CN110826794A - Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine) - Google Patents

Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine) Download PDF

Info

Publication number
CN110826794A
CN110826794A CN201911053400.8A CN201911053400A CN110826794A CN 110826794 A CN110826794 A CN 110826794A CN 201911053400 A CN201911053400 A CN 201911053400A CN 110826794 A CN110826794 A CN 110826794A
Authority
CN
China
Prior art keywords
svm
pso
coal consumption
prediction
power plant
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201911053400.8A
Other languages
Chinese (zh)
Inventor
彭道刚
车权
李一琨
潘永旗
赵慧荣
李军
王丹豪
孔雯
黄雨薇
梁伟
李甫
朱琪
杨梦雪
李建芳
高义民
钟鸿舟
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.)
Shanghai Houshang Electric Power Technology Co ltd
Shanghai Qianhuan Intelligent Technology Co ltd
State Grid Chongqing Electric Power Co Ltd
Shanghai University of Electric Power
Original Assignee
Shanghai Houshang Electric Power Technology Co ltd
Shanghai Qianhuan Intelligent Technology Co ltd
State Grid Chongqing Electric Power Co Ltd
Shanghai University of Electric Power
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 Shanghai Houshang Electric Power Technology Co ltd, Shanghai Qianhuan Intelligent Technology Co ltd, State Grid Chongqing Electric Power Co Ltd, Shanghai University of Electric Power filed Critical Shanghai Houshang Electric Power Technology Co ltd
Priority to CN201911053400.8A priority Critical patent/CN110826794A/en
Publication of CN110826794A publication Critical patent/CN110826794A/en
Pending legal-status Critical Current

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
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Health & Medical Sciences (AREA)
  • Educational Administration (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a power plant coal consumption reference value rolling prediction method and a device based on a PSO (particle swarm optimization) SVM (support vector machine), wherein the method comprises the following steps: acquiring daily coal consumption data of a power plant as input data; carrying out filtering smoothing treatment on input data by adopting a nine-point quadratic exponential smoothing method; establishing a rolling updating prediction model based on the SVM; performing parameter optimization on the SVM model through a PSO algorithm to obtain a PSO-SVM model; the input data smoothed in step S2 is input to the PSO-SVM model in step S4 to perform rolling prediction of the coal consumption reference value, and a prediction result is obtained. Compared with the prior art, the method can accurately and effectively carry out rolling test and prediction updating on the coal consumption reference value required by power grid dispatching.

Description

Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine)
Technical Field
The invention relates to a power grid dispatching side prediction method, in particular to a power plant coal consumption reference value rolling prediction method and device based on a PSO (particle swarm optimization) SVM (support vector machine).
Background
In recent years, with the transformation and upgrading of national economic structures, the implementation of energy revolution strategies and the continuous promotion of electric power market transformation, an optimal scheduling strategy capable of meeting electric power and electric quantity balance is urgently needed by a power grid, and the optimal economic distribution of the load of a power plant unit is realized. The power grid dispatching method can use a coal consumption reference value in research and formulation of power and coal supply grading early warning and load economic distribution of each power plant, and has an important effect on monitoring the number of days for storing and using coal in the future of the power plant and formulating a reasonable power generation dispatching plan.
At present, no accurate algorithm aiming at the coal consumption reference value exists at home and abroad. For the time sequence model for establishing the coal consumption reference value prediction, a time sequence method and a neural network deep learning algorithm are mainly adopted. However, the daily coal consumption of the thermal power plant under the power grid is influenced by various uncontrollable factors, the fluctuation of the variation trend is high, an uncontrollable error result is easily caused by adopting a fitting algorithm commonly used in engineering, a large error is caused by simply selecting historical data to perform single-step prediction, and a complex neural network algorithm is difficult to realize in the programming of actual software.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a power plant coal consumption benchmark rolling prediction method and device based on a PSO optimization SVM.
The purpose of the invention can be realized by the following technical scheme:
a power plant coal consumption reference value rolling prediction method based on a PSO (particle swarm optimization) SVM (support vector machine) comprises the following steps:
s1, acquiring daily coal consumption data of the power plant as input data;
s2, performing filtering smoothing processing on input data by adopting a nine-point quadratic exponential smoothing method;
s3, establishing a rolling updating prediction model based on the SVM;
s4, carrying out parameter optimization on the SVM model through a PSO algorithm to obtain a PSO-SVM model;
and S5, inputting the input data smoothed in the step S2 into the PSO-SVM model in the step S4 to perform rolling prediction of the coal consumption reference value, and acquiring a prediction result.
Further, in step S2, before the filtering and smoothing process, the value of the discontinuity in the input data is modified to be an average value of 5 days before and 5 days after the discontinuity.
Further, in step S2, the quadratic exponential smoothing formula is:
in the formula (I), the compound is shown in the specification,and
Figure BDA0002255912280000023
respectively the secondary exponential smoothing values of the t period and the t-1 period; a is a smoothing coefficient, in
Figure BDA0002255912280000024
And
Figure BDA0002255912280000025
under known conditions, the prediction model of the quadratic exponential smoothing method is as follows:
Figure BDA0002255912280000026
Figure BDA0002255912280000027
Figure BDA0002255912280000028
in the formula, at、btIs a prediction model parameter consisting of a smoothing coefficient and a quadratic index of t phase and t-1 phaseCalculating a number smoothing value; t is the number of lapsed periods backwards from T; y ist+TIs the predicted value of T + T phase.
Furthermore, in the rolling updating prediction model, n days are taken as a time standard for evaluating the coal consumption standard, wherein n is more than or equal to 18 and less than or equal to 25; and taking k days as the current day, taking input data as the daily coal consumption of k-49 days to k days, and taking corresponding output data as the average value of n days after the k +1 th day.
Further, in step S4, a PSO algorithm is introduced to the built SVM model to optimize the penalty factor c and the kernel parameter g, so as to obtain an optimized PSO-SVM model.
Further, the specific optimization process is to predict the data samples by using the SVM corresponding to each particle, calculate the current position value and prediction error of each particle, and use them as the fitness value of the particle, compare the fitness value of the current position of the particle with the fitness value of itself, and the optimal value is the current optimal position of the particle, and continuously update the latest position and velocity of the particle.
Further, the velocity and position of the particle are updated as follows:
Figure BDA00022559122800000210
in the formula, k is iteration times; omega is the inertial weight; d is dimension; c. C1,c2Is a learning factor; r is1,r2∈[0,1]Is a random number;
Figure BDA00022559122800000211
for the d-dimensional component of the velocity vector of the particle i at the kth iteration,
Figure BDA00022559122800000212
for the d-dimensional component of the location vector of the particle i at the k-th iteration,
Figure BDA00022559122800000213
the d-dimensional component of the best position that particle i has experienced after the k-th iteration,the d-dimension component of the best position the population has experienced after the kth iteration.
A power plant coal consumption benchmark rolling prediction device based on a PSO optimization SVM comprises:
the acquisition module is used for acquiring daily coal consumption data of the power plant as input data;
the preprocessing module is used for carrying out filtering smoothing processing on input data by adopting a nine-point quadratic exponential smoothing method;
the modeling module is used for establishing a rolling updating prediction model based on the SVM and carrying out parameter optimization on the SVM model through a PSO algorithm to obtain a PSO-SVM model;
and the prediction module is used for inputting the input data after the smoothing processing in the preprocessing module into a PSO-SVM model in the modeling module to perform rolling prediction of a coal consumption reference value, and acquiring a prediction result.
Compared with the prior art, the invention has the following advantages:
1. according to the invention, a rolling updating model of the PSO-SVM is built, the latest historical data is combined, and the coal consumption reference value is predicted in a rolling mode, so that data support can be provided for a power grid dispatching department to guess the number of usable coal storage days of each power plant, establish a power plant coal storage early warning mechanism and make a power generation dispatching plan; meanwhile, the model is simple and easy to understand, the prediction precision is high, and the practicability is high.
2. In the prediction process, the data filtering processing is carried out on the extracted daily coal consumption by adopting nine-point quadratic exponential smoothing processing, so that the influence of the mutation point on the instability of the algorithm is effectively reduced, and the prediction precision is improved.
Drawings
FIG. 1 is a flow chart of the PSO-SVM model prediction in the present invention;
FIG. 2 is a diagram illustrating the monthly coal consumption of the A power plant 2015-2018;
FIG. 3 is a diagram illustrating the monthly coal consumption of the B power plant 2015-2018 in an example;
FIG. 4 is a diagram of the monthly coal consumption of the C power plant 2015-2018 in the example;
FIG. 5 is a flow chart of SVM model prediction in the present invention;
FIG. 6 is a flow chart of a PSO optimization SVM model of the present invention;
FIG. 7 is a comparison graph of the daily coal consumption before and after smoothing in the example;
FIG. 8 is a graph of fitness evolution for A, B, C three power plants in an example;
FIG. 9 is a graph of predicted results for plant A in the example;
FIG. 10 is a graph of predicted results for an example B plant;
FIG. 11 is a graph of predicted results for example C plant.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
As shown in fig. 1, the embodiment provides a power grid dispatching power plant coal consumption reference value rolling prediction method based on a PSO optimization SVM, which includes extracting historical daily coal consumption used as a test from a large database in a background of a dispatching center of a power grid, performing secondary exponential smoothing on the historical daily coal consumption, and removing invalid data. Grouping the processed data to meet the initial preparation requirement for establishing the rolling model, wherein the specific grouping mode is that X1-X50 is a first group, X2-X51 is a second group, X3-X52 is a third group, and so on, so that a data set of the rolling model is formed. And (3) carrying out normalization processing on the data set, inputting the data set into a built prediction model of a PSO optimization SVM (PSO-SVM) for model prediction, and establishing a coal consumption reference value prediction model capable of being updated in a rolling manner. The latest daily coal consumption is read from the power grid data background and input into the built prediction model, the latest coal consumption reference value can be obtained, and the latest coal consumption reference value is fed back to the power grid big data scheduling platform to provide a reference basis for scheduling operation.
The method comprises the following specific steps:
step S1, extracting required daily coal consumption data from the big data platform as input data;
step S2, performing filtering smoothing processing on all data by adopting a nine-point quadratic exponential smoothing method;
step S3, establishing a rolling updating prediction model based on SVM;
s4, carrying out parameter optimization on the SVM model by using a PSO algorithm;
and step S5, sending the processed data to a PSO-SVM model to carry out rolling prediction on the coal consumption reference value.
As shown in fig. 2-4, the coal consumption of plant a, plant B and plant C2015-2018 per month is shown. The daily coal consumption change of the unit is influenced by various factors such as the running state of equipment, load change, coal quality and the like, wherein the running state of the equipment is related to the environmental temperature, the scheduling plan of the day and the like, and the regularity is not strong; the historical trend of the load change is strong, and the influence effect of the load change is included when the actual value of the daily coal consumption is considered; the coal quality is related to the actual situation of the coal mine and market trading, and the coal quality mixing situation is complex. The objective factors are difficult to be used as input quantities of mathematical modeling together, so that the historical daily coal consumption value is used as the input quantity, the coal consumption reference value can be accurately predicted, and modeling is convenient to achieve. Because the time sequence of the daily coal consumption is strong in change and large in fluctuation, and the time sequence has obvious difference between a peak period and a stationary period, and a single-input single-output one-time prediction can cause a large error, the embodiment provides the following time sequence updatable rolling model:
considering that the minimum coal storage days set for a power plant by a power grid in practice are 7 days, the normal coal storage days are 18 to 25 days, and generally 18 to 25 days are taken as a time standard for evaluating coal consumption reference, the time standard is set as a parameter n (n belongs to [18,25]), and the model can be adapted to various requirements. In order to realize the coal consumption reference value of n days after the rolling prediction every day, the selection rule is that the daily coal consumption of 50 days before k is taken as the input quantity every time k is assumed as the current day date, and the average value of n days after the k +1 th day is output correspondingly. That is, assuming that the current is No. one, n is 20, the daily coal consumption of 50 days before the No. one is taken as an input set, and the average daily coal consumption of 20 days after the No. one is taken as an output value. The method comprises the steps of comprehensively considering accuracy and realizability, establishing a time sequence model of SVM coal consumption benchmark rolling measurement based on PSO optimization, wherein the establishing process comprises the following steps:
as shown in fig. 5, in the initial setting section, the influence factors of the coal consumption reference value are analyzed, and the daily coal consumption is subjected to the second-order exponential smoothing and validation processing in combination with the actual demand. Grouping the processed data to meet the initial preparation requirement of establishing a rolling model, wherein the specific grouping mode is that X1-X50 is a first group, X2-X51 is a second group, and X3-X52 is a third group, and so on, so that a data set of the rolling model is formed, and finally, the data set is divided into a training set and a testing set. In the regression prediction part, a PSO optimization model is established, a data set is input to carry out initialization constant, penalty factor and kernel function setting, high-dimensional operation in a multilayer perception kernel function and a low-dimensional space is calculated, and an optimization problem error is solved. And then calculating the increment of the optimized variable as error correction, updating the increment of the optimized variable by combining the kernel function, further updating an optimal solution, and judging whether the accuracy requirement of the error is met: if not, the optimal solution is substituted again to calculate the error of the optimization problem; and if so, generating a rolling SVM coal consumption reference value prediction model.
In step S2, after the required daily coal consumption sample data is read from the big data reporting platform of the power grid, smoothing the read data and removing invalid data. The method comprises reducing the mutation points to the average value of the first 5 days and the last 5 days, and performing nine-point second exponential smoothing on the processed data to reduce noise and reduce burrs. And finally, carrying out normalization processing, and dividing the processed data into a training set part and a prediction set part.
The quadratic exponential smoothing formula is:
Figure BDA0002255912280000051
in the formula:the quadratic exponential smoothing values of the t period and the t-1 period respectively, α is a smoothing coefficientAnd
Figure BDA0002255912280000054
under known conditions, the prediction model of the quadratic exponential smoothing method is as follows:
Figure BDA0002255912280000055
Figure BDA0002255912280000056
Figure BDA0002255912280000057
wherein a is 0.9 and at、btThe prediction model parameters are calculated by a smoothing coefficient and quadratic exponential smoothing values of a t period and a t-1 period; t is the number of lapsed periods backwards from T; y ist+TIs the predicted value of T + T phase.
In step S3, the SVM has specific advantages for solving the problems of small number of samples, high variable dimension, non-linear relationship, local optimal solution, and the like, and has better generalization capability than a general learning machine. Firstly, a single-output support vector machine model is adopted to establish a basic coal consumption reference regression prediction model.
The basic idea of SVM estimation regression is to map the data x of the input space into the high-dimensional feature space G by a non-linear mapping and perform linear regression in this space.
Given a sample data value of { xi,yi},i＝1,2,3…，s(xi∈Rn,yiE.g. R). Wherein, yiIs the expected value of u and s is the total number of data points.
SVM solves by introducing a loss functionAnd solving the regression prediction problem. Using functions
Figure BDA0002255912280000067
And taking an extreme value for the optimization target.
Figure BDA0002255912280000061
Figure BDA0002255912280000062
Wherein C is a penalty factor ξiAnd ζiIs a relaxation factor; ε is the loss function. The loss function may represent the decision function with sparse data points. Introducing a loss function with good effect:
Figure BDA0002255912280000063
introducing lagrange multiplier aiAnd biThe convex optimization problem is simplified to a maximized quadratic form.
Figure BDA0002255912280000064
Figure BDA0002255912280000065
Figure BDA0002255912280000066
In the formula: c is a penalty factor and is used for controlling the complexity of the model and compromising the approximation error, and the larger C is, the better the fitting degree is; ε is used to control the regression approximation error and the generalization ability of the model.
When the nonlinear problem is solved by using support vector regression, the mainly adopted procedure is to map nonlinear data variables to a high-dimensional data space through a certain transformation function, and then perform linear regression on the high-dimensional feature space. And the conversion from the low dimension to the high dimension is realized by a kernel function. The kernel function of the SVM has a crucial influence on the output result of the model. Considering the nonlinear relation that the time-sequence influence of the selected sample is large and the input coal consumption and output coal consumption reference values are crossed and complicated, the analysis learning capability is improved by adopting a multilayer perception function, and the kernel function expression is as follows:
K(xi,x)＝tanh(kxi·x+θ)
the final regression prediction function is:
Figure BDA0002255912280000071
as shown in fig. 6, a PSO optimization model is established, the position and the velocity of the particle are initialized, and the coefficients and the penalty factors of the kernel function are updated in combination with the preprocessed daily coal consumption data. And substituting the training set into an SVM model for training, and calculating the prediction error of the SVM coal consumption reference value. And (3) carrying out fitness calculation on the input set by using a PSO algorithm, optimizing the position and the speed of the particles, updating, calculating the fitness of each particle by combining with the prediction error of the SVM, searching for an individual extreme value and a global extreme value, and if the accuracy requirement is met, obtaining the optimal values of the kernel function coefficient and the penalty factor so as to obtain the coal consumption reference value prediction model of the PSO optimization SVM.
In step S4, the SVM algorithm needs to optimize two important parameters, a penalty factor c and a kernel parameter g. The PSO algorithm has the advantages of easiness in operation, high calculation speed, good convergence effect and the like, and the PSO algorithm is introduced to the built SVM model to optimize two parameters of the model, so that the SVM model has better purpose in parameter selection, and the performance of the SVM model is improved. The particle swarm optimization algorithm is a global optimization method based on a population, and the population evolution is realized through the competition and cooperation of particles. Namely, the speed and the position of each particle in the population are initialized, and then iteration and adjustment are carried out on the initial velocity and the position, and the optimal solution is found through iteration.
The specific optimization process is to predict the data samples by using the SVM corresponding to each particle, calculate the current position value and prediction error of each particle, regard the position value and prediction error as the fitness value of the particle, compare the fitness value of the current position of the particle with the fitness value of the particle, determine which is more optimal as the current optimal position of the particle, and continuously update the latest position and speed of the particle. And (3) aiming at the coal consumption reference value model, sending the data processed in the first step into a PSO optimization algorithm, obtaining the punishment coefficient C of the SVM model and the optimal value of the nuclear parameter g, and training the samples of the training set by substituting the optimal values for the SVM regression model.
The velocity and position update formula for the particles is as follows:
Figure BDA0002255912280000073
in the formula, k is iteration times; omega is the inertial weight; d is dimension; c. C1,c2Is a learning factor; r is1,r2∈[0,1]Is a random number;for the d-dimensional component of the velocity vector of the particle i at the kth iteration,
Figure BDA0002255912280000075
for the d-dimensional component of the location vector of the particle i at the k-th iteration,the d-dimensional component of the best position that particle i has experienced after the k-th iteration,
Figure BDA0002255912280000077
the d-dimension component of the best position the population has experienced after the kth iteration.
The embodiment performs experiments on total 2190 daily coal consumption data of A, B, C three power plants from 2017 to 2018. The daily coal consumption data of each power plant in 2017 is selected as a training set, and the daily coal consumption data of each power plant in 2018 is selected as a testing set.
As shown in fig. 1, the selected test set data is input into the trained PSO-SVM regression prediction model in groups to predict the required coal consumption reference value corresponding to each group, and the model can realize a rolling update function, that is, the coal consumption reference value of n days in the future can be obtained in any day.
Fig. 7 is a graph showing a comparison of the effects of the daily coal consumption data of the power plant before and after smoothing. The PSO optimized SVM coal consumption benchmark model main parameter settings in the example are shown in the table 1.
TABLE 1 coal consumption benchmark model Main parameter settings
Figure BDA0002255912280000081
Fig. 8 shows an iterative process of the power plant A, B, C, and fig. 9, 10, and 11 show rolling updated coal consumption benchmark prediction contrast maps. As can be seen from the iteration process diagram, the optimal fitness of each of the three power plants with different capacities is reached before the iteration times are 70, wherein the fitness of the power plant A tends to be stable and optimal when the iteration times are less than 10. The PSO-SVM model can quickly enter a reasonable parameter region for selection, and the model has a small fitness value and the best adaptability. The evaluation index of the model performance can be obtained from mean square error mse and correlation coefficient R2Is represented by R2And the more the correlation between the measured value and the prediction is, the closer the value of the correlation is to the value of 1, the better the coincidence between the predicted value of the sample and the data to be tested is, and the better the fitting effect of the model is. The mse can represent the variation degree between sample data, and the smaller the value of the mse is, the better the prediction accuracy of the prediction model is shown. It can be seen that A, B, C power plants with different sizes can all show extremely high correlation when the PSO is applied to the optimization SVM model, and the mse values are very small, which shows that the prediction performance and the accuracy of the PSO-optimized SVM model are very good.
Common coal consumption reference measuring and calculating methods for a power grid include methods of obtaining an average value of coal consumption of the first 7 days and the first 30 days, obtaining a maximum day coal consumption of 85 percent, and obtaining a coal consumption reference value of the last 20 days by fitting a coal consumption characteristic curve through a least square method. And the PSO optimizes the performance of the SVM model on the coal consumption benchmark measurement for more obvious and direct comparison. The four methods are compared with the PSO optimization SVM model for the error of the peak period and the stationary period in one year, and the relative error comparison results are shown in the following tables 2-4.
TABLE 2 relative error contrast diagram for A power plant with four methods
Figure BDA0002255912280000091
TABLE 3 relative error comparison chart of four methods for B power plant
Figure BDA0002255912280000092
TABLE 4 relative error contrast diagram for C power plant with four methods
Figure BDA0002255912280000093
The PSO optimization SVM model has great advantages compared with the traditional coal consumption benchmark measuring and calculating method after training and learning of training sample data, the data fluctuation is large in the peak period, the maximum relative error of the traditional coal consumption benchmark measuring and calculating method comprises a least square method commonly used in engineering reaches more than 60%, and the change is large and is difficult to stabilize although the relative error is small in the stationary period. The coal consumption reference value with very small relative error can be predicted no matter in the peak period or the stationary period by the PSO optimization SVM model, and the high accuracy of the PSO optimization SVM model is fully explained.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.

Claims (8)

1. A power plant coal consumption reference value rolling prediction method based on a PSO (particle swarm optimization) SVM (support vector machine) is characterized by comprising the following steps:
s1, acquiring daily coal consumption data of the power plant as input data;
s2, performing filtering smoothing processing on input data by adopting a nine-point quadratic exponential smoothing method;
s3, establishing a rolling updating prediction model based on the SVM;
s4, carrying out parameter optimization on the SVM model through a PSO algorithm to obtain a PSO-SVM model;
and S5, inputting the input data smoothed in the step S2 into the PSO-SVM model in the step S4 to perform rolling prediction of the coal consumption reference value, and acquiring a prediction result.
2. The power plant coal consumption benchmark rolling prediction method based on the PSO optimization SVM as claimed in claim 1, wherein in the step S2, before the filtering smoothing process, the value of the catastrophe point in the input data is modified to be an average value of 5 days before and 5 days after the catastrophe point.
3. The power plant coal consumption reference value rolling prediction method based on the PSO optimization SVM as claimed in claim 1, wherein in the step S2, the quadratic exponential smoothing formula is as follows:
Figure FDA0002255912270000011
in the formula (I), the compound is shown in the specification,
Figure FDA0002255912270000012
and
Figure FDA0002255912270000013
respectively the secondary exponential smoothing values of the t period and the t-1 period; a is a smoothing coefficient, in
Figure FDA0002255912270000014
And
Figure FDA0002255912270000015
under known conditions, the prediction model of the quadratic exponential smoothing method is as follows:
Figure FDA0002255912270000016
Figure FDA0002255912270000017
Figure FDA0002255912270000018
in the formula, at、btThe prediction model parameters are calculated by a smoothing coefficient and quadratic exponential smoothing values of a t period and a t-1 period; t is the number of lapsed periods backwards from T; y ist+TIs the predicted value of T + T phase.
4. The power plant coal consumption reference value rolling prediction method based on the PSO optimization SVM according to claim 1, characterized in that in the rolling updating prediction model, n days are taken as a time standard for evaluating a coal consumption reference, wherein n is more than or equal to 18 and less than or equal to 25; and taking k days as the current day, taking input data as the daily coal consumption of k-49 days to k days, and taking corresponding output data as the average value of n days after the k +1 th day.
5. The power plant coal consumption reference value rolling prediction method based on the PSO optimization SVM according to claim 1, characterized in that in step S4, a PSO algorithm is introduced to a built SVM model to optimize a penalty factor c and a kernel parameter g, and a PSO-SVM model is obtained through optimization.
6. The power plant coal consumption reference value rolling prediction method based on the PSO optimization SVM according to claim 5, characterized in that a specific optimization process is to respectively predict data samples by using the SVM corresponding to each particle, calculate the current position value and prediction error of each particle, and use the current position value and prediction error as the fitness value of the particle, compare the fitness value of the current position of the particle with the fitness value of the particle, the optimal value is the current optimal position of the particle, and continuously update the latest position and speed of the particle.
7. The power plant coal consumption benchmark rolling prediction method based on the PSO optimization SVM is characterized in that the speed and position updating formula of the particles is as follows:
Figure FDA0002255912270000022
in the formula, k is iteration times; omega is the inertial weight; d is dimension; c. C1,c2Is a learning factor; r is1,r2∈[0,1]Is a random number;for the d-dimensional component of the velocity vector of the particle i at the kth iteration,
Figure FDA0002255912270000024
for the d-dimensional component of the location vector of the particle i at the k-th iteration,
Figure FDA0002255912270000025
the d-dimensional component of the best position that particle i has experienced after the k-th iteration,
Figure FDA0002255912270000026
the d-dimension component of the best position the population has experienced after the kth iteration.
8. A power plant coal consumption benchmark rolling prediction device based on a PSO optimization SVM is characterized by comprising the following components:
the acquisition module is used for acquiring daily coal consumption data of the power plant as input data;
the preprocessing module is used for carrying out filtering smoothing processing on input data by adopting a nine-point quadratic exponential smoothing method;
the modeling module is used for establishing a rolling updating prediction model based on the SVM and carrying out parameter optimization on the SVM model through a PSO algorithm to obtain a PSO-SVM model;
and the prediction module is used for inputting the input data after the smoothing processing in the preprocessing module into a PSO-SVM model in the modeling module to perform rolling prediction of a coal consumption reference value, and acquiring a prediction result.
CN201911053400.8A 2019-10-31 2019-10-31 Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine) Pending CN110826794A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911053400.8A CN110826794A (en) 2019-10-31 2019-10-31 Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911053400.8A CN110826794A (en) 2019-10-31 2019-10-31 Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine)

Publications (1)

Publication Number Publication Date
CN110826794A true CN110826794A (en) 2020-02-21

Family

ID=69551782

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911053400.8A Pending CN110826794A (en) 2019-10-31 2019-10-31 Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine)

Country Status (1)

Country Link
CN (1) CN110826794A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111612232A (en) * 2020-05-13 2020-09-01 云南电网有限责任公司电力科学研究院 Power distribution network line re-jump probability prediction optimization method and device based on gradient descent
CN111738585A (en) * 2020-06-16 2020-10-02 国网重庆市电力公司 Thermal power plant coal storage quantity grading early warning method for power grid dispatching
CN116307075A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Intelligent algorithm-based electricity-measuring coal-burning cost optimization method and system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106503833A (en) * 2016-09-30 2017-03-15 天津大学 Photovoltaic generation short term power rolling forecast method based on algorithm of support vector machine
CN107908638A (en) * 2017-09-26 2018-04-13 国网甘肃省电力公司 The operation of power networks efficiency rating method and system matched somebody with somebody are excavated based on big data
CN109003145A (en) * 2018-08-17 2018-12-14 华北电力大学 A kind of power coal price prediction technique
US20190171187A1 (en) * 2016-05-09 2019-06-06 StrongForce IoT Portfolio 2016, LLC Methods and systems for the industrial internet of things
CN110119747A (en) * 2019-05-14 2019-08-13 安徽理工大学 A kind of coal rock detection method based on radioscopic image
CN110390436A (en) * 2019-07-25 2019-10-29 上海电力大学 Short-term prediction method for coal storage amount of power plant based on SSA and LSTM deep learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190171187A1 (en) * 2016-05-09 2019-06-06 StrongForce IoT Portfolio 2016, LLC Methods and systems for the industrial internet of things
CN106503833A (en) * 2016-09-30 2017-03-15 天津大学 Photovoltaic generation short term power rolling forecast method based on algorithm of support vector machine
CN107908638A (en) * 2017-09-26 2018-04-13 国网甘肃省电力公司 The operation of power networks efficiency rating method and system matched somebody with somebody are excavated based on big data
CN109003145A (en) * 2018-08-17 2018-12-14 华北电力大学 A kind of power coal price prediction technique
CN110119747A (en) * 2019-05-14 2019-08-13 安徽理工大学 A kind of coal rock detection method based on radioscopic image
CN110390436A (en) * 2019-07-25 2019-10-29 上海电力大学 Short-term prediction method for coal storage amount of power plant based on SSA and LSTM deep learning

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111612232A (en) * 2020-05-13 2020-09-01 云南电网有限责任公司电力科学研究院 Power distribution network line re-jump probability prediction optimization method and device based on gradient descent
CN111612232B (en) * 2020-05-13 2023-08-18 云南电网有限责任公司电力科学研究院 Power distribution network line re-jump probability prediction optimization method and device based on gradient descent
CN111738585A (en) * 2020-06-16 2020-10-02 国网重庆市电力公司 Thermal power plant coal storage quantity grading early warning method for power grid dispatching
CN116307075A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Intelligent algorithm-based electricity-measuring coal-burning cost optimization method and system
CN116307075B (en) * 2023-02-01 2024-04-05 华能国际电力股份有限公司上海石洞口第二电厂 Intelligent algorithm-based electricity-measuring coal-burning cost optimization method and system

Similar Documents

Publication Publication Date Title
CN111783953B (en) 24-point power load value 7-day prediction method based on optimized LSTM network
Bracale et al. A probabilistic competitive ensemble method for short-term photovoltaic power forecasting
CN109558975B (en) Integration method for multiple prediction results of power load probability density
CN110826794A (en) Power plant coal consumption reference value rolling prediction method and device based on PSO (particle swarm optimization) SVM (support vector machine)
CN101893884B (en) Soft measurement method of quality index data in rubber mixing process of internal mixer
CN112182720B (en) Building energy consumption model evaluation method based on building energy management application scene
CN106251027B (en) Electric load probability density Forecasting Methodology based on fuzzy support vector quantile estimate
CN105023066A (en) Business expansion analytical prediction system and method based on seasonal adjustment
CN116150897A (en) Machine tool spindle performance evaluation method and system based on digital twin
CN111967183A (en) Method and system for calculating line loss of distribution network area
CN104915736A (en) Method for improving accuracy of wind power combined prediction based on improved entropy weight method
CN110247396A (en) State Estimation for Distribution Network and system based on adaptive robust Unscented kalman filtering
CN114372360A (en) Method, terminal and storage medium for power load prediction
Yu et al. Load forecasting based on smart meter data and gradient boosting decision tree
CN107706938B (en) A kind of wind power waving interval analysis method returned based on quantile
CN116070769A (en) Ultra-short-term wind power plant power multi-step interval prediction modularization method and device thereof
CN115764870A (en) Multivariable photovoltaic power generation power prediction method and device based on automatic machine learning
CN114583767B (en) Data-driven wind power plant frequency modulation response characteristic modeling method and system
CN115545333A (en) Method for predicting load curve of multi-load daily-type power distribution network
CN110533247B (en) Monthly power consumption prediction method adopting temperature data anomaly point compensation
CN113991711B (en) Capacity configuration method for energy storage system of photovoltaic power station
CN117851908B (en) Improved on-line low-voltage transformer area electric energy meter misalignment monitoring method and device
CN113887833A (en) Distributed energy user side time-by-time load prediction method and system
CN117930776A (en) Accurate steam supply pre-regulation method and system for white spirit production based on model prediction
CN117876150A (en) Short-term multiple comprehensive load prediction method based on non-invasive load decomposition algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20200221