CN111009893B - Household user short-term load prediction method based on load decomposition technology - Google Patents

Household user short-term load prediction method based on load decomposition technology Download PDF

Info

Publication number
CN111009893B
CN111009893B CN201911090119.1A CN201911090119A CN111009893B CN 111009893 B CN111009893 B CN 111009893B CN 201911090119 A CN201911090119 A CN 201911090119A CN 111009893 B CN111009893 B CN 111009893B
Authority
CN
China
Prior art keywords
electric equipment
load
electric
formula
decomposition
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911090119.1A
Other languages
Chinese (zh)
Other versions
CN111009893A (en
Inventor
韦杏秋
蒋雯倩
谢雄威
杨舟
李刚
李金瑾
梁捷
卿柏元
唐利涛
林秀清
陈珏羽
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.)
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Original Assignee
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of Guangxi Power Grid Co Ltd filed Critical Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority to CN201911090119.1A priority Critical patent/CN111009893B/en
Publication of CN111009893A publication Critical patent/CN111009893A/en
Application granted granted Critical
Publication of CN111009893B publication Critical patent/CN111009893B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/12—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load
    • H02J3/14—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load by switching loads on to, or off from, network, e.g. progressively balanced loading
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B70/00—Technologies for an efficient end-user side electric power management and consumption
    • Y02B70/30—Systems integrating technologies related to power network operation and communication or information technologies for improving the carbon footprint of the management of residential or tertiary loads, i.e. smart grids as climate change mitigation technology in the buildings sector, including also the last stages of power distribution and the control, monitoring or operating management systems at local level
    • Y02B70/3225—Demand response systems, e.g. load shedding, peak shaving
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
    • Y04S20/20—End-user application control systems
    • Y04S20/222—Demand response systems, e.g. load shedding, peak shaving

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention discloses a method for predicting short-term load of a home user based on a load decomposition technology, which relates to the technical field of decomposition and detection of electric power data, and is characterized in that electric equipment of the home user is classified by a non-intrusive electric power load monitoring technology, load electric equipment is selected, the electric equipment is decomposed, and the running state of the electric equipment is obtained; and then converting the running state of the electric equipment into data information, inputting the data information converted from the running state of the electric equipment and the original load time sequence into an immune neuron for training to obtain an electric quantity value used by the electric equipment at a certain future moment, thereby predicting the electric quantity used by the household equipment. The invention can ensure that the prediction error is within an acceptable range, and the improvement of the prediction performance is more feasible and cost-effective.

Description

Household user short-term load prediction method based on load decomposition technology
Technical Field
The invention relates to the technical field of decomposition and detection of power data, in particular to a short-term load prediction method for a household user based on a load decomposition technology.
Background
The non-invasive power load monitoring technology is a brand-new power load electricity consumption information acquisition and analysis technology. According to the technology, the power utilization state information and the power utilization law of each or every type of electric equipment in the total load can be known through measuring and real-time analyzing the single-point total quantity of the electric information of the electric load, a sensor with a digital communication function is not required to be arranged for each interested electric equipment like the traditional monitoring technology, and the real-time power consumption proportion of different electric equipment in the load can be obtained only through measuring and analyzing the voltage, the current and the power information at the inlet of the electric load, so that the electric load decomposition is realized.
Load prediction is a traditional research topic of an electric power system, and is to explore the internal relation and development rule of the electric power system through analysis and research on historical data and predict a future electric power load value. The goal of the short-term load prediction of the household users based on the non-invasive power load monitoring technology is to fully master the electric energy utilization condition of the household users and realize the balance of supply and demand, so the accurate load prediction has important significance for improving the coordination of the power grid, reducing the cost and maintaining the safety and stability of the power grid. Currently, the commonly used load prediction methods can be divided into two categories: a conventional prediction method and an intelligent prediction method. The conventional prediction methods include a regression analysis method, an autoregressive moving average model, a gray model, and the like. The intelligent prediction method comprises an artificial neural network, a support vector machine, a particle swarm optimization algorithm, fuzzy logic and the like.
However, in the existing methods, few people combine the non-intrusive power load monitoring technology with the load prediction technology to improve the performance of the load prediction.
Disclosure of Invention
The invention aims to provide a method for predicting the short-term load of a household user based on a load decomposition technology, thereby overcoming the defect that the accuracy of the conventional power load prediction is not high enough.
In order to achieve the purpose, the invention provides a method for predicting the short-term load of a home user based on a load decomposition technology, which comprises the following steps:
s11, classifying the electric equipment of the household user, and selecting load electric equipment;
s12, constructing a load current equation set according to the formula (1) and the formula (2), solving to obtain the electric energy consumption proportion of the electric equipment, and decomposing the electric equipment;
ia(t)＝Ia1·cos(wt+θa1)+...Iak·cos(kwt+θak)+... (1)
Iak＝αak·Ia1 (2)
in the formulae (1) and (2), ia(t) is the operating current of a device at time t; i isa1Is the magnitude of the fundamental component of the operating current; w is the angular frequency of the fundamental component of the operating current; thetaa1Is the initial phase angle of the fundamental component of the working current; k is a positive integer; i isakIs the amplitude of the kth harmonic component in the operating current; kw is the angular frequency of the kth harmonic component in the operating current; thetaakIs the initial phase angle of the kth harmonic component in the working current; alpha is alphaakIs a representation IakAnd Ia1The coefficient of proportionality therebetween;
s13, judging the running state of the electric equipment according to the decomposition results of the electric equipment at two adjacent moments;
s14, forming an operation rule according to the electric equipment operation state obtained by the judgment of the S13, performing data representation to obtain an equipment operation state time sequence, and inputting the time sequence and the original load time sequence into an immune neuron;
s15, training the immune neurons by using the equipment running state time sequence and the original load time sequence, and predicting the power consumption of the household equipment;
and S16, outputting a prediction result through the S15.
Further, in S11, the electric devices of the home users are classified into air conditioners, resistors, fluorescent lamps, computers, and motors.
Further, in S13, the operating status of the electric device includes: start, stop and run.
Further, in S14, 1, 0, and 2 are used to respectively indicate the start, stop, and operation of the electric device.
Further, in S12, a load current equation set is constructed and solved according to the formula (1) and the formula (2), so as to obtain a power consumption ratio of the electric device, and decomposing the electric device includes the following steps:
according to the formula (1), when n types of main electric equipment are contained in the power load, the linear superposition of the currents of the main electric equipment for the total current is used for obtaining a formula (3), and a unit current parameter matrix and a load current vector are obtained;
il'(t)＝β1·i'a1(t)+β2·i'a2(t)...+βn·i'an(t) (3)
in the formula (3), i'l(t) is the total cell current of the electrical load, the corresponding eigenvector is FPl；i’a1(t)、i’a2(t)、…、i’an(t) are unit currents of 1 st, 2 nd, … th and n kinds of electric equipment respectively, and corresponding characteristic vectors are FP respectivelya1，FPa2And FPn；β1、β2、…、βnE [0, + ∞) ] is the current weight coefficient of the 1 st, 2 nd, … th and n kinds of electric equipment respectively, and beta is [ beta ]1，β2，…，βn]T；
And (3) processing the equation (3) according to a phasor method to obtain a non-invasive power load monitoring estimation equation set:
FPl＝Ha·β (4)
in the formula (4), FPlIs i'l(t) corresponding feature vectors; hαThe unit current parameter matrix of the electric equipment is obtained;
to the said NOTOptimal solution is carried out on an invasive power load monitoring estimation equation set to obtain beta1，β2，…，βn；
Current weight coefficient beta based on kth electric equipmentkCalculating the proportional coefficient of the active power and the reactive power of the fundamental wave of each main electric device according to the formula (5) and the formula (6):
Figure GDA0003548046820000031
Figure GDA0003548046820000032
in the formulae (5) and (6), betapkThe ratio of the active power of the fundamental wave of the kth electric equipment is obtained; beta is aQkThe fundamental wave reactive power proportion of the kth electric equipment is shown;
and obtaining the power consumption proportion of the electric equipment according to the formula (5) and the formula (6).
Further, the step S13 of determining the operating state of the electric device includes:
s1301, comparing the decomposition results of the electric equipment at two adjacent moments, wherein the decomposition result at the previous moment has no solution of the electric equipment, and the decomposition result at the later moment has the solution of the electric equipment, so that the running state of the electric equipment is starting and is used for detecting the complete starting process of the electric equipment;
s1302, comparing the decomposition results of the electric equipment at two adjacent moments, wherein the decomposition results at the two moments before and after have the solution of the electric equipment, and the running state of the electric equipment is running and is used for tracking and detecting the running process of the electric equipment;
s1303, comparing the decomposition results of the electric equipment at two adjacent moments, wherein the decomposition result at the former moment has the solution of the electric equipment, and the decomposition result at the latter moment has no solution of the electric equipment, so that the running state of the electric equipment is stopped and the process of stopping the electric equipment is detected.
Further, in S15, the training using the original load time sequence and the device operating state time sequence of the immune neuron S14 includes the following steps:
s1501, for m training samples, d (i) is an expected output corresponding to input x (i);
s1502, defining an error function E of the m training samples by adopting a batch updating method:
Figure GDA0003548046820000041
in equation (7), e (i) is the training error of a single sample:
Figure GDA0003548046820000042
in equation (8), k is the kth time of a time series of a certain sample, yk(i) The actual output value corresponding to the kth moment of the ith sample;
s1503, substituting formula (8) into formula (7) to obtain:
Figure GDA0003548046820000043
s1504, the weights and the bias of the neural network are updated according to the formula (10) and the formula (11) by the iteration of the immune neurons;
Figure GDA0003548046820000044
Figure GDA0003548046820000045
in the formulae (10) and (11),
Figure GDA0003548046820000046
is the jth of the l-1 th layerThe connection weight of the neuron and the ith neuron of the l-th layer,
Figure GDA0003548046820000047
the bias of the ith neuron of the l layer is shown, alpha is the learning rate, and the value range is (0, 1);
s1505, the partial derivatives of the weight and bias of the l-th layer are:
Figure GDA0003548046820000048
Figure GDA0003548046820000049
in the formulas (12) and (13), L is more than or equal to 2 and less than or equal to L-1;
Figure GDA0003548046820000051
Figure GDA0003548046820000052
further, in S1504, updating the weights and biases of the neural network includes the following steps:
s15041, setting DeltaW for all layers(l)＝0，△b(l)＝0，△W(l)And Δ b(l)Respectively an all-zero matrix and an all-zero vector;
s15042, setting i to be 1: m, and calculating the weight and biased gradient matrix of each layer of neurons by using a back propagation algorithm
Figure GDA0003548046820000053
And
Figure GDA0003548046820000054
then calculate
Figure GDA0003548046820000055
Inverse gradient weight matrix of(l)And
Figure GDA0003548046820000056
reverse gradient bias matrix Δ b(l)；
S15043, updating weight and bias, calculating
Figure GDA0003548046820000057
And
Figure GDA0003548046820000058
W(l)is the updated weight value, b(l)Is an updated bias.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a household user short-term load forecasting method based on a load decomposition technology, which comprises the steps of classifying household user electric equipment through a non-entrance type power load monitoring technology, selecting load electric equipment, decomposing the electric equipment and obtaining the running state of the electric equipment; and then converting the running state of the electric equipment into data information, inputting the data information converted from the running state of the electric equipment and the original load time sequence into an input immune neuron for training so as to obtain an electric quantity value used by the electric equipment at a certain time in the future, and predicting the electric quantity used by the household equipment. The invention can ensure that the prediction error is within an acceptable range, provides a basis for the work of dispatching and the like of the power department, and enables the improvement of the prediction performance to have more feasibility and cost benefit.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a flow chart of a method for predicting short-term load of a home subscriber based on load decomposition technology according to the present invention;
FIG. 2 is a comparison graph of predicted results provided by an embodiment of the present invention.
Detailed Description
The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the method for predicting the short-term load of the home user based on the load decomposition technology, provided by the invention, obtains the operation rule of the device based on the non-intrusive power load monitoring technology, and subjects the operation rule to datamation and input into a short-term load prediction model for prediction, and includes the following steps:
and S11, according to the working principle of the power load, dividing the electric equipment of the household user into air conditioners, resistors, fluorescent lamps, computers and motors, and selecting load electric equipment and the like. In order to make the study systematic and simple, this embodiment uses these types of loads as study targets.
In particular, the air conditioner refers to air conditioning equipment of various brands such as a vertical hanging type air conditioner, and the equipment has a compressor inside, so that the starting transient characteristic and the steady-state characteristic are obvious; the resistors refer to various electric devices which mainly rely on resistance heating to realize functions, such as: the device comprises an electric water heater, an electric heater, an incandescent lamp and the like, and the device has ideal resistance characteristics to the outside, mainly represented by that the power factor is close to 1, and the current harmonic distortion is very small; fluorescent lamps refer to the existing common lighting equipment and the like, and the equipment not only has obvious starting transient characteristics, but also has obvious harmonic distortion characteristics; computers comprise electric equipment with complex characteristics such as desktop computers, notebook computers, televisions and the like, and the electric equipment works by depending on large-scale integrated circuits, so that the electric equipment has richer harmonic characteristics than fluorescent lamp equipment; the electric machines are devices which mainly depend on the rotation of the electric machine to realize functions, such as a refrigerator, a fan and the like, wherein the electric machines are not obvious as the air-conditioning equipment, but the electric machines can absorb a large amount of reactive power when running, so that the electric machines have obvious steady-state reactive power characteristics.
S12, according to formula ia(t)＝Ia1·cos(wt+θa1)+...Iak·cos(kwt+θak) +.. and Iak＝αak·Ia1And constructing a load current equation set and solving to obtain the electric energy consumption proportion of the equipment, and decomposing the electric equipment.
S12 includes the steps of:
s1201, regarding a certain electric device, taking a steady-state current harmonic wave of normal operation of the electric device as an imprinting characteristic, and expressing the characteristic as follows:
ia(t)＝Ia1·cos(wt+θa1)+...Iak·cos(kwt+θak)+... (1)
Iak＝αak·Ia1 (2)
in the formulae (1) and (2), ia(t) is the operating current of a device at time t; i isa1Is the amplitude of the fundamental component of the operating current; w is the angular frequency of the fundamental component of the operating current; thetaa1Is the initial phase angle of the fundamental component of the working current; k is a positive integer; i isakRepresenting the magnitude of the kth harmonic component in the operating current; kw denotes the angular frequency of the kth harmonic component in the operating current; theta.theta.akRepresenting an initial phase angle of a kth harmonic component in the working current; alpha is alphaakIs a representation IakAnd Ia1The coefficient of proportionality therebetween.
After the per unit value is adopted, the formula (1) is substituted as follows:
i'a(t)＝1·cos(wt+θa1)+...+αakcos(kwt+θak)+... (16)
of formula (16), i'aAnd (t) is the unit current of the electric equipment, namely the current when the per unit value of the amplitude of the fundamental wave of the working current is 1 is called the unit current of the electric equipment.
According to the phasor description, the extracted feature vector FPa representing each type of electric equipment is as follows:
FPa＝[1·∠θa1,...,αak∠θak,...]T (17)。
when n types of primary consumers are contained inside the power load (the load may also contain other types of consumers, but the proportion in the load current is very small and negligible), the total current inside the load is obtained by linearly adding the n types of primary consumer currents:
il'(t)＝β1·i'a1(t)+β2·i'a2(t)...+βn·i'an(t) (3)
in the formula (3), i'l(t) represents the total cell current of the electrical load, the corresponding feature vector being FPl；i’a1(t)、i’a2(t)、…、i’an(t) represents the unit currents of the 1 st, 2 nd, … th and n-th electric devices, and the corresponding characteristic vectors are FPa1，FPa2And FPn；β1、β2、…、βnE [0, + ∞) ] represents the current weighting factors of the 1 st, 2 nd, … th and n-th electric devices, respectively, and β [ β ═ β1，β2，…，βn]TAnd obtaining the negative charge current vector.
And S1202, processing the formula (5) into the following form according to the description of a phasor method:
FPl＝[FPa1,FPa2,...,FPan]·β (18)
in the formula (18), n is the number of types of main electric devices in the load.
The matrix form is used as a non-intrusive power load monitoring estimation equation set and is simplified as follows:
FPl＝Ha·β (4)
in the formula (4), HαIs a unit current parameter matrix of the electric equipment.
S1203, estimating equation system with constraint (beta) due to non-intrusive power load monitoring1、β2、…、βnE [0, + ∞)) and therefore a general optimization algorithm is selected to solve the above formula (4), such as a particle swarm algorithm, a genetic algorithm, a differential evolution algorithm and the like; and obtaining current weight coefficients of various main electric equipment in the power load through optimization solution.
S1204, obtaining an optimal solution beta in the solution1，β2，…,βk，βnThen, the proportional coefficient of the active power and the reactive power of the fundamental wave is calculated by the following equations (5) and (6):
that is, the power load is decomposed according to the main electric equipment.
Figure GDA0003548046820000081
Figure GDA0003548046820000082
In the formulae (5) and (6), betapkThe ratio of the active power of the fundamental wave of the kth electric equipment is obtained; beta is aQkThe fundamental wave reactive power proportion of the kth electric equipment is shown.
Furthermore, the fundamental wave active power P of the kth electric equipmentkAnd fundamental reactive power QkThe calculations of equations (19) and (20) allow the power of the consumer to be estimated by integrating power and time as in equation (21).
Pk＝βPk·PΣ (19)
Qk＝βQk·PΣ (20)
In formulae (19) and (20), PΣAs the total active power sum Q of fundamental wavesΣThe fundamental wave active total power and the fundamental wave reactive total power.
Figure GDA0003548046820000083
In the formula (21), Ek(1: T) representsThe total amount of power consumed by the k types of consumers at a time interval Δ T from the 1 st sampling instant (e.g., start-up instant) to the T-th sampling instant (e.g., shutdown instant) is (T-1) Δ T.
S1204, estimating the electric quantity of the electric equipment according to the integral of the power and the time of the formula (21), and obtaining the electric energy consumption proportion of the electric equipment.
S13, judging the operation state of the electric equipment according to the decomposition results of the electric equipment at two adjacent moments, wherein the operation state of the electric equipment comprises the following steps: starting, stopping and running; the judgment of the running state of the electric equipment comprises the following steps:
and S1301, judging whether the electric equipment is in a starting stage.
If a certain electric device(s) in the result of the previous moment is not operated in the two adjacent total load decomposition results, namely, the result of the load decomposition does not have a solution about the electric device(s), and the result of the later moment is in an operation state, namely, the result of the load decomposition has the solution of the electric device(s), judging that the starting process of the electric device(s) occurs, and determining the previous moment as the starting moment of the electric device(s); from this rule, can detect the complete start-up process of consumer.
And S1302, judging that the electric equipment is in the operation stage.
If a certain electric device (class) in the results of two adjacent total load decomposition results at the previous and next moments is always in an operating state, namely the solutions of the electric device (classes) exist in the load decomposition results all the time, the electric device (classes) is in an operating stage, at this time, if the operating power of the electric device (classes) is obviously changed, a corresponding decomposition result exists in the load decomposition results, and the power change of the electric device (classes) can be clearly displayed in the results; according to the rule, the operation process of the detection electric equipment can be tracked.
And S1303, judging that the power utilization equipment is in a shutdown stage.
If a certain electric device(s) in the results of the previous time is in the running state in the two adjacent total load decomposition results, that is, the value of the electric device(s) exists in the load decomposition results, and if the electric device(s) in the results of the latter time is in the shutdown state, that is, the value of the electric device(s) does not exist in the results of the load decomposition, it is determined that the shutdown process of the electric device(s) occurs, and the latter time is determined as the shutdown time of the electric device(s). From this rule, the process of the shutdown of the electric device can be detected.
The obtained starting and stopping time is about the estimation of the real starting or stopping time of the electric equipment, the accuracy depends on the sampling precision of the whole system, and the flexible load electric equipment can be flexibly adjusted according to the specific requirements of a real scene.
And S14, forming an operation rule according to the operation state of the electric equipment obtained by the judgment of S13, performing data representation, namely, adopting '1' to represent the starting of the electric equipment, '0' to represent the starting of the electric equipment, and '2' to represent the operation of the electric equipment, obtaining an equipment operation state time sequence, and inputting the equipment operation state time sequence into the immune neuron together with the original load time sequence.
S15, training the immune neurons by using the equipment running state time sequence and the original load time sequence, and predicting the power consumption of the household equipment;
the original load time sequence of the household electrical equipment has volatility and regularity, the immune neurons are used for learning and predicting to obtain prediction data of the power load value at the historical time, and meanwhile, the prediction precision is guaranteed;
error analysis can be carried out according to the result obtained by training and the expected result every time, and then the weight and the threshold are modified, so that a model which can be output and is consistent with the expected result is obtained in one step.
Therefore, decomposing the original load time series and the equipment running state time series comprises the following steps:
s1501, for m training samples, where d (i) is the expected output corresponding to input x (i). By optimizing the input weight and bias of each layer of neurons, the output of the neural network is close to the expected output as much as possible, so as to achieve the aim of training (or learning).
S1502, defining an error function E for the m training samples by adopting a batch updating method:
Figure GDA0003548046820000101
in equation (7), e (i) is the training error of a single sample:
Figure GDA0003548046820000102
in equation (8), k is the kth time of a time series of a certain sample, yk(i) The actual output value corresponding to the kth moment of the ith sample;
s1503, substituting formula (8) into formula (7) to obtain:
Figure GDA0003548046820000103
s1504, the weights and the bias of the neural network are updated according to the formula (10) and the formula (11) by the iteration of the immune neurons;
Figure GDA0003548046820000111
Figure GDA0003548046820000112
in the formulae (10) and (11),
Figure GDA0003548046820000113
the connection weight of the jth neuron of the l-1 layer and the ith neuron of the l layer is taken as the connection weight,
Figure GDA0003548046820000114
the bias of the ith neuron of the l layer is shown, alpha is the learning rate, and the value range is (0, 1);
s1505, the partial derivatives of the weight and bias of the l-th layer are:
Figure GDA0003548046820000115
Figure GDA0003548046820000116
in the formulas (12) and (13), L is more than or equal to 2 and less than or equal to L-1;
Figure GDA0003548046820000117
Figure GDA0003548046820000118
in S1504, updating the weights and biases of the neural network by using a batch updating method includes the following steps:
s15041, for all layers (L is not less than 2 and not more than L), setting Delta W(l)＝0，△b (l)0, where Δ W(l)Is an all-zero matrix, Δ b(l)An all-zero vector;
s15042, for i ═ 1: m, calculating gradient matrix of neuron weight values of each layer by using back propagation algorithm
Figure GDA0003548046820000119
And a biased gradient matrix
Figure GDA00035480468200001110
Then calculate
Figure GDA00035480468200001111
Inverse gradient weight matrix of(l)And
Figure GDA00035480468200001112
reverse gradient bias matrix Δ b(l)；
S15043, update rightValue and offset, calculating
Figure GDA00035480468200001113
And
Figure GDA00035480468200001114
W(l)is the updated weight value, b(l)Is an updated bias.
And S15043, repeating S15041-S15043 to train the BP neural network prediction model. And then, inputting test sample data into the trained BP prediction model to acquire an electric quantity value used by the electric power equipment at a future moment.
S16: and outputting a prediction result.
In order to verify the feasibility and the effectiveness of the method, a comparative test is set for the method for predicting the short-term load of the home user based on the load decomposition technology, so that the method can be better understood by the technical personnel in the field:
the BP neural network is compared with the household user short-term load prediction method based on the load decomposition technology in an experiment, and the prediction precision is evaluated by adopting MAE (mean absolute error) and RMSE (mean square root error).
MAE is expressed as:
Figure GDA0003548046820000121
RMSE is expressed as:
Figure GDA0003548046820000122
in equations (22) and (23), N is the number of prediction data, y (t) is the prediction value, and ya(t) is the desired value of time t.
The experimental results obtained by adopting the two methods are shown in fig. 2, and it can be intuitively seen from fig. 2 that the errors of the MAE and the RMSE of the invention are smaller than the errors of prediction by directly adopting the BP neural network, so that the invention has obvious prediction advantages and smaller prediction errors.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.

Claims (7)

1. A method for predicting short-term load of a home user based on a load decomposition technology is characterized by comprising the following steps: the method comprises the following steps:
s11, classifying the electric equipment of the household user, and selecting load electric equipment;
s12, constructing a load current equation set according to the formula (1) and the formula (2), solving to obtain the electric energy consumption proportion of the electric equipment, and decomposing the electric equipment;
ia(t)＝Ia1·cos(wt+θa1)+...Iak·cos(kwt+θak)+... (1)
Iak＝αak·Ia1 (2)
in the formulae (1) and (2), ia(t) is the operating current of a device at time t; i isa1Is the magnitude of the fundamental component of the operating current; w is the angular frequency of the fundamental component of the operating current; thetaa1Is the initial phase angle of the fundamental component of the working current; k is a positive integer; I.C. AakIs the amplitude of the kth harmonic component in the operating current; kw is the angular frequency of the kth harmonic component in the operating current; thetaakIs the initial phase angle of the kth harmonic component in the working current; alpha is alphaakIs a representation IakAnd Ia1The coefficient of proportionality therebetween;
s13, judging the running state of the electric equipment according to the decomposition results of the electric equipment at two adjacent moments;
s14, forming an operation rule according to the operation state of the electric equipment obtained by the judgment of the S13, performing data representation to obtain an equipment operation state time sequence, and inputting the time sequence and the original load time sequence into an immune neuron;
s15, training the immune neurons by using the equipment running state time sequence and the original load time sequence, and predicting the power consumption of the household equipment;
in the S15, the training using the original load time series and the device operating state time series of the immune neuron S14 comprises the following steps:
s1501, for m training samples, d (i) is an expected output value corresponding to input x (i);
s1502, defining an error function E of the m training samples by adopting a batch updating method:
Figure FDA0003548046810000011
in equation (7), e (i) is the training error of a single sample:
Figure FDA0003548046810000012
in equation (8), k is the kth time of a time series of a certain sample, yk(i) The actual output value corresponding to the kth moment of the ith sample;
s1503, substituting formula (8) into formula (7) to obtain:
Figure FDA0003548046810000021
s1504, the weights and the bias of the neural network are updated according to the formula (10) and the formula (11) by the iteration of the immune neurons;
Figure FDA0003548046810000022
Figure FDA0003548046810000023
in the formulae (10) and (11),
Figure FDA0003548046810000024
the connection weight of the jth neuron of the l-1 layer and the ith neuron of the l layer is taken as the connection weight,
Figure FDA0003548046810000025
the bias of the ith neuron of the l layer is shown, alpha is the learning rate, and the value range is (0, 1);
s1505, the weight and offset partial derivatives of the l-th layer are:
Figure FDA0003548046810000026
Figure FDA0003548046810000027
in the formulas (12) and (13), L is more than or equal to 2 and less than or equal to L-1;
Figure FDA0003548046810000028
Figure FDA0003548046810000029
and S16, outputting the prediction result through the S15.
2. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 1, wherein: in S11, the electric devices of the home users are classified into air conditioners, resistors, fluorescent lamps, computers, and motors.
3. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 1, wherein: in S13, the operating state of the electric device includes: start, stop and run.
4. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 3, wherein: in S14, 1, 0, and 2 are used to respectively indicate the start, stop, and operation of the electric device.
5. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 1, wherein: in the step S12, a load current equation set is constructed and solved according to the formula (1) and the formula (2), so as to obtain a power consumption ratio of the electric device, and the step of decomposing the electric device includes the following steps:
according to the formula (1), when n types of main electric equipment are contained in the power load, the linear superposition of the currents of the main electric equipment for the total current obtains a formula (3), and a unit current parameter matrix and a load current vector are obtained;
i′l(t)＝β1·i'a1(t)+β2·i'a2(t)...+βn·i'an(t) (3)
in the formula (3), i'l(t) is the total cell current of the electrical load, the corresponding eigenvector is FPl；i’a1(t)、i’a2(t)、…、i’an(t) the unit currents of the 1 st, 2 nd, … th and n kinds of electric equipment respectively, and the corresponding characteristic vectors are FP respectivelya1，FPa2And FPn；β1、β2、…、βnE [0, + ∞) ] is the current weight coefficient of the 1 st, 2 nd, … th and n kinds of electric equipment respectively, and beta is [ beta ]1，β2，…，βn]T；
And (3) processing the equation (3) according to a phasor method to obtain a non-invasive power load monitoring estimation equation set:
FPl＝Ha·β (4)
in the formula (4), FPlIs i'l(t) corresponding feature vectors; hαThe unit current parameter matrix of the electric equipment is obtained;
to the non-invasivenessThe power load monitoring estimation equation set is optimized and solved to obtain beta1，β2，…，βn；
Current weight coefficient beta based on kth electric equipmentkCalculating the proportional coefficient of the active power and the reactive power of the fundamental wave of each main electric device according to the formula (5) and the formula (6):
Figure FDA0003548046810000031
Figure FDA0003548046810000032
in the formulae (5) and (6), betapkThe ratio of the active power of the fundamental wave of the kth electric equipment is obtained; beta is aQkThe fundamental wave reactive power proportion of the kth electric equipment is shown;
and obtaining the power consumption proportion of the electric equipment according to the formula (5) and the formula (6).
6. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 3, wherein: the step S13 of determining the operating state of the electric device includes:
s1301, comparing decomposition results of the electric equipment at two adjacent moments, wherein the decomposition result at the previous moment does not have the solution of the electric equipment, and the decomposition result at the later moment has the solution of the electric equipment, so that the running state of the electric equipment is starting and the electric equipment is used for detecting the complete starting process of the electric equipment;
s1302, comparing the decomposition results of the electric equipment at two adjacent moments, wherein the decomposition results at the two moments before and after have the solution of the electric equipment, and the running state of the electric equipment is running and is used for tracking and detecting the running process of the electric equipment;
s1303, comparing the decomposition results of the electric equipment at two adjacent moments, wherein the decomposition result at the former moment has the solution of the electric equipment, and the decomposition result at the latter moment has no solution of the electric equipment, so that the running state of the electric equipment is stopped and the process of stopping the electric equipment is detected.
7. The method for predicting the short-term load of the home subscriber based on the load decomposition technology as claimed in claim 1, wherein: in S1504, updating the weight and bias of the neural network includes the following steps:
s15041, setting DeltaW for all layers(l)＝0，△b(l)＝0，△W(l)And Δ b(l)Respectively an all-zero matrix and an all-zero vector;
s15042, setting i to be 1: m, and calculating the weight and biased gradient matrix of each layer of neurons by using a back propagation algorithm
Figure FDA0003548046810000041
And
Figure FDA0003548046810000042
then calculate
Figure FDA0003548046810000043
Inverse gradient weight matrix of(l)And
Figure FDA0003548046810000044
reverse gradient bias matrix Δ b(l)；
S15043, updating weight and bias, calculating
Figure FDA0003548046810000045
And
Figure FDA0003548046810000046
W(l)as updated weight, b(l)Is the updated bias.
CN201911090119.1A 2019-11-08 2019-11-08 Household user short-term load prediction method based on load decomposition technology Active CN111009893B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911090119.1A CN111009893B (en) 2019-11-08 2019-11-08 Household user short-term load prediction method based on load decomposition technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911090119.1A CN111009893B (en) 2019-11-08 2019-11-08 Household user short-term load prediction method based on load decomposition technology

Publications (2)

Publication Number Publication Date
CN111009893A CN111009893A (en) 2020-04-14
CN111009893B true CN111009893B (en) 2022-07-01

Family

ID=70111608

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911090119.1A Active CN111009893B (en) 2019-11-08 2019-11-08 Household user short-term load prediction method based on load decomposition technology

Country Status (1)

Country Link
CN (1) CN111009893B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111600304B (en) * 2020-06-17 2021-11-26 广东工业大学 Building power scheduling method, device and equipment
CN111815152B (en) * 2020-07-06 2023-02-28 安徽建筑大学 Fine-grained power load modeling method based on power consumption behavior rule and minimum state of electric appliance current
CN113379005B (en) * 2021-08-12 2021-10-29 新风光电子科技股份有限公司 Intelligent energy management system and method for power grid power equipment
DE102022105138A1 (en) * 2022-03-04 2023-09-07 Dehn Se Method and system for monitoring a power grid, method for training artificial intelligence for forecasting a future state of a power grid, computer program and computer-readable data carrier

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105404784A (en) * 2015-12-07 2016-03-16 河南许继仪表有限公司 Non-invasive power load decomposition method
CN107341736A (en) * 2016-11-07 2017-11-10 天津求实智源科技有限公司 A kind of electric load decomposition method and system based on Current Waveform Characteristics
CN109145949A (en) * 2018-07-19 2019-01-04 山东师范大学 Non-intrusive electrical load monitoring and decomposition method and system based on integrated study
CN110137942A (en) * 2019-04-23 2019-08-16 河海大学 Multiple Time Scales flexible load rolling scheduling method and system based on Model Predictive Control

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105404784A (en) * 2015-12-07 2016-03-16 河南许继仪表有限公司 Non-invasive power load decomposition method
CN107341736A (en) * 2016-11-07 2017-11-10 天津求实智源科技有限公司 A kind of electric load decomposition method and system based on Current Waveform Characteristics
CN109145949A (en) * 2018-07-19 2019-01-04 山东师范大学 Non-intrusive electrical load monitoring and decomposition method and system based on integrated study
CN110137942A (en) * 2019-04-23 2019-08-16 河海大学 Multiple Time Scales flexible load rolling scheduling method and system based on Model Predictive Control

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"基于稳态特征的非侵入式负荷辨识系统的设计与实现";张志豪;《中国优秀硕士学位论文全文数据库 工程科技II辑》;20180915(第9期);C042-183 *

Also Published As

Publication number Publication date
CN111009893A (en) 2020-04-14

Similar Documents

Publication Publication Date Title
CN111009893B (en) Household user short-term load prediction method based on load decomposition technology
Liu et al. Energy disaggregation of appliances consumptions using ham approach
Harell et al. Wavenilm: A causal neural network for power disaggregation from the complex power signal
CN107591844B (en) Active power distribution network robust reconstruction method considering node injection power uncertainty
CN106786534B (en) A kind of non-intrusive electrical load transient process discrimination method and system
CN107425520B (en) Active power distribution network three-phase interval state estimation method containing node injection power uncertainty
CN109145949A (en) Non-intrusive electrical load monitoring and decomposition method and system based on integrated study
Schneider et al. Evaluating conservation voltage reduction: An application of GridLAB-D: An open source software package
Gao et al. An iterative optimization and learning-based IoT system for energy management of connected buildings
CN109829560A (en) A kind of power distribution network renewable energy power generation cluster access planing method
CN116707331B (en) Inverter output voltage high-precision adjusting method and system based on model prediction
Zhang et al. High spatial granularity residential heating load forecast based on Dendrite net model
Pang et al. A composite state of charge estimation for electric vehicle lithium-ion batteries using back-propagation neural network and extended kalman particle filter
CN110880055A (en) Building intelligent ammeter system
Yang et al. State of health estimation based on GAN-LSTM-TL for lithium-ion batteries
Ma et al. Adaptive energy management strategy based on a model predictive control with real-time tuning weight for hybrid energy storage system
Trinh Smart energy systems, infrastructure financing, and the wider economy
Azad et al. Grid serving deployment of smart meter data in the context of distribution grid automation
Jiang et al. Two‐level area‐load modelling for OPF of power system using reinforcement learning
CN109948678A (en) A kind of long-term electricity demand forecasting method based on Fuzzy Bayesian Theory
Sharma et al. Renewable energy systems energy modeling using deep learning techniques
Wang et al. Improved long short-term memory: statistical regression model for high precision SOC estimation of lithium-ion batteries adaptive to complex current variation conditions
Chen et al. Variation-cognizant probabilistic power flow analysis via multi-task learning
Nichiforov et al. Embedded on-line system for electrical energy measurement and forecasting in buildings
Sun et al. Power load disaggregation of households with solar panels based on an improved long short-term memory network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant