CN113570164A - Method and storage medium for predicting demand response potential of industrial user - Google Patents
Method and storage medium for predicting demand response potential of industrial user Download PDFInfo
- Publication number
- CN113570164A CN113570164A CN202111026169.0A CN202111026169A CN113570164A CN 113570164 A CN113570164 A CN 113570164A CN 202111026169 A CN202111026169 A CN 202111026169A CN 113570164 A CN113570164 A CN 113570164A
- Authority
- CN
- China
- Prior art keywords
- demand response
- load
- industrial
- user
- industrial user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000004044 response Effects 0.000 title claims abstract description 166
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000003860 storage Methods 0.000 title claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000006399 behavior Effects 0.000 claims description 20
- 238000003062 neural network model Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 14
- 238000012216 screening Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 9
- 230000005284 excitation Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 230000001932 seasonal effect Effects 0.000 claims description 3
- 238000007667 floating Methods 0.000 claims description 2
- 230000005611 electricity Effects 0.000 description 7
- 230000009467 reduction Effects 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000009776 industrial production Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 108090000672 Annexin A5 Proteins 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- 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
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting demand response potential of an industrial user and a storage medium, comprising the following steps of: step S1, acquiring historical load data of an industrial user and preprocessing the data; step S2, modeling the response behavior of the industrial user, and solving to obtain a demand response load curve based on the preprocessed historical load data; step S3, quantifying the demand response potential of the industrial user by using the baseline load curve and the demand response load curve; step S4, extracting features influencing the demand response potential of industrial users; and step S5, establishing a dual-output neural network probability prediction model with characteristics as input, and predicting the demand response potential of the industrial user. The method can simulate the response behavior of industrial users and expand historical demand response potential data, thereby improving the accuracy of the prediction model. Meanwhile, the invention provides a real-time demand response potential day-ahead probability prediction model of the industrial user, and provides reference information for the industrial user in the electric power market transaction.
Description
Technical Field
The invention belongs to the field of demand response potential prediction, and particularly relates to an industrial user demand response potential quantification and day-ahead probability prediction method and a storage medium based on response behavior modeling.
Background
In recent years, with the development of smart grid technology and the wide access of various novel loads such as electric vehicles and 5G base station energy storage, demand response gets more and more attention and application. The management modes of demand response mainly include three types: peak clipping, valley filling, and peak shifting valley filling. Peak clipping refers to reducing the load during peak electricity consumption, valley filling refers to increasing the load during valley electricity consumption, and peak shifting and valley filling refers to adjusting the load usage pattern during peak valley period so that the load during peak period is shifted to the valley period. The demand response can relieve the situation that the power supply of the power system is short, and the operating pressure of the power system is reduced.
Demand response potential refers to the ability of a user to adjust the load during a demand response period. On one hand, the power consumption of most industrial users is large, the peak load time of the industrial users can be quickly and accurately adjusted to participate in demand response, and the stable operation of a power system is guaranteed in an emergency; on the other hand, most industrial equipment is already equipped with measurement, control and communication infrastructure, and the relevant data can be collected and processed more easily. Therefore, the industrial users are suitable for participating in demand response and have large response potential. A part of large industrial users can directly participate in the transaction of the auxiliary service market, and the response potential needs to be reported in advance in the transaction process of the auxiliary service market, so that the industrial users need to predict the response potential of the next day, and a market bidding strategy is formulated based on the prediction. If the actual response potential of the next day is greatly different from the reported response potential, the industrial user is subjected to economic punishment. Therefore, the day-ahead probability prediction of the demand response potential can provide reference information for industrial users in the transaction process of the auxiliary service market, and the decision risk is reduced, so that benefit maximization is realized.
At present, the policy related to demand response in China is still in a pilot stage, and the obtained historical demand response potential data is very little. Insufficient data for training can lead to insufficient information extraction of the prediction model, and the overfitting problem is easy to occur, so that the precision and the generalization capability of the prediction model are seriously influenced. Based on the method, response behaviors of the industrial users in an incentive type demand response project can be modeled, the response behaviors of the industrial users in the demand response project are simulated, the demand response potential is reasonably quantized, the area enclosed by a baseline load curve and a load curve after response is creatively provided as a demand response potential quantization index, and the characteristics influencing the demand response potential of the industrial users are screened, so that training data are expanded, and the precision of a demand response potential prediction model is improved. Meanwhile, the invention provides a dual-output neural network based probability prediction model for carrying out the day-ahead probability prediction on the demand response potential of the industrial user. The probability prediction model directly uses the upper limit and the lower limit of the demand response potential interval as the output result of the dual-output neural network model, and can effectively overcome the defects of carrying out massive data statistics and priori assumption in the traditional probability interval prediction method.
Disclosure of Invention
In order to solve the problems of less historical demand response potential data and low accuracy of a prediction model, the invention provides a response behavior modeling-based method for quantifying demand response potential of an industrial user and predicting the future probability, which comprises the following steps:
step S1, acquiring historical load data of an industrial user and preprocessing the data;
step S2, modeling the response behavior of the industrial user, and solving to obtain a demand response load curve based on the preprocessed historical load data;
step S3, quantifying the demand response potential of the industrial user by using the baseline load curve and the demand response load curve;
step S4, extracting features influencing the demand response potential of industrial users;
and step S5, establishing a dual-output neural network probability prediction model with characteristics as input, and predicting the demand response potential of the industrial user.
In the foregoing technical solution, further, in the step S1, acquiring historical load data of the industrial user and preprocessing the data, specifically, acquiring the historical load data of the industrial user, filling default data by using a piecewise linear interpolation method, and processing the resolution of the acquired data into 15 minutes.
The basic idea of the piecewise linear interpolation method is to divide two adjacent data into a small interval, and the data in each small interval is fitted by approximating a straight line, so as to reasonably fill the default data.
Further, the modeling of the response behavior of the industrial user in the step S2 specifically includes the following steps:
step S201, maximum income of an industrial user in an incentive type demand response project is an objective function, and the constraint that the economic expenditure can be reduced when the user participates in the demand response project is satisfied;
step S202, dividing the industrial load into a transferable load, reducing the load and an unadjustable load, and respectively modeling the response behaviors of the three types of loads in the incentive type demand response project.
Further, the step S3 specifically includes the following steps:
step S301, estimating the baseline load of the industrial user by adopting a High X of Y method;
step S302, an area enclosed by the baseline load curve and the responded load curve is used as a quantitative index of demand response potential.
Further, step S4 extracts features affecting the demand response potential of the industrial user, the extracted features including features affecting daily load size and features affecting user load adjustment ability;
further, the characteristics that affect the daily load magnitude include: weekday characteristics, weekend characteristics, holiday characteristics, seasonal characteristics; features that affect the ability of a user to adjust the load include: excitation signature, baseline load signature, time signature.
Between the steps S4 and S5, there is further included step S6: and calculating the maximum information coefficient between each characteristic and the demand response potential of the industrial user, and screening out partial characteristics with smaller maximum information coefficients.
Further, the method comprises the following specific steps:
601, respectively comparing the maximum information coefficient MIC between each feature and the demand response potential of the industrial user with a first preset threshold m1, and counting to obtain the feature with the MIC smaller than the first preset threshold m 1;
step 602, if the number of the features smaller than the first preset threshold m1 is less than or equal to a second preset threshold m2, screening out the features smaller than the first preset threshold m 1; if the number of features smaller than the first preset threshold m1 is greater than the second preset threshold m2, only m2 features with the minimum MIC coefficient are screened out.
Further, the step S5 specifically includes the following steps:
step S501, taking the characteristics based on the demand response potential as the input of a dual-output neural network model, carrying out small-amplitude up-and-down floating based on the demand response potential to obtain upper and lower limit values, and directly taking the upper and lower limit values as two outputs of the dual-output neural network model;
step S502, dividing data into a training set and a test set, setting initialization parameters based on the dual-output neural network model and training on the training set;
and S503, verifying the probability prediction effect of the dual-output neural network model in the test set.
The embodiment of the invention also provides a storage medium for storing a computer program, and the computer program is executed by a processor to realize the method for predicting the demand response potential of the industrial user.
The invention has the beneficial effects that:
(1) modeling the response behavior of the industrial user in the incentive type demand response project, simulating the response behavior of the industrial user in the demand response project, expanding training data, and solving the problems that the obtained historical demand response potential data is less, the insufficient data for training can cause insufficient information extraction of a prediction model, the overfitting problem is easy to occur, and the precision and the generalization capability of the prediction model are seriously influenced;
(2) the area enclosed by the baseline load curve and the responded load curve is used as a quantitative index of the demand response potential, so that the demand response potential of an industrial user can be reflected comprehensively and accurately;
(3) the response behavior modeling needs to meet the constraint that the user can reduce the economic expenditure when participating in the demand response project, can improve the enthusiasm of the industrial user for participating in the demand response project, and guarantees the stable operation of the power system while reducing the burden of an enterprise;
(4) screening the demand response potential features to screen out the features with low correlation, if only adopting the technical scheme of screening out a plurality of features with the lowest maximum information coefficients, possibly screening out the features with low correlation but later sequencing, and adopting the screening-out strategy in the invention can avoid mistakenly screening out the features with low correlation but later sequencing;
(5) the dual-output neural network probability prediction model directly uses the upper limit and the lower limit of the demand response potential interval as output results, and can effectively overcome the defect of carrying out massive data statistics and prior hypothesis in the traditional probability interval prediction method.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of a dual-output neural network of the present invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
Referring to fig. 1, fig. 1 is a flowchart of a method for predicting a demand response potential based on an industrial user according to an embodiment of the present invention, as shown in fig. 1, including the following steps:
step S1, acquiring historical load data of an industrial user and preprocessing the data;
step S2, modeling the response behavior of the industrial user, and solving to obtain a demand response load curve based on the preprocessed historical load data;
step S3, quantifying the demand response potential of the industrial user by using the baseline load curve and the demand response load curve;
step S4, extracting features influencing the demand response potential of industrial users;
and step S5, establishing a dual-output neural network probability prediction model with characteristics as input, and predicting the demand response potential of the industrial user.
And step S1, acquiring historical load data of the industrial user and preprocessing the data. Including acquiring historical load data of industrial users, populating default data with piecewise linear interpolation, and processing the acquired data to 15 minutes in resolution.
The basic idea of the piecewise linear interpolation method is to divide two adjacent data into a small interval, and the data in each small interval is fitted by approximating a straight line, so as to reasonably fill the default data.
Step S2, modeling the response behavior of the industrial user, and solving to obtain a demand response load curve, wherein the method comprises the following steps:
step S201, maximum income of an industrial user in an incentive type demand response project is an objective function, and the constraint that the economic expenditure can be reduced when the user participates in the demand response project is satisfied; the method specifically comprises the following steps:
in the formula: λ represents daily electricity price; wt,i,kRepresenting the consumed electric quantity of a kth demand response day T (T is equal to {1, 2.. multidot.T }, and T is the total time period of one day) period of a user i; inc represents a monetary reward per unit of power reduction; pen represents the unit currency punishment when the user does not meet the electric quantity reduction requirement agreed by the contract;dcontthe power consumption reduction of contract agreement is shown; s (μ) is a logical variable for judging whether the used amount of electricity meets the contract requirement, and is calculated by equation (2).
In the formula: for user i on the kth demand response day, Δ d represents the amount of power consumption reduction of the user during the demand response period:
wherein: Δ t represents a demand response period;
representing the baseline load of the user i at the kth demand response day t; pt,i,kRepresenting the after-response load of the user i at the time of the kth demand response day t. Equation (4) is used to determine whether the user's participation in the demand response program reduces the economic cost. If participation in the demand response program does not reduce the economic expenditure, the user may decline to participate in the program.
In the formula: req (W)t,i,k) Representing the consumed electric quantity when the user i does not participate in the demand response at the time of k days and t days on the demand response day; if F2If < 0, it indicates that the user participates in the demand response item, and if F, it can reduce the economic expenditure2And more than or equal to 0 indicates that the user can not reduce the economic expenditure when participating in the demand response project.
Step S202, dividing the industrial load into a transferable load, reducing the load and an unadjustable load, and respectively modeling the response behaviors of the three types of loads in the incentive type demand response project.
One is a transferable load, the load behavior of which has certain fluctuation and uncertainty, such as discontinuous industrial heat, and time-adjustable industrial production power consumption, and the power consumption time period of the load can move forwards or backwards in a certain time period without causing great influence on the production and the comfort of industrial users. When modeling it, the optimized transferable load should meet the following initial requirements:
in the formula: Δ T represents the time interval that can be shifted back and forth at a certain excitation, and the time intervals for different excitations will also differ; t is tstartRepresents the load on time; t is tendRepresents the end time of the load;
representing the power of the historical transferable load of the user i at the kth demand response day t;
indicating that user i may transfer the power of the load after responding at time t on the kth demand response day.
The second type of load is a load which can be reduced and has strong load behavior regulation capacity, such as the power consumption of the industrial production which is not needed urgently, and the load can reduce partial load within a certain time without causing great influence on the production and the comfort of industrial users. When modeling it, the optimized reducible load should meet the following initial requirements:
in the formula:
representing the power of the reducible load of the history of the user i at the kth demand response day t; gamma is a reduction coefficient;
the power of the reducible load after the user i responds at the time t of the kth demand response day.
The third type of load is an unadjustable load, the load behavior is more regular in electricity consumption and is not easy to schedule, for example, continuous industrial production electricity consumption, once the load electricity consumption is inconsistent with a plan, the normal production requirements of industrial users can be influenced, almost no elasticity exists in an incentive type demand response project, and the original load is still kept. The constraints of the non-adjustable load modeling are:
in the formula
The power of the historical non-adjustable load at the kth demand response day t is given to the user i;
the power of the unregulated load after the response of the user i at the kth demand response day t.
And constructing an optimization model according to the objective function and the constraint condition, reasonably setting a demand response time period and excitation, namely simulating the response behavior of the industrial user in a demand response project based on historical load data, and solving the optimization model to obtain a load curve after the industrial user responds in the demand response project.
Step S3, quantifying the demand response potential of the industrial user using the baseline load curve and the demand response load curve, comprising:
step S301, estimating the baseline load of the industrial user by adopting a High X of Y method. The baseline load of the user is the load that the user would otherwise consume if there were no demand response items. Since there is no historic baseline load after the demand response project is implemented, a reasonable estimate of the baseline load is needed. There are many methods to estimate the baseline load, and the High X of Y method is selected for estimation in this embodiment. The baseline load estimated by user i at the kth demand response day t is:
in the formula: high (X, Y, k, t) represents the X highest loads before Y non-demand response days t time before the kth demand response day. Wherein the demand response day refers to a demand response implementation day, and the non-demand response day refers to a non-implementation-demand response day.
Step S302, an area enclosed by the baseline load curve and the post-demand-response load curve is used as a demand response potential quantization index. In the embodiment, only the demand response during the peak period is considered, and the demand response potential at this time may be the area enclosed by the baseline load curve and the post-response load curve, that is:
in the formula: f. ofT,i,kRepresenting the demand response potential of user i during the kth demand response day, T.
And step S4, extracting characteristics influencing the demand response potential of the industrial user. Specifically, the extracted features need to satisfy two basic requirements: firstly, effectiveness, the extracted features need to influence the demand response potential of industrial users; secondly, easy availability. The extracted feature data needs to be easily acquired. Therefore, specific industrial loads and features that are not easily accessible, such as the volume of industrial equipment, the number of operators, etc., are not suitable as extracted features. The features extracted by the embodiment mainly have two parts: the first part is characteristics influencing the load magnitude of each day, including working day characteristics, weekend characteristics, holiday characteristics and season characteristics; the second component is a feature that affects the user's ability to adjust the load, including an excitation feature, a baseline load feature, and a time feature.
The encoding mode of the seasonal characteristics is as follows: the four seasons are coded as numbers 1,2,3 and 4 in spring, summer, autumn and winter; the encoding mode of week characteristics is as follows: encoding the Monday to the Sunday into numbers 1-7 respectively; the encoding mode of the weekend characteristics is as follows: saturday and Sunday are coded as the number 1, and Monday through Friday are coded as the number 0; the coding mode of the holiday features is as follows: the holiday code is a number 1, and the non-holiday code is a number 0; the coding mode of the time characteristics adopts one-hot coding. One-hot encoding uses an m-bit status register to encode m states, each having its own independent register bit and only one of which is active at any one time.
Between the steps S4 and S5, there is further included step S6: and calculating the maximum information coefficient between each characteristic and the demand response potential of the industrial user, and screening out partial characteristics with smaller maximum information coefficients.
The maximum information coefficient is derived on the basis of mutual information, the strength of the correlation between variables can be measured, the calculation complexity is low, and the robustness is high. For the random variable x and the random variable y, the mutual information is calculated as follows:
in the formula: p (x, y) is the joint probability between x, y. Carrying out grid division (a, b) on a scatter diagram formed by x and y data, respectively calculating mutual information in each grid, and selecting the maximum value of the mutual information under different grid division modes to obtain an MIC coefficient, namely:
in the formula: MIC (x; y) represents the maximum information coefficient of the random variable x and the random variable y; a and b are the number of the divided cells in the x and y directions; and B is the upper limit value of the grid division. The MIC coefficient range is [0, 1], and the closer the MIC coefficient is to 1, the stronger the correlation of the two variables is represented; conversely, the closer the MIC coefficient is to 0, the weaker the correlation between the two variables is.
601, respectively comparing the maximum information coefficient MIC between each feature and the demand response potential of the industrial user with a first preset threshold m1, and counting to obtain the feature with the MIC smaller than the first preset threshold m 1;
step 602, if the number of the features smaller than the first preset threshold m1 is less than or equal to a second preset threshold m2, screening out the features smaller than the first preset threshold m 1; if the number of features smaller than the first preset threshold m1 is greater than the second preset threshold m2, only m2 features with the minimum MIC coefficient are screened out.
In order to avoid screening out the characteristics with low relevance, m1 is 0.1 in the embodiment after repeated tests; since a total of 7 features are extracted in this embodiment, the information is greatly reduced by screening out too many features, and therefore m2 is 2 in this embodiment.
Step S5, establishing a dual-output neural network based probability prediction model to carry out the day-ahead probability prediction on the real-time demand response potential of the industrial user, wherein the method comprises the following steps:
step S501, please refer to fig. 2, taking the characteristics based on the demand response potential as the input of the dual-output neural network model, performing small fluctuation based on the demand response potential in step S3 to obtain upper and lower limit values, and directly taking the upper and lower limit values as two outputs of the dual-output neural network model; the embodiment takes the fluctuation of 25% of the demand response potential as the initial upper and lower limits of the dual-output neural network model. In the embodiment, the upper limit and the lower limit of the demand response potential interval are directly used as the output result of the dual-output neural network model, so that the defect of carrying out massive data statistics in the traditional probability interval prediction method can be effectively overcome, and the prior assumption on the probability distribution of point prediction errors is not needed.
Step S502, dividing the data into a training set and a testing set, wherein the training set accounts for 70% and the testing set accounts for 30%. Setting initialization parameters based on the dual-output neural network model and training on a training set;
step S503, verifying the probability prediction effect of the dual-output neural network model in the test set, specifically, calculating the reliability probability prediction evaluation index and the definition probability prediction evaluation index of the test set sample, and verifying the prediction effect of the model.
The selected reliability evaluation index is PICP, represents the probability that the actual observed value falls in the probability prediction interval, and has the formula as follows:
in the formula: n is the number of samples, ε (N) is an indicative function defined as follows:
in the formula: f. ofT,i,kRepresenting a demand response potential observed value of a user i in a kth demand response day T time period; l isTI, k, representing the lower limit of a demand response potential probability prediction interval in the kth demand response day T time period of the user i; u shapeT,i,kAnd representing the upper limit of the demand response potential probability prediction interval in the kth demand response day T time period of the user i. The greater the PICP value is, the greater the probability that the actual observation value falls in the probability prediction interval is, and the more accurate the probability prediction is.
The selected definition evaluation index is the average bandwidth PINAW of the prediction interval, can reflect the width of the prediction interval, and has the formula as follows:
in the formula: r is the variation range of the predicted target value, namely the difference between the maximum value of the demand response potential and the minimum value of the demand response potential; u shapejRepresenting the upper limit of the probability prediction interval of the jth sample; l isjRepresents the lower bound of the probability prediction interval for the jth sample. The smaller the PINAW value is, the wider the probability prediction interval is represented
The smaller the probability prediction, the more accurate the information provided.
The embodiment of the invention also provides a storage medium for storing a computer program, and the computer program is executed by a processor to realize the method for predicting the demand response potential of the industrial user.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application 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 in the embodiments of the present application.
Claims (10)
1. The method for predicting the demand response potential of the industrial user and the storage medium are characterized by comprising the following steps of:
step S1, acquiring historical load data of an industrial user and preprocessing the data;
step S2, modeling the response behavior of the industrial user, and solving to obtain a demand response load curve based on the preprocessed historical load data;
step S3, quantifying the demand response potential of the industrial user by using the baseline load curve and the demand response load curve;
step S4, extracting features influencing the demand response potential of industrial users;
and step S5, establishing a dual-output neural network probability prediction model with characteristics as input, and predicting the demand response potential of the industrial user.
2. The method according to claim 1, wherein in step S1, the data is preprocessed, specifically, the default data is padded by piecewise linear interpolation, and the resolution of the acquired data is processed to 15 minutes.
3. The method according to claim 1, wherein the step S2 of modeling the response behavior of the industrial user specifically comprises the steps of:
step S201, maximum income of an industrial user in an incentive type demand response project is an objective function, and the constraint that the economic expenditure can be reduced when the user participates in the demand response project is satisfied;
step S202, dividing the industrial load into a transferable load, reducing the load and an unadjustable load, and respectively modeling the response behaviors of the three types of loads in the incentive type demand response project.
4. The method of claim 1, wherein: the step S3 specifically includes the following steps:
step S301, estimating the baseline load of the industrial user by adopting a High X of Y method;
step S302, an area enclosed by the baseline load curve and the responded load curve is used as a quantitative index of demand response potential.
5. The method of claim 1, wherein: the features extracted at step S4 include features that affect the daily load size and features that affect the ability of the user to adjust the load.
6. The method of claim 5, wherein the characteristics that affect the daily load size comprise: weekday characteristics, weekend characteristics, holiday characteristics, seasonal characteristics; features that affect the ability of a user to adjust the load include: excitation signature, baseline load signature, time signature.
7. The method of claim 1, further comprising, between the steps S4 and S5, step S6: and calculating the maximum information coefficient between each characteristic and the demand response potential of the industrial user, and screening out partial characteristics with smaller maximum information coefficients.
8. The method according to claim 7, characterized in that it comprises in particular the steps of:
601, respectively comparing the maximum information coefficient MIC between each feature and the demand response potential of the industrial user with a first preset threshold m1, and counting to obtain the feature with the MIC smaller than the first preset threshold m 1;
step 602, if the number of the features smaller than the first preset threshold m1 is less than or equal to a second preset threshold m2, screening out the features smaller than the first preset threshold m 1; if the number of features smaller than the first preset threshold m1 is greater than the second preset threshold m2, only m2 features with the minimum MIC coefficient are screened out.
9. The method according to claim 1, 7 or 8, wherein the step S5 specifically includes the following steps:
step S501, taking the characteristics based on the demand response potential as the input of a dual-output neural network model, carrying out small-amplitude up-and-down floating based on the demand response potential to obtain upper and lower limit values, and directly taking the upper and lower limit values as two outputs of the dual-output neural network model;
step S502, dividing data into a training set and a test set, setting initialization parameters based on the dual-output neural network model and training on the training set;
and S503, verifying the probability prediction effect of the dual-output neural network model in the test set.
10. A storage medium storing a computer program which, when executed by a processor, implements the method of predicting demand response potential of an industrial user as claimed in any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111026169.0A CN113570164A (en) | 2021-09-02 | 2021-09-02 | Method and storage medium for predicting demand response potential of industrial user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111026169.0A CN113570164A (en) | 2021-09-02 | 2021-09-02 | Method and storage medium for predicting demand response potential of industrial user |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113570164A true CN113570164A (en) | 2021-10-29 |
Family
ID=78173473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111026169.0A Pending CN113570164A (en) | 2021-09-02 | 2021-09-02 | Method and storage medium for predicting demand response potential of industrial user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113570164A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114169802A (en) * | 2021-12-31 | 2022-03-11 | 佰聆数据股份有限公司 | Power grid user demand response potential analysis method, system and storage medium |
-
2021
- 2021-09-02 CN CN202111026169.0A patent/CN113570164A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114169802A (en) * | 2021-12-31 | 2022-03-11 | 佰聆数据股份有限公司 | Power grid user demand response potential analysis method, system and storage medium |
| CN114169802B (en) * | 2021-12-31 | 2022-08-16 | 佰聆数据股份有限公司 | Power grid user demand response potential analysis method, system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111091232B (en) | Power load prediction method considering power demand change trend | |
| CN116646933A (en) | Big data-based power load scheduling method and system | |
| CN102509173B (en) | A kind of based on markovian power system load Accurate Prediction method | |
| CN105405061A (en) | Evaluation method of power supply reliability of distribution network based on demand response | |
| CN109558626B (en) | Dynamic control method for cascade reservoir flood period operation water level based on time-varying design flood | |
| CN111680841A (en) | Short-term load prediction method and system based on principal component analysis and terminal equipment | |
| CN111724210A (en) | Power demand response regulation and control method based on price elastic coefficient matrix | |
| CN113469585A (en) | Carbon management data adjusting method, carbon management platform, system and storage medium | |
| CN113570164A (en) | Method and storage medium for predicting demand response potential of industrial user | |
| CN116050866A (en) | Combined user missing electric quantity fitting method and system based on typical load curve | |
| Zhou et al. | Eliciting private user information for residential demand response | |
| CN111210260A (en) | Electricity price data processing method and device, computer equipment and storage medium | |
| CN110162403A (en) | A kind of hardware resource distribution method and system based on artificial neural network | |
| CN117495056A (en) | Power consumption data monitoring and optimizing method and system | |
| CN111915211A (en) | Power resource scheduling method and device and electronic equipment | |
| CN116249186A (en) | Data processing method and device of wireless network equipment, storage medium and electronic equipment | |
| CN115511230A (en) | Electric energy substitution potential analysis and prediction method | |
| CN111797924B (en) | Three-dimensional garden portrait method and system based on clustering algorithm | |
| CN108615091A (en) | Electric power meteorology load data prediction technique based on cluster screening and neural network | |
| CN111404146B (en) | Power distribution method, system, terminal and medium based on user load transfer comfort | |
| CN111415040A (en) | Method and device for calculating suggested purchase amount based on PET purchase model | |
| CN112085394A (en) | User side energy storage assessment method and system considering demand side response | |
| CN111626494A (en) | Power load prediction method, power load prediction device, computer equipment and storage medium | |
| CN114649820B (en) | Multi-link capacity allocation method and system for wind-solar energy storage station | |
| CN115409427B (en) | Method and device for evaluating multiple states of resident load available regulation capacity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |