capacitance sensor that coral reef algorithm Recognition with Recurrent Neural Network is improved based on eliminative mechanism optimisation strategy
Load setting prediction model carries out optimizing to Optimality Criteria, updates neural network weight threshold value by improving coral reef algorithm；
Step 5 establishes neural network according to optimal weight threshold, and interval prediction is carried out to micro-grid load.
Its further technical solution is that the step 2 specifically includes：
Calculate forecast interval coverage rate δPISCPFor：
In formula (1), N is total sample number；I is sample serial number；
Parameter ciFor：
In formula (2), ζiFor practical micro-grid load, LiFor forecast interval lower bound, UiFor the forecast interval upper bound；
Calculate forecast interval bandwidth root mean square ψRPIWFor：
In formula (3), R is maximum predicted interval width；
Calculate mean deviation index φMOFor：
Construct width coverage criterion TCCWCAs Optimization goal function；Width coverage criterion TCCWCFor：
Formula has in (5)：
In formula (5), (6), μ is the confidence interval of (1- α) confidence level, η δPISCPPunishing when not up to confidence interval μ
The amount of penalizing.
Its further technical solution is that the improved method method of the step 3 is specifically：
In formula (7), α and α ' is iterations, and α ≠ α ', θ indicate the standard deviation of fitness value；When | | cα-cα'| |≤S,
That is when Euclidean distance is less than preset value S between two individuals, | fit (cα)-fit(cα’) |≤θ indicates that shown algorithm is absorbed in stagnation at this time；
ε ' andFor superseded probability and superseded quantitative proportion are recycled after improvement every time.
Its further technical solution is that the step 4 specifically includes：
Step 1 is obtained micro-grid load data normalizing to [0,1] section by step 41, is divided into training set and test set；
Step 42, Recognition with Recurrent Neural Network initialization；Recognition with Recurrent Neural Network structure is set；
Step 43 assumes that initial coral reef has U × V node to adhere to for coral polyp, the coral reef being attached at this time
The ratio for accounting for all corals is ρ；If coral polyp zoogamy ratio is ξ, schizogamy ratio is υ, and filial generation coral polyp is attempted attached
It is τ threshold number, and the superseded probability of cycle is ε every time, and superseded quantitative proportion isMaximum iteration is σ；
Step 44, dioecious parental generation coral polyp generate filial generation coral polyp by formula below (8), formula (9), i.e., each
Node weight threshold；
In formula (7), C1,α、C2,αFor dioecious parental generation coral polyp, c1,α、c2,αFor filial generation coral polyp, α is iteration time
Number, ω is the stochastic variable generated by formula (9)；
In formula (9), ι is the random number on section (0,1), and κ is to intersect constant；
It is remainingThe hermaphroditic parental generation coral polyp C of quantityαA filial generation coral is generated according to formula (10)
Coral worm cα：
In formula (9),Respectively parental generation coral polyp CαIn maximum value and minimum value；
Step 45, according to filial generation coral polyp, current width coverage criterion T is calculated by formula (5)CCWCValue, pass through
Compare the value T of the anthozoic width coverage criterion of filial generationCCWC, judge that its success adheres to, if it is possible to successfully adhere to, then it is fixed
Justice is advantage coral polyp；
The division of advantage coral polyp generates filial generation, and repeats current procedures, judges whether it can successfully adhere to；
Step 46 calculates the superseded probability of current filial generation coral polyp and superseded ratio by the improved method of the step 3, washes in a pan
Eliminate qualified coral polyp；
Step 47 judges whether to reach maximum iteration or meet default output to require, if not satisfied, then repeating to walk
Rapid 43；If meeting the requirements, current optimal coral polyp is exported, as best initial weights threshold value.
The method have the benefit that：
The present invention preferably overcomes the basic slow precision of coral reef algorithm the convergence speed not by the improvement of coral reef algorithm
High disadvantage effectively increases micro-grid load prediction level.
Description of the drawings
Fig. 1 is the flow diagram of short-term micro-grid load interval probability prediction technique of the present invention.
Fig. 2 is the micro-grid load interval prediction result in one embodiment.
Specific implementation mode
As shown in FIG. 1, FIG. 1 is the flow diagram of short-term micro-grid load interval probability prediction technique of the invention, this hairs
It is bright to comprise the steps of (1~step 5 of following steps is corresponding in turn to S101~S105 in Fig. 1)：
Step 1 obtains micro-grid load data as sample.
By taking the micro-capacitance sensor of Northern Europe as an example, which includes 168 family residents, 1 wood-working factory and other business
It is mating etc..The micro-capacitance sensor historical load data includes continuous two months practical micro-grid load data, resolution ratio 1h.
Step 2, in conjunction with forecast interval coverage rate and forecast interval bandwidth root mean square, and introduce mean deviation index, construct
Width coverage criterion is as Optimization goal function.Specifically have：
Calculate forecast interval coverage rate δPISCPFor：
In formula (1), N is total sample number；I is sample serial number；
Parameter ciFor：
In formula (2), ζiFor practical micro-grid load data, LiFor forecast interval lower bound, UiFor the forecast interval upper bound.
Calculate forecast interval bandwidth root mean square ψRPIWFor：
In formula (3), R is maximum predicted interval width.
Calculate mean deviation index φMOFor：
Construct width coverage criterion TCCWCAs Optimization goal function.Width coverage criterion TCCWCFor：
Formula has in (5)：
In formula (5), (6), μ is the confidence interval of (1- α) confidence level, η δPISCPPunishing when not up to confidence interval μ
The amount of penalizing.
Step 3 is improved basic coral reef algorithm using eliminative mechanism optimisation strategy；
In formula (7), α and α ' is iterations, and α ≠ α ', θ indicate the standard deviation of fitness value.When | | cα-cα'| |≤S,
That is when Euclidean distance is less than preset value S between two individuals, | fit (cα)-fit(cα’) |≤θ indicates that shown algorithm is absorbed in stagnation at this time.
ε ' andFor superseded probability and superseded quantitative proportion are recycled after improvement every time.
Step 4 establishes the short-term micro-capacitance sensor that coral reef algorithm Recognition with Recurrent Neural Network is improved based on eliminative mechanism optimisation strategy
Load setting prediction model carries out optimizing to Optimality Criteria, updates neural network weight threshold value by improving coral reef algorithm.
It is comprised the concrete steps that：
Step 1 is obtained micro-grid load data normalizing to [0,1] section by step 41, is divided into training set and test set；
Step 42, Recognition with Recurrent Neural Network initialization；Recognition with Recurrent Neural Network structure is set.
Step 43 assumes that initial coral reef has U × V node to adhere to for coral polyp, the coral reef being attached at this time
The ratio for accounting for all corals is ρ.If coral polyp zoogamy (hatching and hatching in vitro in occlusion body) ratio is ξ, schizogamy
Ratio is υ, and it is τ that filial generation coral polyp, which attempts limit of adhesion number, and the superseded probability of cycle is ε every time, and superseded quantitative proportion is
Maximum iteration is σ.
Step 44, dioecious parental generation coral polyp generate filial generation coral polyp by formula below (8), formula (9), i.e., each
Node weight threshold.
In formula (7), C1,α、C2,αFor dioecious parental generation coral polyp, c1,α、c2,αFor filial generation coral polyp, α is iteration time
Number, ω is the stochastic variable generated by formula (8).
In formula (9), ι is the random number on section (0,1), and κ is to intersect constant.
It is remainingThe hermaphroditic parental generation coral polyp C of quantityαA filial generation coral is generated according to formula (9)
Worm cα：
In formula (9),Respectively parental generation coral polyp CαIn maximum value and minimum value.
Step 45 calculates current width covering standard by formula (5) according to filial generation coral polyp with training set data
Then TCCWCValue.Specifically, formula (8), the calculated value c of formula (9)α、cα,1、cα,2Be vector, i.e. the weights threshold of neural network
Value calculates the bound of forecast interval, i.e. forecast interval lower bound L in formula (4) with neural networkiWith forecast interval upper bound Ui, so
Mean deviation index φ is calculated according to formula (4) afterwardsMO, bring formula (5) into, finally calculate current width coverage criterion TCCWC。
By comparing the value T of the anthozoic width coverage criterion of filial generationCCWC, judge that its success adheres to, as it can
Success is adhered to, then is defined as advantage coral polyp.
The division of advantage coral polyp generates filial generation, and repeats current procedures, and judgement can successfully adhere to.
Step 46, by recycling superseded probability ε ' and superseded quantity every time after formula (7) calculated improvement in step 3
RatioEliminate partial width coverage criterion TCCWCLarger coral polyp.
Step 47 judges whether to reach maximum iteration or meet default output to require, if not satisfied, then repeating to walk
Rapid 43；If meeting the requirements, current optimal coral polyp is exported, as best initial weights threshold value.
Step 5, the weight threshold for setting the calculated best initial weights threshold value of step 4 to neural network, with prediction data
Input of multiple data as neural network before point, predicts output of the section as neural network up and down, to micro-grid load sequence
The test set of row carries out interval prediction.
Fig. 2 is that a kind of short-term micro-grid load interval probability prediction technique carries out section to micro-grid load test set data
Prediction result.As shown in Fig. 2, the load prediction upper bound and load prediction lower bound form a section, actual load this section it
It is interior, fully demonstrate the forecasting accuracy of the present invention.
What has been described above is only a preferred embodiment of the present invention, and present invention is not limited to the above embodiments.It is appreciated that this
The other improvements and change that field technology personnel directly export or associate without departing from the spirit and concept in the present invention
Change, is considered as being included within protection scope of the present invention.
Claims (4)
1. a kind of short-term micro-grid load interval probability prediction technique, it is characterised in that including：
Step 1 obtains micro-grid load data；
Step 2, in conjunction with forecast interval coverage rate and forecast interval bandwidth root mean square, and introduce mean deviation index, construct width
Coverage criterion is as Optimization goal function；
Step 3 is improved basic coral reef algorithm using eliminative mechanism optimisation strategy；
Step 4 establishes the short-term micro-grid load that coral reef algorithm Recognition with Recurrent Neural Network is improved based on eliminative mechanism optimisation strategy
Interval prediction model carries out optimizing to Optimality Criteria, updates neural network weight threshold value by improving coral reef algorithm；
Step 5 establishes neural network according to optimal weight threshold, and interval prediction is carried out to micro-grid load.
2. short-term micro-grid load interval probability prediction technique according to claim 1, which is characterized in that the step 2
It specifically includes：
Calculate forecast interval coverage rate δPISCPFor：
In formula (1), N is total sample number；I is sample serial number；
Parameter ciFor：
In formula (2), ζiFor practical micro-grid load, LiFor forecast interval lower bound, UiFor the forecast interval upper bound；
Calculate forecast interval bandwidth root mean square ψRPIWFor：
In formula (3), R is maximum predicted interval width；
Calculate mean deviation index φMOFor：
Construct width coverage criterion TCCWCAs Optimization goal function；Width coverage criterion TCCWCFor：
Formula has in (5)：
In formula (5), (6), μ is the confidence interval of (1- α) confidence level, η δPISCPPunishment when not up to confidence interval μ
Amount.
3. short-term micro-grid load interval probability prediction technique according to claim 1, which is characterized in that the step 3
Improved method method be specifically：
In formula (7), α and α ' is iterations, and α ≠ α ', θ indicate the standard deviation of fitness value；When | | cα-cα'| |≤S, i.e., two
When Euclidean distance is less than preset value S between individual, | fit (cα)-fit(cα') |≤θ indicates that shown algorithm is absorbed in stagnation at this time；ε ' andFor superseded probability and superseded quantitative proportion are recycled after improvement every time.
4. short-term micro-grid load interval probability prediction technique according to claim 1, which is characterized in that the step 4
It specifically includes：
Step 1 is obtained micro-grid load data normalizing to [0,1] section by step 41, is divided into training set and test set；
Step 42, Recognition with Recurrent Neural Network initialization；Recognition with Recurrent Neural Network structure is set；
Step 43 assumes that initial coral reef has U × V node to adhere to for coral polyp, and the coral reef being attached at this time accounts for institute
It is ρ to have the ratio of coral；If coral polyp zoogamy ratio is ξ, schizogamy ratio is υ, and filial generation coral polyp attempts attachment pole
Limit number is τ, and the superseded probability of cycle is ε every time, and superseded quantitative proportion isMaximum iteration is σ；
Step 44, dioecious parental generation coral polyp generate filial generation coral polyp, i.e., each node by formula below (8), formula (9)
Weight threshold；
In formula (7), C1,α、C2,αFor dioecious parental generation coral polyp, c1,α、c2,αFor filial generation coral polyp, α is iterations, ω
The stochastic variable generated for the formula of pressing (9)；
In formula (9), i is the random number on section (0,1), and κ is to intersect constant；
It is remainingThe hermaphroditic parental generation coral polyp C of quantityαA filial generation coral polyp c is generated according to formula (10)α：
In formula (9),Respectively parental generation coral polyp CαIn maximum value and minimum value；
Step 45, according to filial generation coral polyp, current width coverage criterion T is calculated by formula (5)CCWCValue, by comparing son
For the value T of anthozoic width coverage criterionCCWC, judge that its success adheres to, if it is possible to successfully adhere to, then be defined as excellent
Gesture coral polyp；
The division of advantage coral polyp generates filial generation, and repeats current procedures, judges whether it can successfully adhere to；
Step 46 calculates the superseded probability of current filial generation coral polyp and superseded ratio by the improved method of the step 3, eliminates symbol
The coral polyp of conjunction condition；
Step 47 judges whether to reach maximum iteration or meet default output to require, if not satisfied, then repeating step
43；If meeting the requirements, current optimal coral polyp is exported, as best initial weights threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810227968.6A CN108448572A (en) | 2018-03-20 | 2018-03-20 | A kind of short-term micro-grid load interval probability prediction technique |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810227968.6A CN108448572A (en) | 2018-03-20 | 2018-03-20 | A kind of short-term micro-grid load interval probability prediction technique |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108448572A true CN108448572A (en) | 2018-08-24 |
Family
ID=63195837
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810227968.6A Pending CN108448572A (en) | 2018-03-20 | 2018-03-20 | A kind of short-term micro-grid load interval probability prediction technique |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108448572A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113704319A (en) * | 2021-07-29 | 2021-11-26 | 南京邮电大学 | Mobile terminal service prediction method combined with context information |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104571262A (en) * | 2015-01-16 | 2015-04-29 | 江南大学 | Method for predicating interval probability of short-term wind power |
| CN104809528A (en) * | 2015-05-11 | 2015-07-29 | 江南大学 | Ultrashort-term wind power forecasting method based on Mycielski method and Markov chain |
| CN107317323A (en) * | 2017-06-16 | 2017-11-03 | 江南大学 | Short-term micro-grid load power interval probability forecasting method |
-
2018
- 2018-03-20 CN CN201810227968.6A patent/CN108448572A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104571262A (en) * | 2015-01-16 | 2015-04-29 | 江南大学 | Method for predicating interval probability of short-term wind power |
| CN104809528A (en) * | 2015-05-11 | 2015-07-29 | 江南大学 | Ultrashort-term wind power forecasting method based on Mycielski method and Markov chain |
| CN107317323A (en) * | 2017-06-16 | 2017-11-03 | 江南大学 | Short-term micro-grid load power interval probability forecasting method |
Non-Patent Citations (2)
| Title |
|---|
| 于昕妍等: "考虑概率区间的微电网短期负荷多目标预测方法", 《电子学报》 * |
| 陆欣等: "考虑气象特征提取的短期风速预测方法", 《控制工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113704319A (en) * | 2021-07-29 | 2021-11-26 | 南京邮电大学 | Mobile terminal service prediction method combined with context information |
| CN113704319B (en) * | 2021-07-29 | 2023-05-30 | 南京邮电大学 | Mobile terminal service prediction method combined with context information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107590567A (en) | A kind of Recognition with Recurrent Neural Network short-term load forecasting method based on comentropy cluster and notice mechanism | |
| CN107358060A (en) | A kind of method estimated wind power prediction error burst based on HMM | |
| CN106315319A (en) | Intelligent pre-dispatching method and system for elevator | |
| CN110516840A (en) | Short term prediction method based on the wind light generation power output for improving random forest method | |
| CN107730031A (en) | A kind of ultra-short term peak load forecasting method and its system | |
| CN110232483A (en) | Deep learning load forecasting method, device and terminal device | |
| CN111343650B (en) | Urban scale wireless service flow prediction method based on cross-domain data and loss resistance | |
| CN104571262B (en) | Short-term wind-electricity power interval probability Forecasting Methodology | |
| CN103023065A (en) | Wind power short-term power prediction method based on relative error entropy evaluation method | |
| EP3640869A1 (en) | Method for predicting an energy demand, data processing system and renewable power plant with a storage | |
| CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| Kolhe et al. | GA-ANN for short-term wind energy prediction | |
| CN110535149A (en) | A kind of area Gong Biantai electric load three-phase imbalance prediction technique | |
| CN109636054A (en) | Solar energy power generating amount prediction technique based on classification and error combination prediction | |
| CN109886567A (en) | A kind of short-term load forecasting method considering sendible temperature and radiation intensity | |
| CN110070228A (en) | BP neural network wind speed prediction method for neuron branch evolution | |
| CN112819189A (en) | Wind power output prediction method based on historical predicted value | |
| CN110414725A (en) | The integrated wind power plant energy-storage system dispatching method of forecast and decision and device | |
| Cetinkaya | Long-term electrical load forecasting based on economic and demographic data for Turkey | |
| CN114648176A (en) | Wind-solar power consumption optimization method based on data driving | |
| CN104219682B (en) | The network layer construction method and system of farmland hybrid power supply wireless sensor network | |
| CN109858668B (en) | Coordination prediction method for power load region in thunder and lightning climate | |
| CN113344283B (en) | Energy internet new energy consumption capability assessment method based on edge intelligence | |
| KR20210026447A (en) | Apparatus and method for Deep neural network based power demand prediction | |
| CN109934396A (en) | A kind of regional network for the load fining prediction technique based on two stages reduction |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180824 |
|
| RJ01 | Rejection of invention patent application after publication |