CN108365608B - Uncertain optimization scheduling method and system for regional energy Internet - Google Patents

Abstract

A regional energy Internet uncertain optimization scheduling method and system comprises the following steps: acquiring random variable data; generating a plurality of scenes based on an error probability distribution of the random variables; cutting down the plurality of scenes to obtain typical scenes; processing the typical scene to obtain a discrete combination optimization scene; and optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene, and correcting an optimization result according to a short-time rolling correction model. The technical scheme of the invention is beneficial to improving the permeability of the distributed energy, coping with the uncertain fluctuation of the intermittent energy and reducing the scheduling pressure of the next-stage time scale.

Description

Uncertain optimization scheduling method and system for regional energy Internet

Technical Field

The invention relates to the field of operation and control of power systems, in particular to a regional energy Internet uncertain optimization scheduling method and system.

Background

The active power distribution network has the advantages that the tide distribution can be actively adjusted, various distributed energy sources are managed, and the utilization efficiency of the distributed energy sources is improved. The flexible direct current device is used for carrying out feeder interconnection, power exchange among interconnected networks can be completed, flexible regulation and control of controllable resources in a larger range are achieved, and the consumption capacity of intermittent energy is further improved. With the access of high-permeability distributed energy to an active power distribution network, the randomness and the volatility of the distributed energy bring a plurality of uncertain factors to the scheduling optimization of the power distribution network: firstly, the uncertainty of output inevitably leads to a certain prediction error, and the economy and the safety can not be ensured by a scheduling strategy directly generated according to a predicted value; secondly, the intermittent energy output condition is greatly influenced by weather, and when the output fluctuates sharply due to sudden environmental change, the power grid dispatching of the next time scale is stressed greatly. In this context, the conventional deterministic scheduling optimization strategy is no longer applicable. At present, a multi-scenario technology is partially researched and applied to describing intermittent distributed power sources of a power grid or uncertain load fluctuation, wherein most of the multi-scenario technology only discusses uncertainty on a certain time section or considers a plurality of sections but ignores the relevance of power fluctuation on adjacent time nodes, and the uncertainty problem that calculation is complex and description is difficult exists.

In summary, for an active power distribution network with high permeability distributed energy, a scheduling strategy capable of effectively describing uncertainty of the active power distribution network is urgently needed to cope with uncertain fluctuation in the network and further improve the consumption of intermittent energy.

Disclosure of Invention

In order to solve the defects in the prior art, the invention provides a regional energy Internet uncertain optimization scheduling method and system.

The technical scheme provided by the invention is as follows:

an uncertain optimization scheduling method for regional energy Internet comprises the following steps:

acquiring random variable data;

generating a plurality of scenes based on an error probability distribution of the random variables;

cutting down the plurality of scenes to obtain typical scenes;

processing the typical scene to obtain a discrete combination optimization scene;

and optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene, and correcting an optimization result according to a short-time rolling correction model.

Preferably, the generating a plurality of scenes based on the error probability distribution of the random variable includes:

discretizing the error into n intervals, wherein the area enclosed by each interval is S_1,t,S_2,t…S_n,tAnd forming an error state vector at the current moment:

wherein S is_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval;

correcting the error state vector of the random variable based on a Markov theory;

and randomly sampling the error state vector, and converting an error interval corresponding to the selected error state vector into a plurality of scenes.

Preferably, the state vector of the random variable is corrected based on markov theory and calculated according to the following formula:

wherein E represents a one-step state transition matrix,

denotes t ═ t₀Error state vector at time of day.

Preferably, the randomly sampling the error state vector, and converting an error interval corresponding to the selected error state vector into a plurality of scenes, includes:

randomly sampling the error state vector according to the probability of each error state at the time t;

after N times of sampling, form the state matrix

Wherein the selection of each interval constitutes a binary vector (X) of length N_t)_N×1The binary vector (X)_t)_N×1Zhongling x_i(i ═ 1 … n) represents the selected state of the error interval: if selected, x_i1, otherwise x_i＝0；

For the state matrix

Forming the state matrix into t by Monte Carlo sampling₀-a scene of a period T.

Preferably, the cutting down the plurality of scenes results in a typical scene, including:

and carrying out scene reduction on the plurality of scenes by adopting a fuzzy c-means clustering algorithm, and calculating the typical scene probability.

Preferably, the scene reduction is performed on the plurality of scenes by using a fuzzy c-means clustering algorithm according to the following formula:

in the formula, U: membership matrix, V is central vector matrix, C: number of typical scenes, M₀: number of original scenes, V_c: center vector, X, of the c-th clustered scene set_i: ith original scene vector, μ_ci: membership function from ith scene vector to c clustering scene set, m: convergence factor, J: the degree of similarity of the scene inside each category to the center vector.

Preferably, the c-thCenter vector V of clustering scene set_cThe probability is calculated as follows:

in the formula, N_c: can be clustered into a central vector V_cThe number of scenes in the scene (c).

Preferably, the processing of the typical scene probability to obtain a discrete combination optimization scene includes:

the probability of the discrete combination optimization scene is calculated according to the following formula:

in the formula, I: number of optimal typical scenes of a photovoltaic, K: number of optimal typical scenes of the fan, J: the load is optimal, the number of typical scenes is optimal, and M is the number of combined scenes;

the probability of the photovoltaic power source occurring in the ith scene,

the probability of the fan occurring in the k-th scene,

probability of load occurrence of jth scene.

Preferably, the objective function of the long-time uncertainty optimization scheduling model includes:

the global total cost is lowest and the network node voltage is ideally distributed;

the total cost is calculated as follows:

in the formula, λ₁: the total cost; epsilon_s: combining the probability of occurrence of scene S;c₁: the cost of the distributed power supply can be adjusted; p is a radical of_DG,t: generating power of the distributed power supply at the time t; Δ t: a unit time; c. C₂: flexible direct current device scheduling cost; p is a radical of_VSC,t: the power of the flexible direct current device at the moment t; c. C_IL: interruptible load costs; p is a radical of_IL,t: the load power can be interrupted at the moment t; c. C_grid: the cost of purchasing electricity from the upper-level power grid; p is a radical of_grid,t: the transmission power of the upper-level power grid at the moment t; c. C_e: power supply benefits; p is a radical of_L,t: power supply load power at time t; c. C_loss: running loss cost; p is a radical of_{line_loss,s}: network loss power when combining scene S; p is a radical of_{VSC_loss,s}: the power loss of the flexible direct current device when the scene S is combined; m: combining the number of scenes; t: optimizing the duration of the long time scale;

the network node voltage is calculated as follows:

in the formula, λ₂: an average voltage deviation amount; n is a radical of_node: the number of nodes; u. of_i,t,s: voltage deviation of the ith node at the moment t under the S scene; t: optimizing the duration of the long time scale; u. of_s,t: voltage offset at time t, u_i ^N: the node is rated for voltage.

Preferably, the short time rolling correction objective function is as follows:

wherein, g: the deviation of the adjustable resource from the reference value; t is the optimization duration of the short time scale; u is the number of tunable resources, p_i,t ^res: an ith adjustable resource contribution reference value; p is a radical of_i,t: the optimization result of the ith adjustable resource short time scale; p_Gi: the ith adjustable resource power rating.

Preferably, the acquiring random variable data includes: and acquiring wind power output, photovoltaic output and fluctuating load.

Another objective of the present invention is to provide a regional energy internet uncertain optimization scheduling system, including: the device comprises an acquisition module, a generation module, a reduction module, a processing combination module and a determination module;

the acquisition module is used for acquiring random variable data;

the generating module is used for generating a plurality of scenes based on the error probability distribution of the random variable;

the cutting module is used for cutting the scenes to obtain typical scenes;

the processing combination module is used for processing the typical scene probability to obtain a discrete combination optimization scene;

and the determining module is used for optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene and correcting the optimization result according to a short-time rolling correction model.

Preferably, the generating module includes: dividing a submodule, a correction submodule and a sampling submodule;

a division submodule for discretizing the error into n intervals, the area enclosed by each interval is S_1,t,S_2,t…S_n,tAnd forming an error state vector at the current moment:

wherein S is_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval;

the correction submodule is used for correcting the error state vector of the random variable based on the Markov theory;

and the sampling submodule is used for randomly sampling the error state vector and converting an error interval corresponding to the selected error state vector into a plurality of scenes.

Preferably, the reduction module comprises a clustering submodule and a calculating submodule;

the clustering submodule is used for carrying out scene reduction on the plurality of scenes by adopting a fuzzy c-means clustering algorithm according to the following formula:

in the formula, U: membership matrix, V: center vector matrix, C: number of typical scenes, M₀: number of original scenes, V_c: center vector, X, of the c-th clustered scene set_i: ith original scene vector, μ_ci: membership function from ith scene vector to c clustering scene set, m: convergence factor, J: the degree of similarity of the scenes inside each category to the central vector;

the calculation submodule is used for calculating a central vector V of the c-th clustering scene set according to the following formula_cProbability:

in the formula, N_c: can be clustered into a central vector V_cThe number of scenes in the scene (c).

Preferably, the processing combination module comprises,

a combining submodule, configured to calculate a probability of the discrete combination optimization scenario according to the following formula:

in the formula, I: number of optimal typical scenes of a photovoltaic, K: number of optimal typical scenes of the fan, J: number of load-optimized typical scenes, M: combining the number of scenes;

the probability of the photovoltaic power source occurring in the ith scene,

the probability of the fan occurring in the k-th scene,

probability of load occurrence of jth scene.

Preferably, the determining module includes: a long-time uncertain optimization scheduling model and a short-time rolling correction model;

the long-time uncertain optimization scheduling model is used for calculating an objective function with the lowest overall cost and the ideal distribution of the network node voltage;

and the short-time rolling correction model is used for calculating a target function of the deviation of the adjustable resource relative to the reference value.

Compared with the prior art, the invention has the beneficial effects that:

the technical scheme of the invention generates a plurality of scenes by the error probability distribution of the random variable based on the acquired random variable data; cutting down a plurality of scenes to obtain a typical scene; processing the typical scene to obtain a discrete combination optimization scene; the method has the advantages that the optimization is carried out through the long-time uncertain optimization scheduling model according to the discrete combination optimization scene, the optimization result is corrected according to the short-time rolling correction model, the problem of random uncertainty is solved by using a multi-scene technology, the problem of complex uncertainty which is difficult to describe can be converted into a plurality of possible certainty scenes, the solving difficulty is simplified, the improvement of the permeability of distributed energy resources is facilitated, the uncertain fluctuation of intermittent energy resources is responded, and the scheduling pressure of the next-stage time scale is reduced.

The technical scheme of the invention can take account of the error correlation of the uncertain variables among a plurality of time sections, effectively describe the uncertain problems into a plurality of deterministic scenes and simplify the original problems.

Drawings

FIG. 1 is a flow chart of a multi-time scale scheduling method of the Markov chain-multi-scenario technique according to the present invention;

FIG. 2 is a flow chart of scene cuts of the present invention;

FIG. 3 is a flowchart illustrating scene generation steps according to the present invention.

FIG. 4 is a flow chart of a regional energy Internet uncertain optimization scheduling method of the present invention;

Detailed Description

For a better understanding of the present invention, reference is made to the following description taken in conjunction with the accompanying drawings and examples.

Aiming at the uncertain scheduling problem of the current active power distribution network, the invention provides a multi-time scale uncertain scheduling model based on a Markov chain-multi-scenario technology, and the multi-time scale uncertain scheduling model takes the time relevance of the output error of uncertain units in the active power distribution network into consideration. The method comprises the following steps:

firstly, establishing a scene generation model combined with a Markov chain, and carrying out uncertain sampling on wind power output, photovoltaic output and uncertain load to generate a large number of scenes; then establishing a scene reduction model based on a fuzzy c-means clustering algorithm, and reducing the uncertain scene to obtain a typical scene; and finally, establishing a multi-time scale scheduling optimization model, namely a long-time scale active power distribution network uncertain optimization scheduling model and a deterministic short-time rolling correction model based on a typical scene.

The technical scheme shown in fig. 4 is as follows:

firstly, acquiring random variable data;

the involved uncertainty random variables include wind turbine output, photovoltaic output, and fluctuating loads.

Generating a plurality of scenes based on the error probability distribution of the random variables;

scene generation a large number of scenes are generated based on the error probability distribution of random variables, describing the uncertainty.

Research has shown that under a certain time section t, the prediction errors of the renewable energy output and the fluctuating load can be approximately considered to be in accordance with normal distribution. Discretizing the error into n intervals, wherein the area enclosed by each interval is S_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval, and forming an error state vector at the current moment:

on a time scale, there is a certain correlation between the uncertain deviations at each time instant, thus affecting the original prediction error probability distribution. Because the Markov chain shows good performance in the simulation of wind power and photovoltaic output sequences, the change process of random prediction errors along with time can be regarded as the Markov process, namely at the moment t_kWith the error state known, the process is at time t (t > t)_k) The state of (c) is only equal to t_kThe state of the moment of time is related to t_kThe previous state is irrelevant. The uncertain output error state model can be expressed as:

wherein, X_tError state at time t, E_ijThe one-step state transition probability representing the transition of the error from state i to state j can be obtained from statistical data, as shown in equation (3):

wherein x is_ijThe number of times of occurrence of the state j is changed from the state i of the time interval t to the state j of the time interval t +1 by statistically analyzing historical data and numerical weather forecast data.

Considering the time correlation, the error state vector m at the time t_tThe correction can be as follows:

wherein E is a one-step state transition matrix having E ═ E_ij]_n×nAnd is and

is t ═ t₀Error state vector at time of day.

And scene simulation is carried out on the state vector corrected based on the Markov theory, so that the effectiveness and the calculation efficiency of subsequent scene reduction are improved.

And randomly sampling the error state probabilities at the time t according to the error state probabilities. Let x_i(i ═ 1 … n) represents the selected state of the error interval: if selected, x_i1, otherwise x_iA sampling scene may be represented by a set of binary numbers, 0. After N times of sampling, the selected condition of each interval forms a binary vector (X) with the length of N_t)_N×1。

Obtaining error state interval sample X by the above process_tForming a state matrix

Monte Carlo sampling was then performed. For the ith sample, from X_sSampling in w-th column to obtain sample value X_iwForm t₀-a scene of a period T. The whole scene generation process is summarized as shown in fig. 3.

Thirdly, reducing the plurality of scenes to obtain typical scenes;

and (3) reducing the large-scale scenes by adopting a fuzzy c-means clustering method to ensure the calculation efficiency, and clustering the scenes by taking the formula (5) as a target.

Wherein U is a membership matrix, V is a central vector matrix, C represents the number of typical scenes, and M₀Representing the number of original scenes, V_cCenter vector, X, representing the c-th set of clustered scenes_iRepresenting the ith original scene vector, μ_ciAnd (3) representing a membership function from the ith scene vector to the c clustering scene set, wherein m is a convergence factor. Will M₀Dividing original scenes into C sets, and replacing all fields in the C sets with the central vectors of the clusters as typical scenesAnd the scene, the typical scene probability, is the probability sum of all scenes in the cluster. J represents the similarity degree of scenes in each category and the central vector, the optimization of J is completed through the following steps, and the central vector V is determined_c。

From fig. 2, a scene clustering flow chart can be seen, which is described in detail as follows:

step 1: the number of iterations h is 0 and the initial membership function μ is determined according to equation (5)_ci ⁽⁰⁾，U⁽⁰⁾＝[μ_ci ⁽⁰⁾]；

Step 2: let h be h +1, calculate the central vector

And Step3, updating the membership function of each scene:

step 4: judging whether U is satisfied^(h)-U^(h-1)If < epsilon, then output the center vector V_cOtherwise, go to Step 2.

The center vector V can be obtained after clustering_cThe probability is:

in the formula, N_cCan be clustered into a central vector V_cThe number of scenes in the scene (c). Delta_cCenter vector V of the c-th clustering scene set_cProbability.

The invention constructs a fuzzy clustering validity index PS, and the expression is shown as formula (7). The first term in the formula can represent the degree of compactness in the class, and the second term can represent the degree of separation between the classes. The larger the value, the more compact the c-th cluster and the larger the difference between the c-th cluster and other clusters. In the clustering process, the optimal typical scene number C is determined by the PS value^*As shown in formula (8).

Wherein the content of the first and second substances,

is the average of all typical scenes. V_kAnd representing the central vector of the kth clustering scene set.

Fourthly, processing the typical scene to obtain a discrete combination optimization scene;

and (4) arranging and combining the typical scenes corresponding to the random variables to obtain a series of discrete combination optimized scenes as subsequent research objects. The combined scene occurrence probability is the probability product of the corresponding typical scene, namely:

the number of the combined scenes is equal to the number of the photovoltaic typical scenes, the number of the fan typical scenes is equal to the number of the fan typical scenes, the number of the load typical scenes is equal to the number of the fan typical scenes, and the number of the load typical scenes is equal to the number of the combined scenes.

Representing the probability of the photovoltaic power source occurring in the ith scene,

the same is true.

And fifthly, optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene, and correcting the optimization result according to a short-time rolling correction model.

The invention relates to two time scale scheduling optimization models: the long-time scale global scheduling is based on a typical scene and corresponding probability obtained by a Markov chain-multi-scene technology, the planned output of a tie line and a VSC device is obtained, and the lowest global total cost and the ideal distribution of network node voltage are realized; and (4) keeping the long-time scale result unchanged in the short-time scale, and correcting other adjustable resources to ensure that the adjustment quantity is minimum relative to the reference value.

The long-time-scale scheduling objective is to achieve the lowest global total cost and the ideal distribution of network node voltages, expressed as equations (10) - (11).

λ₁: the total cost; epsilon_s: combining the probability of occurrence of scene S; c. C₁: the cost of the distributed power supply can be adjusted; p is a radical of_DG,t: generating power of the distributed power supply at the time t; Δ t: a unit time; c. C₂: flexible direct current device scheduling cost; p is a radical of_VSC,t: the power of the flexible direct current device at the moment t; c. C_IL: interruptible load costs; p is a radical of_IL,t: the load power can be interrupted at the moment t; c. C_grid: the cost of purchasing electricity from the upper-level power grid; p is a radical of_grid,t: the transmission power of the upper-level power grid at the moment t; c. C_e: power supply benefits; p is a radical of_L,t: power supply load power at time t; c. C_loss: running loss cost; p is a radical of_{line_loss,s}: network loss power when combining scene S; p is a radical of_{VSC_loss,s}: the power loss of the flexible direct current device when the scene S is combined; m: combining the number of scenes; t: optimizing the duration of the long time scale;

λ₂: an average voltage deviation amount; n is a radical of_node: the number of nodes; u. of_i,t,s: voltage deviation of the ith node at the moment t under the S scene; t: the long time scale optimizes the duration.

The expression (10) is a total probability expression form of the expected total cost, and the first five terms are adjustable DG cost and flexible direct current respectivelyDevice scheduling cost, interruptible load cost, cost of purchasing electricity from the upper level grid and power supply revenue, the last term being the operating cost, epsilon, due to grid loss, VSC loss_sIs the probability of occurrence of scene s. Equation (11) represents the average offset of the node voltage. Wherein u is_s,tIs the voltage deviation at time t, u, under S scene_i ^NThe voltage rating for the ith node.

The long-time scale constraint conditions comprise power balance constraint, power unbalance constraint, output constraint of each adjustable unit, node voltage constraint, flexible direct current device constraint and the like.

And (3) keeping the power of the tie line and the output of the flexible direct current device unchanged in a short time scale, optimizing the adjustable resource by taking the output value of the adjustable resource under the long time scale as a reference value, setting the target function to be the minimum deviation of the adjustable resource relative to the reference value as shown in the formula (12), g is the deviation of the adjustable resource relative to the reference value, and T is the optimization duration of the short time scale.

U is the number of tunable resources, p_i,t ^resFor the ith adjustable resource contribution reference, i.e. the long timescale result, p_i,tFor the optimization result of the ith adjustable resource short time scale, P_GiAnd rated for the ith adjustable resource.

The short time scale constraint is similar to the long time scale constraint and, in addition, a maximum value constraint for the amount of power adjustment for each cell is added.

An embodiment of the process of the present invention is described in detail below with reference to fig. 1.

Step1 in fig. 1 illustrates that the uncertain random variables involved in the present invention include wind turbine output, photovoltaic output, and fluctuating loads. And for the t time section, based on the original probability distribution of random variables, correcting the probability distribution by combining the Markov chain principle, and randomly sampling to generate a state sample. Turning to step 2;

step2 in fig. 1 describes that whether the current state sampling time reaches the long time scale control period is judged, if yes, the process goes to step3, otherwise, the process goes to step 9;

step3 in FIG. 1 illustrates the sampling of t₀State sample X of the T period_t(t＝t₀… T), constituting a state matrix

Monte carlo sampling was performed: for the ith sample, from X_sSampling in w-th column to obtain sample value X_iwForm t₀-original set of scenes for T period, go to step 4;

step4 in fig. 1 describes that, after an original scene is obtained, scene reduction is performed according to a fuzzy c-means clustering principle to obtain a typical scene, with formula (5) as a target, an optimal typical scene number is determined by formula (7) and formula (8), scene clustering is performed according to the steps of step1-step4 to obtain the typical scene and corresponding probabilities of each scene, then a combined scene and corresponding probabilities are determined by formula (9), and then the process goes to step 5;

step 5 in fig. 1 describes that the long-time scale is set to 4h, the scheduling process of the long-time scale global coordination power flow distribution is performed by taking the formula (10) and the formula (11) as targets, the constraint conditions include power balance constraint, power unbalance constraint, node voltage constraint, each adjustable unit and adjustable load constraint and flexible direct current device constraint, the tie line power and the flexible direct current device output can be obtained through long-time scale global optimization, and the process goes to step 6;

step 6 in fig. 1 illustrates that the short time scale is set to be 1h, in order to ensure that power fluctuation in the region does not affect an external network connected with the region, the tie line power and the output of the flexible direct current device are kept unchanged under the long time scale, the output of other adjustable units is taken as a reference value, the formula (12) is taken as a target, the process of automatically controlling the adjustable units in the short time scale region is carried out, the power of a distributed power supply, an energy storage device and an adjustable load in a period of time is obtained, and the process is issued, and the step 7 is shifted;

step 7 in fig. 1 describes that whether short time scale optimization is performed to a control cycle is judged, if yes, step 8 is turned to, otherwise step 6 is turned to continue short time scale optimization in the next time period, and the invention sets that optimization is performed once every 1h to realize rolling optimization;

step 8 in fig. 1 describes that the historical data is updated, the actual power of the wind power, the photovoltaic power and the load is added into the historical database, the prediction of the wind power, the photovoltaic power and the load in the next control period is performed, and the process goes to step 9;

step 9 of fig. 1 describes that a new error transfer matrix is solved from the statistical data according to equation (3), the error transfer matrix is transferred to the scene generation process, and the process goes to step 1.

Another objective of the present invention is to provide a regional energy internet uncertain optimization scheduling system, including: the device comprises an acquisition module, a generation module, a reduction module, a processing combination module and a determination module;

the four modules are further described below:

the acquisition module is used for acquiring random variable data;

a generating module for generating a plurality of scenes based on an error probability distribution of the random variables;

the reduction module is used for reducing the scenes to obtain typical scenes;

the processing combination module is used for processing the typical scene probability to obtain a discrete combination optimization scene;

and the determining module is used for optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene and correcting the optimization result according to a short-time rolling correction model.

A generation module comprising: dividing a submodule, a correction submodule and a sampling submodule;

a division submodule for discretizing the error into n intervals, the area enclosed by each interval is S_1,t,S_2,t…S_n,tAnd forming an error state vector at the current moment:

wherein S is_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval;

the correction submodule is used for correcting the error state vector of the random variable based on the Markov theory;

and the sampling submodule is used for randomly sampling the error state vector and converting an error interval corresponding to the selected error state vector into a plurality of scenes.

A trimming module comprising: a clustering submodule and a calculating submodule;

the clustering submodule is used for carrying out scene reduction on the plurality of scenes by adopting a fuzzy c-means clustering algorithm according to the following formula:

in the formula, U: membership matrix, V: center vector matrix, C: number of typical scenes, M₀: number of original scenes, V_c: center vector, X, of the c-th clustered scene set_i: ith original scene vector, μ_ci: membership function from ith scene vector to c clustering scene set, m: convergence factor, J: the degree of similarity of the scenes inside each category to the central vector;

a calculation submodule for calculating a central vector V of the c-th clustering scene set_cThe probability is calculated as follows:

in the formula, N_c: can be clustered into a central vector V_cThe number of scenes in the scene (c).

Processing combined modules, including combined submodules

A combination submodule, configured to calculate a probability of the discrete combination optimization scenario according to the following formula:

in the formula, I: number of optimal typical scenes of a photovoltaic, K: number of optimal typical scenes of the fan, J: number of load-optimized typical scenes, M: combining the number of scenes;

the probability of the photovoltaic power source occurring in the ith scene,

the probability of the fan occurring in the k-th scene,

probability of load occurrence of jth scene.

A determination module comprising: a long-time uncertain optimization scheduling model and a short-time rolling correction model;

the long-time uncertain optimization scheduling model is used for calculating an objective function with the lowest overall cost and the ideal distribution of the network node voltage;

and the short-time rolling correction model is used for calculating an objective function of the deviation of the adjustable resource relative to the reference value.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The present invention is not limited to the above embodiments, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are included in the scope of the claims of the present invention which are filed as the application.

Claims (14)

1. An uncertain optimization scheduling method for regional energy Internet is characterized by comprising the following steps:

acquiring random variable data;

generating a plurality of scenes based on an error probability distribution of the random variables;

cutting down the plurality of scenes to obtain typical scenes;

processing the typical scene to obtain a discrete combination optimization scene;

optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene, and correcting an optimization result according to a short-time rolling correction model;

generating a plurality of scenes based on the probability distribution of errors for the random variables, including:

discretizing the error into n intervals, wherein the area enclosed by each interval is S_1,t,S_2,t…S_n,tAnd forming an error state vector at the current moment: m is_t＝[S_1,t,S_2,t…S_n,t],

Wherein S is_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval;

correcting the error state vector of the random variable based on a Markov theory;

randomly sampling the error state vector, and converting an error interval corresponding to the selected error state vector into a plurality of scenes;

the randomly sampling the error state vector, and converting an error interval corresponding to the selected error state vector into a plurality of scenes, including:

randomly sampling the error state vector according to the probability of each error state at the time t;

after N times of sampling, form the state matrix

Wherein the selection of each interval constitutes a binary vector (X) of length N_t)_N×1The binary vector (X)_t)_N×1Zhongling x_i(i ═ 1 … n) represents the selected state of the error interval: if selected, x_i1, otherwise x_i＝0；

For the state matrix

Forming the state matrix into t by Monte Carlo sampling₀-a scene of a period T.

2. The regional energy internet uncertainty optimization scheduling method of claim 1, wherein the state vector of the random variable is corrected based on the markov theory and is calculated according to the following formula:

wherein E represents a one-step state transition matrix,

denotes t ═ t₀Error state vector at time of day.

3. The regional energy internet uncertainty optimization scheduling method of claim 1, wherein the pruning of the plurality of scenes into typical scenes comprises:

and carrying out scene reduction on the plurality of scenes by adopting a fuzzy c-means clustering algorithm, and calculating the typical scene probability.

4. The regional energy internet uncertain optimization scheduling method of claim 3, wherein the plurality of scenes are scene cut using a fuzzy c-means clustering algorithm according to the following formula:

1≤m≤∞

in the formula, U: membership matrix, V is central vector matrix, C: number of typical scenes, M₀: number of original scenes, V_c: center vector, X, of the c-th clustered scene set_i: ith original scene vector, μ_ci: membership function from ith scene vector to c clustering scene set, m: convergence factor, J: scenes and medians within each categoryThe degree of similarity of the cardiac vectors.

5. The regional energy internet uncertainty optimization scheduling method of claim 4, wherein the c-th clustering scene set center vector V_cThe probability is calculated as follows:

in the formula, N_c: can be clustered into a central vector V_cThe number of scenes in the scene (c).

6. The regional energy internet uncertainty optimization scheduling method of claim 5, wherein the processing of the typical scenario probabilities to obtain discrete combination optimization scenarios comprises:

the probability of the discrete combination optimization scene is calculated according to the following formula:

in the formula, I: number of optimal typical scenes of a photovoltaic, K: number of optimal typical scenes of the fan, J: the load is optimal, the number of typical scenes is optimal, and M is the number of combined scenes; delta_i ^PV: the probability of the photovoltaic power source occurring in the ith scene,

the probability of the fan occurring in the k-th scene,

probability of load occurrence of jth scene.

7. The regional energy internet uncertainty optimization scheduling method of claim 6, wherein the objective function of the long-time uncertainty optimization scheduling model comprises:

the global total cost is lowest and the network node voltage is ideally distributed;

the total cost is calculated as follows:

in the formula, λ₁: the total cost; epsilon_s: combining the probability of occurrence of scene S; c. C₁: the cost of the distributed power supply can be adjusted; p is a radical of_DG,t: generating power of the distributed power supply at the time t; Δ t: a unit time; c. C₂: flexible direct current device scheduling cost; p is a radical of_VSC,t: the power of the flexible direct current device at the moment t; c. C_IL: interruptible load costs; p is a radical of_IL,t: the load power can be interrupted at the moment t; c. C_grid: the cost of purchasing electricity from the upper-level power grid; p is a radical of_grid,t: the transmission power of the upper-level power grid at the moment t; c. C_e: power supply benefits; p is a radical of_L,t: power supply load power at time t; c. C_loss: running loss cost; p is a radical of_{line_loss,s}: network loss power when combining scene S; p is a radical of_{VSC_loss,s}: the power loss of the flexible direct current device when the scene S is combined; m: combining the number of scenes; t: optimizing the duration of the long time scale;

the network node voltage is calculated as follows:

in the formula, λ₂: an average voltage deviation amount; n is a radical of_node: the number of nodes; u. of_i,t,s: voltage deviation of the ith node at the moment t under the S scene; t: optimizing the duration of the long time scale; u. of_s,t: voltage offset at time t, u_i ^N: the node is rated for voltage.

8. The regional energy Internet uncertainty optimization scheduling method of claim 7,

the short time rolling correction objective function is as follows:

wherein, g: the deviation of the adjustable resource from the reference value; t is the optimization duration of the short time scale; u is the number of tunable resources, p_i,t ^res: an ith adjustable resource contribution reference value; p is a radical of_i,t: the optimization result of the ith adjustable resource short time scale; p_Gi: the ith adjustable resource power rating.

9. The regional energy internet uncertainty optimization scheduling method of claim 1, wherein the obtaining random variable data comprises: and acquiring wind power output, photovoltaic output and fluctuating load.

10. An optimized dispatching system for the regional energy internet uncertainty optimized dispatching method as claimed in any one of claims 1-9, characterized by comprising: the device comprises an acquisition module, a generation module, a reduction module, a processing combination module and a determination module;

the acquisition module is used for acquiring random variable data;

the generating module is used for generating a plurality of scenes based on the error probability distribution of the random variable;

the cutting module is used for cutting the scenes to obtain typical scenes;

the processing combination module is used for processing the typical scene probability to obtain a discrete combination optimization scene;

and the determining module is used for optimizing through a long-time uncertain optimization scheduling model according to the discrete combination optimization scene and correcting the optimization result according to a short-time rolling correction model.

11. The optimized scheduling system of claim 10, wherein the generating module comprises: dividing a submodule, a correction submodule and a sampling submodule;

a division submodule for discretizing the error into n intervals, the area enclosed by each interval is S_1,t,S_2,t…S_n,tAnd forming an error state vector at the current moment:

wherein S is_1,t,S_2,t…S_n,tRepresenting the probability of occurrence of each interval;

the correction submodule is used for correcting the error state vector of the random variable based on the Markov theory;

and the sampling submodule is used for randomly sampling the error state vector and converting an error interval corresponding to the selected error state vector into a plurality of scenes.

12. The optimized scheduling system of claim 10, wherein the pruning module includes a clustering submodule and a computation submodule;

the clustering submodule is used for carrying out scene reduction on the plurality of scenes by adopting a fuzzy c-means clustering algorithm according to the following formula:

1≤m≤∞

in the formula, U: membership matrix, V: center vector matrix, C: number of typical scenes, M₀: number of original scenes, V_c: center vector, X, of the c-th clustered scene set_i: ith original scene vector, μ_ci: membership function from ith scene vector to c clustering scene set, m: convergence factor, J: the degree of similarity of the scenes inside each category to the central vector;

the calculation submodule is used for calculating a central vector V of the c-th clustering scene set according to the following formula_cProbability:

in the formula, N_c: can be clustered into a central vector V_cThe number of scenes in the scene (c).

13. The optimal scheduling system of claim 12 wherein the processing combination module comprises,

a combining submodule, configured to calculate a probability of the discrete combination optimization scenario according to the following formula:

in the formula, I: number of optimal typical scenes of a photovoltaic, K: number of optimal typical scenes of the fan, J: number of load-optimized typical scenes, M: combining the number of scenes; delta_i ^PV: the probability of the photovoltaic power source occurring in the ith scene,

the probability of the fan occurring in the k-th scene,

probability of load occurrence of jth scene.

14. The optimized scheduling system of claim 10, wherein said determining module comprises: a long-time uncertain optimization scheduling model and a short-time rolling correction model;

the long-time uncertain optimization scheduling model is used for calculating an objective function with the lowest overall cost and the ideal distribution of the network node voltage;

and the short-time rolling correction model is used for calculating a target function of the deviation of the adjustable resource relative to the reference value.

Images