CN112836885B - Combined load prediction method, combined load prediction device, electronic equipment and storage medium - Google Patents

Combined load prediction method, combined load prediction device, electronic equipment and storage medium Download PDF

Info

Publication number
CN112836885B
CN112836885B CN202110179158.XA CN202110179158A CN112836885B CN 112836885 B CN112836885 B CN 112836885B CN 202110179158 A CN202110179158 A CN 202110179158A CN 112836885 B CN112836885 B CN 112836885B
Authority
CN
China
Prior art keywords
load
response
day
value
prediction
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.)
Active
Application number
CN202110179158.XA
Other languages
Chinese (zh)
Other versions
CN112836885A (en
Inventor
马振祺
马志程
张光儒
杨军亭
苏娟
邢广进
方舒
张家午
温定筠
吴建军
拜润卿
沈渭程
梁有珍
张秀斌
高磊
朱亮
张艳丽
蒋臣
智勇
张睿
张凯
李亚昕
田阔
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.)
STATE GRID GASU ELECTRIC POWER RESEARCH INSTITUTE
Original Assignee
STATE GRID GASU ELECTRIC POWER RESEARCH INSTITUTE
China Agricultural University
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 STATE GRID GASU ELECTRIC POWER RESEARCH INSTITUTE, China Agricultural University filed Critical STATE GRID GASU ELECTRIC POWER RESEARCH INSTITUTE
Priority to CN202110179158.XA priority Critical patent/CN112836885B/en
Publication of CN112836885A publication Critical patent/CN112836885A/en
Application granted granted Critical
Publication of CN112836885B publication Critical patent/CN112836885B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06Q10/067—Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

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

Abstract

The embodiment of the invention provides a combined load prediction method, a combined load prediction device, electronic equipment and a storage medium, wherein the method comprises the following steps: determining the value of a forecast meteorological factor, a basic load, a response load, a load before and after response and an electricity price of a day to be predicted; inputting the value of a forecast meteorological factor, a basic load, a response load, a load before and after response and electricity price of a day to be forecasted into a forecasting model, and outputting a power grid load forecasting value; the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices of a plurality of historical days; the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load on the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data. The invention realizes accurate prediction of the load participating in demand response.

Description

Combined load prediction method, combined load prediction device, electronic equipment and storage medium
Technical Field
The present invention relates to the field of power system load prediction technologies, and in particular, to a combined load prediction method and apparatus, an electronic device, and a storage medium.
Background
With the continuous growth of social economy, the power demand of people increases, the power load increases rapidly, and huge impact is brought to the planning and operation of a power distribution network.
With the market deep reformation and the rapid development of intelligent demand response, the generalized load is accessed to a power distribution network system in a large scale, and great influence is caused on the aspects of system operation, resource allocation and the like. Different from the traditional load, the active load is a load which can be used by a power consumer in a certain period of the future according to the change of the electricity price in the power market, the electricity economy of the power consumer and the safety and stability of a system are realized, the original electricity load curve is changed, and obviously, the traditional load prediction method cannot accurately predict the load participating in demand response.
At present, active load prediction at home and abroad mainly comprises analysis of corresponding influence factors and analysis of user electricity utilization characteristics, and new requirements of an active power distribution network on load prediction are provided. In the face of uncertainty of load operation and complexity of load influence factors, a single load prediction method is difficult to accurately predict short-term load.
Disclosure of Invention
The embodiment of the invention provides a combined load prediction method, a combined load prediction device, electronic equipment and a storage medium, which are used for solving the problem that the load participating in demand response cannot be accurately predicted by the conventional load prediction method.
In a first aspect, an embodiment of the present invention provides a combined load prediction method, including:
determining the value of a forecast meteorological factor, a basic load, a response load, a load before and after response and an electricity price of a day to be predicted;
inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices of a plurality of historical days;
the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
Preferably, the prediction model comprises a base load prediction model, a response load prediction model and a superposition model;
inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model, wherein the power grid load forecasting value comprises the following steps:
inputting the value of the forecast environment factor of the day to be forecasted and the basic load into the basic load forecasting model, and outputting a basic load forecasting value;
inputting the response load of the day to be predicted, the load before and after response and the electricity price into the response load prediction model, and outputting a predicted value of the demand response load;
and inputting the basic load predicted value and the demand response load predicted value into the superposition model, and outputting a power grid load predicted value of a day to be predicted.
Preferably, the basic load prediction model is obtained by performing short-term basic load prediction training based on the values of the actual meteorological factors and the sample data of the basic load on the plurality of historical days;
completing short-term base load prediction training based on the actual meteorological factor values and the base load in the sample data, and comprising the following steps of:
s1, inputting the actual meteorological factor value and the basic load in the sample data into a Cuckoo optimization support vector machine (CS-SVM) model set by initial parameters including a penalty coefficient, the width of a radial basis kernel function, an initial probability, a bird nest position and the maximum iteration number to obtain the optimal fitness value of the bird nest;
s2, determining the current optimal bird nest position according to the optimal fitness value of the bird nest, updating other bird nest positions according to the current optimal bird nest position to generate a group of new bird nest positions, and comparing the current optimal bird nest position with the group of new bird nest positions to obtain a prediction error;
s3, obtaining a better bird nest position according to the prediction error, comparing the probability of eliminating the brook bird eggs with the value of the random number, replacing the current optimal bird nest position with a new group of bird nest positions when the probability of eliminating the brook bird eggs is smaller than the random number, judging whether the current iteration end condition is met, jumping out to find the value of the optimal bird nest position if the iteration end condition is met, and returning to S2 to continue to find the optimal bird nest position if the iteration end condition is not met; wherein the iteration end condition comprises a maximum iteration number or an optimal fitness value determined by the width of a radial basis kernel function;
and S4, updating parameters in the CS-SVM model of the Cuckoo optimization support vector machine to establish a basic load prediction model for the optimal bird nest position, and completing short-term basic load prediction training.
Preferably, the response load prediction model is obtained by performing short-term response load prediction training based on the response loads of the plurality of historical days, the sample data of the loads before and after response and the electricity prices;
completing short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data, and comprising the following steps:
constructing an error minimum objective function according to the response load, the load before and after response and the electricity price in the sample data;
solving the target function with the minimum error based on a gray wolf optimization algorithm for setting the number of wolf clusters, the maximum number of iterations, the direction correction probability and the initialized coordinate space of the wolf clusters to obtain a demand response elastic matrix;
and establishing the response load prediction model based on the demand response elastic matrix, and completing short-term response load prediction training.
Preferably, the constructing an objective function with a minimum error according to the response load, the load before and after the response and the electricity price in the sample data includes:
respectively obtaining the load demand variation delta q of the i-th day period of the M day based on the load before and after the 24-time-period response in the M days and the electricity pricei(m) and corresponding rate of change of electricity price Δ pi(m)；
A load demand variation Δ q based on the m-th day i periodi(m) and corresponding rate of change of electricity price Δ pi(m) obtaining a vector of products made between the demand amount of the responsive preload participating in the demand response at the i-th day period and the rate of change in the electricity prices at the 24 i-th day period
Figure BDA0002940941580000041
Based on the product vector
Figure BDA0002940941580000042
And the elastic coefficient of the demand response price to obtain the variable quantity of the participation demand response in the i-th time period of the mth day
Figure BDA0002940941580000043
Estimated value of Δ q'i m；
Participating in demand response variation based on the i-th day
Figure BDA0002940941580000045
Estimated value of Δ q'i mAnd said product vector qi mAnd constructing an error minimum objective function.
Preferably, the solving the objective function with the minimum error by the gray wolf optimization algorithm based on the set number of wolf clusters, the maximum number of iterations, the direction correction probability and the initialized coordinate space of the wolf clusters to obtain the demand response elastic matrix includes:
updating the wolf pack position according to a preset rule to obtain a new wolf pack characteristic feature formed by a new generation of grey wolf individuals, and judging whether the maximum iteration times is reached currently;
if the maximum iteration times are reached, judging whether 24 time periods are met, if so, outputting a demand response elastic matrix, otherwise, re-optimizing the demand price elastic coefficient of the next time period, and outputting the demand response elastic matrix until 24 time periods are executed;
if the iteration maximum number is not reached, updating the position of the wolf pack according to the preset rule again, adding 1 to the iteration number, and continuously executing the step of updating the position of the wolf pack according to the preset rule until the iteration maximum number is reached.
Preferably, the updating the wolf pack position according to the predetermined rule to obtain the new generation wolf individuals to form the new wolf pack characteristics includes: updating the position of the wolf group according to rules including surrounding, hunting and updating when the gray wolf is hunted, generating a new generation gray wolf individual, and combining the father generation and the generated new offspring generation through optimization to form a new wolf group characteristic.
In a second aspect, an embodiment of the present invention provides a combined load prediction apparatus, including:
the data determining unit is used for determining the value of the forecast meteorological factor, the basic load, the response load, the load before and after the response and the electricity price of the day to be predicted;
the load prediction unit is used for inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be predicted into a prediction model to obtain a power grid load prediction value output by the prediction model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices on a plurality of historical days;
and the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the combined load prediction method according to any one of the foregoing first aspects when executing the program.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the combined load prediction method according to any one of the above-mentioned first aspects.
According to the combined load forecasting method, the combined load forecasting device and the combined load forecasting medium, the value of the forecast meteorological factor of the day to be forecasted, the basic load, the response load, the load before and after response and the electricity price are input into the forecasting model, the power grid load forecasting value is output, accurate forecasting of the load participating in demand response is achieved, and the problems that uncertainty of load operation and complexity of load influence factors are faced, and short-term load is difficult to accurately forecast by a single load forecasting method are solved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a combined load prediction method provided by the present invention;
FIG. 2 is a block diagram of a predictive model provided by the present invention;
FIG. 3 is a schematic diagram of basic load prediction model training provided by the present invention;
FIG. 4 is a schematic diagram of the response load prediction model training provided by the present invention;
FIG. 5 is a comparison of the effect of short term base load prediction provided by the present invention;
FIG. 6 is a graph comparing a predicted response load curve to an actual load curve provided by the present invention;
FIG. 7 is a graph comparing the predicted load value and the actual load value after response provided by the present invention;
FIG. 8 is a schematic structural diagram of a combined load prediction device provided in the present invention;
fig. 9 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A combined load prediction method, apparatus, electronic device and storage medium provided by the present invention are described below with reference to fig. 1 to 9.
The embodiment of the invention provides a combined load prediction method. Fig. 1 is a schematic flowchart of a combined load prediction method according to an embodiment of the present invention, as shown in fig. 1, the method includes:
step 110, determining the value of the forecast meteorological factor, the basic load, the response load, the load before and after the response and the electricity price of the day to be forecasted;
specifically, values of the forecast temperature and humidity of the day to be forecasted, load data, the load before and after response and the electricity price are obtained, and the load data of the day to be forecasted are decomposed into a basic load and a response load based on a fractal characteristic correction similar day method; the base load is independent of a price factor and the response load is dependent on a price factor.
Step 120, inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after the response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices on a plurality of historical days;
the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
Specifically, after sampling the actual temperature and humidity values, the load data and the load data before and after response of a plurality of historical days, the load data in the sample data is decomposed into a basic load and a response load based on a fractal characteristic correction similar day method, the basic load in the sample data is irrelevant to a price factor, and the response load in the sample data is relevant to the price factor.
According to the method provided by the embodiment of the invention, the prediction model is trained on the basis of the actual meteorological factor values, the basic load, the response load, the loads before and after response and the sample data of the electricity prices of a plurality of historical days, and the power grid load prediction value of the day to be predicted is obtained by inputting the forecast meteorological factor values, the basic load, the response load, the loads before and after response and the electricity prices of the day to be predicted into the prediction model, so that the load participating in the demand response in the power grid system can be accurately predicted.
Based on any of the above embodiments, as shown in fig. 2, the prediction model includes a base load prediction model 210, a response load prediction model 220, and an overlay model 230;
inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after the response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model, wherein the method comprises the following steps:
inputting the value of the forecast environment factor and the basic load of the day to be forecasted into the basic load forecasting model 210, and outputting a basic load forecasting value;
inputting the response load of the day to be predicted, the load before and after response and the electricity price to the response load prediction model 220, and outputting a predicted value of the demand response load;
and inputting the basic load predicted value and the demand response load predicted value into the superposition model 230, and outputting a power grid load predicted value of the day to be predicted.
Specifically, the basic load predicted value and the demand response load predicted value are added to obtain a final power grid load predicted value.
Based on any of the above embodiments, the base load prediction model 210 is obtained by performing short-term base load prediction training based on the actual meteorological factor values and the sample data of the base load for the plurality of historical days;
as shown in fig. 3, the basic load prediction model training schematic diagram completes short-term basic load prediction training based on the actual meteorological factor value and the basic load in the sample data, and includes the following steps:
inputting the actual meteorological factor value and the basic load in the sample data into a CS-SVM model of the cuckoo optimization support vector machine set by initial parameters including a penalty coefficient, the width of a radial basis kernel function, an initial probability, a nest position and the maximum iteration number to obtain the optimal fitness value of the nest;
specifically, the following 3 ideal states are assumed to simulate the behavior of cuckoo in finding nests to show the cuckoo optimization search CS algorithm, which specifically includes:
1) assuming that each cuckoo produces only one bird egg at each egg laying, the hatching locations of the bird eggs are placed in random fashion into any of the bird nests.
2) And randomly extracting a group of bird nest positions by adopting a random method, and transmitting the bird nest positions to the next generation as the optimal bird nest (optimal solution).
3) Setting the number of cuckoo nests to be N, wherein the probability of finding cuckoo eggs is P in the nest searching process of the cuckoo nestsa。
Based on the assumed ideal state, the path and the position of the cuckoo nest are updated by adopting a formula (1):
Figure BDA0002940941580000091
in the formula:
Figure BDA0002940941580000092
representing the position of the descendant nest of the tth generation cuckoo in the ith nest;
Figure BDA0002940941580000093
the step control amount is shown; x is a radical of a fluorine atomt,bestFor the optimal solution of the t-th generation cuckoo nest iteration, α0Is a constant, take 0.01;
Figure BDA0002940941580000094
representing a dot product; l (λ) represents a search path, which obeys a Levy distribution:
Levy～u＝t-λ(1≤λ≤3) (2)
the symmetrical Levy stable distribution is obtained by calculation according to a Mantegna algorithm, and a specific calculation formula is shown as a formula (3):
Figure BDA0002940941580000095
in the formula: u, V are all normal distributions and λ is 1.5.
Wherein σ2Is calculated byThe following (4):
Figure BDA0002940941580000096
in the formula: Γ is the standard Gamma function.
According to the assumed third ideal state, the probability that the brook bird nest finds the bird egg in the nest searching process is PaThe nest owner can discard or create a new nest, which is equivalent to discarding a part of the solutions, i.e., iteratively generating new solutions as follows:
Figure BDA0002940941580000097
where γ means a scaling factor, is a standard (0,1) distribution, used to monitor the probability that a newly generated solution is replaced,
Figure BDA0002940941580000101
two random solutions during the t-th iteration.
Determining the current optimal bird nest position according to the optimal fitness value of the bird nest, updating other bird nest positions according to the current optimal bird nest position to generate a group of new bird nest positions, and comparing the current optimal bird nest position with the group of new bird nest positions to obtain a prediction error;
obtaining a better bird nest position according to the prediction error, comparing the probability of eliminating cuckoo eggs with the numerical value of the random number, replacing the current optimal bird nest position with a group of new bird nest positions when the probability of eliminating cuckoo eggs is smaller than the random number, judging whether the current iteration end condition is met, jumping out to find the value of the optimal bird nest position if the current iteration end condition is met, and otherwise, executing to continuously find the optimal bird nest position; wherein the iteration end condition comprises a maximum iteration number or an optimal fitness value determined by the width of a radial basis kernel function;
and updating parameters in the CS-SVM model of the cuckoo optimization support vector machine to establish a basic load prediction model for the optimal bird nest position, and completing short-term basic load prediction training.
The method and the device have the advantages that the value of the actual meteorological factor and the basic load in the sample data are substituted into the Cuckoo optimization support vector machine CS-SVM model, the optimal bird nest position corresponding to the optimal adaptive value is iteratively searched, and the CS-SVM model is introduced to achieve short-term basic load prediction training.
It should be noted that the short-term base load prediction model training based on the CS-SVM includes the following steps:
(1) inputting model data and preprocessing operation;
(2) setting parameters of the CS optimization SVM: penalty coefficient C, width sigma of radial basis kernel function and initial probability PaThe position (N) of the bird nest, the maximum iteration number and the like;
(3) determining the current optimal solution (optimal bird nest position) according to the bird nest optimal fitness value;
(4) updating the positions of the bird nests of other cuckoos through the optimal solution to generate a group of new positions of the bird nests of the cuckoos, and calculating the position of a better bird nest;
(5) comparing the nest position of the last generation of cuckoo with the nest position of the new generation of cuckoo according to the predicted error value to obtain a better nest position;
(6) to Pa(probability of eliminating cuckoo egg) is compared with the value of r (random number), when P isaIf the result is less than r, replacing the original (abandoned) position with the newly generated brook bird nest position, comparing the fitness function values before and after the adjustment, and finally selecting the optimal brook bird nest position and the fitness value;
(7) judging whether an iteration ending condition (maximum iteration times or an optimal fitness function) is reached, if the iteration ending condition is met, jumping out to seek optimization, otherwise, returning to continue to seek optimization;
(8) and (C, sigma) values of the optimal bird nest position are used as parameter set values of the support vector machine, and a short-term basic load prediction model based on the CS-SVM is established.
Based on any of the above embodiments, the response load prediction model 220 is obtained by performing short-term response load prediction training based on the response loads of the plurality of historical days, the sample data of the loads before and after response and the electricity prices;
as shown in fig. 4, the response load prediction model training diagram completes short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data, and includes the following steps:
410, constructing an objective function with the minimum error according to the response load, the load before and after response and the electricity price in the sample data;
solving the target function with the minimum error based on a grey wolf optimization algorithm for setting the number of wolf clusters, the maximum iterative times, the direction correction probability and the initialized coordinate space of the wolf clusters to obtain a demand response elastic matrix;
430, establishing the response load prediction model based on the demand response elastic matrix, and completing short-term response load prediction training.
The embodiment of the invention substitutes the response load, the load before and after response and the electricity price in the sample data into the gray wolf optimization algorithm, constructs the objective function with the minimum error, continuously and iteratively updates the wolf cluster aiming at the initialized coordinate space of the wolf cluster, and finally outputs the demand response elastic matrix, namely, the gray wolf optimization algorithm is introduced to realize the short-term response load prediction training.
It should be noted that the training of the response load prediction model for elastic matrix modification based on the gray wolf optimization algorithm includes the following steps:
1) reading electricity price and load data before and after 24h of response every day for M days;
2) the load demand variation Δ q in the i-th day period is calculated according to the following formula (6) and formula (7), respectivelyi(m) and rate of change of electricity price Δ p for each periodi(m)：
Figure BDA0002940941580000121
Figure BDA0002940941580000122
Wherein i is 1,2, …,24, M is 1,2, …, M,
Figure BDA0002940941580000123
3) obtaining the original electricity consumption in the ith time period of M days and the electricity price change rate delta p in all time periods (24 time periods) of the day through the following formula (8)i(m) a product vector qi(M×24)；
Figure BDA0002940941580000124
Wherein,
Figure BDA0002940941580000125
Pj0 mrepresents the electricity rate before the j time period of the m day to participate in the user response of the demand response,
Figure BDA0002940941580000126
indicating the acceptable electricity rates of the users participating in the demand response in the j period of the mth day, namely the response electricity rates of the users,
Figure BDA0002940941580000127
the product vector is made between the power load demand amount before the response of the power consumer participating in the demand response in the i-th day and the power price change rate of the m-th day in the whole period (24 hours);
4) obtaining the participation demand response variable quantity of the ith time period of M days according to the demand response price elasticity
Figure BDA0002940941580000128
Estimated value of Δ q'i mAnd can be represented by formula (9):
Δq′i m＝e(i，j)·(qi m)T i＝j＝1，2，…，24 (9)
5) the minimum reconstruction error of the i-period participation demand response variation is taken as an objective function, and the objective function can be expressed by a formula (10):
Figure BDA0002940941580000132
6) solving the objective function by adopting a gray wolf optimization algorithm in an intelligent optimization algorithm, and setting the following parameters: d is the number of wolfs, ymaxRepresenting the maximum number of iterations, PVIndicating the directional correction probability, Ei(1×24)The 24 elements of (1) are set as the initialized coordinate space of the wolf group;
7) updating the wolf group position according to rules of surrounding, hunting, updating and the like when the grey wolf is hunted, generating a new generation of grey wolf individual, and combining the parent generation and the generated new offspring generation through a traditional elite-maintained preferred strategy to form a new wolf group characteristic which is marked as D;
8) judging whether a population iteration end condition y is reachedmaxIf yes, carrying out the next step, if the condition is not met, executing the previous step, and executing the operation of adding 1 for the iteration times;
9) judging whether i is more than or equal to 24, if so, outputting a demand response elastic matrix; if not, the loop is re-entered, the demand price elasticity coefficient of the next time interval is optimized, i +1 operation is executed, and when i is 24, all steps are finished, and the demand response elasticity matrix is output.
Based on any one of the above embodiments, the constructing an objective function with a minimum error according to the response load, the load before and after response, and the electricity price in the sample data includes:
respectively obtaining the load demand variation delta q of the i period of the M day based on the load before and after the response of 24 periods of each day and the electricity price in the M daysi(m) and corresponding rate of change of electricity price Δ pi(m)；
A load demand variation Δ q based on the m-th day i periodi(m) and corresponding rate of change of electricity price Δ pi(m) obtaining a vector of products made between the demand amount of the responsive preload participating in the demand response at the i-th day period and the rate of change in the electricity prices at the 24 i-th day period
Figure BDA0002940941580000133
Based on the product vector
Figure BDA0002940941580000134
And the elastic coefficient of the demand response price to obtain the variable quantity of the participation demand response in the i-th time period of the mth day
Figure BDA0002940941580000135
Estimated value of Δ q'i m；
Participating in demand response variation based on the m-th day period i
Figure BDA0002940941580000137
Estimated value of Δ q'i mAnd the product vector
Figure BDA0002940941580000139
And constructing an error minimum objective function.
Based on any of the above embodiments, the solving the objective function with the minimum error based on the gray wolf optimization algorithm that sets the number of wolf clusters, the maximum number of iterations, the direction correction probability, and the initialized coordinate space of the wolf clusters to obtain the demand response elastic matrix includes:
updating the wolf pack position according to a preset rule to obtain a new wolf pack characteristic formed by a new generation of grey wolf individuals, and judging whether the maximum iteration times is reached currently;
if the maximum number of iterations is reached, judging whether the 24 time intervals are met, if so, outputting a demand response elastic matrix, otherwise, re-optimizing the demand price elastic coefficient of the next time interval, and outputting the demand response elastic matrix until 24 time intervals are executed;
if the iteration maximum number is not reached, updating the position of the wolf pack according to the preset rule again, adding 1 to the iteration number, and continuously executing the step of updating the position of the wolf pack according to the preset rule until the iteration maximum number is reached.
Specifically, an intelligent optimization algorithm is adopted for the objective function with the minimum errorSolving the grey wolf optimization algorithm, and setting the following parameters: d is the number of wolf groups, ymaxRepresenting the maximum number of iterations, PVIndicating the directional correction probability, Ei(1×24)The 24 elements of (1) are set as the initialized coordinate space of the wolf group;
updating the wolf group position according to rules of surrounding, hunting, updating and the like when the gray wolf is hunted, creating a new generation of gray wolf individuals, and combining the father generation and the generated new offspring through a traditional elite-maintained preferred strategy to form a new wolf group characteristic which is marked as D.
Judging whether a population iteration end condition y is reachedmaxIf yes, carrying out the next step, if the condition is not met, executing the previous step, and executing the operation of adding 1 for the iteration times;
judging whether i is more than or equal to 24, if so, outputting a demand response elastic matrix; if not, the loop is re-entered, the demand price elasticity coefficient of the next time interval is optimized, i +1 operation is executed, and when i is 24, all steps are ended, and the demand response elasticity matrix is output.
Based on any of the above embodiments, the updating the wolf pack position according to the predetermined rule to obtain a new generation of grey wolf individuals to form a new wolf pack characteristic includes: updating the wolf group position according to the rules including surrounding, hunting and updating when the grey wolf is hunted, generating a new generation of grey wolf individual, and combining the parent generation and the generated new offspring generation preferentially to form new wolf group characteristics.
Any of the above embodiments of the present invention are described below with particular reference to applications as follows:
as shown in fig. 5, historical base load data, temperature and humidity data of 24 hours per day in 2016, 6, 7 and a year in load prediction of the nation a are selected as training sets, and 24-hour base load in 2016, 8, 1 and a year is predicted by adopting the CS-SVM and other 2 prediction models provided by the invention. In the model of the CS optimization SVM, the parameters are set as follows: n-20, Pa＝0.01、C＝50、σ＝0.05。
The invention adopts indexes of Root Mean Square Error (RMSE), average relative percentage error (MAPE) and relative error percentage (e) to judge the quality of the model prediction result provided by the invention, and the specific calculation formula of each index is as follows:
Figure BDA0002940941580000151
Figure BDA0002940941580000152
Figure BDA0002940941580000153
wherein: x is a radical of a fluorine atomiRepresenting the actual load value of the ith hour of the day to be predicted;
Figure BDA0002940941580000154
the predicted load value of the ith hour of the day to be predicted is represented, the number of samples is n, and n is 1,2, …, 24.
The average error values of the 3 predictions of the above 3 base load prediction models for 8 months and 1 day are shown in table 1.
TABLE 1 comparison table of 3-time average prediction errors of each model
Figure BDA0002940941580000155
As can be seen, the prediction errors RMSE and MAPE of the CS-SVM are the smallest, so that the prediction accuracy of the method is the highest.
As shown in FIG. 6, the load before and after response and the electricity price data before and after response of two weeks (14 days) before the prediction day (8 months and 1 day) are selected, and the demand price elasticity and the actual value of the basic load are optimized through the gray wolf algorithm to obtain a comparison graph of the response load prediction curve and the actual load curve. The error ratio of the response load predicted value to the actual value is shown in table 2.
TABLE 2 error comparison table for predicted value and actual value of response load
Figure BDA0002940941580000161
Therefore, the predicted value curve of the active demand response load obtained by the prediction method provided by the invention is compared with the actual value curve and is fit. In Table 2, 9:00 am, the relative error percentage reaches a maximum value of emax32.1157 pm at 8:00 pm, the relative error percentage appears to be the minimum, emin0.7276, the overall error is small, and the proposed gray wolf algorithm corrects the demand price elasticity-based response load prediction effectively.
As shown in fig. 7, based on the prediction of the previous base load and the response load, the predicted base load and the response load are superimposed to finally obtain a load prediction curve after response. The error ratio of the predicted value to the actual value of the after-response load is shown in table 3.
TABLE 3 comparison table of error between predicted value and actual value of load after response
Figure BDA0002940941580000171
Figure BDA0002940941580000181
Therefore, the predicted load curve of the power consumer after active response obtained by the prediction method provided by the invention is relatively fitted with the actual load curve, and the predicted value curve of the active demand response load is fitted with the actual value curve. The effectiveness of the method of the invention is demonstrated herein. In Table 3, the maximum relative error percentage emax-2.0910, occurring in the morning at 7: 00; minimum relative error percentage emin-0.0354, which occurs at 8:00 pm. The error index RMSE is 6.8935, the MAPE is 0.95, the total error is small, and the response load prediction method based on the gray wolf algorithm for correcting the demand price elasticity is effective.
In summary, compared with the prior art, the invention has the beneficial effects that: the combined load forecasting method and the combined load forecasting device deeply excavate the autocorrelation of the electricity price difference, the load difference, the temperature and humidity sequence and the load, and excavate the relation between the historical value and the value to be forecasted by means of the demand-price elasticity and the support vector machine tool in the economics; by comparing and analyzing the CS-SVM, the PSO-SVM, the SVM and the BP prediction model which take the demand price elasticity into consideration, the method provided by the invention has the advantages of lower prediction error and higher speed; meanwhile, the CS-SVM prediction model considering demand price elasticity introduced by the electric price difference and the load difference has lower prediction error than the same model not introducing demand price elasticity, and the prediction method considering demand price elasticity designed by the invention improves the prediction precision.
The following describes a combined load prediction device provided by the present invention, and the following description and the above-described combined load prediction method may be referred to in correspondence.
Fig. 8 is a schematic structural diagram of a combined load prediction apparatus according to an embodiment of the present invention, as shown in fig. 8, the apparatus includes a data determination unit 810 and a load prediction unit 820;
a data determination unit 810 for determining a value of a forecasted weather factor for a day to be predicted, a base load, a response load, a load before and after the response, and a power rate;
the load prediction unit 820 is used for inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be predicted into a prediction model to obtain a power grid load prediction value output by the prediction model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices of a plurality of historical days;
and the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
According to the device provided by the embodiment of the invention, the prediction model is trained on the basis of the actual meteorological factor values, the basic load, the response load, the loads before and after response and the sample data of the electricity price on a plurality of historical days, and the power grid load prediction value on the day to be predicted is obtained by inputting the forecast meteorological factor values, the basic load, the response load, the loads before and after response and the electricity price on the day to be predicted into the prediction model, so that the load participating in demand response in the power grid system can be accurately predicted.
Based on any of the above embodiments, the load prediction unit 820 includes a prediction model composed of a basic load prediction module, a response load prediction module, and a superposition module;
the basic load prediction module is used for inputting the value of the forecast environmental factor of the day to be predicted and the basic load and outputting a basic load prediction value;
the response load prediction module is used for inputting the response load of the day to be predicted, the load before and after response and the electricity price and outputting a predicted value of the demand response load;
and the superposition model is used for inputting the basic load predicted value and the demand response load predicted value and outputting the power grid load predicted value of the day to be predicted.
Based on any one of the above embodiments, the basic load prediction module is obtained by performing short-term basic load prediction training on the optimization module based on the actual meteorological factor values and basic load sample data of the plurality of historical days;
the optimization module comprises a CS-SVM module, a prediction error module, a position iteration module and a parameter updating module;
the CS-SVM module is used for inputting the value of an actual meteorological factor and a basic load in the sample data and outputting the optimal fitness value of the bird nest after setting initial parameters including a penalty coefficient, the width of a radial basis kernel function, an initial probability, the bird nest position and the maximum iteration frequency;
the prediction error module is used for determining the current optimal bird nest position according to the optimal adaptability value of the bird nest, updating other bird nest positions according to the current optimal bird nest position to generate a group of new bird nest positions, and comparing the current optimal bird nest position with the group of new bird nest positions to obtain a prediction error;
the position iteration module is used for obtaining a better bird nest position according to the prediction error, comparing the probability of eliminating the brook bird eggs with the numerical value of the random number, replacing the current optimal bird nest position with a group of new bird nest positions when the probability of eliminating the brook bird eggs is smaller than the random number, judging whether the current iteration end condition is met, jumping out to find the value of the optimal bird nest position if the iteration end condition is met, and continuously searching the optimal bird nest position if the iteration end condition is not met; wherein the iteration end condition comprises a maximum iteration number or an optimal fitness value determined by the width of a radial basis kernel function;
and the parameter updating module is used for updating parameters in the CS-SVM model of the Cuckoo optimization support vector machine to establish a basic load prediction model for the optimal bird nest position and finish short-term basic load prediction training.
Based on any one of the embodiments, the response load prediction module is obtained by performing short-term response load prediction training on the target module based on the response loads of the plurality of historical days, the sample data of the loads before and after response and the sample data of the electricity price;
the target module comprises a target function module, an optimization algorithm module and a model construction module;
the objective function module is used for constructing an objective function with the minimum error according to the response load, the load before and after response and the electricity price in the sample data;
the optimization algorithm module is used for solving the target function with the minimum error based on a gray wolf optimization algorithm for setting the number of wolf groups, the maximum number of iterations, the direction correction probability and the initialized coordinate space of the wolf groups to obtain a demand response elastic matrix;
and the model building module is used for building the response load prediction model based on the demand response elastic matrix and completing short-term response load prediction training.
Based on any one of the embodiments above, the constructing an objective function with a minimum error according to the response load, the load before and after response, and the electricity price in the sample data includes:
respectively obtaining the load demand variation delta q of the i-th day period of the M day based on the load before and after the 24-time-period response in the M days and the electricity pricei(m) and corresponding rate of change of electricity price Δ pi(m)；
A load demand variation Δ q based on the m-th day i periodi(m) and corresponding rate of change of electricity price Δ pi(m) obtaining a vector of products made between the response preload demand amount participating in demand response at the i-th period on the m-th day and the electricity price change rates of the 24 periods on the m-th day
Figure BDA0002940941580000211
Based on the product vector
Figure BDA0002940941580000212
And demand response price elastic coefficient to obtain the m-th day i time interval participation demand response variable quantity
Figure BDA0002940941580000213
Estimated value of Δ q'i m；
Participating in demand response variation based on the i-th day
Figure BDA0002940941580000215
Estimated value of Δ q'i mAnd the product vector
Figure BDA0002940941580000217
And constructing an error minimum objective function.
Based on any of the above embodiments, the solving the objective function with the minimum error based on the gray wolf optimization algorithm that sets the number of wolf clusters, the maximum number of iterations, the direction correction probability, and the initialized coordinate space of the wolf clusters to obtain the demand response elastic matrix includes:
updating the wolf pack position according to a preset rule to obtain a new wolf pack characteristic formed by a new generation of grey wolf individuals, and judging whether the maximum iteration times is reached currently;
if the maximum iteration times are reached, judging whether 24 time periods are met, if so, outputting a demand response elastic matrix, otherwise, re-optimizing the demand price elastic coefficient of the next time period, and outputting the demand response elastic matrix until 24 time periods are executed;
if the maximum iteration times are not reached, updating the wolf pack position according to the preset rule again, adding 1 to the iteration times, and continuously executing the step of updating the wolf pack position according to the preset rule until the maximum iteration times are reached.
Based on any of the above embodiments, the updating the wolf pack position according to the predetermined rule to obtain a new generation of grey wolf individuals to form a new wolf pack characteristic includes: updating the wolf group position according to the rules including surrounding, hunting and updating when the grey wolf is hunted, generating a new generation of grey wolf individual, and combining the parent generation and the generated new offspring generation preferentially to form new wolf group characteristics.
Fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 9, the electronic device may include: a processor (processor)910, a communication Interface (Communications Interface)920, a memory (memory)930, and a communication bus 940, wherein the processor 910, the communication Interface 920, and the memory 930 communicate with each other via the communication bus 940. Processor 910 may invoke logic instructions in memory 930 to perform a combined load prediction method comprising: determining the value, the basic load, the response load, the load before and after the response and the electricity price of a forecast weather factor of a day to be predicted; inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model; the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices on a plurality of historical days; the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
Furthermore, the logic instructions in the memory 930 may be implemented in software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute 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), a magnetic disk, or an optical disk, and various media capable of storing program codes.
In another aspect, an embodiment of the present invention further provides a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute the combined load prediction method provided by the above methods, where the method includes: determining the value of a forecast meteorological factor, a basic load, a response load, a load before and after response and an electricity price of a day to be predicted; inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model; the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices of a plurality of historical days; the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
In yet another aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the combined load prediction method provided in the foregoing, the method including: determining the value, the basic load, the response load, the load before and after the response and the electricity price of a forecast weather factor of a day to be predicted; inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model; the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices of a plurality of historical days; the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing the short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (6)

1. A combined load prediction method, comprising:
determining the value of a forecast meteorological factor, a basic load, a response load, a load before and after response and an electricity price of a day to be predicted;
inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices on a plurality of historical days;
the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data;
the prediction model comprises a basic load prediction model, a response load prediction model and a superposition model; inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be forecasted into a forecasting model to obtain a power grid load forecasting value output by the forecasting model, wherein the power grid load forecasting value comprises the following steps: inputting the value of the forecast environmental factor of the day to be forecasted and the basic load into the basic load forecasting model, and outputting a basic load forecasting value; inputting the response load of the day to be predicted, the load before and after response and the electricity price into the response load prediction model, and outputting a predicted value of the demand response load; inputting the basic load predicted value and the demand response load predicted value into the superposition model, and outputting a power grid load predicted value of a day to be predicted;
the response load prediction model is obtained by performing short-term response load prediction training on the basis of the response loads of the plurality of historical days, the loads before and after response and the sample data of the electricity price; completing short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data, and comprising the following steps of: constructing an error minimum objective function according to the response load, the load before and after response and the electricity price in the sample data; solving the target function with the minimum error based on a gray wolf optimization algorithm for setting the number of wolf clusters, the maximum number of iterations, the direction correction probability and the initialized coordinate space of the wolf clusters to obtain a demand response elastic matrix; establishing the response load prediction model based on the demand response elastic matrix to complete short-term response load prediction training;
the constructing of the objective function with the minimum error according to the response load, the load before and after the response and the electricity price in the sample data comprises the following steps: respectively obtaining the load demand variation of the i-th day period on the basis of the load and the electricity price before and after 24 periods of each day in the M days
Figure 40999DEST_PATH_IMAGE002
And corresponding rate of change of electricity prices
Figure 284898DEST_PATH_IMAGE004
(ii) a Load demand variation based on the m-th day i period
Figure 38353DEST_PATH_IMAGE002
And corresponding rate of change of electricity prices
Figure 734913DEST_PATH_IMAGE004
Obtaining a product vector made between the response front load demand quantity participating in demand response in the i period of the m day and the electricity price change rate of 24 periods of the m day
Figure 160079DEST_PATH_IMAGE006
(ii) a Based on the product vector
Figure 25529DEST_PATH_IMAGE008
And demand response price elastic coefficient to obtain the m-th day i time interval participation demand response variable quantity
Figure 182840DEST_PATH_IMAGE010
Is estimated value of
Figure 897856DEST_PATH_IMAGE012
(ii) a Participating in demand response variation based on the m-th day period i
Figure 64395DEST_PATH_IMAGE010
Is estimated by
Figure 518772DEST_PATH_IMAGE012
And the product vector
Figure 174881DEST_PATH_IMAGE008
Constructing an error minimum objective function;
the grey wolf optimization algorithm based on the set wolf pack number, the maximum iterative times, the direction correction probability and the initialized coordinate space of the wolf pack is used for solving the objective function with the minimum error to obtain a demand response elastic matrix, and the method comprises the following steps: updating the wolf pack position according to a preset rule to obtain the new wolf pack characteristics formed by the new generation of wolf individuals, and judging whether the maximum iteration times is reached currently; if the maximum number of iterations is reached, judging whether the 24 time intervals are met, if so, outputting a demand response elastic matrix, otherwise, re-optimizing the demand price elastic coefficient of the next time interval, and outputting the demand response elastic matrix until 24 time intervals are executed; if the iteration maximum number is not reached, updating the position of the wolf pack according to the preset rule again, adding 1 to the iteration number, and continuously executing the step of updating the position of the wolf pack according to the preset rule until the iteration maximum number is reached.
2. The combined load forecasting method according to claim 1, wherein the base load forecasting model is obtained by performing short-term base load forecasting training based on the actual meteorological factor values and base load sample data of the plurality of historical days;
and completing short-term base load prediction training based on the actual meteorological factor value and the base load in the sample data, wherein the method comprises the following steps:
s1, inputting the actual meteorological factor value and the basic load in the sample data into a CS-SVM (support vector machine-support vector machine) model of the brook optimization support vector machine set by initial parameters including a penalty coefficient, the width of a radial basis kernel function, an initial probability, a nest position and the maximum iteration number to obtain the optimal fitness value of the nest;
s2, determining the current optimal bird nest position according to the optimal fitness value of the bird nest, updating other bird nest positions according to the current optimal bird nest position to generate a group of new bird nest positions, and comparing the current optimal bird nest position with the group of new bird nest positions to obtain a prediction error;
s3, obtaining a better bird nest position according to the prediction error, comparing the probability of eliminating cuckoo eggs with the numerical value of the random number, replacing the current optimal bird nest position with a group of new bird nest positions when the probability of eliminating cuckoo eggs is smaller than the random number, judging whether the current iteration end condition is met, jumping out to find the value of the optimal bird nest position if the current iteration end condition is met, and returning to S2 to continuously find the optimal bird nest position if the current iteration end condition is not met; wherein the iteration end condition comprises a maximum iteration number or an optimal fitness value determined by the width of a radial basis kernel function;
and S4, updating parameters in the CS-SVM model of the Cuckoo optimization support vector machine to establish a basic load prediction model for the optimal bird nest position, and completing short-term basic load prediction training.
3. The method of claim 1, wherein the updating the wolf pack location according to the predetermined rule to obtain a new generation of wolf individuals forming a new wolf pack characteristic comprises: updating the wolf group position according to the rules including surrounding, hunting and updating when the grey wolf is hunted, generating a new generation of grey wolf individual, and preferably combining the parent generation and the generated new offspring generation to form new wolf group characteristics.
4. A combined load prediction device, comprising:
the data determining unit is used for determining the value of the forecast meteorological factor, the basic load, the response load, the load before and after the response and the electricity price of the day to be predicted;
the load prediction unit is used for inputting the value of the forecast meteorological factor, the basic load, the response load, the load before and after response and the electricity price of the day to be predicted into a prediction model to obtain a power grid load prediction value output by the prediction model;
the prediction model is obtained by training sample data of actual meteorological factor values, basic loads, response loads, loads before and after response and electricity prices on a plurality of historical days;
the prediction model is used for completing short-term basic load prediction training based on the value of the actual meteorological factor and the basic load in the sample data, and predicting the power grid load of the day to be predicted after completing short-term response load prediction training based on the response load, the load before and after response and the electricity price in the sample data;
the load prediction unit comprises a prediction model consisting of a basic load prediction module, a response load prediction module and a superposition module;
the basic load prediction module is used for inputting the value of the forecast environment factor of the day to be predicted and the basic load and outputting a basic load prediction value;
the response load prediction module is used for inputting the response load of the day to be predicted, the load before and after response and the electricity price and outputting a predicted value of the demand response load;
the superposition module is used for inputting the basic load predicted value and the demand response load predicted value and outputting a power grid load predicted value of a day to be predicted;
the response load prediction module is obtained by performing short-term response load prediction training on a target module based on the response loads of the plurality of historical days, and sample data of loads before and after response and electricity prices;
the target module comprises a target function module, an optimization algorithm module and a model construction module;
the objective function module is used for constructing an objective function with the minimum error according to the response load, the load before and after response and the electricity price in the sample data;
the optimization algorithm module is used for solving the target function with the minimum error based on a gray wolf optimization algorithm for setting the number of wolf groups, the maximum number of iterations, the direction correction probability and the initialized coordinate space of the wolf groups to obtain a demand response elastic matrix;
the model building module is used for building the response load forecasting module based on the demand response elastic matrix and completing short-term response load forecasting training;
the method for constructing the minimum objective function of the error according to the response load, the load before and after the response and the electricity price in the sample data comprises the following steps: respectively obtaining the load demand variation of the i-th day period on the basis of the load and the electricity price before and after 24 periods of each day in the M days
Figure 846034DEST_PATH_IMAGE014
And corresponding rate of change of electricity prices
Figure 816264DEST_PATH_IMAGE016
(ii) a Load demand variation based on the m-th day i period
Figure 402446DEST_PATH_IMAGE014
And corresponding rate of change of electricity prices
Figure 432719DEST_PATH_IMAGE004
Obtaining a product vector made between the response front load demand quantity participating in demand response in the i period of the m day and the electricity price change rate of 24 periods of the m day
Figure 591168DEST_PATH_IMAGE006
(ii) a Based on the product vector
Figure 630668DEST_PATH_IMAGE008
And the elastic coefficient of the demand response price to obtain the variable quantity of the participation demand response in the i-th time period of the mth day
Figure DEST_PATH_IMAGE018
Is estimated value of
Figure DEST_PATH_IMAGE020
(ii) a Participating in demand response variation based on the i-th day
Figure 653113DEST_PATH_IMAGE018
Is estimated by
Figure 57549DEST_PATH_IMAGE020
And the product vector
Figure 703294DEST_PATH_IMAGE008
Constructing an objective function with minimum error;
the solving of the error minimum objective function is performed by the gray wolf optimization algorithm based on the set wolf pack number, the maximum iterative times, the direction correction probability and the initialization coordinate space of the wolf pack to obtain the demand response elastic matrix, and the method comprises the following steps: updating the wolf pack position according to a preset rule to obtain the new wolf pack characteristics formed by the new generation of wolf individuals, and judging whether the maximum iteration times is reached currently; if the maximum iteration times are reached, judging whether 24 time periods are met, if so, outputting a demand response elastic matrix, otherwise, re-optimizing the demand price elastic coefficient of the next time period, and outputting the demand response elastic matrix until 24 time periods are executed; if the iteration maximum number is not reached, updating the position of the wolf pack according to the preset rule again, adding 1 to the iteration number, and continuously executing the step of updating the position of the wolf pack according to the preset rule until the iteration maximum number is reached.
5. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the combined load prediction method according to any of claims 1 to 3 are implemented when the program is executed by the processor.
6. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the combined load prediction method according to any one of claims 1 to 3.
CN202110179158.XA 2021-02-09 2021-02-09 Combined load prediction method, combined load prediction device, electronic equipment and storage medium Active CN112836885B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110179158.XA CN112836885B (en) 2021-02-09 2021-02-09 Combined load prediction method, combined load prediction device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110179158.XA CN112836885B (en) 2021-02-09 2021-02-09 Combined load prediction method, combined load prediction device, electronic equipment and storage medium

Publications (2)

Publication Number Publication Date
CN112836885A CN112836885A (en) 2021-05-25
CN112836885B true CN112836885B (en) 2022-07-15

Family

ID=75933205

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110179158.XA Active CN112836885B (en) 2021-02-09 2021-02-09 Combined load prediction method, combined load prediction device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN112836885B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113505534B (en) * 2021-07-07 2023-09-15 南京工程学院 Load prediction method considering demand response

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102592246A (en) * 2010-11-30 2012-07-18 通用电气公司 System and method for estimating demand response in electric power systems
CN104318329A (en) * 2014-10-20 2015-01-28 国家电网公司 Power load forecasting method of cuckoo search algorithm improved support vector machine
CN207339413U (en) * 2017-08-29 2018-05-08 南京新联电能云服务有限公司 A kind of automatic demand response control system of air-conditioning
CN111860979A (en) * 2020-07-01 2020-10-30 广西大学 Short-term load prediction method based on TCN and IPSO-LSSVM combined model

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10739742B2 (en) * 2017-04-27 2020-08-11 Johnson Controls Technology Company Building energy system with stochastic model predictive control

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102592246A (en) * 2010-11-30 2012-07-18 通用电气公司 System and method for estimating demand response in electric power systems
CN104318329A (en) * 2014-10-20 2015-01-28 国家电网公司 Power load forecasting method of cuckoo search algorithm improved support vector machine
CN207339413U (en) * 2017-08-29 2018-05-08 南京新联电能云服务有限公司 A kind of automatic demand response control system of air-conditioning
CN111860979A (en) * 2020-07-01 2020-10-30 广西大学 Short-term load prediction method based on TCN and IPSO-LSSVM combined model

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Improved gray wolf optimization algorithm for solving placement and sizing of electrical energy storage system in micro-grids;Di Miao 等;《ISA Transactions》;20200215;第102卷;第376-387页 *
考虑安全策略的需求响应实时电价机制研究;庞泽宇;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20210115(第01期);第C042-108页 *
计及需求响应的主动配电网短期负荷预测;苏小林 等;《电力系统自动化》;20180307(第10期);第60-66页 *
需求侧响应下的两级优化调度策略研究;郭崇 等;《信息通信》;20171115(第11期);第165-166页 *

Also Published As

Publication number Publication date
CN112836885A (en) 2021-05-25

Similar Documents

Publication Publication Date Title
Wang et al. A seasonal GM (1, 1) model for forecasting the electricity consumption of the primary economic sectors
CN111027732B (en) Method and system for generating multi-wind power plant output scene
CN113361761A (en) Short-term wind power integration prediction method and system based on error correction
CN113592144A (en) Medium-and-long-term runoff probability forecasting method and system
CN112884236B (en) Short-term load prediction method and system based on VDM decomposition and LSTM improvement
CN112163671A (en) New energy scene generation method and system
CN112307672A (en) BP neural network short-term wind power prediction method based on cuckoo algorithm optimization
CN112381315A (en) LS-SVM intelligent platform area load prediction method and system based on PSO optimization
CN112836885B (en) Combined load prediction method, combined load prediction device, electronic equipment and storage medium
CN112884238A (en) Photovoltaic power generation power prediction method and device
CN117909888B (en) Intelligent artificial intelligence climate prediction method
CN112700057A (en) Power load prediction method based on modular recurrent neural network
CN114154753A (en) Load prediction method and system
CN117557375A (en) Transaction evaluation method and related device based on virtual power plant
CN109685242B (en) Photovoltaic ultra-short term combined prediction method based on Adaboost algorithm
CN116760006A (en) Wind power prediction method, device, equipment and storage medium
CN116562455A (en) Air temperature forecast data processing method and device of wind driven generator and computer equipment
CN109840308B (en) Regional wind power probability forecasting method and system
CN112581311B (en) Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants
CN114066068A (en) Short-term power load prediction method, device, equipment and storage medium
CN112686770A (en) Load prediction method and device based on demand price elastic correction
CN112950033A (en) Reservoir dispatching decision method and system based on reservoir dispatching rule synthesis
CN111612419A (en) Method and device for processing power declaration data and computer equipment
CN110910164A (en) Product sales forecasting method, system, computer device and storage medium
CN116362418B (en) Online prediction method for application-level manufacturing capacity of intelligent factory of high-end battery

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
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20230525

Address after: 730070 no.249 Wanxin North Road, Anning District, Lanzhou City, Gansu Province

Patentee after: STATE GRID GASU ELECTRIC POWER Research Institute

Address before: Qilihe District of Gansu city in Lanzhou Province, 730050 West East Road No. 648

Patentee before: STATE GRID GASU ELECTRIC POWER Research Institute

Patentee before: CHINA AGRICULTURAL University