CN117252288A - Regional resource active support capacity prediction method and system - Google Patents

Regional resource active support capacity prediction method and system Download PDF

Info

Publication number
CN117252288A
CN117252288A CN202310983608.XA CN202310983608A CN117252288A CN 117252288 A CN117252288 A CN 117252288A CN 202310983608 A CN202310983608 A CN 202310983608A CN 117252288 A CN117252288 A CN 117252288A
Authority
CN
China
Prior art keywords
features
active power
prediction
regional
predicted
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
Application number
CN202310983608.XA
Other languages
Chinese (zh)
Inventor
付宇
郝树青
李跃
蔡永翔
刘安茳
肖小兵
李新皓
李前敏
苗宇
郑友卓
王扬
吴鹏
刘念
张宽
张洋
王悦婧
谈竹奎
熊楠
窦陈
余立文
周波
何心怡
何肖蒙
陈宇
张恒荣
宋子宏
王卓月
班诗雪
吴亚龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guizhou Power Grid Co Ltd
Original Assignee
Guizhou Power Grid Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guizhou Power Grid Co Ltd filed Critical Guizhou Power Grid Co Ltd
Priority to CN202310983608.XA priority Critical patent/CN117252288A/en
Publication of CN117252288A publication Critical patent/CN117252288A/en
Pending legal-status Critical Current

Links

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
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • 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
    • H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
    • 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
    • H02J3/007—Arrangements for selectively connecting the load or loads to one or several among a plurality of power lines or power sources
    • H02J3/0075—Arrangements for selectively connecting the load or loads to one or several among a plurality of power lines or power sources for providing alternative feeding paths between load and source according to economic or energy efficiency considerations, e.g. economic dispatch

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • General Health & Medical Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Operations Research (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Game Theory and Decision Science (AREA)
  • Computing Systems (AREA)
  • Power Engineering (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Quality & Reliability (AREA)
  • Biophysics (AREA)
  • Development Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of power resource allocation, in particular to a method and a system for predicting active supporting capacity of regional resources, comprising the steps of constructing an active power prediction model based on local characteristics, constructing a sample set based on local meteorological data, and inputting a stacking model based on training effect of a base learner to obtain regional active power distribution condition and a total active power predicted value based on the local characteristics; constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features; and (3) weighting and integrating the two types of total active power predicted values, optimizing the weights of the two types of total active power predicted values in real time by adopting a balying search algorithm, outputting a final predicted value of the regional aggregate active power, and obtaining a predicted result of the regional active supporting capability. The method and the device comprehensively consider the influence of the combined algorithm on the active power, and improve the prediction accuracy on the premise of ensuring the training efficiency.

Description

Regional resource active support capacity prediction method and system
Technical Field
The invention relates to the technical field of power resource allocation, in particular to a method and a system for predicting active supporting capacity of regional resources.
Background
To meet the increasing energy demand, solving the problem of climate warming, global distribution networks are integrating renewable energy and configuring energy storage systems on a large scale. Structural reform of the energy supply side pushes the energy development from rough to fine, forming a decentralized and distributed power system. The distributed resources mainly use clean energy, have the advantages of environmental protection, small occupied area and nearby consumption, can realize cascade utilization of energy, further stabilize load peak-valley difference and relieve peak regulation and frequency modulation pressure of a power grid. However, due to the intermittence and randomness of renewable energy power generation, the uncertainty variable of the power distribution network is increased to a plurality of variables from a single load, the difficulty of active power prediction of regional aggregate resources is increased, and the active supporting capability of the power grid is difficult to accurately and quantitatively evaluate.
Active power prediction has a number of classification schemes according to different classification dimensions. Based on general principles of awareness, systematicness and the like, and following basic principles of continuity, similarity, relativity and the like, active power prediction is mainly divided into a time sequence prediction problem and a multi-factor prediction problem; the modeling mode is divided into a physical modeling method based on the power generation principle and a data driving modeling method based on the data intrinsic law; the predicted content is mainly divided into a long term, a medium term, a short term, an ultra-short term and the like according to a predicted period; the prediction results are mainly divided into point prediction, interval prediction and probability density prediction according to the information detail degree. In recent years, with the development of artificial intelligence technology, a deep learning algorithm becomes an important research direction of a power prediction technology by virtue of the excellent nonlinear fitting capability and the data high-order feature extraction capability.
In the problem of prediction of the active supporting capacity of the power grid, most of the existing power prediction research is based on a data-driven modeling method, a traditional machine learning algorithm, a deep learning algorithm and a combination algorithm are adopted, power history data and external features represented by meteorological data are taken as inputs, and power output at a future moment is predicted. The multi-model fusion stacking integrated learning mode and the LSTM model are the first choice of algorithms of researchers.
Disclosure of Invention
This section is intended to outline some aspects of embodiments of the invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description summary and in the title of the application, to avoid obscuring the purpose of this section, the description summary and the title of the invention, which should not be used to limit the scope of the invention.
The present invention has been made in view of the above-described problems.
Therefore, the technical problems solved by the invention are as follows: the existing aggregate resource power prediction problem is mostly treated as a time sequence prediction problem, and the model training efficiency is low because the time sequence prediction problem highly depends on a large amount of historical data; the invention provides a stacking power prediction model based on the training effect of the base learner and an LSTM power prediction model focused on the sample training difference, and the model training task resources are distributed to each class of learner or sample according to the learning effect in proportion, so that the model feature extraction capability is enhanced, and the model training efficiency is improved. The deep learning algorithm model comprises a large number of super parameters, the problem of over fitting is easy to cause in the learning and training process, so that the generalization capability of the model is reduced. The power prediction of the existing aggregate resources only considers the integral characteristics of the total power sequence or the local characteristics of the individual sequences, and does not consider the two types of characteristics at the same time, so that the power prediction precision is lower.
In order to solve the technical problems, the invention provides the following technical scheme: constructing an active power prediction model based on local characteristics, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the distribution condition of the regional active power and the total active power predicted value based on the local characteristics; constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features; and (3) weighting and integrating the two types of total active power predicted values, optimizing the weights of the two types of total active power predicted values in real time by adopting a balying search algorithm, outputting a final predicted value of the regional aggregate active power, and obtaining a predicted result of the regional active supporting capability.
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the active power prediction model based on the local features comprises the steps of dividing feature variables influencing resource power prediction into time features, electricity price features, weather features and historical power features, wherein the time features are composed of hour, week, month and holiday information corresponding to a sequence to be predicted, the weather features are composed of local meteorological data of a subarea where the sequence to be predicted is located, discretization processing is carried out on the time features by adopting independent thermal coding, standardization processing is carried out on continuous variables of the electricity price features, the weather features and the historical power features by adopting maximum and minimum normalization, and a normalization formula is shown as formula (1):
wherein X is an actual characteristic value,for normalized values of features, X max ，X min Respectively maximum and minimum of features, X max -X min To the utmost extent, the numerical range of the normalized characteristic is [0,1]。
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the stacking model comprises a base learner which selects a XGBoost, lightGBM, GBDT, KNN, RF, SVM, lasso model as a stacking integrated model, and selects a square root error RMSE as a prediction effect evaluation index, wherein a load sequence calculation error calculation formula is shown as follows:
wherein N is the total number of samples, y i For the actual power to be available,to predict power, C i The starting capacity of the photovoltaic panel and the wind turbine generator is obtained.
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the stacking model further comprises determining a primary learner and a meta learner, respectively inputting training set data into each basic learner for fitting training, and calculating error association degree R of each learner by using Pearson correlation coefficient AB The calculation formula is as follows:
wherein,and->Average of vectors A and B, R AB The value range is [ -1,1]，|R AB I.apprxeq.1, the higher the correlation degree of the two,
screening XGBoost, KNN, RF, SVM, lasso models with large error distribution differences according to calculation results to serve as primary learners of a first layer of a stacking integrated model, and selecting XGBoost as a meta learner of a second layer of the stacking integrated model;
weights for each primary learner are obtained based on the prediction effect. Calculating the accuracy k of the prediction task on each primary learner i Obtaining a prediction effect evaluation index by an exponential smoothing average methodThe formula is as follows:
k i ＝1-E rmse
wherein, the subscript i=1, 2,3,4,5 respectively uses different initial learners, τ represents the iteration number,the average value of the accuracy rate of tau iterations is the adjustable factor for balancing the past state and the current state;
a Focal Loss function is introduced to measure the difficulty level D of each primary learner to train the same prediction task i Obtaining weight lambda of each initial learner i The calculation formula is as follows:
wherein, gamma is an adjustable factor for measuring the weight reduction proportion, and M is the number of initial learners;
performing feature weighting to obtain input data set of element learner, and outputting data S of each primary learner i Multiplying each learner weight lambda i Post-combining to obtain a new datasetAs input data to the meta learner.
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the active power prediction model based on global features comprises the steps of integrating all resources to be predicted in a region to obtain a region total power sequence, dividing feature variables influencing resource power prediction into time features, electricity price features, weather features and historical power features, wherein the time features are composed of hour, week, month and holiday information corresponding to the sequences to be predicted, the weather features are composed of local weather data of a subarea where the sequences to be predicted are located, discretizing is carried out on the time features by adopting independent heat codes, and standardization is carried out on the electricity price features, the weather features and the historical power feature continuous variables by adopting maximum and minimum normalization;
constructing a loss function focused on differences of sample training, and aggregating resource active power prediction model loss function L based on LSTM model 0 The calculation formula is as follows:
wherein N is the total number of samples, T is the predicted time sequence length, y ti Andthe actual power and the predicted power of the sample i at the time t are respectively.
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the LSTM model comprises the steps that after a predicted data training set is substituted into an improved LSTM model to be trained, a total active power sequence to be predicted is substituted into the trained model to obtain a relatively accurate regional total active power prediction result, and then the active supporting capacity of the power distribution network is quantitatively evaluated.
As a preferable scheme of the method for predicting the active supporting capacity of the regional resource, the invention comprises the following steps: the bald-eagle search algorithm comprises the steps of constructing a time-varying weight combination model based on a bald-eagle optimization algorithm to conduct aggregate resource active power prediction, determining weight coefficients of a local prediction model and a global prediction model at each moment by adopting the bald-eagle optimization algorithm, and setting a fitness function as follows:
wherein omega l And omega g Weight coefficients, P, of the local prediction model and the global prediction model at the moment t respectively l (t) and P g (t) the predicted powers at the time t of the local prediction model and the global prediction model, P true (t) is the real power at time t;
initializing parameter omega l And omega g Determining the number N of the population, wherein the position of each bald eagle in the population is P= (P) 1 ,P 2 ,P 3 ,…,P M ) Judging the quality degree by an fitness function, and determining the maximum iteration times and initial value boundary conditions;
the search space is selected based on the number of prey items and the location is updated continuously according to the following formula. Sequencing each calculated bald eagle fitness value, selecting an optimal fitness value and a corresponding optimal position, and calculating the following formula:
wherein,selecting an updated position for the ith bald eagle in the kth iteration, if the updated position is better than the original position, updating the position, otherwise, keeping the original position unchanged, and carrying out +.>For the current optimal position, alpha controls the position change of the balding, the value range is (1.5, 2), r random Representing random number, uniformly distributed in the value range (0, 1), and +.>Represents the average position of all bald hawks of the current population after the k-1 iteration, +.>The current position of the i-th bald eagle is shown;
in the determined search space, the bald hawk flies around the current position according to the track of the Archimedes spiral line to find the optimal dive position, and the position is further updated, wherein the specific process comprises the following formula:
wherein,and->For the polar angle and polar diameter of the ith bald eagle spiral flight in the kth iterative space prey searching stage, a is the parameter control flight polar angle, the value range is (0, 5), R is the parameter control flight polar diameter, the value range is (0.5, 2), and the value range is (0, 2)>And->Represents the polar position of the ith bald eagle, < >>Represents the ith onlyThe bald hawk selects an updated position in the k-th iterative space prey searching stage, if the updated position is better than the original position, the position is updated, otherwise, the original position is kept unchanged, and the bald hawk is added>Represents the position of the (i+1) th bald eagle;
according to the position of the prey searched in the previous stage, the prey is captured by spiral flight and dive from the optimal search space, and the specific formula is as follows:
wherein,and->The polar angle and polar diameter of the ith bald eagle flight for the kth iteration capture prey stage,and->Represents the polar position of the ith bald eagle, < >>Representing the updated position of the ith bald eagle at the stage of obtaining prey in the kth iteration c 1 And c 2 Representing the flying movement intensity, wherein the value range is (1, 2);
judging whether the ending condition is reached, and selecting the optimal parameters as the weight coefficients of each model after the ending condition is reached, otherwise, repeating the continuous iteration until the constraint condition is met.
The invention further aims to provide a regional resource active supporting capability prediction system, which can provide a stacking power prediction model based on the training effect of a base learner and an LSTM power prediction model focused on the training difference of samples by constructing the regional resource active supporting capability prediction system, so that model training task resources are distributed to each type of learner or sample proportionally according to the learning effect, the model feature extraction capability is enhanced, and the model training efficiency is improved.
As a preferable scheme of the regional resource active supporting capability prediction system of the present invention, the method comprises the following steps: the system comprises a local feature prediction module, a global feature prediction module and an optimization output module;
the local feature prediction module comprises the steps of constructing an active power prediction model based on local features, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the distribution condition of the regional active power and the total active power prediction value based on the local features;
the global feature prediction module comprises the steps of constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features;
the optimizing output module comprises the steps of integrating the local characteristic and the global characteristic of the total active power predicted values in a weighting mode, adopting a balying search algorithm to optimize the weights of the local characteristic and the global characteristic in real time, outputting the final active power predicted value of the regional aggregate, and obtaining the regional active supporting capacity predicted result.
The invention also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the method is characterized in that the processor realizes the steps of the method for predicting the active supporting capability of the regional resource when executing the computer program.
The present invention also provides a computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the method for predicting active support capacity of a regional resource.
The invention has the beneficial effects that: the invention provides a stacking power prediction model based on a training effect of a base learner and an LSTM power prediction model focused on a sample training difference, model training task resources are distributed to each class of learner or sample according to the learning effect in proportion, model feature extraction capacity is enhanced, model training efficiency is improved, elastic regular terms are introduced into an LSTM model loss function, model migration capacity is improved from different layers, the influence of the model training task resources on active power is comprehensively considered by using a combination algorithm, weight is optimized in real time by using a balying search algorithm, and prediction precision is improved on the premise of guaranteeing training efficiency.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
fig. 1 is a diagram of an active support capacity prediction structure of a method for predicting active support capacity of regional resources according to the present invention.
Fig. 2 is a structure diagram of a stacking model based on training effect of a base learner in the prediction method of active supporting capability of regional resources.
Fig. 3 is a system structure diagram of a regional resource active supporting capability prediction system provided by the invention.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
While the embodiments of the present invention have been illustrated and described in detail in the drawings, the cross-sectional view of the device structure is not to scale in the general sense for ease of illustration, and the drawings are merely exemplary and should not be construed as limiting the scope of the invention. In addition, the three-dimensional dimensions of length, width and depth should be included in actual fabrication.
Also in the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper, lower, inner and outer", etc. are based on the orientation or positional relationship shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
The terms "mounted, connected, and coupled" should be construed broadly in this disclosure unless otherwise specifically indicated and defined, such as: can be fixed connection, detachable connection or integral connection; it may also be a mechanical connection, an electrical connection, or a direct connection, or may be indirectly connected through an intermediate medium, or may be a communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Example 1
Referring to fig. 1 and 2, for a first embodiment of the present invention, a method for predicting active supporting capability of a regional resource is provided.
S1, constructing an active power prediction model based on local characteristics, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the regional active power distribution condition and the total active power predicted value based on the local characteristics;
specifically, the active power prediction model based on local features includes that feature variables affecting resource power prediction are divided into time features, electricity price features, weather features and historical power features, the time features are composed of hour, week, month and holiday information corresponding to a sequence to be predicted, the weather features are composed of local weather data of a subarea where the sequence to be predicted is located, discretization processing is carried out on the time features by adopting independent heat codes, standardization processing is carried out on continuous variables of the electricity price features, the weather features and the historical power features by adopting maximum and minimum normalization, and a normalization formula is shown as follows:
wherein X is an actual characteristic value,for normalized values of features, X max ，X min Respectively maximum and minimum of features, X max -X min To the utmost extent, the numerical range of the normalized characteristic is [0,1]。
Further, the stacking model includes selecting a XGBoost, lightGBM, GBDT, KNN, RF, SVM, lasso model as a base learner of the stacking integrated model, selecting a square root error RMSE as a prediction effect evaluation index, and calculating an error calculation formula of a load sequence, wherein the calculation formulas of the distributed photovoltaic and wind-electricity errors are as follows:
wherein N is the total number of samples, y i For the actual power to be available,to predict power, C i The starting capacity of the photovoltaic panel and the wind turbine generator is obtained.
Furthermore, the stacking model further comprises determining a primary learner and a meta learner, respectively inputting training set data into each basic learner for fitting training, and calculating the error association degree R of each learner by using Pearson correlation coefficient AB The calculation formula is as follows:
wherein,and->Average of vectors A and B, R AB The value range is [ -1,1]，|R AB When the I is approximately equal to 1, the higher the correlation degree of the two is,
screening XGBoost, KNN, RF, SVM, lasso models with large error distribution differences according to calculation results to serve as primary learners of a first layer of a stacking integrated model, and selecting XGBoost as a meta learner of a second layer of the stacking integrated model;
weights for each primary learner are obtained based on the prediction effect. Calculating the accuracy k of the prediction task on each primary learner i Obtaining a prediction effect evaluation index by an exponential smoothing average methodThe formula is as follows:
wherein, the subscript i=1, 2,3,4,5 respectively uses different initial learners, τ represents the iteration number,the average value of the accuracy rate of tau iterations is the adjustable factor for balancing the past state and the current state;
a Focal Loss function is introduced to measure the difficulty level D of each primary learner to train the same prediction task i Obtaining weight lambda of each initial learner i The calculation formula is as follows:
wherein, gamma is an adjustable factor for measuring the weight reduction proportion, and M is the number of initial learners;
performing feature weighting to obtain input data set of element learner, and outputting data S of each primary learner i Multiplying each learner weight lambda i Post-combining to obtain a new datasetAs input data to the meta learner.
S2, constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features;
the active power prediction model based on global features comprises the steps of integrating all resources to be predicted in a region to obtain a region total power sequence, dividing feature variables influencing resource power prediction into time features, electricity price features, weather features and historical power features, wherein the time features are composed of hour, week, month and holiday information corresponding to the sequences to be predicted, the weather features are composed of local weather data of a subarea where the sequences to be predicted are located, discretizing is carried out on the time features by adopting independent heat codes, and standardization is carried out on the electricity price features, the weather features and the historical power feature continuous variables by adopting maximum and minimum normalization;
constructing a loss function focused on differences of sample training, and aggregating resource active power prediction model loss function L based on LSTM model 0 The calculation formula is as follows:
wherein N is the total number of samples, T is the predicted time sequence length, y ti Andrespectively are provided withThe actual power and the predicted power of the sample i at the time t are obtained.
And S3, weighting and integrating the two types of total active power predicted values, optimizing the weights of the two types of total active power predicted values in real time by adopting a balying search algorithm, outputting a final predicted value of the regional aggregate active power, and obtaining a predicted result of the regional active supporting capability.
Specifically, the embodiment simultaneously excavates local features and global features of the aggregate resources, considers the time variability of the precision of the prediction model, and constructs a time-varying weight combination model based on a balying optimization algorithm to predict the active power of the aggregate resources.
In the embodiment, a balk optimization algorithm is adopted to determine the weight coefficients of the local prediction model and the global prediction model at each moment, and the set fitness function is as follows:
wherein omega l And omega g Weight coefficients, P, of the local prediction model and the global prediction model at the moment t respectively l (t) and P g (t) the predicted powers at the time t of the local prediction model and the global prediction model, P true And (t) is the real power at time t.
Initializing parameter omega l And omega g Determining the number N of the population, wherein the position of each bald eagle in the population is P= (P) 1 ,P 2 ,P 3 ,…,P M ) The quality degree is judged by the fitness function, and the maximum iteration number and the initial value boundary condition are determined.
The search space is selected based on the number of prey items and the location is updated continuously according to the following formula. Sequencing each calculated bald eagle fitness value, selecting an optimal fitness value and a corresponding optimal position, and calculating the following formula:
wherein,selecting an updated position for the ith bald eagle in the kth iteration, if the updated position is better than the original position, updating the position, otherwise, keeping the original position unchanged, and carrying out +.>For the current optimal position, alpha controls the position change of the balding, the value range is (1.5, 2), r random Representing random number, uniformly distributed in the value range (0, 1), and +.>Represents the average position of all bald hawks of the current population after the k-1 iteration, +.>Table i current position of bald hawk only.
In the determined search space, the bald hawk flies around the current position according to the track of the Archimedes spiral line to find the optimal dive position, and the position is further updated, wherein the specific process comprises the following formula:
wherein,and->For the polar angle and polar diameter of the ith bald eagle spiral flight in the kth iterative space prey searching stage, a is the parameter control flight polar angle, the value range is (0, 5), R is the parameter control flight polar diameter, the value range is (0.5, 2), and the value range is (0, 2)>And->Represents the polar position of the ith bald eagle, < >>Selecting updated position representing ith bald eagle in the kth iterative space hunting phase, if the updated position is better than the original position, updating the position, otherwise, keeping the original position unchanged, and selecting the position>Represents the position of the i+1 bald eagle.
According to the position of the prey searched in the previous stage, the prey is captured by spiral flight and dive from the optimal search space, and the specific formula is as follows:
wherein,and->The polar angle and polar diameter of the ith bald eagle flight for the kth iteration capture prey stage,and->Represents the polar position of the ith bald eagle, < >>Representing the updated position of the ith bald eagle at the stage of obtaining prey in the kth iteration c 1 And c 2 Representing the flying movement intensity, and the value range is (1, 2).
Judging whether the ending condition is reached, and selecting the optimal parameters as the weight coefficients of each model after the ending condition is reached, otherwise, repeating the continuous iteration until the constraint condition is met.
Example 2
Referring to fig. 3, for a second embodiment of the present invention, a regional resource active support capability prediction system is provided.
Specifically, the system comprises a local feature prediction module, a global feature prediction module and an optimization output module.
The local feature prediction module comprises the steps of constructing an active power prediction model based on local features, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the distribution condition of the regional active power and the total active power predicted value based on the local features.
The global feature prediction module comprises the steps of constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features.
The optimizing output module comprises the steps of integrating the local characteristic and the global characteristic of the total active power predicted values in a weighting mode, adopting a balying search algorithm to optimize the weights of the local characteristic and the global characteristic in real time, outputting the final active power predicted value of the regional aggregate, and obtaining the regional active supporting capacity predicted result.
Example 3
The third embodiment of the invention provides a loss function construction method of sample training difference of a regional resource active support capacity prediction method.
Specifically, the loss function of the active power prediction model of the aggregate resource based on the LSTM model is shown as follows,
wherein N is the total number of samples, T is the predicted time sequence length, y ti Andthe actual power and the predicted power of the sample i at the time t are respectively.
According to the method, training samples are divided into two types of training samples to be reinforced and common samples according to the prediction error ratio, the training samples to be reinforced are given higher weight, the feature extraction capacity of the model to the samples is enhanced, and the loss function is as follows:
/>
α i for each sample weight, focus the prediction task on the harder-to-train sample, and the value of c is related to the sample class proportion.
Introducing L1 and L2 regular terms into the loss function to form a final loss function as shown in the following formula, and adopting a grid search method to obtain the final loss function [0.00,0.30 ]]For the search interval, 0.01 determines the optimal regularization parameter lambda for the step size 1 And lambda (lambda) 2 ；
Wherein ω is the parameter to be estimated of LSTM model, λ 1 And lambda (lambda) 2 The regularization coefficient is used for balancing the balance relation between the model loss function and the regularization term, and improving the generalization capability of the model.
Substituting the predicted data training set into the improved LSTM model for training, substituting the total active power sequence to be predicted into the trained model to obtain a relatively accurate regional total active power prediction result, and further quantitatively evaluating the active supporting capacity of the power distribution network.
Example 4
A fourth embodiment of the present invention, which is different from the previous embodiment, is:
the functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only memory (ROM), a random access memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium may even be paper or other suitable medium on which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
Furthermore, in an effort to provide a concise description of the exemplary embodiments, all features of an actual implementation may not be described (i.e., those not associated with the best mode presently contemplated for carrying out the invention, or those not associated with practicing the invention).
It should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort might be complex and time consuming, but would nevertheless be a routine undertaking of design, fabrication, and manufacture for those of ordinary skill having the benefit of this disclosure.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.

Claims (10)

1. A prediction method for the active supporting capacity of regional resources is characterized by comprising the following steps: comprising the steps of (a) a step of,
constructing an active power prediction model based on local characteristics, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the distribution condition of the regional active power and the total active power predicted value based on the local characteristics;
constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features;
and (3) weighting and integrating the two types of total active power predicted values, optimizing the weights of the two types of total active power predicted values in real time by adopting a balying search algorithm, outputting a final predicted value of the regional aggregate active power, and obtaining a predicted result of the regional active supporting capability.
2. The method for predicting the active supporting capacity of regional resources according to claim 1, wherein the method comprises the following steps: the active power prediction model based on the local features comprises the steps of dividing feature variables influencing resource power prediction into time features, electricity price features, weather features and historical power features, wherein the time features are composed of hour, week, month and holiday information corresponding to a sequence to be predicted, the weather features are composed of local meteorological data of a subarea where the sequence to be predicted is located, discretization processing is carried out on the time features by adopting independent thermal coding, standardization processing is carried out on continuous variables of the electricity price features, the weather features and the historical power features by adopting maximum and minimum normalization, and a normalization formula is shown as formula (1):
wherein X is an actual characteristic value,for normalized values of features, X max ，X min Respectively maximum and minimum of features, X max -X min To the utmost extent, the numerical range of the normalized characteristic is [0,1]。
3. The method for predicting the active supporting capacity of regional resources according to claim 2, wherein the method comprises the following steps: the stacking model comprises a base learner which selects XGBoost, lightGBM, GBDT, KNN, RF, SVM and Lasso models as stacking integrated models, square root error RMSE as a prediction effect evaluation index, and a load sequence calculation error calculation formula, wherein the distributed photovoltaic and wind power error calculation formula is as follows:
wherein N is the total number of samples, y i For the actual power to be available,to predict power, C i The starting capacity of the photovoltaic panel and the wind turbine generator is obtained.
4. A method for predicting active support capacity of a regional resource as claimed in claim 3, wherein: the stacking model further comprises determining a primary learner and a meta learner, respectively inputting training set data into each basic learner for fitting training, and calculating error association degree R of each learner by using Pearson correlation coefficient AB The calculation formula is as follows:
wherein,and->Average of vectors A and B, R AB The value range is [ -1,1]，|R AB When the I is approximately equal to 1, the correlation degree of the two is high,
screening XGBoost, KNN, RF, SVM, lasso models with large error distribution differences according to calculation results to serve as primary learners of a first layer of a stacking integrated model, and selecting XGBoost as a meta learner of a second layer of the stacking integrated model;
obtaining primary based on predictive effectThe weight of the learner calculates the accuracy ki of the prediction task on each primary learner, and obtains the prediction effect evaluation index through an exponential smoothing average methodThe formula is as follows:
k i ＝1-E rmse
wherein, the subscript i=1, 2,3,4,5 respectively uses different initial learners, τ represents the iteration number,the average value of the accuracy rate of tau iterations is the adjustable factor for balancing the past state and the current state;
the FocalLoss function is introduced to measure the difficulty level D of each primary learner to train the same prediction task i Obtaining weight lambda of each initial learner i The calculation formula is as follows:
wherein, gamma is an adjustable factor for measuring the weight reduction proportion, and M is the number of initial learners;
performing feature weighting to obtain input data set of element learner, and outputting data S of each primary learner i Multiplying each learner weight lambdaj and merging to obtain new data setAs input data to the meta learner.
5. The method for predicting the active supporting capacity of regional resources according to claim 4, wherein: the active power prediction model based on global features comprises the steps of integrating all resources to be predicted in a region to obtain a region total power sequence, dividing feature variables influencing resource power prediction into time features, electricity price features, weather features and historical power features, wherein the time features are composed of hour, week, month and holiday information corresponding to the sequences to be predicted, the weather features are composed of local weather data of a subarea where the sequences to be predicted are located, discretizing is carried out on the time features by adopting independent heat codes, and standardization is carried out on the electricity price features, the weather features and the historical power feature continuous variables by adopting maximum and minimum normalization;
constructing a loss function focused on differences of sample training, and aggregating resource active power prediction model loss function L based on LSTM model 0 The calculation formula is as follows:
wherein N is the total number of samples, T is the predicted time sequence length, y ti Andthe actual power and the predicted power of the sample i at the time t are respectively.
6. The method for predicting the active supporting capacity of regional resources according to claim 5, wherein the method comprises the following steps: the LSTM model comprises the steps that after a predicted data training set is substituted into an improved LSTM model to be trained, a total active power sequence to be predicted is substituted into the trained model to obtain a relatively accurate regional total active power prediction result, and then the active supporting capacity of the power distribution network is quantitatively evaluated.
7. The method for predicting the active supporting capacity of regional resources according to claim 6, wherein: the bald-eagle search algorithm comprises the steps of constructing a time-varying weight combination model based on a bald-eagle optimization algorithm to conduct aggregate resource active power prediction, determining weight coefficients of a local prediction model and a global prediction model at each moment by adopting the bald-eagle optimization algorithm, and setting a fitness function as follows:
wherein omega l And omega g Weight coefficients, P, of the local prediction model and the global prediction model at the moment t respectively l (t) and P g (t) the predicted powers at the time t of the local prediction model and the global prediction model, P true (t) is the real power at time t;
initializing parameter omega l And omega g Determining the number N of the population, wherein the position of each bald eagle in the population is P= (P) 1 ,P 2 ,P 3 ,…,P M ) Judging the quality degree by an fitness function, and determining the maximum iteration times and initial value boundary conditions;
selecting a search space according to the number of the hunting objects, continuously updating the positions according to the following formula, sequencing each calculated bald eagle fitness value, selecting an optimal fitness value and a corresponding optimal position, and calculating the following formula:
wherein,selecting an updated position for the ith bald eagle in the kth iteration, if the updated position is better than the original position, updating the position, otherwise, keeping the original position unchanged, and carrying out +.>Is the best at presentPosition, alpha controls the position change of bald hawk, the value range is (1.5, 2), r random Representing random number, uniformly distributed in the value range (0, 1), and +.>Represents the average position of all bald hawks of the current population after the k-1 iteration, +.>The current position of the i-th bald eagle is shown;
in the determined search space, the bald hawk flies around the current position according to the track of the Archimedes spiral line to find the optimal dive position, and the position is further updated, wherein the specific process comprises the following formula:
wherein,and->For the polar angle and polar diameter of the ith bald eagle spiral flight in the kth iterative space hunting stage, a is the parameter control flight polar angle, the value range is (0, 5), R is the parameter control flight polar diameter, the value range is (0.5, 2),and->Represents the polar position of the ith bald eagle, < >>Representing the ith bald eagle, selecting an updated position in the kth iterative space prey searching stage, if the updated position is better than the original position, carrying out position updating, otherwise, keeping the original position unchanged,represents the position of the (i+1) th bald eagle;
according to the position of the prey searched in the previous stage, the prey is captured by spiral flight and dive from the optimal search space, and the specific formula is as follows:
wherein,and->Polar angle and polar diameter of the ith bald eagle flight for the kth iteration capture prey stage, +.>Andrepresents the polar position of the ith bald eagle, < >>Representing the updated position of the ith bald eagle at the stage of obtaining prey in the kth iteration c 1 And c 2 Representing the flying movement intensity, wherein the value range is (1, 2);
judging whether the ending condition is reached, selecting the optimal parameters as the weight coefficients of each model after the ending condition is reached, and repeating iteration until the constraint condition is met.
8. A system employing a regional resource active support capability prediction method as claimed in any one of claims 1 to 7, wherein: the system comprises a local feature prediction module, a global feature prediction module and an optimization output module;
the local feature prediction module comprises the steps of constructing an active power prediction model based on local features, constructing a sample set based on local meteorological data, and inputting a stacking model based on the training effect of a base learner to obtain the distribution condition of the regional active power and the total active power prediction value based on the local features;
the global feature prediction module comprises the steps of constructing an active power prediction model based on global features, constructing a sample set based on global meteorological data, and inputting an LSTM model focused on sample training differences to obtain a total active power prediction value based on the global features;
the optimizing output module comprises the steps of integrating the local characteristic and the global characteristic of the total active power predicted values in a weighting mode, adopting a balying search algorithm to optimize the weights of the local characteristic and the global characteristic in real time, outputting the final active power predicted value of the regional aggregate, and obtaining the regional active supporting capacity predicted result.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of a regional resource active support capacity prediction method according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of a regional resource active support capacity prediction method according to any of claims 1 to 7.
CN202310983608.XA 2023-08-07 2023-08-07 Regional resource active support capacity prediction method and system Pending CN117252288A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310983608.XA CN117252288A (en) 2023-08-07 2023-08-07 Regional resource active support capacity prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310983608.XA CN117252288A (en) 2023-08-07 2023-08-07 Regional resource active support capacity prediction method and system

Publications (1)

Publication Number Publication Date
CN117252288A true CN117252288A (en) 2023-12-19

Family

ID=89133930

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310983608.XA Pending CN117252288A (en) 2023-08-07 2023-08-07 Regional resource active support capacity prediction method and system

Country Status (1)

Country Link
CN (1) CN117252288A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117578465A (en) * 2024-01-16 2024-02-20 山东建筑大学 Multi-scale interpretable micro-grid power load prediction method and system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117578465A (en) * 2024-01-16 2024-02-20 山东建筑大学 Multi-scale interpretable micro-grid power load prediction method and system
CN117578465B (en) * 2024-01-16 2024-04-12 山东建筑大学 Multi-scale interpretable micro-grid power load prediction method and system

Similar Documents

Publication Publication Date Title
Wang et al. Deep learning method based on gated recurrent unit and variational mode decomposition for short-term wind power interval prediction
Huang et al. Short‐term wind power forecasting and uncertainty analysis using a hybrid intelligent method
CN110969290B (en) Runoff probability prediction method and system based on deep learning
JP5888640B2 (en) Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program
Huang et al. One‐day‐ahead hourly forecasting for photovoltaic power generation using an intelligent method with weather‐based forecasting models
CN107944648B (en) Large ship speed and oil consumption rate prediction method
CN110443417A (en) Multiple-model integration load forecasting method based on wavelet transformation
CN116596044B (en) Power generation load prediction model training method and device based on multi-source data
CN111311001B (en) Bi-LSTM network short-term load prediction method based on DBSCAN algorithm and feature selection
CN117252288A (en) Regional resource active support capacity prediction method and system
CN115374995A (en) Distributed photovoltaic and small wind power station power prediction method
CN109858665A (en) Photovoltaic short term power prediction technique based on Feature Selection and ANFIS-PSO
CN116644970A (en) Photovoltaic power prediction method based on VMD decomposition and lamination deep learning
CN114742307A (en) Wave element prediction method and system
CN113268929B (en) Short-term load interval prediction method and device
Sun et al. Interval forecasting for wind speed using a combination model based on multiobjective artificial hummingbird algorithm
Abdoos et al. An intelligent hybrid method based on Monte Carlo simulation for short-term probabilistic wind power prediction
Zhu et al. Wind power interval and point prediction model using neural network based multi-objective optimization
Xu et al. A novel intelligent deep learning-based uncertainty-guided network training in market price
Abdelaziz et al. Convolutional Neural Network With Genetic Algorithm for Predicting Energy Consumption in Public Buildings
CN116894383A (en) Random simulation method and device for annual weather scene
CN116561569A (en) Industrial power load identification method based on EO feature selection and AdaBoost algorithm
Yan et al. Transferability and robustness of a data-driven model built on a large number of buildings
CN115345303A (en) Convolutional neural network weight tuning method, device, storage medium and electronic equipment
CN113191526B (en) Method and system for optimizing and predicting short-term wind speed interval based on random sensitivity

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