CN-122000999-A - Multi-objective rolling optimization operation method for micro-grid considering voltage stabilization and energy storage cooperation

CN122000999ACN 122000999 ACN122000999 ACN 122000999ACN-122000999-A

Abstract

A micro-grid multi-target rolling optimization operation method considering voltage stabilization and energy storage cooperation belongs to the technical field of power system operation and control. The method aims to solve the problems of low precision, poor robustness and high running cost of the existing micro-grid energy management strategy when the renewable energy uncertainty is processed. Preprocessing original photovoltaic output and meteorological data, smoothing by adopting an improved Akima interpolation method, screening relevant input characteristics by adopting a pearson correlation coefficient, decomposing a wavelet packet into sub-signals of different frequency bands, training an LSTM predictor model, performing linear weighted fusion output, establishing a micro-grid system mathematical model, taking power generation cost and environmental cost as optimization targets, solving based on an improved NSGA2 algorithm, acquiring a pareto optimal solution set by introducing an adaptive learning mechanism and a constraint meeting strategy, deciding to select an optimal scheduling scheme by adopting a TOPSIS method, and executing rolling time domain optimization. For microgrid sustainable energy management.

Inventors

SU YANHAO
LU GENGRU
JIN YANSONG
MA QI
FENG YAN
LIU BAOLIN

Assignees

国网青海省电力公司海南供电公司
国网青海省电力公司

Dates

Publication Date: 20260508
Application Date: 20251030

Claims (10)

1. The micro-grid multi-target rolling optimization operation method considering voltage stabilization and energy storage is characterized by comprising the following steps of: The original photovoltaic output and meteorological data are preprocessed, the missing data are smoothed by adopting an improved Akima interpolation method, input features related to the photovoltaic output are screened by pearson correlation coefficients, and sequence data of the screened input features are decomposed into a plurality of sub-signals with different frequency bands by adopting wavelet packet decomposition; Constructing a mixed prediction model, namely respectively training corresponding LSTM prediction sub-models for sub-signals of each frequency band, wherein each sub-model takes a variable sub-signal under a specific frequency band as input and takes the contribution of the frequency band to the photovoltaic output as output; establishing a multi-objective optimization model, namely establishing a mathematical model of each device in a micro-grid system based on a photovoltaic output prediction result, defining a multi-objective function taking the generation cost and the environmental cost as optimization targets, and introducing key constraint conditions to form the multi-objective optimization model; And executing rolling time domain optimization, namely applying an optimal scheduling scheme to the micro-grid system according to a model predictive control mechanism, rolling the system state to the next moment, and repeatedly executing the steps to realize closed-loop optimization operation of the whole scheduling period.
2. The method for optimizing the operation of the micro-grid multi-objective rolling taking voltage stabilization and energy storage synergy into consideration according to claim 1, wherein the specific method for smoothing missing data by adopting the improved Akima interpolation method comprises the following steps: The data was smoothed using modified Akima segmentation three Hermite interpolations: ; Wherein, the Representing the function of the interpolation as such, Representing interpolation variables, the interpolation interval being , 、、 And All represent constants.
3. The method for optimizing operation of the micro-grid multi-objective rolling taking voltage stabilization and energy storage synergy into consideration according to claim 1, wherein the specific method for screening input characteristics related to photovoltaic output through pearson correlation coefficients comprises the following steps: the raw dataset contains features of temperature, relative humidity, total radiation, scattered radiation, precipitation, global oblique irradiance, and scattered oblique irradiance; correlation of these features with photovoltaic output was screened by pearson correlation coefficient: ; Wherein, the Representing the correlation coefficient between each feature and the photovoltaic output, The number of features is represented and, The photovoltaic output data is represented by the graph, The characteristic data is represented by a representation of the characteristic data, Represent the first The data of the plurality of data, Representing the total number of data.
4. The method for optimizing operation of a micro-grid multi-objective rolling taking voltage stabilization and energy storage synergy into consideration according to claim 1, wherein the specific method for decomposing the sequence data of the screened input features into a plurality of sub-signals with different frequency bands by wavelet packet decomposition comprises the following steps: the wavelet basis function is "db3", and the decomposition level is three; Discrete wavelet packet decomposition is expressed as: ; Wherein, the Represent the first Level of individual wavelet Is used for the coefficient of (a), Which means that the low-pass filter is present, A high-pass filter is represented as such, And Each representing a discrete time index; The reconstruction process is expressed as: ; The dataset was normalized to: ; Wherein, the The normalized result is represented by a graph of the normalized result, Representing the maximum value of the target normalized range, 1, The minimum value of the target normalization range is-1; The original value is represented and, The maximum original value is indicated and the value of the original value is, Representing the minimum original value.
5. The micro-grid multi-objective rolling optimization operation method considering voltage stabilization and energy storage coordination according to claim 1, wherein the LSTM predictor model specifically comprises: the loss function of LSTM is: ; Wherein, the , Indicating the length of the predicted segment, The true value is represented by a value that is true, Representing the predicted value.
6. The method for multi-objective rolling optimization operation of a micro-grid taking voltage stabilization and energy storage cooperation into consideration according to claim 1, wherein the specific method for defining the multi-objective function with the power generation cost and the environmental cost as optimization targets comprises the following steps: Cost objective function for power generation The method comprises the following steps: ; Wherein, the The time is represented by the time period of the day, Represents an optimized time domain of the representation, Representing the cost of power generation of the distributed photovoltaic PV, Time of presentation Active power of the time-distributed photovoltaic PV, Representing the cost of power generation by the distributed wind turbines WT, Time of presentation Active power of the time distributed wind turbine WT, And Representing the active power generation cost and the reactive power generation cost of the schedulable generator DG respectively, Time of presentation Active power exchanged with the main grid at the time, Time of presentation Reactive power exchanged with the main grid at the time, And Representing the purchase of active and reactive power costs from the main grid respectively, Is that , Is that , And The active power price and the reactive power price are sold to the main grid respectively, Is that , Is that , And The cost of operation of the energy storage device ES when charged and the cost of operation when discharged, Time of presentation The charging power of the energy storage device ES at that time, Time of presentation The discharge power of the energy storage equipment ES; Environmental cost goal The method comprises the following steps: ; Wherein, the Representing the equivalent environmental cost factor.
7. The micro-grid multi-objective rolling optimization operation method considering voltage stabilization and energy storage coordination according to claim 1, wherein the introduced key constraint conditions comprise power flow constraint, voltage and power balance constraint, specifically: Wherein, the And Respectively represent time Time injection node Is used for the control of the active power and the reactive power, ; Is connected with the node A set of adjacent nodes; Is the time of Time node Is satisfied by the voltage amplitude of , And Is the lower and upper limits of the voltage amplitude, Voltage amplitudes of all nodes; Is the time of Time node Is set to the voltage amplitude of (1); And Respectively represent slave nodes To the node Is a conductivity and susceptance of (a); Is the time of Time slave node To the node Is a phase difference of (2); indicating the number of schedulable generators, The amount of stored energy is indicated, The number of the photovoltaic power stations is represented, Representing the number of wind farms, The number of coincidences is indicated, A serial number representing the power plant/energy storage/load, Representing the loss of the network and, Representing the load power of each node; Indicating number of Is capable of scheduling the active power exchanged by the generator DG with the main grid, Indicating number of Is used for discharging power of the energy storage device ES, Indicating number of Active power of the distributed photovoltaic PV, Indicating number of Active power of the distributed wind turbine WT, Indicating number of Is provided for the charging power of the energy storage device ES.
8. The micro-grid multi-objective rolling optimization operation method considering voltage stabilization and energy storage coordination according to claim 1, wherein the multi-objective optimization model is specifically: ; Wherein, the Is an optimization variable that is used to determine the optimum state of the plant, Is a feasible region; the optimization problem is limited by equipment operation constraints, power flow constraints, and power balance constraints.
9. The micro-grid multi-objective rolling optimization operation method considering voltage stabilization and energy storage coordination according to claim 1, wherein the specific method for solving the multi-objective optimization model based on the improved NSGA2 algorithm comprises the following steps: Two operators in NSGA2 algorithm are fast non-dominant ordering and congestion distance calculation; the specific process of the rapid non-dominant ordering is as follows: determining a set of individuals of a first ranking level: Is provided with As a population of the population, To at the same time Middle quilt individual ( ) The collection of all the individuals that are in possession, Is that Center control Is a total number of individuals; For the following Each individual of (3) Performing: For the sum of each individual An initialization operation is performed such that the data of the data packet, , ; For the following Each individual of (3) Performing: If it is Then will be subjected to Overruled individuals Logging into In the process, the ; If it is Then every time a dominant is found Is a group of the individuals of the group (a), The number of the elements is increased by 1, ; If it is Then individual Rank of (2) of the order of (3) And store the individuals in a first order Is set in the collection of (2); determining a set of individuals for the remaining ranking levels: Is provided with Is the first Individual sets of rank ordering, order ; When (when) At that time, execution: , wherein, Representing storing temporary variables of the next level of individuals; For the following Each individual of (3) Performing: For the following Each individual of (3) Performing: If (3) Then , , Representing individuals Is a ranking of (3); updating the sequence number of the sequencing level: ; Obtain the first Level ordered set of individuals ; All individuals in (a) are assigned to different levels of the ordered collection; The specific process of the congestion distance calculation is as follows: for each ordered set Performing: , wherein, Representing a sorted set The number of individuals in (a); For each of Order-making , Representation storage Temporary variables of crowded distance for each individual; for each objective function Performing: Will be The individuals in the model are stored in a way of arranging the corresponding objective functions from small to large In (a): ; ; For the following To the point of : First, the The crowding distance of individual is equivalent to The sum of the distances of the two individuals in the direction of each objective function, namely: ; Represent the first Individual first The value of the objective function is set, Is a divisor for unified dimension; Obtaining Crowded distance of all individuals: 。
10. The method for achieving multi-objective rolling optimization operation of the micro-grid taking voltage stabilization and energy storage synergy into consideration according to claim 1, wherein the specific method for obtaining the pareto optimal solution set by introducing an adaptive learning mechanism and constraint satisfaction strategy comprises the following steps: for a cost-based index, the solution in pareto front should be normalized to: ; Wherein, the Is the first Solution No. A normalized value of the individual objective function, Is the first Solution No. The original value of the individual objective function value, And Respectively represent the first Minimum and maximum values of the respective objective function values; The standard deviation was calculated as: ; Wherein, the Is the first The standard deviation of the individual targets is calculated, Is the size of the pareto front, Is the first Average value of individual targets; Calculating a target And The correlation coefficient of (2) is: ; Wherein, the Is the object of And Is a correlation coefficient of (2); First, the In the solution of the first The information content of each target is calculated as: ; Wherein, the Is the content of the information which is to be processed, The number of targets; weights of each target The calculation is as follows: ; Wherein, the In order to be a target sequence number, Indicating the target sequence number as Information content of the target of (a); Then constructing a weighted standardized decision matrix as follows: ; Wherein, the Is the first Solution No. A weighted normalized value of the individual indicators; then positive and negative ideal solutions And The method comprises the following steps: ; Wherein, the Representing a positive ideal solution of the last object, , ; The distance between the solution and the ideal solution is calculated as: ; Wherein, the Is the distance between the solution and the ideal solution, Is the distance between the solution and the negative ideal solution; Relative proximity of each solution The calculation is as follows: ; The larger the solution The better.

Description

Multi-objective rolling optimization operation method for micro-grid considering voltage stabilization and energy storage cooperation Technical Field The invention relates to a micro-grid multi-target rolling optimization operation method considering voltage stabilization and energy storage cooperation, and belongs to the technical field of operation and control of power systems. Background With the deep advancement of global energy structure transformation, the application scale of renewable energy sources represented by photovoltaic and wind power in modern power systems is continuously expanding. The energy source has the advantages of cleanliness, sustainability and the like, but also brings significant challenges to stable operation and accurate scheduling of the power grid due to the characteristics of intermittent performance, volatility and low inertia. Unlike conventional synchronous generators, most renewable energy sources are connected into the system through a power electronic converter, the output of the renewable energy sources is obviously influenced by weather conditions, and the uncertainty is high, so that the conventional scheduling method is difficult to meet the operation requirement under high-proportion renewable energy source connection. To cope with the uncertainty of renewable energy sources, predictive technology is a key support means. The existing prediction method is mainly divided into a mechanism model and a data driving model. The mechanism model depends on high-precision numerical weather forecast, and is often limited by the reliability of meteorological data in practical application, and the data driving method, particularly a time sequence prediction model (such as RNN, LSTM, GRU, TCN and the like) based on deep learning, shows excellent performance in the aspect of extracting nonlinear time sequence characteristics. In recent years, the physical-data hybrid deep learning model further improves the robustness of the prediction. However, the existing research focuses on network structure improvement or physical information embedding, and deep excavation and comprehensive utilization of time-frequency domain features in renewable energy output are still insufficient, so that further improvement of prediction precision and scheduling effect is limited. In terms of power system scheduling, existing strategies mainly include rule-based methods and optimization-based methods. Although the policy based on rules is simple and easy to implement, it is difficult to realize economic operation of the system and multi-objective coordination. The optimal scheduling scheme is solved by constructing a mathematical model based on the optimization method, wherein the robust optimization can effectively process uncertainty and has good interpretability, but the conservation is strong, non-standard constraint is often required to be relaxed in modeling, and the linear matrix inequality is involved in the solving process, so that the calculation burden is heavy. On the other hand, deep reinforcement learning is receiving a great deal of attention due to its excellent real-time decision capability, and policy learning is mostly implemented by adopting an Actor-Critic architecture. However, DRLs have significant limitations in handling hard constraints, adapting to complex grid environments, and coordinating multi-objective optimization. The existing safe DRL method generally converts hard constraint into soft constraint by means of Lagrangian multiplier or punishment function, and constraint violation cannot be thoroughly avoided, while in multi-objective processing, a weighted summation mode is often adopted, and subjectivity of dimension difference and weight setting among objectives is ignored. Model predictive control has been applied to power distribution network and micro-grid energy management due to its roll optimization and feedback correction characteristics. However, most of the existing MPC researches are directed to a single economic goal, and safety constraints such as voltage quality in the running process are not fully considered, and the cooperative optimization of economic and environmental protection goals under the 'carbon neutralization' goal is not achieved. Therefore, how to construct a multi-objective optimal scheduling strategy capable of simultaneously taking voltage safety, economic operation and low carbon benefit into consideration in a renewable energy source high-permeability micro-grid becomes a key problem to be solved urgently. Disclosure of Invention The invention aims to solve the problems of low precision, poor robustness and high operation cost existing in the existing micro-grid energy management strategy containing distributed renewable energy sources when the renewable energy sources are uncertainty treated, and provides a micro-grid multi-target rolling optimization operation method considering voltage stabilization and energy storage cooperation. Th