CN113657032A - Low-frequency load shedding method and system for pre-centralized coordination and real-time distributed control - Google Patents
Low-frequency load shedding method and system for pre-centralized coordination and real-time distributed control Download PDFInfo
- Publication number
- CN113657032A CN113657032A CN202110924308.5A CN202110924308A CN113657032A CN 113657032 A CN113657032 A CN 113657032A CN 202110924308 A CN202110924308 A CN 202110924308A CN 113657032 A CN113657032 A CN 113657032A
- Authority
- CN
- China
- Prior art keywords
- load
- time
- average
- future
- line
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000005520 cutting process Methods 0.000 claims abstract description 28
- 230000009471 action Effects 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 49
- 238000013528 artificial neural network Methods 0.000 claims description 32
- 230000008859 change Effects 0.000 claims description 28
- 238000011156 evaluation Methods 0.000 claims description 21
- 210000002569 neuron Anatomy 0.000 claims description 20
- 238000005070 sampling Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000035945 sensitivity Effects 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 230000035508 accumulation Effects 0.000 claims description 3
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 230000009467 reduction Effects 0.000 description 6
- 238000011161 development Methods 0.000 description 5
- 230000001932 seasonal effect Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000007123 defense Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid 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]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
- Y02B70/00—Technologies for an efficient end-user side electric power management and consumption
- Y02B70/30—Systems integrating technologies related to power network operation and communication or information technologies for improving the carbon footprint of the management of residential or tertiary loads, i.e. smart grids as climate change mitigation technology in the buildings sector, including also the last stages of power distribution and the control, monitoring or operating management systems at local level
- Y02B70/3225—Demand response systems, e.g. load shedding, peak shaving
-
- 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
-
- 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
- Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
- Y04S20/20—End-user application control systems
- Y04S20/222—Demand response systems, e.g. load shedding, peak shaving
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a low-frequency load shedding method for pre-centralized coordination and real-time distributed control, which comprises the following steps: predicting to obtain the future T of the load linefPredicting the moment load; searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping when the sum of the predicted values of the moment load reaches the low-frequency required cutting amountSearching for future TfSwitching load lines at any time; will be T in futurefThe circuit information of the load to be cut at the moment is converted into a low-frequency tripping outlet fixed value and is issued to the device for the device to switch to the T statefAnd when the low-frequency fault is judged to occur at any moment, the corresponding load circuit is cut off according to the fixed-value action of the low-frequency trip outlet by combining a local anti-misoperation strategy. The invention can predict the predicted value of the load, improve the load shedding precision as much as possible, avoid under-shedding and over-shedding and realize accurate load control.
Description
Technical Field
The invention belongs to the technical field of safety and stability of an electric power system, and particularly relates to a low-frequency load shedding method for pre-centralized coordination and real-time distributed control, and a low-frequency load shedding system for pre-centralized coordination and real-time distributed control.
Background
The scale and the range of the interconnected power grid are multiplied, the new energy ratio is rapidly improved, and the characteristics of the power grid are fundamentally changed. The electric power market makes the operation mode change various, and the electric wire netting stability level change surveys more and more, because the probability that successive incident leads to interconnected network to lose the safety and stability greatly increases.
For a long time, the low-frequency load reduction device is used as important technical equipment for a third safety and stability line, and plays an important role in ensuring the safe and stable operation of a power system and preventing the occurrence of a blackout accident. The traditional low-frequency load reduction device only reflects the change of local electric quantity to cut off a locally determined load line, which is beneficial to ensuring the reliability of the action of the device on one hand, but is not beneficial to the coordination and optimization of the overall control action of the system on the other hand, even in the event of a power failure of a foreign power system, the situation that the low-frequency load reduction device cannot be timely and correctly reflected in the dynamic state of the system is found, and the situation is also one embodiment of the complicated system stability caused by the development of the power system.
At present, a large number of low-frequency load shedding devices are deployed in an Anhui power grid, and include a novel centralized low-frequency load shedding device (hereinafter, referred to as a centralized device) and a distributed device (hereinafter, referred to as a distributed device) integrating a low-frequency function in a protection device, but all currently used low-frequency load shedding devices generally have the defects that the overall cooperation degree is not high, a load line cannot be accurately cut off, the measures of the low-frequency load shedding devices are not sufficiently coordinated with a second defense line power grid safety and stability control device, and the like, and are mainly expressed in that:
(1) with the wide access of novel energy utilization equipment such as distributed power supplies, micro-grids, energy storage and electric vehicles, the power supply and demand forms present diversified characteristics, and the load characteristics present obvious difference and complementarity. The proportion of the original traditional load circuit with the dual characteristics of source load is continuously increased, and the conventional low-frequency device does not detect the power direction of the circuit, so that the circuit cut off in action cannot be guaranteed to be a load circuit. When the distributed photovoltaic output is larger, the power of a line where the photovoltaic power supply is located is reversely transmitted to the system side, and the reduction amplitude of the low-frequency load shedding load control rate is larger; with the increase of the scale of the distributed photovoltaic grid connection, the low-frequency load shedding load control rate is gradually reduced in a large photovoltaic output period.
(2) The low-frequency load shedding measures are configured according to the proportion of the total load of the power grid, the device cannot monitor the line power information of the transformer substation, only can cut off specific lines according to a preset fixed sequence, and cannot accurately master the load shedding of each turn. The device adopts blind cutting during action, the actual load cutting amount is difficult to be accurately counted, and the over-cutting or under-cutting condition exists. The method can not accurately evaluate whether the actual load shedding amount meets the requirement of the load shedding proportion when the extreme fault of the power grid occurs, and the risk of power grid breakdown exists.
The new conditions cause the inadaptability of the third line defense low-frequency load reduction measure of the power grid, and the method can be mainly summarized into two points, namely, the load line and the power line are difficult to distinguish when the device acts; secondly, the cuttable load can not be guaranteed to meet the requirement of low-frequency action cutting, and the two points seriously affect the frequency and voltage stability of the system.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a low-frequency load shedding method for pre-centralized coordination and real-time distributed control, and solves the technical problem that the device in the prior art cannot calculate in real time according to the power grid condition because the total load shedding amount is preset.
In order to solve the technical problems, the technical scheme of the invention is as follows.
In a first aspect, the present invention provides a low frequency load shedding method for pre-centralized coordination of real-time distributed control, which comprises the following steps:
obtaining the active power of a load line, and calculating to obtain the low-frequency required tangent;
predicting to obtain the future T of the load linefPredicting the moment load;
searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping searching when the sum of the predicted values of the moment load reaches the low-frequency required cutting amount to obtain the future TfSwitching load lines at any time;
will be T in futurefThe circuit information of the load to be cut at the moment is converted into a low-frequency tripping outlet fixed value and is issued to the device for the device to switch to the T statefAnd when the low-frequency fault is judged to occur at any moment, the corresponding load circuit is cut off according to the fixed-value action of the low-frequency trip outlet by combining a local anti-misoperation strategy.
Optionally, the prediction yields a future TfThe moment load predicted value comprises the following steps:
based on future TfThe time information is related to the time, and the future T of the load line is obtained by predictionfPredicted load at different time scales including future TfThe annual average predicted load, the monthly average predicted load and the time-interval average predicted load of the moment load;
the obtained annual average predicted load P of the load circuit at the future Tf momentav_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK) Inputting the load prediction model obtained by training the BP artificial neural network together with the meteorological factor data of the area to obtain the future T of the load linefAnd (4) predicting the moment load.
Optionally, the meteorological factor data includes air temperature data, solar intensity, haze index, and cloud cover.
Optionally, the base is based on future TfThe time information is related to the time, and the future T of the load line is obtained by predictionfThe predicted load of different time scales at different moments comprises the following steps:
s1, obtaining load line prediction TfYear T of timeYYear-averaged load sequence of previous i years { P }av_Y(tn) 1-i, wherein the annual average load Pav_Y(tn) Calculated as follows:
wherein, p (T) is a power average value of a certain line in a time interval of delta T after time T, the value of delta T is 1 year, and p (T) is real-time sampling data in the time interval of delta T;
considering the sampling values as a discrete sequence, the formula discretization is expressed as:
wherein, p (t)n) The sequence is discretized at equal intervals on a time axis by power p (T) in a time interval of delta T, wherein the delta T is a sampling interval, and the value of the delta T is 1 year;
s2, obtaining load line prediction TfMonth T of timeMMonthly average load sequence { P) of m months beforeav_M(tn) 1-m, wherein the average monthly load Pav_M(tn) Calculated according to the following formula:
wherein, p (t)n) The sequence is discretized at equal intervals on a time axis by power p (T) in a time interval of delta T, wherein the delta T is a sampling interval, and the value of the delta T is 1 month;
s3, predicting T according to the load line to be predictedfAttribute of time TWDetermine if it is a weekend, based on the date TDPushing forward the date of which the day is the weekend or the weekday, skipping the attribute TWA date of inconsistency; dividing one day into a plurality of time intervals, wherein the difference between every two time intervals is the interval time of two predictions, and the time interval T is the time interval from the current timeK’Push forward a set of P slots of { T }k1-P }; first, the average value of the load for a certain period is calculated:
wherein, p (t)n) For a sequence in which the power p (T) is discretized at equal intervals on the time axis within a period of Δ T, Δ T being the sampling interval, Δ T here being taken to be (T)k+1-Tk);
Next, the attribute T is calculatedWSame q days same period TkAverage value P of average load ofTk_DqThen attribute TWSame q days same period TkThe average load of (a) is taken as the average of the column-wise accumulations in the table, e.g. T over q days1The average of the load over the time period is:
calculating the average value of the load of all the time periods in q days to obtain a time period average load sequence { P }av_TK(tn)|n=1~p}。
Then, respectively carrying out polynomial fitting on the annual average load sequence, the monthly average load sequence and the time-interval average load sequence to obtain a corresponding annual average load curve function, monthly average load curve function and time-interval average load curve function;
respectively combine T withfYear T of timeYThe month of TMAnd the period TKSubstituting the average annual load curve function, the average monthly load curve function and the average time interval load curve function to obtain the average annual predicted load P at the future Tf momentav_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK)。
Optionally, the performing polynomial fitting on the annual average load sequence, the monthly average load sequence and the time-interval average load sequence respectively to obtain corresponding annual average load curve function, monthly average load curve function and time-interval average load curve function includes:
s1, fitting an annual average load curve by using a (i-1) th-order polynomial, and setting an annual average load curve function determined by the annual average load of i as follows:
Pav_Y(t)=a0ti-1+a1ti-2+…+ai-1
the previously obtained annual average load series { P }av_Y(tn) Substituting | n | -1 to i } into the above formula, each coefficient a of the function can be determined0…ai-1;
S2. similarly, obtaining a monthly average load curve function as:
Pav_M(t)=b0tm-1+b1tm-2+…+bm-1
the previously obtained monthly average load sequence { P }av_M(tn) Substituting | n | -1 to m } into the above formula, each coefficient b of the function can be determined0…bi-1;
S3, similarly, obtaining a time interval average load curve function as follows:
Pav_TK(t)=c0tp-1+c1tp-2+…+cp-1
will be obtained beforeTime-lapse average load sequence { P }av_TK(tn) Substituting | n | -1 to p } into the above equation, each coefficient c of the function can be determined0…ci-1。
Optionally, the BP artificial neural network comprises forward delivery of information and back propagation of errors; in the forward propagation process, input information is transmitted to an output layer from an input layer through a hidden layer by layer calculation, the state of each layer of neuron only affects the state of the next layer of neuron, if expected output is not obtained in the output layer, an error change value of the output layer is calculated, then the error is transmitted in a reverse direction, and an error signal is transmitted back along an original connecting path through a network to modify the weight of each neuron until an expected target is reached;
let the input node be xj(j is 1 … r), r is the number of input nodes, and hidden nodes are yi(i=1…S1),S1For the number of output nodes, the output node is ok(k=1…S2),S2Is the number of output nodes; the network weight between the input node and the hidden node is w1ijThe connection weight between the hidden node and the output node is w2kiThe threshold value of the hidden node is theta1iThe threshold value of the output contact is theta2kThe activation functions of the hidden layer and the output layer areThe desired output is tk;
The algorithm steps of the BP neural network are as follows:
(1) forward propagation of information
The output of the ith neuron in the hidden layer is:
the output of the kth neuron in the output layer is:
defining an error function as:
(2) solving weight value change and threshold value change by using a gradient descent method;
the weight change of the output layer has the following weight from the ith input to the kth output:
the threshold change of the output layer has the following threshold values from the ith input to the kth output:
the weight change of the hidden layer has the following weight values from the jth input to the ith output:
wherein, γiThe output of the ith hidden layer node;
the threshold change of the hidden layer has the following threshold values from the jth input to the ith output:
and training the BP neural network by using the annual average predicted load, the monthly average predicted load, the time-interval average predicted load and meteorological factor data in the historical load data as the input of the BP neural network and the load predicted value as the output of the neural network.
Optionally, the searching for a line that can be switched, comprises:
s1, sorting the cutting priorities of the load lines according to the multi-dimensional attribute comprehensive evaluation score of each load line, determining the level fixed value of each load line according to the sorting priorities, and cutting the lines with high priority preferentially;
s2, dividing each load line to a corresponding logic substation according to the substation to which the load line actually belongs, and setting a priority fixed value for each logic substation;
s3, sequentially searching the switchable load lines according to the sequence of the two-dimensional list of the priority levels of the first logic substation and the second logic substation until the searched switchable load lines reach the future TfThe sum of the predicted values of the moment load reaches the low-frequency required cut PTAnd stopping searching to obtain the load to be cut.
Optionally, the search process is:
firstly, all load lines in level 1 are searched for the switchable load lines according to the priority of the logic substation from high to low, and when the switchable load lines are searched for in the future TfWhen the sum of the predicted values of the loads at the moment is larger than the low-frequency required switching value, the search is stopped, and all the searched switchable load lines are the future TfA line is required to be cut at any time;
if all the load lines in the level 1 are searched, the searched future T of the switchable load linefWhen the sum of the predicted values of the loads is still smaller than the low-frequency required switching value, the next level is continuously searched, and the process is repeated until the searched future T of the load-cutting-able line is metfThe sum of the predicted values of the loads at the moment is greater than the low-frequency load shedding requirement condition, the search is stopped, and all the searched load-shedding lines which can be cut are the future TfA line is required to be cut at any time;
before the search is stopped, the load circuit in the last logic substation of the last hierarchy searched is according to the future T because the load circuit in the last logic substation in the last hierarchy is partially cut offfAnd searching the moment load predicted values in the descending order.
Optionally, the load circuit cutting priorities are ranked according to the multi-dimensional attribute comprehensive evaluation score of each load circuit, and a level fixed value of each load circuit is determined according to the ranking;
counting attributes of interruption loss, an affiliated power utilization department, sensitivity and an electrical distance in four dimensions, and taking the attributes as evaluation indexes; for a load line, the lower the score of a certain evaluation index is, the lower the cutting priority of the load line in the dimension is;
distributing a weight to each evaluation index, and multiplying each evaluation index score of the load line by the weight of the evaluation index score to obtain a weighted comprehensive score of each load line;
and sequencing according to the weighted comprehensive scores of the load lines so as to determine the level definite value of each load line.
Optionally, the priority rating of the logic substation is determined according to an average value of the comprehensive evaluation scores of all load circuits administered by the logic substation.
Optionally, the local error prevention policy is that when the device is operated, the real-time power P of the line is required to be cut onlyr>0, the outlet is allowed.
In a second aspect, the present invention further provides a low frequency load shedding system for pre-centralized coordination and real-time distributed control, comprising:
the device is used for acquiring active power of a load line, uploading the active power to a main station, receiving a low-frequency trip exit fixed value issued by the main station at intervals, and sending the low-frequency trip exit fixed value to the main station at TfWhen a low-frequency fault is judged to occur, a local anti-misoperation strategy is combined, and a corresponding load circuit is cut off according to the fixed-value action of a low-frequency trip outlet;
the master station is used for receiving the active power of the load line and calculating to obtain the low-frequency required switching value based on the active power of the load line; predicting to obtain the future T of the load linefPredicting the moment load; searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping searching when the sum of the predicted values of the moment load reaches the low-frequency required cutting amount to obtain the future TfSwitching load lines at any time; will be T in futurefAnd converting the information of the line needing to be switched into a low-frequency tripping outlet fixed value at the moment and sending the low-frequency tripping outlet fixed value to the device.
Compared with the prior art, the invention has the following beneficial effects: the invention calculates the low-frequency required load shedding amount in real time, so that the load shedding amount can be matched with the operation mode and the fault condition of the power grid as much as possible; comprehensive coordination control of load shedding can be performed in the region, the requirement of the required shedding amount can be met to the maximum extent, meanwhile, the minimum over-shedding can be guaranteed, and the problem of insufficient low-frequency control amount under the condition of photovoltaic heavy-duty is effectively solved; monitoring real-time power of a load line, and avoiding cutting off a line with a source attribute; the safety and the reliability of the system are ensured to the maximum extent by means of pre-coordination of the master station end and distributed control of the device end.
Drawings
FIG. 1 is a diagram of a typical architecture of a regional low frequency load shedding apparatus;
FIG. 2 is a main flow chart of the method of the present invention;
fig. 3 is a schematic diagram of a neural network structure.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
The invention provides a low-frequency load shedding method of 'pre-centralized coordination and real-time distributed control', wherein the pre-centralized coordination refers to the power shortage condition when a system operation mode and a power flow condition calculation system in a comprehensive area have low-frequency faults; receiving real-time electric quantity information and line switching and outage information sent by a low-frequency load reduction device (hereinafter referred to as a device), decomposing load data in multiple time scales according to a curve obtained by fitting historical load data, and predicting results of multiple time scales and future TfThe weather factor data at the moment is used as the input of BP artificial neural network, and T in the futurefThe load of the line at the moment is used as output, and historical data is used for training to realize prediction of the load of the line at the future Tf moment; each line TfSumming the predicted power at a moment to obtain total cutting capacity, generating two-dimensional load sequencing according to a load level and an area to which the load level belongs, dynamically distributing the cutting capacity of the load when each transformer substation breaks down, and controlling a master station (hereinafter referred to as the master station) to cut the load at a certain intervalThe load shedding amount is issued to the low-frequency device in a mode of a low-frequency trip outlet fixed value. The real-time distribution control means that the device receives T transmitted by the master station at intervalsfConstant value of low-frequency trip outlet at moment is stored, and when T is reachedfAt that moment, the device updates the real-time trip outlet fixed value to the fixed value. And when a low-frequency fault occurs, the fault is judged only based on local information, and a certain local anti-error strategy is combined, so that the corresponding load circuit is cut off according to a trip outlet fixed value action issued by the main station in advance.
The method can realize the coordination control of the load between stations and in the stations, improve the load shedding precision as much as possible, avoid under-shedding and over-shedding, avoid the line with the attribute of 'source' and important sensitive users, realize accurate load control and reduce the cost influence of stability control measures; meanwhile, the real-time distributed control does not depend on a channel and only judges a fault and acts an outlet based on local information, so that the reliability of the real-time distributed control is also guaranteed.
Example 1
The invention discloses a low-frequency load shedding method for pre-centralized coordination and real-time distributed control, which is described by taking a certain round of load shedding amount as an example, and is shown in figure 2, and comprises the following steps:
step 1: the device monitors the active power P of the line in real time and sends the active power P to a master station; the master station calculates the low-frequency required tangent P according to the current power grid operation mode and the current power flow section conditionT。
As shown in a typical architecture diagram of the regional low-frequency load shedding device in fig. 1, a master station is used as an internal application of a power system control system at a master station scheduling end, and communicates with a distributed/centralized low-frequency load shedding device in a substation through a scheduling data network. The distributed low-frequency device is connected with a remote machine set through an industrial Ethernet (for a protection and test integrated device without an Ethernet port, the protection and test integrated device needs to be switched through a serial server), and is communicated with the master station through a scheduling data network through the remote machine; the centralized low frequency device communicates with the master station directly through a dispatch data network. The device sends the active power P of the load line to the main station. The system operation mode in the master station comprehensive area and the power shortage condition when the active power P of the load line calculates the low-frequency fault of the system are obtainedSystem low frequency cut-off requirement P when specific low frequency fault occursTI.e. the predetermined amount of load shedding of the low frequency shedding.
Step 2: the Master station obtains the future TfObtaining historical power data and predicting the data to obtain the future T of the load linefPredicted load at different time scales including future TfThe annual average predicted load, the monthly average predicted load and the time-interval average predicted load of the moment load;
in said step 2, the primary station first obtains TfTime information associated with time of day, including TfYear T of timeYThe month of TMDate of, TDDay of week TWAnd the period TK. Week TWThe value range is Monday to Sunday; time interval TKAnd TK+1The difference is the interval time of two predictions, the shorter the time is, the higher the load prediction precision is, and the larger the occupation of communication resources and calculation resources is, the larger the occupation can be selected according to the actual requirements of the engineering.
The main station acquires historical power data and predicts the data to obtain the future TfThe load time sequence with different time scales at different moments specifically comprises the following steps:
s1, obtaining prediction T of certain load linefYear T of timeYYear-averaged load sequence of previous i years { P }av_Y(tn) 1-i, wherein the annual average load Pav_Y(tn) Calculated as follows:
wherein, p (T) is a power average value in a time interval of Δ T after a time T of a certain line, where Δ T is 1 year, and p (T) is real-time sampling data in the time interval of Δ T.
Considering the sampling values as a discrete sequence, the formula discretization is expressed as:
wherein, p (t)n) The power p (T) is discretized at equal intervals on a time axis in a delta T period, delta T is a sampling interval, and delta T takes 1 year.
The annual average load represents the annual average level of the load, the influence of seasonal changes on the load is filtered by smoothing the load by taking the year as a unit, and the development trend of the annual average load reflects the development trend of national economy.
S2, obtaining a certain load line prediction TfMonth T of timeMMonthly average load sequence { P) of m months beforeav_M(tn) And | n is 1-m, and the value of m is 12 in consideration of covering complete seasonal changes. Average monthly load Pav_M(tn) Calculated according to the following formula:
wherein, p (t)n) For a sequence with power p (T) discretized at equal intervals on the time axis within a period of Δ T, Δ T being the sampling interval, Δ T here being 1 month.
The monthly average load generated by smoothly generating the load by taking a month as a unit filters out daily load fluctuation and only reflects the influence of seasonal change on the load.
S3, predicting T according to the line to be predictedfAttribute of time TWDetermine if it is a weekend, based on the date TDPushing forward the date of which the day is the weekend or the weekday, skipping the attribute TWInconsistent dates. Dividing one day into a plurality of time intervals, wherein the difference between every two time intervals is the interval time of two predictions, and the time interval T is the time interval from the current timeK’Push forward a set of P slots of { T }k1-P }. First, the average value of the load for a certain period is calculated:
wherein, p (t)n) For a sequence in which the power p (T) is discretized at equal intervals on the time axis within a period of Δ T, Δ T being the sampling interval, Δ T here being taken to be (T)k+1-Tk)。
Next, the attribute T is calculatedWSame q days same period TkAverage value P of average load ofTk_DqThe dates and periods are shown in the following table:
then attribute TWSame q days same period TkThe average load of (a) is taken as the average of the column-wise accumulations in the table, e.g. T over q days1The average of the load over the time period is:
calculating the average value of the load of all the time periods in q days to obtain a time period average load sequence { P }av_TK(tn)|n=1~p}。
Selecting and predicting TfThe load of a certain period is accumulated and averaged on q days which are weekends or working days, the fluctuation of the load of the working days or weekends is considered, the influence of random weather factors on the load (distributed photovoltaic) is filtered, and the fluctuation of the load at different periods within one day is only reflected.
Then, polynomial fitting is respectively carried out on the annual average load sequence, the monthly average load sequence and the time interval average load sequence to obtain the future TfThe annual average predicted load, the monthly average predicted load and the time-interval average predicted load of the moment load,
the fitting method specifically comprises the following steps:
s1, fitting an annual average load curve by using a (i-1) th-order polynomial, and setting an annual average load curve function determined by the annual average load of i years as follows:
Pav_Y(t)=a0ti-1+a1ti-2+…+ai-1
the previously obtained annual average load series { P }av_Y(tn) Substituting | n | -1 to i } into the above formula, each coefficient a of the function can be determined0…ai-1. In practical engineering, the larger the value of the data point i of the annual average load used for fitting the curve is, the longer the time required for calculating the function coefficient is, the better the curve fitting effect is, and generally, the value of i should be selected as small as possible as long as the judgment of the annual average load trend is not influenced.
S2. similarly, obtaining a monthly average load curve function as:
Pav_M(t)=b0tm-1+b1tm-2+…+bm-1
the previously obtained monthly average load sequence { P }av_M(tn) Substituting | n | -1 to m } into the above formula, each coefficient b of the function can be determined0…bi-1。
S3, similarly, obtaining a time interval average load curve function as follows:
Pav_TK(t)=c0tp-1+c1tp-2+…+cp-1
averaging the previously obtained time interval load sequence { Pav_TK(tn) Substituting | n | -1 to p } into the above equation, each coefficient c of the function can be determined0…ci-1。
S4, respectively combining TfYear T of timeYThe month of TMAnd the period TKSubstituting the average annual load curve function, the average monthly load curve function and the average time interval load curve function to obtain the average annual predicted load P at the future Tf momentav_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK)。
And step 3: the annual average predicted load P of the load circuit at the future Tf moment obtained in the step 2av_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK) And the weather factor data (including air temperature data, sunlight intensity, haze index and cloud amount) of the area where the device is positioned is used as input, and T is used in the futurefThe predicted value of the moment load is used as output, and a load prediction model obtained by BP artificial neural network training is utilized to realize future TfAnd predicting the moment load.
The load prediction model is obtained by training a BP artificial neural network by using historical data, wherein the number of neurons in an input layer of the BP artificial neural network is 7, and the number of neurons in an output layer of the BP artificial neural network is 1.
When the BP artificial neural network is trained, firstly, normalization processing is carried out on each dimension of input data, and a specific normalization processing formula is as follows:
where minA and maxA are the minimum and maximum values in the dimension A data, respectively, and one original value x of dimension A is mapped to a value x' over the interval [0,1 ].
Is determined byWherein m is the number of neurons in the input layer, n is the number of neurons in the output layer, and a is a constant of 1-10, and the number of neurons in the hidden layer is 3 according to the formula and the trial and error method, so as to obtain a BP network with a structure of 7-3-1, as shown in FIG. 3.
The BP neural network is composed of two parts: the forward transfer of information and the back propagation of errors. In the forward propagation process, input information is transmitted from an input layer to an output layer through hidden layer-by-layer calculation, the state of each layer of neuron only affects the state of the next layer of neuron, if expected output is not obtained in the output layer, the error change value of the output layer is calculated, and then the steering error is transmitted reverselyAnd broadcasting, and reversely transmitting the error signals along the original connecting path through the network to modify the weight of each neuron until the desired target is reached. Let the input node be xj(j is 1 … r), r is the number of input nodes, and hidden nodes are yi(i=1…S1),S1For the number of output nodes, the output node is ok(k=1…S2),S2Is the number of output nodes; the network weight between the input node and the hidden node is w1ijThe connection weight between the hidden node and the output node is w2kiThe threshold value of the hidden node is theta1iThe threshold value of the output contact is theta2kThe activation functions of the hidden layer and the output layer areThe desired output is tk。
The algorithm steps of the BP artificial neural network are as follows:
(1) forward propagation of information
The output of the ith neuron in the hidden layer is:
the output of the kth neuron in the output layer is:
defining an error function as:
(2) solving weight value change and threshold value change by using a gradient descent method;
the weight change of the output layer has the following weight from the ith input to the kth output:
the threshold change of the output layer has the following threshold values from the ith input to the kth output:
the weight change of the hidden layer has the following weight values from the jth input to the ith output:
wherein, γiIs the output of the ith hidden layer node.
The threshold change of the hidden layer has the following threshold values from the jth input to the ith output:
and training the BP neural network by using the annual average predicted load, the monthly average predicted load, the time-interval average predicted load and meteorological factor data in the historical load data as the input of the BP neural network and the load predicted value as the output of the neural network.
The degree of influence of historical load and meteorological factors on the load of a certain area is almost the same, so that an artificial neural network can be generated in one area, and the prediction result of the BP artificial neural network is closer to the true value through training of a large amount of historical data.
After completing the training of the BP artificial neural network, only the future T is needed to be obtainedfWeather factor data of time, including air temperature data, sunlight intensity, haze index, cloud amount, and future TfLoad predicted values of different time scales obtained by time prediction are input into a BP artificial neural network to obtain future TfLoad prediction value P at timef。
The fluctuation of the load comprises continuous and regular fluctuation, such as load fluctuation caused by national economic development, seasonal change, change of working days and weekends and the like; meanwhile, the method also comprises random fluctuation, mainly the fluctuation of distributed photovoltaic output and the fluctuation of load caused by the change of meteorological factors. The fluctuation of the continuity regularity is suitable for being fitted by using a mathematical expression, and the BP artificial neural network has good effect on the identification and prediction of the randomness fluctuation. Compared with the method that the artificial neural network is trained by only using meteorological factors, the development trend of the load can be better reflected by introducing historical load data, and the prediction is more accurate; and the prediction results of the fitting of the annual average load, the monthly average load and the time-interval average load function are introduced into the input nodes of the BP artificial neural network, so that the input dimension of the artificial neural network can be reduced, the convergence is accelerated, and the training time is shortened.
And 4, step 4: the main station searches the switchable load line at certain intervals according to the load level and logic substation priority combination cutting principle until the searched switchable load line is T in the futurefThe sum of the predicted values of the moment load reaches the low-frequency required cut PTAnd stopping searching, converting the information of the circuit to be cut into a low-frequency trip outlet fixed value of the device, and transmitting the fixed value to the device.
In step 4, searching the switchable load line according to the load level and logic substation priority combination switching principle comprises the following specific steps:
s1, sorting the load lines according to the multi-dimensional attribute comprehensive evaluation score of each load line, determining the level fixed value of each load line according to the sorting, and preferentially cutting off the line with a smaller level fixed value.
The multi-dimensional attributes include, but are not limited to, outage loss, affiliated electrical sector, sensitivity, electrical distance, frequency-voltage characteristics of the load, and the like.
And the master station counts attributes of the interruption loss, the affiliated power utilization department, the sensitivity and the electrical distance in four dimensions and takes the attributes as evaluation indexes. For a load line, a lower rating index score represents a lower shedding priority for that load line in that dimension. And each evaluation index is assigned with a weight, and then each evaluation index score of the load line is multiplied by the weight of the evaluation index score to obtain a weighted comprehensive score of each load line. And sequencing according to the weighted comprehensive scores of all the load lines so as to determine the cutting priority of each load line. And preferentially cutting off the load circuit with high priority.
If the load circuit score is lower, the load circuit is considered as an important sensitive load, and the load shedding priority is lower.
And S2, dividing each load line to a corresponding logic substation according to the substation to which the load line actually belongs, and setting a priority fixed value for each logic substation. The priority rating of the logic substation can be determined according to the average value of the comprehensive evaluation scores of all the load circuits administered by the logic substation.
And S3, the main station sequentially searches the switchable load lines according to the sequence of the two-dimensional list of the priority levels of the first logic substation and the second logic substation. Until searched future T of line capable of switching loadfThe sum of the predicted values of the moment load reaches the low-frequency required cut PTAnd stopping searching to obtain the load to be cut.
The switchable load line is a load line with positive power and an exit trip pressure plate state of exit.
The searching process comprises the following steps: firstly, all load lines in level 1 are searched for the switchable load lines according to the priority of the logic substation from high to low, and when the switchable load lines are searched for in the future TfWhen the sum of the predicted load values at the moment is larger than the low-frequency required cutting amount, the circuit to be cut is enough, so that the search is stopped, and all the searched switchable load circuits are the future TfA line is required to be cut at any time; if all the load lines in the level 1 are searched, the searched future T of the switchable load linefWhen the sum of the predicted values of the loads is still smaller than the low-frequency required switching value, the next level (level 2) is continuously searched, and the processes are repeated until the searched future T of the switchable load line is metfThe sum of the predicted values of the loads at the moment is greater than the low-frequency load shedding requirement condition, the search is stopped, and all the searched load-shedding lines which can be cut are the future TfThe line is required to be cut at all times.
Before the search stops, due to the last layerThe load circuit in the last logic substation in the level is partially cut off, and the searched load circuit in the last logic substation in the last level is according to the future TfAnd searching the moment load predicted values in the descending order.
S4, after a required tangent path is obtained, the load coordination control master station converts the required tangent path into the future T of the devicefAnd issuing the constant value of the time low-frequency trip outlet to the device.
In the invention, the searching process of the line to be cut comprehensively considers the multi-dimensional attribute of the load and the regional distribution thereof, the load coordination control between the transformer substation and the transformer substation is automatically carried out according to the low-frequency required cut and the current line power condition to be cut, and the over-cut amount is only a certain load line in the last logic substation of the last level, thereby realizing the minimum over-cut. The method reduces over-cut or under-cut, realizes coordinated optimization control of loads in stations, between stations and among each turn, improves the fine level of load control, reduces the influence and economic loss on normal production and life, and has remarkable economic benefit and social benefit.
And 5: the device receives T transmitted by the master station at intervalsfConstant value of low-frequency trip outlet at moment is stored, and when T is reachedfAt that moment, the device updates the real-time trip outlet fixed value to the fixed value. And when a low-frequency fault occurs, the fault is judged only based on local information, and a certain local anti-error strategy is combined, so that the corresponding load circuit is cut off according to a trip outlet fixed value action issued by the main station in advance.
When the device has low-frequency faults, the faults are judged only based on local information and do not depend on a remote command, and the channel can still act according to a preset tripping outlet fixed value when the channel has faults, so that the reliability is guaranteed.
The local error-preventing strategy is that when the device acts, the real-time power P of a line to be cut is issued only by the main stationr>0, the outlet is allowed. Here real time power Pr>0 indicates that the power direction is positive, positive indicating a load line, and negative indicating a power line.
The traditional low-frequency load reducing device cuts off a specified line according to a preset fixed value, and does not know whether the cut-off line is a load or a power supply at the moment or whether the total load cutting amount meets the set value. The low-frequency device of the invention realizes the detection of the power direction of the line, and ensures that the line cut off during the action is a load line.
The invention can realize the coordination control of the load between stations and in stations, improve the load shedding precision as much as possible, realize the minimum over-shedding, avoid the line with the attribute of 'source' and important sensitive users, realize accurate load control and reduce the cost influence of stability control measures; meanwhile, the pre-centralized coordination has a prediction property, so that the dependence on the channel real-time communication is small; and the real-time distributed control does not depend on the channel and only judges the fault and acts an outlet based on the local information, and the reliability of the real-time distributed control is also guaranteed.
Example 2
Based on the same inventive concept as embodiment 1, the low-frequency load shedding system for pre-centralized coordination and real-time distributed control comprises a device and a main station:
the device is used for acquiring active power of a load line, uploading the active power to a main station, receiving a low-frequency trip exit fixed value issued by the main station at intervals, and sending the low-frequency trip exit fixed value to the main station at TfWhen a low-frequency fault is judged to occur, a local anti-misoperation strategy is combined, and a corresponding load circuit is cut off according to the fixed-value action of a low-frequency trip outlet;
the master station is used for receiving the active power of the load line and calculating to obtain the low-frequency required switching value based on the active power of the load line; predicting to obtain the future T of the load linefPredicting the moment load; searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping searching when the sum of the predicted values of the moment load reaches the low-frequency required cutting amount to obtain the future TfSwitching load lines at any time; will be T in futurefAnd converting the information of the line needing to be switched into a low-frequency tripping outlet fixed value at the moment and sending the low-frequency tripping outlet fixed value to the device.
The concrete implementation scheme of each module in the system is shown in the steps and the processes of the method in the embodiment 1.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (10)
1. A low-frequency load shedding method for pre-centralized coordination and real-time distributed control is characterized by comprising the following processes:
obtaining the active power of a load line, and calculating to obtain the low-frequency required tangent;
predicting to obtain the future T of the load linefPredicting the moment load;
searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping searching when the sum of the predicted values of the moment load reaches the low-frequency required cutting amount to obtain the future TfSwitching load lines at any time;
will be T in futurefThe circuit information of the load to be cut at the moment is converted into a low-frequency tripping outlet fixed value and is issued to the device for the device to switch to the T statefAnd when the low-frequency fault is judged to occur at any moment, the corresponding load circuit is cut off according to the fixed-value action of the low-frequency trip outlet by combining a local anti-misoperation strategy.
2. The method of claim 1, wherein the prediction yields a future TfThe moment load predicted value comprises the following steps:
based on future TfThe time information is related to the time, and the future T of the load line is obtained by predictionfPredicted load at different time scales including future TfThe annual average predicted load, the monthly average predicted load and the time-interval average predicted load of the moment load;
the obtained annual average predicted load P of the load circuit at the future Tf momentav_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK) Together with weather factors of the areaInputting data into load prediction model obtained by BP artificial neural network training to obtain future T of load linefAnd (4) predicting the moment load.
3. The method of claim 2, wherein the future T-based low frequency offloading is based onfThe time information is related to the time, and the future T of the load line is obtained by predictionfThe predicted load of different time scales at different moments comprises the following steps:
s1, obtaining load line prediction TfYear T of timeYYear-averaged load sequence of previous i years { P }av_Y(tn) 1-i, wherein the annual average load Pav_Y(tn) Calculated as follows:
wherein, p (T) is a power average value of a certain line in a time interval of delta T after time T, the value of delta T is 1 year, and p (T) is real-time sampling data in the time interval of delta T;
considering the sampling values as a discrete sequence, the formula discretization is expressed as:
wherein, p (t)n) The sequence is discretized at equal intervals on a time axis by power p (T) in a time interval of delta T, wherein the delta T is a sampling interval, and the value of the delta T is 1 year;
s2, obtaining load line prediction TfMonth T of timeMMonthly average load sequence { P) of m months beforeav_M(tn) 1-m, wherein the average monthly load Pav_M(tn) Calculated according to the following formula:
wherein, p (t)n) The sequence is discretized at equal intervals on a time axis by power p (T) in a time interval of delta T, wherein the delta T is a sampling interval, and the value of the delta T is 1 month;
s3, predicting T according to the load line to be predictedfAttribute of time TWDetermine if it is a weekend, based on the date TDPushing forward the date of which the day is the weekend or the weekday, skipping the attribute TWA date of inconsistency; dividing one day into a plurality of time intervals, wherein the difference between every two time intervals is the interval time of two predictions, and the time interval T is the time interval from the current timeK’Push forward a set of P slots of { T }k1-P }; first, the average value of the load for a certain period is calculated:
wherein, p (t)n) For a sequence in which the power p (T) is discretized at equal intervals on the time axis within a period of Δ T, Δ T being the sampling interval, Δ T here being taken to be (T)k+1-Tk);
Next, the attribute T is calculatedWSame q days same period TkAverage value P of average load ofTk_DqThen attribute TWSame q days same period TkThe average load of (a) is taken as the average of the column-wise accumulations in the table, e.g. T over q days1The average of the load over the time period is:
calculating the average value of the load of all the time periods in q days to obtain a time period average load sequence { P }av_TK(tn)|n=1~p}。
Then, respectively carrying out polynomial fitting on the annual average load sequence, the monthly average load sequence and the time-interval average load sequence to obtain a corresponding annual average load curve function, monthly average load curve function and time-interval average load curve function;
respectively combine T withfYear T of timeYThe month of TMAnd the period TKSubstituting the average annual load curve function, the average monthly load curve function and the average time interval load curve function to obtain the average annual predicted load P at the future Tf momentav_Y(TY) Monthly average predicted load Pav_M(TM) And time-interval average predicted load Pav_TK(TK)。
4. The method as claimed in claim 3, wherein the step of performing polynomial fitting on the annual average load sequence, the monthly average load sequence and the time-interval average load sequence to obtain corresponding annual average load curve function, monthly average load curve function and time-interval average load curve function comprises:
s1, fitting an annual average load curve by using a (i-1) th-order polynomial, and setting an annual average load curve function determined by the annual average load of i as follows:
Pav_Y(t)=a0ti-1+a1ti-2+…+ai-1
the previously obtained annual average load series { P }av_Y(tn) Substituting | n | -1 to i } into the above formula, i.e. determining each coefficient a of the function0…ai-1;
S2. similarly, obtaining a monthly average load curve function as:
Pav_M(t)=b0tm-1+b1tm-2+…+bm-1
the previously obtained monthly average load sequence { P }av_M(tn) Substituting | n | -1 to m } into the above equation, i.e. determining the respective coefficients b of the function0…bi-1;
S3, similarly, obtaining a time interval average load curve function as follows:
Pav_TK(t)=c0tp-1+c1tp-2+…+cp-1
averaging the previously obtained time interval load sequence { Pav_TK(tn) Substituting | n | -1 to p } into the above equation, i.e. determining the respective coefficients c of the function0…ci-1。
5. The method of claim 1, wherein the BP artificial neural network comprises forward propagation of information and back propagation of error; in the forward propagation process, input information is transmitted to an output layer from an input layer through a hidden layer by layer calculation, the state of each layer of neuron only affects the state of the next layer of neuron, if expected output is not obtained in the output layer, an error change value of the output layer is calculated, then the error is transmitted in a reverse direction, and an error signal is transmitted back along an original connecting path through a network to modify the weight of each neuron until an expected target is reached;
let the input node be xj(j is 1 … r), r is the number of input nodes, and hidden nodes are yi(i=1…S1),S1For the number of output nodes, the output node is ok(k=1…S2),S2Is the number of output nodes; the network weight between the input node and the hidden node is w1ijThe connection weight between the hidden node and the output node is w2kiThe threshold value of the hidden node is theta1iThe threshold value of the output contact is theta2kThe activation functions of the hidden layer and the output layer areThe desired output is tk;
The algorithm steps of the BP neural network are as follows:
(1) forward propagation of information
The output of the ith neuron in the hidden layer is:
the output of the kth neuron in the output layer is:
defining an error function as:
(2) solving weight value change and threshold value change by using a gradient descent method;
the weight change of the output layer has the following weight from the ith input to the kth output:
the threshold change of the output layer has the following threshold values from the ith input to the kth output:
the weight change of the hidden layer has the following weight values from the jth input to the ith output:
wherein, γiThe output of the ith hidden layer node;
the threshold change of the hidden layer has the following threshold values from the jth input to the ith output:
and training the BP neural network by using the annual average predicted load, the monthly average predicted load, the time-interval average predicted load and meteorological factor data in the historical load data as the input of the BP neural network and the load predicted value as the output of the neural network.
6. The method of claim 1, wherein the searching for the line capable of switching off the load comprises:
s1, sorting the cutting priorities of the load lines according to the multi-dimensional attribute comprehensive evaluation score of each load line, determining the level fixed value of each load line according to the sorting priorities, and cutting the lines with high priority preferentially;
s2, dividing each load line to a corresponding logic substation according to the substation to which the load line actually belongs, and setting a priority fixed value for each logic substation;
s3, sequentially searching the switchable load lines according to the sequence of the two-dimensional list of the priority levels of the first logic substation and the second logic substation until the searched switchable load lines reach the future TfThe sum of the predicted values of the moment load reaches the low-frequency required cut PTAnd stopping searching to obtain the load to be cut.
7. The method of claim 6, wherein the searching process comprises:
firstly, all load lines in level 1 are searched for the switchable load lines according to the priority of the logic substation from high to low, and when the switchable load lines are searched for in the future TfWhen the sum of the predicted values of the loads at the moment is larger than the low-frequency required switching value, the search is stopped, and all the searched switchable load lines are the future TfA line is required to be cut at any time;
if all the load lines in the level 1 are searched, the searched future T of the switchable load linefWhen the sum of the predicted values of the loads is still smaller than the low-frequency required switching value, the next level is continuously searched, and the process is repeated until the searched future T of the load-cutting-able line is metfThe sum of the predicted values of the moment load is greater than the low frequency to be cutThe quantity condition, stopping searching, and the future T being all the available load linesfA line is required to be cut at any time;
before the search is stopped, the load circuit in the last logic substation of the last hierarchy searched is according to the future T because the load circuit in the last logic substation in the last hierarchy is partially cut offfAnd searching the moment load predicted values in the descending order.
8. The method according to claim 6, wherein the cutting priorities of the load lines are ranked according to the multidimensional attribute comprehensive evaluation score of each load line, and the level rating of each load line is determined according to the ranking;
counting attributes of interruption loss, an affiliated power utilization department, sensitivity and an electrical distance in four dimensions, and taking the attributes as evaluation indexes; for a load line, the lower the score of a certain evaluation index is, the lower the cutting priority of the load line in the dimension is;
distributing a weight to each evaluation index, and multiplying each evaluation index score of the load line by the weight of the evaluation index score to obtain a weighted comprehensive score of each load line;
and sequencing according to the weighted comprehensive scores of the load lines so as to determine the level definite value of each load line.
9. The method as claimed in claim 1, wherein the local anti-misoperation strategy is that when the device is in operation, the real-time power P of the required line is only requiredr>0, the outlet is allowed.
10. A low-frequency load shedding system for pre-centralized coordination and real-time distributed control is characterized by comprising a device and a main station:
the device is used for collecting active power of a load line, uploading the active power to a main station and receiving low-frequency tripping output transmitted by the main station at intervalsOral fixed value, at TfWhen a low-frequency fault is judged to occur, a local anti-misoperation strategy is combined, and a corresponding load circuit is cut off according to the fixed-value action of a low-frequency trip outlet;
the master station is used for receiving the active power of the load line and calculating to obtain the low-frequency required switching value based on the active power of the load line; predicting to obtain the future T of the load linefPredicting the moment load; searching the switchable load line at intervals until the searched switchable load line is in the future TfStopping searching when the sum of the predicted values of the moment load reaches the low-frequency required cutting amount to obtain the future TfSwitching load lines at any time; will be T in futurefAnd converting the information of the line needing to be switched into a low-frequency tripping outlet fixed value at the moment and sending the low-frequency tripping outlet fixed value to the device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110924308.5A CN113657032B (en) | 2021-08-12 | 2021-08-12 | Low-frequency load shedding method and system for pre-centralized coordination real-time distribution control |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110924308.5A CN113657032B (en) | 2021-08-12 | 2021-08-12 | Low-frequency load shedding method and system for pre-centralized coordination real-time distribution control |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113657032A true CN113657032A (en) | 2021-11-16 |
CN113657032B CN113657032B (en) | 2023-11-24 |
Family
ID=78491532
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110924308.5A Active CN113657032B (en) | 2021-08-12 | 2021-08-12 | Low-frequency load shedding method and system for pre-centralized coordination real-time distribution control |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113657032B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116415801A (en) * | 2023-06-12 | 2023-07-11 | 山东创宇环保科技有限公司 | Commercial energy load intelligent distribution method and system based on big data |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5295078A (en) * | 1991-05-17 | 1994-03-15 | Best Power Technology Corporation | Method and apparatus for determination of battery run-time in uninterruptible power system |
JP2006204039A (en) * | 2005-01-21 | 2006-08-03 | Chugoku Electric Power Co Inc:The | Method and device for assuming load of distribution system |
CN104242316A (en) * | 2014-09-24 | 2014-12-24 | 广州供电局有限公司 | Low-frequency low-voltage load shedding amount analyzing method and low-frequency low-voltage load shedding amount analyzing system |
CN104466971A (en) * | 2014-12-22 | 2015-03-25 | 南京加伏沃新能源科技有限公司 | Low frequency load shedding cut-off load distribution method with distributed power supply points considered |
CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学(保定) | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
CN107301478A (en) * | 2017-06-26 | 2017-10-27 | 广东电网有限责任公司珠海供电局 | A kind of cable run short-term load forecasting method |
CN107591814A (en) * | 2017-10-25 | 2018-01-16 | 贵州电网有限责任公司电力调度控制中心 | A kind of adaptive accurate cutting load method of low-frequency and low-voltage |
CN109449922A (en) * | 2018-09-30 | 2019-03-08 | 国网浙江省电力有限公司温州供电公司 | Island intelligent electric power utilization system and its with micro-capacitance sensor control centre exchange method |
CN110348592A (en) * | 2019-05-21 | 2019-10-18 | 华电电力科学研究院有限公司 | A kind of load model prediction technique and forecasting system based on artificial neural network |
CN110601208A (en) * | 2019-09-12 | 2019-12-20 | 深圳供电局有限公司 | Accurate load control method and system based on multi-dimensional load attributes |
CN112186782A (en) * | 2020-09-30 | 2021-01-05 | 山东大学 | Accurate low-frequency load shedding system and method proportional to frequency offset |
CN112712203A (en) * | 2020-12-29 | 2021-04-27 | 湖南大学 | Method and system for predicting daily maximum load of power distribution network |
-
2021
- 2021-08-12 CN CN202110924308.5A patent/CN113657032B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5295078A (en) * | 1991-05-17 | 1994-03-15 | Best Power Technology Corporation | Method and apparatus for determination of battery run-time in uninterruptible power system |
JP2006204039A (en) * | 2005-01-21 | 2006-08-03 | Chugoku Electric Power Co Inc:The | Method and device for assuming load of distribution system |
CN104242316A (en) * | 2014-09-24 | 2014-12-24 | 广州供电局有限公司 | Low-frequency low-voltage load shedding amount analyzing method and low-frequency low-voltage load shedding amount analyzing system |
CN104466971A (en) * | 2014-12-22 | 2015-03-25 | 南京加伏沃新能源科技有限公司 | Low frequency load shedding cut-off load distribution method with distributed power supply points considered |
CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学(保定) | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
CN107301478A (en) * | 2017-06-26 | 2017-10-27 | 广东电网有限责任公司珠海供电局 | A kind of cable run short-term load forecasting method |
CN107591814A (en) * | 2017-10-25 | 2018-01-16 | 贵州电网有限责任公司电力调度控制中心 | A kind of adaptive accurate cutting load method of low-frequency and low-voltage |
CN109449922A (en) * | 2018-09-30 | 2019-03-08 | 国网浙江省电力有限公司温州供电公司 | Island intelligent electric power utilization system and its with micro-capacitance sensor control centre exchange method |
CN110348592A (en) * | 2019-05-21 | 2019-10-18 | 华电电力科学研究院有限公司 | A kind of load model prediction technique and forecasting system based on artificial neural network |
CN110601208A (en) * | 2019-09-12 | 2019-12-20 | 深圳供电局有限公司 | Accurate load control method and system based on multi-dimensional load attributes |
CN112186782A (en) * | 2020-09-30 | 2021-01-05 | 山东大学 | Accurate low-frequency load shedding system and method proportional to frequency offset |
CN112712203A (en) * | 2020-12-29 | 2021-04-27 | 湖南大学 | Method and system for predicting daily maximum load of power distribution network |
Non-Patent Citations (2)
Title |
---|
史军;王加澍;熊峰;程维杰;宋俊文;马刚;: "基于智能负载的微电网精准切负荷控制策略", 电力工程技术, no. 02 * |
王传旭;: "基于RBF神经网络的短期电力负荷预测", 供用电, no. 12 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116415801A (en) * | 2023-06-12 | 2023-07-11 | 山东创宇环保科技有限公司 | Commercial energy load intelligent distribution method and system based on big data |
CN116415801B (en) * | 2023-06-12 | 2023-08-29 | 山东创宇环保科技有限公司 | Commercial energy load intelligent distribution method and system based on big data |
Also Published As
Publication number | Publication date |
---|---|
CN113657032B (en) | 2023-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107818385B (en) | Method for predicting real-time operation trend of cascade hydropower station group | |
JP4890920B2 (en) | Power quality maintenance support method and power quality maintenance support system for a distribution system in which a plurality of distributed power sources are connected | |
CN102509173B (en) | A kind of based on markovian power system load Accurate Prediction method | |
KR20160076376A (en) | Method for System Frequency Controlling of Multiple ESS(Energy Storage System) and System thereof | |
CN104167734B (en) | Based on the grid type microgrid reliability estimation method of timing simulation | |
CN110137952B (en) | Optimization model and method for realizing source-load-storage coordinated operation | |
CN104299057A (en) | Method for predicting top layer oil temperature of oil-immersed transformer based on multiple factors | |
CN111049193B (en) | Standby demand dynamic evaluation method for multiple scheduling scenes of wind power system | |
CN111260188B (en) | Voltage sag treatment device configuration method | |
CN107144764A (en) | A kind of user's voltage dip accident detection method based on ammeter data | |
CN110969306A (en) | Power distribution low-voltage distribution area load prediction method and device based on deep learning | |
CN113300403B (en) | Solar micro-grid energy storage capacity configuration optimization processing system and processing method | |
CN107069835B (en) | Real-time active distribution method and device for new energy power station | |
CN116308306B (en) | New energy station intelligent management system and method based on 5G | |
CN115036978A (en) | Operation control method and system for distributed photovoltaic cluster | |
CN113205252A (en) | Aggregated load scheduling method based on demand side load peak regulation potential parameter prediction | |
CN113657032B (en) | Low-frequency load shedding method and system for pre-centralized coordination real-time distribution control | |
CN114386833A (en) | Elasticity evaluation and mobile energy storage regulation and control method for active power distribution network | |
CN114362246A (en) | Power grid accessible distributed photovoltaic capacity measuring and calculating method | |
US20190052096A1 (en) | Control device | |
CN107491866B (en) | Provincial and regional integrated power grid safety and stability comprehensive defense method | |
CN107749631B (en) | Unit primary frequency modulation capability pre-estimating method | |
CN104252647B (en) | Electro-load forecast method based on anti-distance weighting interpolation method | |
CN105162173A (en) | Determination method for reserve capacity of power system accessed with wind power generation | |
CN115392785A (en) | Power distribution network congestion scheduling method considering congestion resistance |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |