CN108647829A - A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest - Google Patents
A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest Download PDFInfo
- Publication number
- CN108647829A CN108647829A CN201810465449.3A CN201810465449A CN108647829A CN 108647829 A CN108647829 A CN 108647829A CN 201810465449 A CN201810465449 A CN 201810465449A CN 108647829 A CN108647829 A CN 108647829A
- Authority
- CN
- China
- Prior art keywords
- period
- library
- decision
- water level
- periods
- 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
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 46
- 238000000605 extraction Methods 0.000 title claims abstract description 16
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000013507 mapping Methods 0.000 claims abstract description 12
- 238000005096 rolling process Methods 0.000 claims abstract description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 104
- 238000012360 testing method Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 18
- 238000003066 decision tree Methods 0.000 claims description 14
- 238000003860 storage Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000011144 upstream manufacturing Methods 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 4
- 230000006378 damage Effects 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000007599 discharging Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 229910052708 sodium Inorganic materials 0.000 claims description 3
- 239000011734 sodium Substances 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 2
- 230000005611 electricity Effects 0.000 claims description 2
- 230000003993 interaction Effects 0.000 claims description 2
- 238000012795 verification Methods 0.000 claims 1
- 238000010248 power generation Methods 0.000 abstract description 5
- 230000009471 action Effects 0.000 abstract description 2
- 230000007774 longterm Effects 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000002790 cross-validation Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000004513 sizing Methods 0.000 description 2
- 241001672694 Citrus reticulata Species 0.000 description 1
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003278 mimic effect Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001755 vocal 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The Hydropower Stations combined dispatching Rules extraction method based on random forest that the invention discloses a kind of, including Hydropower Stations joint optimal operation model is established and solves, generate Optimized Operation sample set；Establish decision variable and the input factor；Design scheduling function structure；Mapping relations are established using random forest；Constraints, which is destroyed, to be corrected；Rolling scheduling simultaneously repeatedly calculates.The present invention can comprehensively consider that " over-fitting " problem can effectively be avoided using " random " and efficiently using for sample information of the ideological guarantee of " integrated " by inputting the factor；The scheduling function of each library synchronization map of multiple-input and multiple-output step can effectively reflect the compensating action between each library, clear in structure, using simplicity；The operational reliability of this method has been effectively ensured for the amendment that constraints is destroyed.Suitable for medium-term and long-term Hydropower Stations power generation dispatching Rulemaking, different when segment lengths can be set in schedule periods according to hydrologic regime, there is stronger versatility and flexibility.
Description
Technical field
The present invention relates to step power station power generation dispatching method, more particularly to a kind of step power station based on random forest
Group's combined dispatching Rules extraction method.
Background technology
Power station scheduling rule refers to the important evidence of water guide library management and running work.Scheduling rule refers to being based on reservoir institute
A set of aerial drainage decision guide that its different function is played conducive to reservoir that the runoff characteristic of control catchment extracts, it is relatively conventional
Scheduling rule form includes three kinds of verbal description, scheduling graph and scheduling function.
Traditional form of expression of scheduling function is math equation group, and math equation group has confined function shape, to a certain degree
On limit non-linear relation expression between variable, the more flexible structure of intelligent algorithm and powerful mapping ability make it be applicable in
Complex nonlinear relationship between refining reservoir scheduling variable.Existing fitting scheduling function method only with subjective judgement or directly
The selection that the input factor is carried out using certain algorithm, so as to cause following problem：(1) part effective information is lost；(2) it is based on
The factor of historical summary screening also needs to be considered for the applicability of future scheduling decision, and changeless factor classification lacks spirit
Active (3) are according to the compatible degree between the input variable after the screening of certain algorithm and the algorithm of fitting scheduling function also without collateral security.
Therefore, how comprehensively and optimization sample information is effectively utilized, with appropriate and efficient method refinement reservoir operation
Rule is the technical barrier of current urgent need to resolve.
Invention content
Goal of the invention：In view of the deficiencies of the prior art, the present invention proposes a kind of step power station based on random forest
Group's combined dispatching Rules extraction method.
Technical solution：The present invention provides a kind of Hydropower Stations combined dispatching Rule Extraction side based on random forest
Method includes the following steps：
(1) establish and solve with gross generation in the Model for Cascade Hydroelectric Stations phase be up to target, with water balance constraint,
Bound restriction of water level, range of stage constraint, the constraint of minimax traffic constraints, flow luffing, load constraint, the first water of schedule periods
Position constraint and the joint optimal operation model that scheduling end of term restriction of water level is constraints；
(2) decision variable and the input factor for establishing scheduling function, pass through water power based on joint optimal operation model result
It adjusts to calculate and generates sample set；
(3) the scheduling function structure of each power station synchronization map of step is established；
(4) mapping relations and decision are established using random forest regression model；
(5) the case where being destroyed to constraints is modified；
(6) it rolling scheduling and repeatedly calculates.
Further, joint optimal operation model described in step (1) is with gross generation in the Model for Cascade Hydroelectric Stations phase
It is up to object function：
In formula, E is gross generation in the Model for Cascade Hydroelectric Stations phase, and n, T are respectively step hydropower station number, schedule periods period
Number；TtFor when segment length；Hi,tThe output of respectively i-th power station t period, generating flow, hair
Electric head.
Further, the constraints of joint optimal operation model described in step (1) is specially；
1) water balance constrains
In formula, Vi,t+1、Vi,tFor the i-th library t period Mos, first reservoir storage capacity；Qi,t、Ji,tAnd Si,tWhen respectively the i-th library t
Section reservoir inflow abandons water flow, loss flow；
2) upper and lower limit restriction of water level and range of stage constraint
|Zi,t+1-Zi,t|≤△Zi；
In formula, Zi,t、Zi,t 、Water level, lower limit water level, upper limit water level respectively at the beginning of the i-th library t periods；△ZiIt is i-th
Library allows range of stage；
3) maximum, minimum discharge constraint and the constraint of flow luffing
|qi,t+1-qi,t|<△qi；
In formula, qi,minMeet the minimum discharge of the comprehensive utilizations such as Downstream Navigation, ecology, water supply for the i-th library；qei,maxIt is i-th
Library water turbine set maximum discharge capacity；△qiAllow letdown flow luffing for the i-th library；qi,t、qi,t+1When respectively the i-th library t
Section, t+1 period storage outflows；
4) load constrains
N i,t≤Ni,t≤NYi；
In formula, Ni,t、N i,tThe minimum load in power station is required for the calculating output of the i-th power station t periods, power grid；NYiFor
I-th installed capacity of power station；
5) schedule periods just, the end of term control restriction of water level
In formula, Zis、Just water level, starting-point detection are calculated for the i-th library schedule periods；Zie、End of term calculating is dispatched for the i-th library
Water level, the end of term control water level；
For the joint optimal operation model to take turns library iterative method as derivation algorithm, result of calculation is each power station in schedule periods
The optimal water level value in day part end.
Further, it is input to grow serial reservoir inflow in step (2) with Hydropower Stations history, utilizes combined optimization
Scheduling model calculates, and obtains in schedule periods the optimal water level value in each library day part end year by year, and calculate by water power tune, that is, passes through water
Equation of equilibrium and Water-sodium disturbance equation calculation generate the sample set of each library length series, which includes test sample and training
Sample；Select the variable for representing reservoir period initial equilibrium state and period water situation as the input factor；Decision variable selection represents
The variable of reservoir period aerial drainage decision is as decision variable.
Further, the input factor includes water level at the beginning of each library faces the period, faces period reservoir inflow, upper a period of time
Section at the beginning of water level, upper period reservoir inflow, subsequent period reservoir inflow, superposition water level, accumulation of energy, enter can and interact item；Wherein, face
Other input factors comprising the two, use in the training stage in interim section reservoir inflow, subsequent period reservoir inflow and calculating
Measured value uses predicted value in test phase；The decision variable is generating flow or the period contributes or period end water level.
Further, using the scheduling function structure of each power station synchronization map of multiple-input and multiple-output in step (3)：
yi,t=fi,t(xm,i,t), i=1,2 ..., n t=1,2 ..., T m=1,2 ..., M；
In formula, n, T indicate hop count when step hydropower station number and schedule periods respectively；yi,tIndicate i-th power station t period
Decision variable；fi,tIndicate the scheduling function of i-th of power station t period；xm,i,tIndicate m-th of i-th power station t period
Input the factor.
Further, the step (4) further comprises：
(41) by k before the sample set for training, rear (K-k) year is for testing；
(42) regression model for establishing random forest by library by the period, that is, establish T × n regression model, random forest
Decision tree number n_estimators and random character variable number max_features, is passed through using grid search tool in parameter
The method of cross validation searches the optimal value of two parameters, and decision tree number is generally searched between 1~1000, and random character becomes
The upper limit for measuring number is input factor sum, so generally being searched between 1~M；Remaining parameter uses default value；
(43) it for the structure of every decision tree, is put back to using bootstrapping sampling techniques and randomly selects part
Training sample, obtained training sample subset are about 2/3rds of original training sample, when decision node is divided using with
The Partial Feature variable that machine extracts；It repeats this step and repeatedly builds more decision trees, combination forms random forest, just establishes in this way
Mapping relations between the input factor and decision variable；
(44) to the input factor of the mapping relations input test sample of foundation, the decision variable value of random forest is using every
The mean value of decision tree decision.
Further, in step (5) constraints destruction is carried out for the decision variable value obtained based on random forest
It corrects：According to different reservoir situations, minimum load constraint amendment, minimum lower aerial drainage are carried out by importance more big sequence more rearward
Amount constraint is corrected and the amendment of range of stage constraint, correction formula are as follows：
In formula, Ni,t、Respectively the i-th reservoir t periods correct before, correct after the period contribute；N i,tWhen the i-th reservoir t
Section minimum load constraint；Zi,t+1、Respectively correct before, correct after the i-th reservoir t periods period end water level；△ZiIt is
The range of stage of i reservoirs constrains；Zi,tWater level at the beginning of period to correct the preceding i-th reservoir t periods；qi,t、Respectively the i-th library
The t periods correct before, correct after storage outflow；qi,minMeet the minimum of the comprehensive utilizations such as Downstream Navigation, ecology, water supply for the i-th library
Flow；
Constraints destroys amendment need to be after using random forest decision by period power station progress one by one, specially：T
Period, from upstream to downstream, power station, which checks, one by one corrects minimum load constraint, minimum aerial drainage constraint, range of stage constraint, the
The t+1 periods and so on.
Further, it is used in step (6) by the period by the method for library rolling scheduling, under the above period decision value is used as
The input value of one period carries out rolling scheduling, until scrolling through entire schedule periods；Specially：It is the t periods, each using the period
The input factor of power station random forest regression model, first input test sample obtains the decision variable value in each power station, then
From upstream to downstream, power station, which checks, one by one corrects minimum load constraint, minimum aerial drainage constraint, range of stage constraint, so far t
The amendment of period finishes, and obtains each power station final decision variate-value of t periods, random due to random forest regression model
Property, test phase is computed repeatedly 5~10 times, chooses the maximum corresponding result of decision of step gross generation as final decision knot
Fruit；The t+1 periods and so on；The result of decision is the power generating value or generating flow value of each period in each power station schedule periods
Or the water level value of period Mo, accordingly obtain step gross generation in schedule periods.
Advantageous effect：Compared with prior art, the present invention has the following advantages and beneficial effect：
1, the method for the present invention can comprehensively consider input the factor, can efficient process higher-dimension input variable, without carry out
Pretreatment screening；
2, the method for the present invention uses " random " and efficiently using for sample information of the ideological guarantee of " integrated ", and can
Effectively avoid " over-fitting " problem；
3, the method for the present invention can be reflected using the scheduling function of each library synchronization map of multiple-input and multiple-output step between each library
Compensating action, it is clear in structure, using simplicity；
4, method reliability of operation has been effectively ensured for the amendment that constraints is destroyed in the method for the present invention；
5, the method for the present invention is suitable for medium-term and long-term Hydropower Stations power generation dispatching Rulemaking, can root in schedule periods
The when segment length of different length is set according to hydrologic regime, there is stronger versatility and flexibility.
Description of the drawings
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the structure and decision flow diagram of random forest；
Fig. 3 is random forest rolling scheduling flow chart；
Fig. 4 is rule-based scheduling and Optimized Operation mimic water-depth process comparison diagram.
Specific implementation mode
Below with reference to the embodiments and with reference to the accompanying drawing being further elaborated with to technical scheme of the present invention.
As shown in Figure 1, a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest, includes mainly
Following steps：
(1) establish and solve with gross generation in the Model for Cascade Hydroelectric Stations phase be up to target, with water balance constraint,
Bound restriction of water level, range of stage constraint, the constraint of minimax traffic constraints, flow luffing, load constraint, the first water of schedule periods
Position constraint and the joint optimal operation model that scheduling end of term restriction of water level is constraints；
The joint optimal operation model is up to object function, formula with gross generation in the Model for Cascade Hydroelectric Stations phase
It is as follows：
In formula, E be the Model for Cascade Hydroelectric Stations phase in gross generation, n, T be respectively step hydropower station number (step reservoir number),
Hop count when schedule periods；TtFor schedule periods when segment length；Hi,tRespectively t-th of period of i-th of power station
Output, generating flow, productive head.
The constraints of the joint optimal operation model is specific as follows；
1) water balance constrains
In formula, Vi,t+1、Vi,tFor the i-th library t period Mos, first reservoir storage capacity；Qi,t、Ji,tAnd Si,tWhen respectively the i-th library t
Section reservoir inflow abandons water flow, loss flow.
2) upper and lower limit restriction of water level and range of stage constraint
|Zi,t+1-Zi,t|≤△Zi(4)；
In formula, Zi,t、Zi,t 、Water level, lower limit water level, upper limit water level respectively at the beginning of the i-th library t periods；△ZiIt is i-th
Library allows range of stage.
3) maximum, minimum discharge constraint and the constraint of flow luffing
|qi,t+1-qi,t|<△qi(6)；
In formula, qi,minMeet the minimum discharge of the comprehensive utilizations such as Downstream Navigation, ecology, water supply for the i-th library；qei,maxIt is i-th
Library water turbine set maximum discharge capacity；△qiAllow letdown flow luffing for the i-th library；qi,t、qi,t+1When respectively the i-th library t
Section, t+1 periods storage outflow (including generating flow and abandon water flow).
4) load constrains
N i,t≤Ni,t≤NYi(7)；
In formula, Ni,t、N i,tThe minimum load in power station is required for the calculating output of the i-th power station t periods, power grid；NYiFor
I-th installed capacity of power station.
5) schedule periods just, the end of term control restriction of water level
In formula, Zis、Just water level, starting-point detection are calculated for the i-th library schedule periods；Zie、End of term calculating is dispatched for the i-th library
Water level, the end of term control water level.
The joint optimal operation model is to take turns library iterative method as derivation algorithm.Result of calculation is each power station in schedule periods
The optimal water level value in day part end.
The storage stream of serial nineteen fifty-one~2014 year is grown with certain 4 libraries (libraries a, the libraries b, the libraries c, the libraries d) Hydropower Stations history
Amount as input, schedule periods are include the reservoir filling phase of 36 periods, when segment length foundation water degree of irregularity be set as
Day, time, the moon etc. different lengths.
(2) decision variable and the input factor for establishing scheduling function, pass through water energy based on joint optimal operation model result
Conversion, which calculates, generates sample set；
Joint optimal operation model result is the optimal water level value in each library day part end year by year in schedule periods, is calculated by water power tune
(passing through water balance formula and Water-sodium disturbance equation calculation) generates the sample set of each library length series, as shown in table 1；
Table 1
Input selecting predictors represent the variable of reservoir period initial equilibrium state and period water situation, and the present invention selects each library to face
Water level at the beginning of period, face period reservoir inflow, water level at the beginning of a upper period, upper period reservoir inflow, subsequent period reservoir inflow,
Superposition water level (reservoir is not in the water level for facing period end reservoir and reaching under the premise of aerial drainage), (reservoir faced in the period for accumulation of energy
The water level as at the beginning of the period, which disappears, drops down onto the theoretical power generation of level of dead water generation), enter can (reservoir last reservoir filling state at the beginning of face the period
Become a mandarin producible theoretical power generation by the period under the premise of constant) and interaction item (enter can and accumulation of energy product)；Face the period
Other input factors of reservoir inflow, subsequent period reservoir inflow and calculating comprising the two, measured value is used in the training stage,
Test phase uses predicted value, since present invention weighing method is probed into, puts aside prediction error, therefore the test rank of the present invention
Duan Caiyong is free of the predicted value of error, i.e. measured value.
Decision variable selects to represent the variable of reservoir period aerial drainage decision, the present invention select generating flow or period contribute or
Period end water level.
(3) the scheduling function structure of each power station synchronization map of step is established；
Using the scheduling function structure of each power station synchronization map of multiple-input and multiple-output：
yi,t=fi,t(xm,i,t), i=1,2 ..., n t=1,2 ..., T m=1,2 ..., M (10)；
In formula, n, T indicate hop count when step hydropower station number (step reservoir number) and schedule periods respectively；yi,tIndicate i-th of water power
It stands the decision variables of t periods；fi,tIndicate the scheduling function of i-th of power station t period；xm,i,tIndicate i-th of power station
M-th of input factor of t periods, such as x1,i,tIndicate the 1st input factor of i-th of power station t period --- at the beginning of facing the period
Water level.
(4) mapping relations and decision are established using random forest regression model；Specifically include following sub-step；
(41) by 51 years before the sample set for training, latter 13 years for testing；
(42) regression model (establishing T × n regression model) of random forest, preferably random forest are established by library by the period
Key parameter --- decision tree number n_estimators and random character variable number max_features, using grid search
Tool searches the optimal value of two key parameters by the method for cross validation, and decision tree number is generally searched between 1~1000
It seeks, sizing grid is set as 50, and the upper limit of random character variable number is input factor sum, so generally searched between 1~M,
Sizing grid is set as 1；Remaining parameter uses default value, default value setting as follows：Criterion=" mse ", max_depth=
None, min_samples_split=2, min_samples_leaf=1, min_weight_fraction_leaf=0,
Max_leaf_nodes=none, min_impurity_split=1e-07, min_impurity_decrease=0,
Bootstrap=true, oob_score=true, n_jobs=1, random_state=none, verbose=0, warm_
Start=false.
(43) it for the structure of every decision tree, is put back to using bootstrapping sampling techniques and randomly selects part
Training sample, obtained training sample subset are about 2/3rds of original training sample, when decision node is divided using with
The Partial Feature variable that machine extracts (number is random character variable number)；It repeats this step and repeatedly builds more decision trees, combine
Random forest is formed, just establishes the mapping relations between the input factor and decision variable in this way.As shown in Figure 2.
(44) to the input factor of the mapping relations input test sample of foundation, the decision variable value of random forest is using every
The mean value of decision tree decision.As shown in Figure 2.
(5) the case where being destroyed to constraints is modified；
Since used random forest is a kind of regression model for establishing mapping relations, gained original analog result very may be used
Constraints can be destroyed, therefore the amendment for carrying out constraint destruction is very necessary.Since the amendment between various boundary conditions may
There are contradiction, such as minimum discharging flow constraint may constrain that there are contradictions with range of stage, thus according to importance it is more big more
Sequence rearward is modified respectively, and this example is corrected by minimum load constraint, minimum discharging flow constraint is corrected and range of stage
Modified sequence is constrained to be modified.
In formula, Ni,t、Respectively the i-th reservoir t periods correct before, correct after the period contribute；N i,tWhen the i-th reservoir t
Section minimum load constraint；Zi,t+1、Respectively correct before, correct after the i-th reservoir t periods period end water level；△ZiIt is
The range of stage of i reservoirs constrains；Zi,tWater level at the beginning of period to correct the preceding i-th reservoir t periods；qi,t、Respectively the i-th library
The t periods correct before, correct after storage outflow；qi,minMeet the minimum of the comprehensive utilizations such as Downstream Navigation, ecology, water supply for the i-th library
Flow.
Constraints, which is destroyed, to be corrected and need to be carried out by library by the period after random forest decision.Specially：The t periods, from upper
To downstream, power station, which checks, one by one corrects minimum load constraint, minimum aerial drainage constraint, range of stage constraint, obtains the t periods for trip
Each power station final decision variate-value, the t+1 periods.
(6) it rolling scheduling and repeatedly calculates；
Since the input factor values of subsequent period are to be determined by decision variable value on last stage, and the present invention wouldn't examine
Consider prediction error, therefore the final decision variate-value for just obtaining regression model on last stage as power station according to the rule-based scheduling
Actual decision situation afterwards, the input as subsequent period carries out rolling scheduling, until traversing entire schedule periods.As shown in figure 3,
Specially：The t periods, using the period each power station random forest regression model, the input factor of first input test sample obtains
To the decision variable value in each power station, then from upstream to downstream, power station checks that the constraint of amendment minimum load, minimum are let out one by one
Stream constraint, range of stage constraint, the so far amendment of t periods finish, and obtain each power station final decision variable of t periods
Value, the t+1 periods and so on.The result of decision is the power generating value or generating flow value of each period in each power station schedule periods
Or the water level value of period Mo, it can accordingly obtain step gross generation in schedule periods；
Due to the randomness of random forest regression model, test phase is computed repeatedly 5~10 times, step is chosen and always generates electricity
The maximum corresponding result of decision of amount is as final decision result.Test phase computes repeatedly selection of times 10 in the embodiment of the present invention
Secondary, numerical results are shown in Fig. 4.
Fig. 4 is certain test year rule-based scheduling and Optimized Operation water level process comparison diagram.Optimized Operation curve refers to utilizing connection
Optimal Operation Model is closed, does not have forecast reservoir inflow (the i.e. actual measurement storage stream of error by all periods in the input scheduling phase
Amount), optimal hydrograph in obtained schedule periods；Rule-based scheduling curve refer to using the present invention method, by by
Period inputs the forecast reservoir inflow (surveying reservoir inflow) of not error, and it is bent to carry out the water level process dispatched by the period
Line.The two the difference is that, although Optimized Operation can obtain optimal hydrograph, its input requirements compared with
Height, it needs entire schedule periods each period to forecast accurate reservoir inflow, however it is often not achieved in mid-and-long term hydrologic forecast
Required precision, but short term hydrological forecasting precision is often higher than mid-and-long term hydrologic forecast, then the present invention uses short-term water
The reservoir inflow of text forecast is as input, and by the outstanding example on period studying history, (outstanding example refers to utilizing combined optimization
Scheduling model calculates the training sample of gained), contacting between input and output is set up using random forest regression model, from
And it is dispatched by the period.Due to Research Methods of the present invention, so the short term hydrological forecasting of input does not consider that prediction error, i.e. input are real
Measurement of discharge, so the validity of assessment method is the curves of water level of the curves of water level and Optimized Operation by measuring rule-based scheduling
Degree of closeness.It can be seen from the figure that the two has higher degree of closeness, it was demonstrated that application effect is preferable.
The present invention is using scheduling function as scheduling rule form.The method used is implicit stochastic optimal regulation, i.e., from base
In the certainty Optimized Operation achievement of history Fuzzy Period of Runoff Series, the quantization sought between reservoir operation decision and reservoir state is closed
System.
The preferred embodiment of the present invention has been described above in detail, still, during present invention is not limited to the embodiments described above
Detail can carry out a variety of equivalents to technical scheme of the present invention within the scope of the technical concept of the present invention, this
A little equivalents all belong to the scope of protection of the present invention.
Claims (9)
1. a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest, which is characterized in that including as follows
Step：
(1) it establishes and solves and target is up to gross generation in the Model for Cascade Hydroelectric Stations phase, is constrained with water balance, up and down
Limit restriction of water level, range of stage constraint, the constraint of minimax traffic constraints, flow luffing, load constraint, the first water level of schedule periods about
Beam and the joint optimal operation model that scheduling end of term restriction of water level is constraints；
(2) decision variable and the input factor for establishing scheduling function, are calculated based on joint optimal operation model result by water power tune
Generate sample set；
(3) the scheduling function structure of each power station synchronization map of step is established；
(4) mapping relations and decision are established using random forest regression model；
(5) the case where being destroyed to constraints is modified；
(6) it rolling scheduling and repeatedly calculates.
2. a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest according to claim 1,
It is characterized in that, joint optimal operation model described in step (1) is up to gross generation in the Model for Cascade Hydroelectric Stations phase
Object function：
In formula, E is gross generation in the Model for Cascade Hydroelectric Stations phase, hop count when n, T are respectively step hydropower station number, schedule periods；TtFor
When segment length；Hi,tThe output of respectively i-th power station t period, generating flow, productive head.
3. a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest according to claim 1,
It is characterized in that, the constraints of joint optimal operation model described in step (1) is specially；
1) water balance constrains
In formula, Vi,t+1、Vi,tFor the i-th library t period Mos, first reservoir storage capacity；Qi,t、Ji,tAnd Si,tRespectively the i-th library t periods entered
Library flow abandons water flow, loss flow；
2) upper and lower limit restriction of water level and range of stage constraint
|Zi,t+1-Zi,t|≤△Zi；
In formula, Zi,t、Zi ,t、Water level, lower limit water level, upper limit water level respectively at the beginning of the i-th library t periods；△ZiAllow for the i-th library
Range of stage；
3) maximum, minimum discharge constraint and the constraint of flow luffing
|qi,t+1-qi,t|<△qi；
In formula, qi,minMeet Downstream Navigation, ecology for the i-th library and the minimum discharge comprehensively utilized that supplies water；qei,maxFor the i-th library water
Take turns unit maximum discharge capacity；△qiAllow letdown flow luffing for the i-th library；qi,t、qi,t+1Respectively the i-th library t periods, t
+ 1 period storage outflow；
4) load constrains
N i,t≤Ni,t≤NYi；
In formula, Ni,t、N i,tThe minimum load in power station is required for the calculating output of the i-th power station t periods, power grid；NYiFor the i-th electricity
It stands installed capacity；
5) schedule periods just, the end of term control restriction of water level
In formula, Zis、Just water level, starting-point detection are calculated for the i-th library schedule periods；Zie、For the i-th library dispatch the end of term calculate water level,
The end of term controls water level；
The joint optimal operation model is to take turns library iterative method as derivation algorithm, and result of calculation is when each power station is each in schedule periods
The optimal water level value in section end.
4. a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest according to claim 1,
It is characterized in that, it is input to grow serial reservoir inflow in step (2) with Hydropower Stations history, joint optimal operation mould is utilized
Type calculates, and obtains in schedule periods the optimal water level value in each library day part end year by year, and calculate by water power tune, that is, it is public to pass through water balance
Formula and Water-sodium disturbance equation calculation generate the sample set of each library length series, which includes test sample and training sample；
Select the variable for representing reservoir period initial equilibrium state and period water situation as the input factor；
Decision variable selects the variable for representing reservoir period aerial drainage decision as decision variable.
5. a kind of Hydropower Stations combined dispatching Rules extraction method based on random forest according to claim 4,
It is characterized in that, the input factor includes water level at the beginning of each library faces the period, faced period reservoir inflow, water at the beginning of a upper period
Position, superposition water level, accumulation of energy, enters energy and interaction item at upper period reservoir inflow, subsequent period reservoir inflow；Wherein, the period is faced
Other input factors comprising the two, measured value is used in the training stage in reservoir inflow, subsequent period reservoir inflow and calculating,
Predicted value is used in test phase；
The decision variable is generating flow or the period contributes or period end water level.
6. the Hydropower Stations combined dispatching Rules extraction method according to claim 1 based on random forest, special
Sign is, using the scheduling function structure of each power station synchronization map of multiple-input and multiple-output in step (3)：
yi,t=fi,t(xm,i,t), i=1,2 ..., n t=1,2 ..., T m=1,2 ..., M；
In formula, n, T indicate hop count when step hydropower station number and schedule periods respectively；yi,tIndicate the decision of i-th of power station t period
Variable；fi,tIndicate the scheduling function of i-th of power station t period；xm,i,tIndicate i-th of power station t period, m-th of input
The factor.
7. the Hydropower Stations combined dispatching Rules extraction method according to claim 1 based on random forest, special
Sign is that the step (4) further comprises：
(41) by k before the sample set for training, rear (K-k) year is for testing；
(42) regression model for establishing random forest by library by the period, that is, establish T × n regression model, the parameter of random forest
Middle decision tree number n_estimators and random character variable number max_features, passes through intersection using grid search tool
The method of verification searches the optimal value of two parameters, and decision tree number is generally searched between 1~1000, random character variable number
The upper limit be input the factor sum, so generally being searched between 1~M；Remaining parameter uses default value；
(43) it for the structure of every decision tree, is put back to using bootstrapping sampling techniques and randomly selects part training
Sample, obtained training sample subset are about 2/3rds of original training sample, are taken out using random when decision node is divided
The Partial Feature variable taken；It repeats this step and repeatedly builds more decision trees, combination forms random forest, just establishes so defeated
Enter the mapping relations between the factor and decision variable；
(44) to the input factor of the mapping relations input test sample of foundation, the decision variable value of random forest is determined using every
The mean value of plan tree decision.
8. the Hydropower Stations combined dispatching Rules extraction method according to claim 1 based on random forest, special
Sign is, for the decision variable value obtained based on random forest in step (5), carries out the amendment of constraints destruction：Foundation
Different reservoir situations carry out minimum load constraint amendment by importance more big sequence more rearward, minimum discharging flow constraint is repaiied
The amendment just constrained with range of stage, correction formula are as follows：
In formula, Ni,t、Respectively the i-th reservoir t periods correct before, correct after the period contribute；N i,tI-th reservoir t periods are most
Small units limits；Zi,t+1、Respectively correct before, correct after the i-th reservoir t periods period end water level；△ZiFor the i-th water
The range of stage in library constrains；Zi,tWater level at the beginning of period to correct the preceding i-th reservoir t periods；qi,t、When respectively the i-th library t
Before Duan Xiuzheng, correct after storage outflow；qi,minMeet Downstream Navigation, ecology for the i-th library and the minimum discharge comprehensively utilized that supplies water；
Constraints destroys amendment need to be after using random forest decision by period power station progress one by one, specially：When t
Section, from upstream to downstream, power station, which checks, one by one corrects minimum load constraint, minimum aerial drainage constraint, range of stage constraint, t+
1 period and so on.
9. the Hydropower Stations combined dispatching Rules extraction method according to claim 1 based on random forest, special
Sign is, using by the period, by the method for library rolling scheduling, the above period decision value is as the defeated of subsequent period in step (6)
Enter value and carry out rolling scheduling, until scrolling through entire schedule periods；Specially：It is the t periods, random using the period each power station
The input factor of forest regression model, first input test sample obtains the decision variable value in each power station, then from upstream under
Trip, power station, which checks, one by one corrects minimum load constraint, minimum aerial drainage constraint, range of stage constraint, the so far amendment of t periods
It finishes, obtains each power station final decision variate-value of t periods, due to the randomness of random forest regression model, to test
Stage computes repeatedly 5~10 times, chooses the maximum corresponding result of decision of step gross generation as final decision result；T+1
Period and so on；The result of decision is the power generating value or generating flow value of each period or period Mo in each power station schedule periods
Water level value, accordingly obtain step gross generation in schedule periods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810465449.3A CN108647829A (en) | 2018-05-16 | 2018-05-16 | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810465449.3A CN108647829A (en) | 2018-05-16 | 2018-05-16 | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108647829A true CN108647829A (en) | 2018-10-12 |
Family
ID=63755789
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810465449.3A Pending CN108647829A (en) | 2018-05-16 | 2018-05-16 | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108647829A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409434A (en) * | 2018-02-05 | 2019-03-01 | 福州大学 | The method of liver diseases data classification Rule Extraction based on random forest |
| CN110533244A (en) * | 2019-08-28 | 2019-12-03 | 重庆大学 | A kind of step dam Optimization Scheduling, system and computer readable storage medium |
| CN111008790A (en) * | 2019-12-23 | 2020-04-14 | 华中科技大学 | Hydropower station group power generation electric scheduling rule extraction method |
| CN111353649A (en) * | 2020-03-06 | 2020-06-30 | 大连理工大学 | KL expansion-based large-scale hydropower station group optimal scheduling dimension reduction method |
| CN113204583A (en) * | 2021-04-14 | 2021-08-03 | 武汉大学 | Reservoir group scheduling decision behavior mining method and reservoir scheduling automatic control device |
| CN113962463A (en) * | 2021-10-25 | 2022-01-21 | 河海大学 | Reservoir optimal scheduling rule extraction method with dynamic time-varying structure |
| CN116186628A (en) * | 2023-04-23 | 2023-05-30 | 广州钛动科技股份有限公司 | APP automatic marking method and system |
| CN116227863A (en) * | 2023-02-22 | 2023-06-06 | 兰州理工大学 | Cascade hydropower station optimal scheduling method based on Harris eagle optimization algorithm |
| CN117421558A (en) * | 2023-10-26 | 2024-01-19 | 华中科技大学 | Cascade reservoir operation rule extraction and model training method thereof |
| CN118153734A (en) * | 2024-01-19 | 2024-06-07 | 水利部水利水电规划设计总院 | Multi-objective optimization scheduling rule extraction method for water network regulation engineering group |
| CN118333242A (en) * | 2024-06-13 | 2024-07-12 | 水利部交通运输部国家能源局南京水利科学研究院 | Cascade reservoir group joint scheduling rule optimization method based on big data samples |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101149822A (en) * | 2007-10-15 | 2008-03-26 | 马志鹏 | Reservoir forecast dispatch grey dynamic programming method |
| CN106022960A (en) * | 2016-05-19 | 2016-10-12 | 武汉大学 | Data mining-based optimal selection method for reservoir multi-target dispatching rule parameters |
| CN106228277A (en) * | 2016-08-16 | 2016-12-14 | 武汉大学 | A kind of reservoir operation forecast information effective accuracy recognition methods based on data mining |
-
2018
- 2018-05-16 CN CN201810465449.3A patent/CN108647829A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101149822A (en) * | 2007-10-15 | 2008-03-26 | 马志鹏 | Reservoir forecast dispatch grey dynamic programming method |
| CN106022960A (en) * | 2016-05-19 | 2016-10-12 | 武汉大学 | Data mining-based optimal selection method for reservoir multi-target dispatching rule parameters |
| CN106228277A (en) * | 2016-08-16 | 2016-12-14 | 武汉大学 | A kind of reservoir operation forecast information effective accuracy recognition methods based on data mining |
Non-Patent Citations (2)
| Title |
|---|
| 严梦佳 等: "基于随机森林的水电站发电调度规则研究", 《水力发电》 * |
| 钟平安 等: "梯级库群水流滞后性影响的日优化调度模型研究", 《水力发电学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409434A (en) * | 2018-02-05 | 2019-03-01 | 福州大学 | The method of liver diseases data classification Rule Extraction based on random forest |
| CN110533244A (en) * | 2019-08-28 | 2019-12-03 | 重庆大学 | A kind of step dam Optimization Scheduling, system and computer readable storage medium |
| CN111008790A (en) * | 2019-12-23 | 2020-04-14 | 华中科技大学 | Hydropower station group power generation electric scheduling rule extraction method |
| CN111353649A (en) * | 2020-03-06 | 2020-06-30 | 大连理工大学 | KL expansion-based large-scale hydropower station group optimal scheduling dimension reduction method |
| CN111353649B (en) * | 2020-03-06 | 2022-03-29 | 大连理工大学 | KL expansion-based large-scale hydropower station group optimal scheduling dimension reduction method |
| CN113204583A (en) * | 2021-04-14 | 2021-08-03 | 武汉大学 | Reservoir group scheduling decision behavior mining method and reservoir scheduling automatic control device |
| CN113204583B (en) * | 2021-04-14 | 2022-11-01 | 武汉大学 | Reservoir group scheduling decision behavior mining method and reservoir scheduling automatic control device |
| CN113962463A (en) * | 2021-10-25 | 2022-01-21 | 河海大学 | Reservoir optimal scheduling rule extraction method with dynamic time-varying structure |
| CN116227863B (en) * | 2023-02-22 | 2024-03-12 | 兰州理工大学 | Cascade hydropower station optimal scheduling method based on Harris eagle optimization algorithm |
| CN116227863A (en) * | 2023-02-22 | 2023-06-06 | 兰州理工大学 | Cascade hydropower station optimal scheduling method based on Harris eagle optimization algorithm |
| CN116186628A (en) * | 2023-04-23 | 2023-05-30 | 广州钛动科技股份有限公司 | APP automatic marking method and system |
| CN117421558A (en) * | 2023-10-26 | 2024-01-19 | 华中科技大学 | Cascade reservoir operation rule extraction and model training method thereof |
| CN118153734A (en) * | 2024-01-19 | 2024-06-07 | 水利部水利水电规划设计总院 | Multi-objective optimization scheduling rule extraction method for water network regulation engineering group |
| CN118153734B (en) * | 2024-01-19 | 2024-09-06 | 水利部水利水电规划设计总院 | Multi-objective optimization scheduling rule extraction method for water network regulation engineering group |
| CN118333242A (en) * | 2024-06-13 | 2024-07-12 | 水利部交通运输部国家能源局南京水利科学研究院 | Cascade reservoir group joint scheduling rule optimization method based on big data samples |
| CN118333242B (en) * | 2024-06-13 | 2024-08-16 | 水利部交通运输部国家能源局南京水利科学研究院 | Cascade reservoir group joint scheduling rule optimization method based on big data samples |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647829A (en) | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest | |
| Wu et al. | Optimal location selection for offshore wind-PV-seawater pumped storage power plant using a hybrid MCDM approach: A two-stage framework | |
| CN108470249B (en) | A kind of Hydropower Stations short-term electricity generation dispatching method coupling clustering and decision tree | |
| Li et al. | Toward cost-oriented forecasting of wind power generation | |
| Ji et al. | Inexact stochastic risk-aversion optimal day-ahead dispatch model for electricity system management with wind power under uncertainty | |
| CN105719015A (en) | PEPSO-basedsiting and sizing method optimization method of distribution type power supply | |
| Tohidi et al. | Multi-regional transmission planning as a non-cooperative decision-making | |
| CN104463356A (en) | Photovoltaic power generation power prediction method based on multi-dimension information artificial neural network algorithm | |
| CN107609716B (en) | A kind of power station load setting prediction technique | |
| CN103904664B (en) | A kind of AGC unit real-time scheduling method based on effective static security territory | |
| CN109934395A (en) | A kind of more small power station areas Short-Term Load Forecasting Method based on timesharing subregion meteorological data | |
| CN110210670A (en) | A kind of prediction technique based on power-system short-term load | |
| CN108734378A (en) | A kind of reservoir operation state of risk groups Decision Method under the conditions of imperfect information | |
| CN106372763A (en) | Improved-support-vector-machine-based power load prediction correction method | |
| CN107292537A (en) | Consider the typhoon day short-term load forecasting method of path trace and the condition of a disaster Risk Evaluation | |
| Toufani et al. | Optimization of pumped hydro energy storage systems under uncertainty: A review | |
| Lee et al. | Probabilistic wind power forecasting based on the laplace distribution and golden search | |
| Scarcelli et al. | Ensemble of Markovian stochastic dynamic programming models in different time scales for long term hydropower scheduling | |
| Zambelli et al. | Long-term hydropower scheduling based on deterministic nonlinear optimization and annual inflow forecasting models | |
| CN105809272A (en) | Step hydropower station group instruction scheduling optimization method based on data mining | |
| CN106910010A (en) | A kind of active distribution network uncertainty tidal current analysis method and device | |
| Makri et al. | Short term net imbalance volume forecasting through machine and deep learning: a UK case study | |
| Montoya et al. | Day-ahead and week-ahead solar PV power forecasting using deep learning neural networks | |
| Oubnaki et al. | Energy Consumption in the Transport Sector: Trends and Forecast Estimates in Morocco | |
| Kim et al. | A multiple objective mathematical model for daily coordinated multi-reservoir operation |
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: 20181012 |
|
| RJ01 | Rejection of invention patent application after publication |