CN115423153A - Photovoltaic energy storage system energy management method based on probability prediction - Google Patents

Photovoltaic energy storage system energy management method based on probability prediction Download PDF

Info

Publication number
CN115423153A
CN115423153A CN202210963607.4A CN202210963607A CN115423153A CN 115423153 A CN115423153 A CN 115423153A CN 202210963607 A CN202210963607 A CN 202210963607A CN 115423153 A CN115423153 A CN 115423153A
Authority
CN
China
Prior art keywords
prediction
energy storage
photovoltaic
storage system
battery
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
CN202210963607.4A
Other languages
Chinese (zh)
Inventor
郑含博
唐钰本
杜齐
胡永乐
郭文豪
陈俊江
万海斌
覃团发
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.)
Runjian Co ltd
Guangxi University
Original Assignee
Runjian Co ltd
Guangxi University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Runjian Co ltd, Guangxi University filed Critical Runjian Co ltd
Priority to CN202210963607.4A priority Critical patent/CN115423153A/en
Publication of CN115423153A publication Critical patent/CN115423153A/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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • 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
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/22—The renewable source being solar energy

Landscapes

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

Abstract

A photovoltaic energy storage system energy management method based on probability prediction is characterized in that a long-term short-term memory (LSTM) neural network is used for obtaining photovoltaic probability prediction, multivariate distribution is sampled to generate a prediction scene based on a copula function model, the maximum benefit cost is the minimum on the premise that load requirements and energy storage characteristics are met according to prediction results, and real-time optimization is provided by model prediction control. The method can quantify the uncertainty of prediction, reduce the robustness optimization loss caused by the traditional point prediction, improve the safety and reliability of the energy storage system, and realize the efficient absorption and utilization of photovoltaic energy.

Description

Photovoltaic energy storage system energy management method based on probability prediction
Technical Field
The invention belongs to the technical field of photovoltaic power generation, and particularly relates to a photovoltaic energy storage system energy management method based on probability prediction.
Background
Solar energy has the advantages of large total energy, easy resource development, cleanness, no pollution and the like, so that photovoltaic power generation is regarded as an important means for solving the energy crisis. However, irregular and random output changes caused by the characteristics of randomness and volatility of photovoltaic power generation bring uncertainty to the planning and operation of the power system, and particularly under the condition of high photovoltaic permeability, energy storage units such as batteries and the like are required to be used for stabilizing fluctuation and storing. In large-scale use of photovoltaic, there is a need to quickly and accurately predict the short-term and long-term output of on-site photovoltaic power generation in order to effectively manage energy demand, energy storage, and energy supply of supplemental and backup energy.
For the estimation and prediction of the maximum photovoltaic power, a data-driven method is mainly adopted, and the method comprises a statistical technique and a machine learning tool. Photovoltaic power is estimated from environmental information, such as array irradiance plane, module temperature, ambient temperature and wind speed conditions and historical data, to build a predictive model of photovoltaic power output with environmental information such as irradiance, ambient temperature and wind speed.
The energy management method mainly comprises a plurality of steps of prediction, day-ahead optimization, real-time optimization and the like. And performing energy management based on the predicted result, and performing real-time optimization. While the stored energy serves as an energy buffer to meet the changing solar radiation and load demands, the stored energy is used to save money by using time-of-use electricity prices and storing energy at low off-peak rates, the stored energy is used to reduce peak demands and obtain revenue, and the controllability, stability and reliability of the whole power system are improved.
The energy management method of the photovoltaic energy storage system for solving the problems of photovoltaic uncertainty and absorption at present has the disadvantages that:
1. the data-driven photovoltaic output prediction method has the advantage that the performance is remarkably reduced under low irradiance and regression coefficients. Due to the seasonality of lighting conditions, the predictive model needs to be updated regularly to maintain the accuracy of the model. And the deterministic prediction of the condition mean point of the prediction signal provides very limited information for the decision of energy management, and the prediction deviation brings large uncertainty influence.
2. Most energy management usually only considers the energy state of the stored energy, neglecting the dynamic characteristics of the system, such as the limit of charging and discharging rate; the service life loss cost of the energy storage system is neglected when the time-of-use electricity price income is considered; rule-based short-term real-time optimization calculations lack the potential impact on subsequent policy choices.
Disclosure of Invention
In view of the technical problems in the background art, the photovoltaic probability prediction is obtained by using a long-short term memory (LSTM) neural network, the multivariate distribution is sampled based on a copula function model to generate a prediction scene, the maximum profit cost is the minimum on the premise of meeting the load requirement and the energy storage characteristic according to the prediction result, real-time optimization is provided by using Model Prediction Control (MPC), the uncertainty of prediction is quantized to reduce the robustness optimization loss caused by the traditional point prediction, the safety and reliability of the energy storage system are improved, and the efficient absorption and utilization of photovoltaic energy is realized.
In order to solve the technical problems, the invention adopts the following technical scheme to realize:
a photovoltaic energy storage system energy management method based on probability prediction comprises the following steps:
a photovoltaic energy storage system energy management method based on probability prediction is characterized by comprising the following steps:
s1, generating a photovoltaic prediction scene, obtaining probability prediction by using an LSTM neural network, taking a plurality of data including temperature, wind speed, cloud cover, cloud layer type and humidity as input variables, and adding historical information of solar radiation values as input variables; the solar radiation value is used as output;
s2, performing normalization processing on all input variables, unifying dimensions, removing units, and periodically inputting according to time;
a statistical parametric model of the prediction error is defined to assume the distribution of uncertainty, and an LSTM neural network is used to predict the parameter-specific distribution,
solving statistical parameters of the prediction result (such as mean, variance, skewness and kurtosis) aiming at the prediction result (solar irradiance);
s3, taking historical data of all input variables as predicted input, dividing a historical data set into a training set by 70 percent and a verification set by 30 percent, adding weighted noise during training to ensure noise information in the data, and setting a network type, a prediction range m, the number of hidden layer layers and the number of neurons;
finishing training by using the training set, and calculating the corresponding degree of the prediction and the actual observed value for the statistical quality of the prediction of the estimation point; using the root mean square error as an error measure, and outputting a designated quantile q of the distribution of the prediction target;
s4, modeling the marginal distribution of each input variable by using a copula function model-based method, collecting historical data of the input variables, and transforming each variable u by using probability integral on the marginal distribution of the univariate i The data points of (A) are converted to D-cube [0,1] D Thereby estimating copula density;
after obtaining the multivariate distribution, the dependency structure based on copula model and the original data is obtained:
u＝(u i ,...,u D )∈[0,1] D ，
generating a multivariate random vector in two steps by using a copula model, generating a dependent random number, and sampling from uniformly distributed U (0,1)
Figure BDA0003794090830000031
Figure BDA0003794090830000032
Representing the first generated number, the others u, of a group 2 、u 3 Etc. are generated (or solved) from the conditional distribution functions; distribution function according to condition
Figure BDA0003794090830000033
In turn generate
Figure BDA0003794090830000034
In the formula, the meanings of U and U are different, the uniformly distributed U is a fixed mathematical expression, and U is self-defined data U i A set of (a);
s5, based on marginal distribution from probability prediction, converting a generated variable of a unit D cube into an original variable dimension by using inverse transformation sampling to obtain a predicted scene photovoltaic power generation scene and a corresponding quantile;
s6, performing energy management on the photovoltaic energy storage system based on an MPC algorithm according to the photovoltaic prediction, the time-of-use electricity price graph and the load change graph;
determining output power P based on fixed photovoltaic system size pv Dependence on input variables including irradiance S
The energy storage system state model comprises a battery and a converter, and calculates energy conversion, charge state and battery cycle loss cost C BE And state of charge SOC based on time of use price R grids Calculating the charging and discharging cost C of the power grid grids Establishing a state space model according to the dynamic characteristics of the energy storage system;
s7, establishing system constraints, wherein the constraint conditions comprise SOC upper and lower limits, battery charge and discharge current limits, and the initial SOC is equal or approximately equal to the ending SOC;
determining the objective function J as:
Figure BDA0003794090830000035
in the formula, alpha and beta are weight coefficients at the point prediction and the 50% probability prediction respectively; c BE For the cost of battery depletion, C grids Is the cost of charging and discharging the power grid;
s8, setting time-of-use electricity price, energy storage equipment parameters and a target initial SOC value, selecting the step length of the MPC, the predicted step number and the control step number to carry out simulation, solving a target function, and outputting a control variable P BE (ii) a Based on the prediction result, the control operation is completed.
Preferably, in the step S2, the photovoltaic output is predicted by taking 10min as a time scale input variable.
Preferably, in step S3, the prediction range is set to 24 hours, and m =144.
Preferably, in step S7, after the system constraint is subjected to soft constraint correction, the following is performed:
Figure BDA0003794090830000041
in the formula, SOC represents the state of charge, SOC min Indicates minimum state of charge, P BE Representing the battery output (discharge) power, P dis_max Represents the maximum discharge power: p is ch_max Representing the maximum charging power.
Preferably, in step S7, according to the energy management objective requirement, the function weight coefficient α is increased to improve the expected benefit and reduce the conservative cost of the system.
Preferably, in step S8, the step lengths of the MPCs are selected to be 15min, 30min and 1h respectively according to the predicted time length, the load change condition and the calculation requirement.
Preferably, in step S6, when the photovoltaic system is dimensioned, itOutput power P pv The relationship with variables such as irradiance is:
P pv ＝η pv η t P PV_nominal S/S nominal
in the formula, P pv To predict photovoltaic power generation output, eta pv Eta for photovoltaic panel conversion efficiency t For temperature-dependent conversion efficiency, P PV_nominal Denotes the nominal case output power, S denotes the predicted irradiance, S nominal Represents the irradiance at nominal;
the energy storage system state model comprises a battery and a converter, and the energy conversion is as follows:
P BE ＝P load -P pv -P grids
in the formula P BE For output of power, P, from the energy storage system grids For exchanging power, P, between photovoltaic energy storage systems and the grid load Consuming power for the load;
the state of charge of the energy storage system is calculated as:
SOC k+1 ＝SOC k -η BE P BE_k /P BE_nominal
in the formula, SOC k+1 Is the state of charge, η, of the battery at time k +1 BE For the energy conversion efficiency of the battery, P BE_nominal Represents the nominal power of the battery at nominal;
energy storage system mainly based on lithium battery and battery cycle loss cost C BE The calculation is as follows:
Figure BDA0003794090830000042
in the formula, C BE_total For the total acquisition cost of the battery, L BE The effective cycle number of the battery in the whole life cycle;
based on time of use price R grids Power grid charging and discharging cost C grids The calculation is as follows:
C grids ＝P grids R grids
according to the dynamic characteristics of the energy storage system, establishing a state space model as follows:
x(k+1)＝f(x(k),u(k))
in the formula, SOC is selected as a state variable x (k), and P is selected BE As the control variable u (k).
This patent can reach following beneficial effect:
1. the method carries out probability prediction based on the neural network, copula samples are carried out to generate a prediction scene, the uncertainty of photovoltaic prediction is effectively reduced, and the prediction uncertainty is quantized to reduce the loss of the traditional point prediction error to decision optimization;
2. according to the invention, the MPC controller is used for simulation, a model is built based on the dynamic characteristics of the system, prediction control is carried out under the constraints of charge-discharge rate and charge state, the maximum voltage and current output is limited, the SOC of the energy storage system is maintained in a safety range, and the system safety is ensured;
3. the energy management method fully considers the influences of photovoltaic output fluctuation, time-of-use electricity price and energy storage system loss, and improves the total yield of the system and increases the photovoltaic consumption utilization rate based on a multi-step prediction and single-step control MPC method.
Drawings
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
fig. 1 is a schematic structural diagram of a grid-connected photovoltaic energy storage system in an embodiment of the present invention;
FIG. 2 is a flow chart of generating a photovoltaic prediction scenario in an embodiment of the present invention;
FIG. 3 is a flow chart of MPC real-time optimization in an embodiment of the present invention;
FIG. 4 is a time-of-use electricity price change diagram according to an embodiment of the present invention;
FIG. 5 is a graph illustrating predicted and actual photovoltaic power generation variation according to an embodiment of the present invention;
FIG. 6 is a diagram illustrating real-time variation of load according to an embodiment of the present invention.
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, for the grid-connected photovoltaic energy storage system, the grid-connected photovoltaic energy storage system includes a photovoltaic power generation assembly, a DC-DC converter including photovoltaic MPPT, an energy storage system, a bidirectional DC-DC converter including MPC control algorithm, a DC bus, an ac/DC load, and a grid-connected inverter;
the photovoltaic module obtains the maximum power after passing through the MPPT algorithm, adjusts the output voltage to a reference value, and provides electric energy through a DC-DC converter; the AC/DC load is connected with a DC bus through a converter to obtain electric energy; the energy storage system provides electric energy for a load side through a DC-DC converter based on an energy management system MPC algorithm or absorbs the electric energy from a photovoltaic place and a grid-connected commercial power place;
the invention provides a photovoltaic energy storage system energy management method based on probability prediction, which comprises the following steps:
s1, generating a photovoltaic prediction scene as shown in FIG. 2, obtaining probability prediction by using an LSTM neural network, and fully sampling by adopting a copula model;
selecting temperature, wind speed, cloud amount, cloud layer type and humidity of a predicted place as input variables, and adding historical information of a sun radiation numerical value as the input variables; the solar radiation value is used as output;
and S2, normalizing all the variables, unifying dimensions and removing units. All input variables have a time periodicity, and are selected to be input at 10min intervals.
Defining a statistical parameter model of a prediction error to assume uncertain distribution, predicting parameter designated distribution by using an LSTM neural network, and solving a statistical parameter of a prediction result according to the prediction result;
and S3, taking historical data of all input variables as predicted input, dividing the historical data set into a training set by 70 percent and a verification set by 30 percent, and adding weighted noise during training to ensure noise information in the data.
Setting a network type, a prediction range m, the number of hidden layers and the number of neurons, selecting a bidirectional propagation neural network in the embodiment, and setting m =144 to indicate that the prediction range is the following 24 hours.
Completing training by using a training set;
to estimate the statistical quality of the point predictions, the degree of correspondence of the predictions to the actual observed values is calculated. Root Mean Square Error (RMSE) is used as the error metric:
Figure BDA0003794090830000061
where n is the number of sample data, y t Is the output of the prediction model, d t Is the actual measurement;
training a neural network, and updating the number of hidden layers and the number of neurons based on the root mean square error;
in this case, the optimal prediction model is set to 4 hidden layers, 10 neurons;
outputting a designated quantile q of the predicted target distribution;
Figure BDA0003794090830000062
when q =0.5, it is an estimate of the conditional median of the output distribution.
S4, as shown in FIG. 3, sampling the multivariate distribution by using a copula function model-based method comprises two substeps:
appropriately modeling the marginal distribution of each variable, constructing a copula model of the dependent structure comprising a multivariate distribution,
firstly, historical data of variables needs to be collected, and each variable u is transformed by using probability integral on single variable marginal distribution i The data points of (A) are converted to D-cube [0,1] D Thereby estimating copula density and completing the model.
After obtaining the multivariate distribution, the dependency structure based on copula model and the original data is obtained:
u＝(u i ,...,u D )∈[0,1] D (3)
generating a multivariate random vector in two steps by using a copula model, generating a dependent random number, and sampling from uniformly distributed U (0,1)
Figure BDA0003794090830000071
Distribution function according to condition
Figure BDA0003794090830000072
In turn generate
Figure BDA0003794090830000073
And S5, based on the marginal distribution from probability prediction, converting the generated variable of the unit D cube into an original variable dimension by using inverse transformation sampling to obtain a predicted scene photovoltaic power generation scene and a corresponding quantile.
S6, as shown in the figures 4-6, performing energy management on the photovoltaic energy storage system based on an MPC algorithm according to photovoltaic prediction, a time-of-use electricity price graph, a load change graph and the like;
when the photovoltaic system is dimensioned, its output power P pv The relationship with variables such as irradiance is:
P pv ＝η pv η t P PV_nominal S/S nominal (4)
in the formula, P pv For predicting photovoltaic power generation output, eta pv For photovoltaic panel conversion efficiency, η t For temperature-dependent conversion efficiency, P PV_nominal Denotes the nominal case output power, S denotes the predicted irradiance, S nominal Represents irradiance at nominal;
the energy storage system state model comprises a battery and a converter, and the energy of the battery is converted into:
P BE ＝P load -P pv -P grids (5)
in the formula P BE For output of power, P, from the energy storage system grids For exchanging power, P, between photovoltaic energy storage systems and the grid load Consuming power for the load;
the state of charge of the energy storage system is calculated as:
SOC k+1 ＝SOC k -η BE P BE_k /P BE_nominal (6)
in the formula, SOC k+1 State of charge of the battery at time k +1, η BE For the energy conversion efficiency of the battery, P BE_nominal Represents the nominal power rating of the battery at nominal;
energy storage system mainly based on lithium battery and battery cycle loss cost C BE The calculation is as follows:
Figure BDA0003794090830000081
in the formula, C BE_total For the total acquisition cost of the battery, L BE The effective cycle number of the battery in the whole life cycle;
based on time of use price R grids Power grid charging and discharging cost C grids The calculation is as follows:
C grids ＝P grids R grids (8)
according to the dynamic characteristics of the energy storage system, establishing a state space model as follows:
x(k+1)＝f(x(k),u(k)) (9)
in the formula, SOC is selected as a state variable x (k), and P is selected BE As control variables u (k);
s7, establishing system constraint, including: the upper and lower limits of SOC, the limit of battery charge-discharge current, the initial SOC and the end SOC are basically equal. In this case, the upper and lower limits of SOC are 20% and 80% respectively for protecting the over-charging and over-discharging of the battery.
Figure BDA0003794090830000082
Determining the objective function J as:
Figure BDA0003794090830000083
in the formula, alpha and beta are weight coefficients at the point prediction and the 50% probability prediction respectively; in the case of the scheme, the values are 1 and 0.1 respectively;
selecting the step length, the predicted step number and the control step number of the MPC according to the time-of-use electricity price change duration and the load change condition; in this case, the step length is set to 1 hour, the control step number is 1, and the prediction step number is 8;
and setting time-of-use electricity price and energy storage equipment parameters. In this case, the time-of-use electricity price is shown in fig. 4, the unit price of the energy storage device is 4 yuan/AH, 24 groups of 1kWH batteries are included, and the cycle life is 10000 times;
s8, setting a target initial SOC value, carrying out simulation, solving an objective function, and outputting a control variable P BE (ii) a Based on the prediction result, the control operation is completed.
The above-described embodiments are merely preferred embodiments of the present invention, and should not be construed as limiting the present invention, and the scope of the present invention is defined by the claims, and equivalents of technical features included in the claims. I.e., equivalent alterations and modifications within the scope hereof, are also intended to be within the scope of the invention.

Claims (7)

1. A photovoltaic energy storage system energy management method based on probability prediction is characterized by comprising the following steps:
s1, generating a photovoltaic prediction scene, obtaining probability prediction by using an LSTM neural network, selecting multiple data including temperature, wind speed, cloud cover type and humidity of a prediction place as input variables, and adding historical information of solar radiation values as the input variables; the solar radiation value is used as output;
s2, normalizing all input variables, unifying dimensions, removing units and periodically inputting according to time;
defining a statistical parametric model of prediction error to assume a distribution of uncertainty and predicting a parameter-specific distribution using an LSTM neural network;
solving the statistical parameters of the prediction result according to the prediction result;
s3, taking historical data of all input variables as predicted input, dividing a historical data set into a training set by 70 percent and a verification set by 30 percent, adding weighted noise during training to ensure noise information in the data, and setting a network type, a prediction range m, the number of hidden layer layers and the number of neurons;
finishing training by using a training set, and calculating the corresponding degree of the prediction and the actual observed value for the statistical quality of the prediction of the estimation point; using the root mean square error as an error measure, and outputting a specified quantile q of the distribution of the prediction target;
s4, modeling the marginal distribution of each input variable by using a copula function model-based method, collecting historical data of the input variables, and transforming each variable u by using probability integral on the marginal distribution of the univariate i The data points of (A) are converted to D-cube [0,1] D Thereby estimating copula density;
after obtaining the multivariate distribution, the dependency structure based on copula model and the original data is obtained:
u＝(u i ,...,u D )∈[0,1] D ，
generating a multivariate random vector in two steps by using a copula model, generating a dependent random number, and sampling from uniformly distributed U (0,1)
Figure FDA0003794090820000011
Representing a group u, namely generating a dependent random number; distribution function according to condition
Figure FDA0003794090820000012
In turn generate
Figure FDA0003794090820000013
S5, based on marginal distribution from probability prediction, converting a generated variable of a unit D cube into an original variable dimension by using inverse transformation sampling to obtain a predicted scene photovoltaic power generation scene and a corresponding quantile;
s6, performing energy management on the photovoltaic energy storage system based on an MPC algorithm according to the photovoltaic prediction, the time-of-use electricity price graph and the load change graph;
determining output power P based on fixed photovoltaic system size pv Dependence on variables such as irradiance S
The energy storage system state model comprises a battery and a converter, and calculates energy conversion, charge state and battery cycle loss cost C BE And state of charge SOC based on time of use price R grids Calculating the charging and discharging cost C of the power grid grids Establishing a state space model according to the dynamic characteristics of the energy storage system;
s7, establishing system constraints, wherein the constraint conditions comprise SOC upper and lower limits, battery charge and discharge current limits, and the initial SOC is equal or approximately equal to the ending SOC;
determining the objective function J as:
Figure FDA0003794090820000021
in the formula, alpha and beta are weight coefficients at the point prediction and the 50% probability prediction respectively; c BE For the cost of battery depletion, C grids Is the cost of charging and discharging the power grid;
s8, setting time-of-use electricity price, energy storage equipment parameters and a target initial SOC value, selecting the step length, the predicted step number and the control step number of the MPC to simulate, solving an objective function, and outputting a control variable P BE (ii) a Based on the prediction result, the control operation is completed.
2. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, wherein: and in the step S2, photovoltaic output prediction is carried out by taking 10min as a time scale input variable.
3. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, wherein: in step S3, the prediction range is set to 24 hours, and m =144.
4. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, wherein: in step S7, the system constraint is modified by soft constraint as follows:
Figure FDA0003794090820000022
in the formula, SOC represents a state of charge, SOC min Indicates minimum state of charge, P BE Representing the battery output power, P dis_max Represents the maximum discharge power: p is ch_max Representing the maximum charging power.
5. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, wherein: in the step S7, according to the energy management objective requirement, the function weight coefficient α is increased to improve the expected profit and reduce the conservative cost of the system.
6. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, characterized in that: in step S8, the step lengths of the MPCs are selected to be 15min, 30min, and 1h, respectively, according to the predicted duration, the load change condition, and the calculation requirement.
7. The photovoltaic energy storage system energy management method based on probability prediction as claimed in claim 1, wherein: in step S6, when the size of the photovoltaic system is determined, the output power P is pv The relationship with variables such as irradiance is:
P pv ＝η pv η t P PV_nominal S/S nominal
in the formula, P pv To predict photovoltaic power generation output, eta pv Eta for photovoltaic panel conversion efficiency t For temperature-dependent conversion efficiency, P PV_nominal Denotes the output power at nominal, S denotes the predicted irradiance, S nominal Represents irradiance at nominal;
the energy storage system state model comprises a battery and a converter, and the energy of the battery is converted into:
P BE ＝P load -P pv -P grids
in the formula P BE For output of power, P, from the energy storage system grids For exchanging power, P, between photovoltaic energy storage systems and the grid load Consuming power for a load;
the state of charge of the energy storage system is calculated as:
SOC k+1 ＝SOC k -η BE P BE_k /P BE_nominal
in the formula, SOC k+1 Is the state of charge, η, of the battery at time k +1 BE For the energy conversion efficiency of the battery, P BE_nominal Represents the nominal power of the battery at nominal;
energy storage system mainly based on lithium battery and battery cycle loss cost C BE The calculation is as follows:
Figure FDA0003794090820000031
in the formula, C BE_total For the total acquisition cost of the battery, L BE The effective cycle number of the battery in the whole life cycle;
based on time of use price R grids Power grid charging and discharging cost C grids The calculation is as follows:
C grids ＝P grids R grids
according to the dynamic characteristics of the energy storage system, establishing a state space model as follows:
x(k+1)＝f(x(k),u(k))
in the formula, SOC is selected as a state variable x (k), and P is selected BE As the control variable u (k).
CN202210963607.4A 2022-08-11 2022-08-11 Photovoltaic energy storage system energy management method based on probability prediction Pending CN115423153A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210963607.4A CN115423153A (en) 2022-08-11 2022-08-11 Photovoltaic energy storage system energy management method based on probability prediction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210963607.4A CN115423153A (en) 2022-08-11 2022-08-11 Photovoltaic energy storage system energy management method based on probability prediction

Publications (1)

Publication Number Publication Date
CN115423153A true CN115423153A (en) 2022-12-02

Family

ID=84198439

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210963607.4A Pending CN115423153A (en) 2022-08-11 2022-08-11 Photovoltaic energy storage system energy management method based on probability prediction

Country Status (1)

Country Link
CN (1) CN115423153A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116454951A (en) * 2023-04-28 2023-07-18 重庆跃达新能源有限公司 Light energy storage control system and method
CN117318111A (en) * 2023-11-29 2023-12-29 南通沃太新能源有限公司 Weather prediction-based dynamic adjustment method and system for light energy storage source

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116454951A (en) * 2023-04-28 2023-07-18 重庆跃达新能源有限公司 Light energy storage control system and method
CN116454951B (en) * 2023-04-28 2023-12-05 重庆跃达新能源有限公司 Light energy storage control system and method
CN117318111A (en) * 2023-11-29 2023-12-29 南通沃太新能源有限公司 Weather prediction-based dynamic adjustment method and system for light energy storage source
CN117318111B (en) * 2023-11-29 2024-03-22 南通沃太新能源有限公司 Weather prediction-based dynamic adjustment method and system for light energy storage source

Similar Documents

Publication Publication Date Title
CN107039975B (en) Energy management method for distributed energy system
CN115693757A (en) Photovoltaic energy optimization regulation and control method based on digital twinning technology
CN115423153A (en) Photovoltaic energy storage system energy management method based on probability prediction
CN112994092B (en) Independent wind-solar storage micro-grid system size planning method based on power prediction
CN116151486B (en) Multi-time-scale random optimization method and device for photovoltaic charging station with energy storage system
CN117833285A (en) Micro-grid energy storage optimization scheduling method based on deep reinforcement learning
CN116581792A (en) Wind-solar energy storage system capacity planning method based on data model driving
Zhuo Microgrid energy management strategy with battery energy storage system and approximate dynamic programming
Gassi et al. Improving real-time energy decision-making model with an actor-critic agent in modern microgrids with energy storage devices
CN107359611B (en) Power distribution network equivalence method considering various random factors
CN112836849A (en) Virtual power plant scheduling method considering wind power uncertainty
CN117996816A (en) Intelligent control method and system for wind, light and firewood storage and team-level energy storage
CN112072643A (en) Light-storage system online scheduling method based on depth certainty gradient strategy
CN117856339A (en) Micro-grid electric energy data control system and method based on big data
CN117578449A (en) Multi-energy micro-energy network optimization scheduling method considering uncertainty of renewable energy sources
Bai et al. An online multi-level energy management system for commercial building microgrids with multiple generation and storage systems
CN115693787B (en) Method for analyzing new energy acceptance of optical storage and distribution power grid in consideration of source load randomness
Ioli et al. A smart grid energy management problem for data-driven design with probabilistic reachability guarantees
CN114186811B (en) Construction method and application of real-time scheduling model of micro-grid system
CN114336597A (en) Light storage inverter energy scheduling based on self-adaptive penalty function particle swarm algorithm
Engel Sizing of a photovoltaic system with battery on the basis of the multi-agent adaptive fuzzy neuronet
Jain et al. Battery optimization in microgrids using Markov decision process integrated with load and solar forecasting
Liu et al. Micro grid energy management based on two-stage robust optimization
Sun et al. Optimal Scheduling of Renewable Energy Power Station with Energy Storage System Based on Long Short-Term Memory Electricity Price Forecasting
CN117833300B (en) Energy storage capacity configuration method based on intelligent power distribution equipment

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