CN105391090B - A kind of intelligent grid multiple agent multiple target uniformity optimization method - Google Patents

Abstract

The invention discloses a smart grid multi-agent multi-objective consistency optimization method, which is characterized in that: according to the characteristics of the system of smart grid multi-agent cooperative control factors, the different typical characteristics of different power components are analyzed, and the respective proposed The objective requires that when the objectives are diverse, select the appropriate objective function to obtain the optimal operation control mode and parameters, ensure the reliability and economy of the system during operation, and verify the effectiveness of the optimal operation strategy. The present invention can not only establish a multi-agent optimization model according to the power network and load characteristics, use the multi-agent theory to study the distributed output optimization algorithm considering partial information sharing, but also analyze the convergence of the algorithm according to different communication topologies, and analyze the algorithm Carry out the simulation analysis of the example and study the relevant technologies to improve the convergence of the distributed algorithm.

Description

A multi-agent multi-objective consistency optimization method for smart grid

technical field

The invention belongs to the technical field of smart grid optimization and coordination dispatching, and relates to a smart grid optimization operation strategy for multi-agent multi-objective coordinated control, in particular to a smart grid multi-agent multi-objective consistency optimization method.

Background technique

Smart grid is an important branch of artificial intelligence, and it is the frontier subject of artificial intelligence in the world from the end of the 20th century to the beginning of the 21st century. With the rapid development of computer technology, artificial intelligence theory, control theory and the continuous exploration of modern science, smart grid has become one of the hot issues in different disciplines. The distributed collaborative control of the smart grid is of great significance to improve the reliability of the distribution network, improve the power quality, improve the operation economy of the distribution network, and optimize the operation arrangement of the distribution network.

Power balance control, that is, real-time economic dispatch, is a basic problem in the operation of power systems. It refers to the optimization problem of maximizing the economic benefits of the entire power system operation under the condition that generators and flexible loads meet a series of operating constraints. . Centralized optimization techniques have traditionally been used to solve economic dispatch problems, including classical optimization methods and modern artificial intelligence methods.

However, when the centralized optimization method is adopted, the system needs the dispatch center to issue instructions to dispatch all generators and flexible loads in the entire system, and the dispatch center needs to exchange information with each dispatch object. Moreover, the widespread penetration of flexible loads and the "plug and play" technology required by power components will make the topology of the power grid and communication network changeable, resulting in a high cost of communication topology construction for centralized optimization methods. Therefore, there is a need for more adaptable optimization algorithms that can operate efficiently under conditions of limited and unreliable communications or even failure of dispatch centers.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art, and provide a multi-objective consistency optimization method for smart grid multi-agents. According to the interaction requirements of the grid and loads, combined with the characteristics of different types of smart grid multi-agents, the smart grid multi-intelligence The coordination control model is established from the perspective of body multi-objective system consistency. The present invention can not only establish a multi-agent optimization model according to the power network and load characteristics, use the multi-agent theory to study the distributed output optimization algorithm considering partial information sharing, but also analyze the convergence of the algorithm according to different communication topologies, and analyze the algorithm Carry out the simulation analysis of the example and study the relevant technologies to improve the convergence of the distributed algorithm.

In order to solve the problems of the technologies described above, the present invention adopts the following technical solutions:

A smart grid multi-agent multi-objective consistency optimization method, characterized in that, according to the characteristics of the smart grid multi-agent collaborative control factor system, analyze the different typical characteristics of different power components, and their respective target requirements, when When the objectives are diverse, select the appropriate objective function to obtain the optimal operation control mode and parameters, ensure the reliability and economy of the system during operation, and verify the effectiveness of the optimal operation strategy; the implementation steps include:

Step 1. According to the network structure of the power system, establish a joint simulation platform based on MATLAB and NETLOGO, in which, the model of the power system components is established in MATLAB, and the general intelligent module representing the power system components is defined in NETLOGO. At the same time, MATLAB and NETLOGO are built The data exchange interface modules among them realize information interaction;

Step 2, for each load type, according to the load reference amount, electricity price, and the target tendency of each target corresponding to the load, respectively, establish a load-electricity price response characteristic model corresponding to each load and power source type; the load Including rigid loads and flexible loads, the power supply includes distributed power sources and energy storage elements; wherein, rigid loads refer to loads that do not participate in grid interaction, and flexible loads refer to loads that participate in grid interaction;

Step 3, according to the load-electricity price response characteristic models respectively corresponding to various load types established in the step 2, obtain the objective functions of each target of each load respectively; Perform weighting processing to obtain the total objective function corresponding to each load;

Step 4, randomly distribute the various loads on the NETLOGO three-dimensional layer to obtain the initial strategy of each load; for the network nodes in the NETLOGO three-dimensional layer, randomly set the electricity price, and establish a load agent;

Step 5, using the initial strategy of each load as the load reference amount, and respectively obtaining the strategy corresponding to each load in the way of +i or -i for the target tendency of each target of each load, and combining the initial strategy of each load The strategy constitutes the strategy set of each load; the i is the iterative step size of each step, and the step means that the strategy of the load changes once every time the electricity price changes;

Step 6, using the smart grid consistency optimization algorithm of multi-agent multi-objective coordinated control, respectively optimize and coordinate the total objective function of each load, and select the strategy to obtain the maximum total objective function value corresponding to each load, as each load load optimization strategy;

Let _xi represent the state of the power element. According to the consensus protocol, if and only if the state values of all nodes in the network topology are equal, the nodes of the network have reached the consensus, that is:

x ₁ =x ₂ =...=x _n

The smart grid consistency optimization algorithm of the multi-agent multi-objective coordinated control includes the following processes:

Assuming that the power generation cost function of the generator set and the power consumption benefit function of the flexible load are both quadratic functions, the power generation cost function of the generator set is as follows:

C _i (P _Gi )＝α _i +β _i P _Gi +γ _i P ² _Gi , i∈S _G

The power consumption benefit function of the flexible load is as follows:

B _i (P _Di )＝a _i +b _i P _Di +c _i P ² _Di ,i∈S _D

The economic dispatch problem refers to the optimization problem of maximizing the economic benefits of the entire power system operation under the conditions of generators and flexible loads satisfying a series of operating constraints, namely:

P _Gi,min ≤P _Gi ≤P _Gi,max , i∈S _G

P _Dj,min ≤P _Dj ≤P _Dj,max ，j∈S _D

Among them, P _Dj represents the demand power of flexible load j, P _Gi represents the output power of generator set i; S _G represents the set of generators, and S _D represents the set of flexible loads; using the classic Lagrangian multiplier method to solve, let λ Represents the Lagrangian multipliers corresponding to the equality constraints, regardless of the constraints, the above equality constraint optimization problem can be transformed into:

The optimality condition is obtained by taking partial derivatives of the variables P _Gi , P _Dj and λ, namely:

The above formula is the coordination equation, according to the coordination equation:

That is, the optimal solution of economic dispatch is to make the incremental cost of the generator equal to the incremental benefit of the flexible load, where m represents the number of generators, and k represents the number of flexible loads;

Assume that all flexible loads and generators operate within their power constraints; in this consensus algorithm, the IC of generators and the IB of flexible loads are defined as follows:

Select IC and IB as the consistency variables, apply the consistency algorithm, and update the IC of the follower generator set (FollowerGenerator) formula as follows:

The update formula of the IB of the follower load is:

In order to meet the power balance constraints in the power system, ΔP is used to represent the difference between the actual demand power of the flexible load and the actual output power of the generator set:

The update formula of the IC of the leader generator set (Leader Generator) is:

The update formula of the IB of the leader load is:

Where ε is the convergence coefficient, which is a positive scalar, which is related to the convergence speed of the distributed optimization of the main generator set and main load;

Step 7, move each load to the corresponding position in the NETLOGO three-dimensional layer respectively according to the target inclination degree of each target in the optimal strategy of each load, and update the target inclination degree of each target of each load; then According to the corresponding load-electricity price response characteristic model, the power of each load at this time is obtained, and combined with the jurisdiction of the load agent for the corresponding load, the total power of each load agent is respectively obtained;

Step 8: Send the total power of each load agent from NETLOGO to MATLAB, obtain the generator output and the electricity price corresponding to each network node in MATLAB, and return it to NETLOGO, and update the corresponding network node in the 3D layer of NETLOGO electricity price;

Step 9, using the electricity price on each network node in the NETLOGO three-dimensional layer as a traction signal, and each load agent releases the electricity price on the corresponding network node to each load under its jurisdiction;

Step 10, according to the position of each load in the NETLOGO three-dimensional layer when the step 9 is completed, and the target inclination of each target of each load, update the initial strategy of each load, and update the initial strategy of each load according to the method in the step 5. Describe the strategy set corresponding to each load, and then according to the total objective function corresponding to each load, combined with the electricity price corresponding to each load, obtain the total objective function value of each load corresponding to each strategy in its strategy set;

Step 11: For each load, judge whether the total objective function value corresponding to the initial policy of the load is greater than the total objective function value corresponding to other policies in its policy set, if yes, the load stops moving; otherwise, return to step 4.

In the step 1, the establishment of the co-simulation platform based on MATLAB and NETLOGO refers to:

A smart grid multi-agent simulation platform composed of MATLAB and NETLOGO, in which the calculation function and programming technology of MATLAB are used to establish the model of the power system components and the complex power network simulation model; and NETLOGO is a natural and social A programmable modeling environment for simulating phenomena, which is suitable for modeling complex systems that evolve over time; the NETLOGO completes the construction of general modules for power system components, and MATLAB performs various calculations for power systems to solve the obtained network parameters Information exchange is realized through the interface program between MATLAB and NETLOGO.

In the step 3, the total objective function corresponding to each flexible load is respectively obtained, and the process is:

Let the economic benefit B _k be the income of the power component, defined as follows:

Where E _k is the sum of net input and output, ρ _k is the price of electricity purchased by the load, D _k is the reference power of the load, B _k is the economic benefit,

μ _k is the tendency of economy, is the inclination degree of comfort, υ _k is the price of distributed power selling electricity, and G _k is the reference power of distributed power;

The comfort level of power components is defined as follows:

Among them, C _k is the comfort degree of power components;

The overall utility of power components is represented by the total objective function weighted by two objective functions, and the total objective function is defined as follows:

where R _k is the overall utility of the power element.

In the step 4, each load is randomly distributed on the NETLOGO three-dimensional layer to form a plurality of load nodes, and the initial target tendency degree of each target of each load is obtained, which is the initial strategy of each load, and the process for:

For the network nodes in the NETLOGO three-dimensional layer, randomly set the electricity price, and establish a load agent according to the load nodes in the NETLOGO three-dimensional layer, the number of the load agent is consistent with the number of load nodes, and the load agent In one-to-one correspondence with load nodes, each of the load agents governs each load, and each of the load agents is used for information transmission between each load under its jurisdiction and MATLAB.

In the step 5, the initial strategy of each load is used as the load reference amount, and the target inclination degree of each target of each load is respectively obtained by using +i or -i to obtain the strategy corresponding to each load, and Combine the initial policies of each load to form a policy set for each load:

Among them, i=1, on the NETLOGO three-dimensional level, there are eight points around each load, and the eight points are That is, each load corresponds to eight different policies, which respectively constitute the policy set of each load.

The realization process of described step 8 is:

The total power of each load agent is sent from NETLOGO to MATLAB through the data exchange interface module between MATLAB and NETLOGO, and the optimal power flow calculation is performed on the total power of each load agent in MATLAB to obtain the generator output And the electricity price corresponding to each network node, and return the electricity price of each network node to NETLOGO through the data exchange interface module between MATLAB and NETLOGO, and update the electricity price on the corresponding network node in the NETLOGO three-dimensional layer.

In the step 9, the use of the electricity price on each network node in the NETLOGO three-dimensional layer as a traction signal, and each load agent releases the electricity price on the corresponding network node to each load under its jurisdiction refers to:

Every time the power system dispatching platform runs in a fixed time period, at the end of each time period, it calculates the real-time electricity price, predicts the short-term electricity price, calculates the grid frequency and node voltage, and sends the electricity price of this period to each load agent and large load. Frequency, voltage, when necessary, send the historical and forecasted electricity price, frequency, and voltage before and after this period at the same time; the electricity price, frequency, and voltage are collectively called the traction signal, which guides each load to adjust its own electricity demand, and maximizes its own interests. At the same time serve the grid.

In said step 10, said respectively obtaining the total objective function value of each load corresponding to each strategy in its strategy set refers to:

For each load, judge whether the total objective function value corresponding to the initial policy of the load is greater than the total objective function value corresponding to other policies in its policy set, and if it is greater, the load stops moving.

Compared with prior art, the present invention contains following advantage and beneficial effect:

(1) According to the network structure of the power system, the present invention establishes a joint simulation platform based on MATLAB and NETLOGO, and builds a data exchange interface module between MATLAB and NETLOGO to realize information interaction. Agent characteristics, establish a coordination control model from the perspective of smart grid multi-agent multi-objective system consistency, select the appropriate objective function to obtain optimal operation control methods and parameters when the objectives are diverse, and ensure the reliability and economy of the system during operation. And verify the effectiveness of the optimization operation strategy;

(2) The present invention considers that the system containing flexible load multi-agent cooperative control has its unique characteristics, and the attributes of such models compete with each other for a game, so the smart grid distributed consistency optimization of multi-agent multi-objective coordinated control is adopted Algorithms and optimal operation strategies, on the basis of ensuring system reliability, enable the system to have a good optimal operation effect, and effectively verify the optimal operation results of multi-agent multi-objective coordinated control;

(3) The present invention can be widely applied to the distributed grid-load interactive multi-agent system control model, and is especially suitable for the smart grid multi-agent multi-objective consistency optimization method under flexible load.

Description of drawings

Fig. 1 is a flowchart of a smart grid multi-agent multi-objective consistency optimization method of the present invention.

Fig. 2 is a schematic diagram of the behavior space of load k normalization in the present invention.

Fig. 3 is the grid multi-agent simulation platform system based on NETLOGO of the present invention.

detailed description

The present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments.

FIG. 1 is a flowchart of a smart grid multi-agent multi-objective consistency optimization method of the present invention. The method of the present invention analyzes the different typical characteristics of different power components and their respective target requirements according to the characteristics of the system of smart grid multi-agent cooperative control factors, and selects an appropriate target function to obtain optimal operation control methods and parameters when the targets are diverse. , to ensure the reliability and economy of the system during operation, and to verify the effectiveness of the optimal operation strategy; its implementation steps include:

Step 1. According to the network structure of the power system, establish a joint simulation platform based on MATLAB and NETLOGO, in which, the model of the power system components is established in MATLAB, and the general intelligent module representing the power system components is defined in NETLOGO. At the same time, MATLAB and NETLOGO are built The data exchange interface modules among them realize information exchange.

The establishment of the joint simulation platform based on MATLAB and NETLOGO refers to: a smart grid multi-agent simulation platform composed of MATLAB and NETLOGO, wherein the calculation function and programming technology of MATLAB are used to establish the model and Establish a complex power network simulation model; and NETLOGO is a programmable modeling environment for simulating natural and social phenomena, suitable for modeling complex systems that evolve over time; the NETLOGO completes the general module of power system components Building, MATLAB performs various calculations of the power system, and the network parameters obtained through the solution realize information exchange through the interface program between MATLAB and NETLOGO.

The power grid multi-agent simulation platform composed of MATLAB and NETLOGO is shown in Figure 3. The power system dispatching platform and the NETLOGO multi-intelligence simulation platform realize multi-agent load control through the MATLAB interface. The power system dispatching platform is mainly responsible for electricity price calculation and forecasting, and at the same time carries out relevant power system dynamic simulation. Taking the load simulation based on the response electricity price as an example, the power system dispatching platform needs to calculate the optimal power flow to obtain the electricity price of the relevant nodes of the power grid at this time, and at the same time send the electricity price to the load agent in NETLOGO through the interface of NETLOGO and MATLAB. The NETLOGO simulation platform mainly completes the construction of the power grid operating environment and the modeling of the power grid components. Specifically, the three-tier power grid operating environment of topology, load agent and load group is built in NETLOGO; at the same time, according to the respective characteristics of the load group, in NETLOGO model its properties. The interface between NETLOGO and MATLAB is mainly for the data communication between NETLOGO and MATLAB of historical electricity price, real-time electricity price, predicted electricity price and related traction signals.

Step 2, for each load type, according to the load reference amount, electricity price, and the target tendency of each target corresponding to the load, respectively, establish a load-electricity price response characteristic model corresponding to each load and power source type; in the power system There are a variety of loads and power sources. The loads include rigid loads and flexible loads. The power sources include distributed power sources and energy storage components. Among them, rigid loads refer to loads that do not participate in grid interaction, and flexible loads refer to Loads that participate in grid interaction.

Different characteristics of load and power supply are modeled separately. The per unit value of electricity price for load power consumption and power generation is [ρ _k , υ _k ] respectively. In the simulation framework proposed by the present invention, the electricity price pair [ρ _k , υ _k ] is the traction signal issued by the Agent to each load. For loads on different buses, the tariffs for each pair may be different.

In the present invention, it is assumed that the loads and power sources used are aimed at pursuing economy and comfort. In the prior art, the electricity price response characteristic of the load and the power generation arrangement of the power source do not consider the comfort level. Therefore, after considering the comfort level, the load demand cannot be accurately predicted according to the traditional model, and the output of the power supply cannot be obtained according to the traditional generation arrangement method. The behaviors of loads and power sources pursuing their respective goals can be equivalent to the corresponding 2-D space: 1) The degree of inclination μ _k towards economy is manifested in avoiding the maximization of consumption cost on the one hand and maximizing economic benefits on the other hand. 2) Tendency towards comfort Representing individuals considering their own desires and wishes, they will use devices or equipment to meet their standard of living (physiological aspects). The target behavioral characteristics for each load are represented by To describe, A _k is given different values in terms of economy (shown in the abscissa of Figure 2) and comfort (shown in the ordinate of Figure 2).

The efficiency that a load obtains from electricity consumption can be quantified in terms of economic benefits and comfort, the value of which depends on the individual's behavior pattern in terms of power input and output. Here, the power input and output of the load and power supply are defined as:

1) Rigid load: the load q _k does not change with the electricity price, that is, the load that does not participate in the grid interaction;

2) Flexible load: refers to the load that participates in the grid interaction, Where d _k is the load demand, D _k is the reference power of the load, and ρ _k is the price of electricity purchased by the load;

3) Distributed power supply: Where g _k is the power generation of distributed power, G _k is the reference power of distributed power, and υ _k is the price of distributed power;

4) Energy storage element: when charging

when discharging

The network output power of the load is calculated from the demand of the reference power. The reference power of the load in this model is constant and does not involve technical aspects. The load-price response characteristic is determined by the state parameter μ _k in the formula, decision, where the rate of decrease in demand and growth rate of production associated with prices ρ _k , υ _k are This results in an elastic change in power. Furthermore, in contrast to traditional fixed-response based models, in our approach social behavior is explicitly modeled, taking into account elastic changes due to social behavior interactions.

The state position of the load in space is shown in Figure 2. If the load k is at the position A _k , it means that the load only considers the economic benefit: the increase of ρ _k (υ _k ) will lead to the decrease or increase of the power output. Position B means that it cannot change the output or input of electric energy according to the price change, it considers the comfort. In contrast to these two cases, the price of the location not on the border is related to a certain degree of economic interest and comfort. For example, load k is located at point C. From the perspective of economy, when the price ρ _k =1, the power demand will be reduced by 0.3, and from the perspective of comfort, the power consumption will be reduced by 0.3. This means that the final demand will decrease by 0.3*(1-0.3) at the price ρ _k =1. Similarly, at point D, the final demand will decrease by 0.7*(1-0.7) when the price ρ _k =1. When the price υ _k =1, the final power generation will be reduced by 0.7*(1-0.7). Theoretically speaking, ρ _k and υ _k can be set independently. But in order to prevent load arbitrage, it is assumed that ρ _k = -υ _k .

As the load gain, the economic benefit _Bk is defined as follows:

Where E _k is the sum of net input and output.

Load comfort is defined as follows:

The overall utility of the load is represented by the total objective function weighted by two objective functions. The total objective function is defined as follows:

Step 3, according to the load-electricity price response characteristic models respectively corresponding to various load types established in the step 2, obtain the objective functions of each target of each load respectively; Perform weighting processing to obtain the total objective function corresponding to each load.

The described total objective function corresponding to each flexible load is respectively obtained, and the process is:

Let the economic benefit B _k be the income of the power component, defined as follows:

Where E _k is the sum of net input and output, ρ _k is the price of electricity purchased by the load, D _k is the reference power of the load, B _k is the economic benefit,

μ _k is the tendency of economy, is the inclination degree of comfort, υ _k is the price of distributed power selling electricity, and G _k is the reference power of distributed power;

The comfort level of power components is defined as follows:

Among them, C _k is the comfort degree of power components;

The overall utility of power components is represented by the total objective function weighted by two objective functions, and the total objective function is defined as follows:

where R _k is the overall utility of the power element.

Step 4. Randomly distribute the various loads on the NETLOGO three-dimensional layer to obtain the initial strategy of each load; for the network nodes in the NETLOGO three-dimensional layer, randomly set the electricity price and establish a load agent. The process is as follows:

For the network nodes in the NETLOGO three-dimensional layer, randomly set the electricity price, and establish a load agent according to the load nodes in the NETLOGO three-dimensional layer, the number of the load agent is consistent with the number of load nodes, and the load agent In one-to-one correspondence with load nodes, each of the load agents governs each load, and each of the load agents is used for information transmission between each load under its jurisdiction and MATLAB.

Step 5, using the initial strategy of each load as the load reference amount, and respectively obtaining the strategy corresponding to each load in the way of +i or -i for the target tendency of each target of each load, and combining the initial strategy of each load The strategy constitutes the strategy set of each load; the i mentioned above is the iteration step size of each step, and the step means that every time the electricity price changes, the strategy of the load changes correspondingly once.

In the step 5, the initial strategy of each load is used as the load reference amount, and the target inclination degree of each target of each load is respectively obtained by using +i or -i to obtain the strategy corresponding to each load, and Combine the initial policies of each load to form a policy set for each load:

Among them, i=1, on the NETLOGO three-dimensional level, there are eight points around each load, and the eight points are That is, each load corresponds to eight different policies, which respectively constitute the policy set of each load.

Step 6, using the smart grid consistency optimization algorithm of multi-agent multi-objective coordinated control, respectively optimize and coordinate the total objective function of each load, and select the strategy to obtain the maximum total objective function value corresponding to each load, as each load Optimal strategy for loading.

Let _xi represent the state of the power element. According to the consensus protocol, if and only if the state values of all nodes in the network topology are equal, the nodes of the network have reached the consensus, that is:

x ₁ =x ₂ =...=x _n

The smart grid consensus optimization algorithm for multi-agent multi-objective coordinated control adopts a distributed economic scheduling strategy, and refers to:

Under flexible load, the goal of power system economic dispatch is to maximize social welfare. From the perspective of distributed optimization, the consensus algorithm is applied, and the incremental cost (IC) of the generator set and the incremental benefit (IB) of the flexible load are taken as the consistency variables, and the economic dispatch problem is solved by distributed optimization. The local controller embedded in each genset and flexible load updates its own incremental cost or incremental benefit according to the neighbor's incremental cost or incremental benefit. Choose a "Host Group" and "Primary Load" decision whether to increase or decrease the global incremental cost and incremental benefit. When the total generating power of the generator is greater than the total demand power of the load, the overall incremental cost will be reduced, and vice versa. When the total demand power of the load is greater than the total power generated by the generator, the overall incremental benefit will be increased, and vice versa.

The algorithm includes the following processes:

Assuming that the power generation cost function of the generator set and the power consumption benefit function of the flexible load are both quadratic functions, the power generation cost function of the generator set is as follows:

C _i (P _Gi )＝α _i +β _i P _Gi +γ _i P ² _Gi , i∈S _G

The power consumption benefit function of the flexible load is as follows:

Bj(P _Dj )＝a _j +b _j P _Dj +c _j P ² _Dj ,j∈S _D

The economic dispatch problem refers to the optimization problem of maximizing the economic benefits of the entire power system operation under the conditions of generators and flexible loads satisfying a series of operating constraints, namely:

P _Gi,min ≤P _Gi ≤P _Gi,max , i∈S _G

P _Dj,min ≤P _Dj ≤P _Dj,max ，j∈S _D

Among them, P _Dj represents the demand power of flexible load j, and P _Gi represents the output power of generator set i. S _G represents the set of generators, and S _D represents the set of flexible loads. Using the classic Lagrange multiplier method to solve, let λ represent the Lagrangian multiplier corresponding to the equality constraint, regardless of the constraint, the above equality constraint optimization problem can be transformed into:

The optimality condition is obtained by taking partial derivatives of the variables P _Gi , P _Dj and λ, namely:

The above formula is the coordination equation, according to the coordination equation:

That is, the optimal solution of economic dispatch is to make the incremental cost of the generator equal to the incremental benefit of the flexible load, where m represents the number of generators and k represents the number of flexible loads.

It is assumed that all flexible loads and generators operate within their power constraints. In this consensus algorithm, the IC of the generating set and the IB of the flexible load are defined as follows:

Select IC and IB as the consistency variables, apply the consistency algorithm, and update the IC of the follower generator set (FollowerGenerator) formula as follows:

The update formula of the IB of the follower load is:

In order to meet the power balance constraints in the power system, ΔP is used to represent the difference between the actual demand power of the flexible load and the actual output power of the generator set:

The update formula of the IC of the leader generator set (Leader Generator) is:

The update formula of the IB of the leader load is:

Where ε is the convergence coefficient, which is a positive scalar. It is related to the convergence speed of the distributed optimization of main generating units and main loads.

Step 7, move each load to the corresponding position in the NETLOGO three-dimensional layer respectively according to the target inclination degree of each target in the optimal strategy of each load, and update the target inclination degree of each target of each load; then According to the corresponding load-electricity price response characteristic model, the power of each load at this time is obtained, and combined with the jurisdiction of the load agent for the corresponding load, the total power of each load agent is respectively obtained.

Step 8: Send the total power of each load agent from NETLOGO to MATLAB, obtain the generator output and the electricity price corresponding to each network node in MATLAB, and return it to NETLOGO, and update the corresponding network node in the 3D layer of NETLOGO electricity price. Its implementation process is:

The total power of each load agent is sent from NETLOGO to MATLAB through the data exchange interface module between MATLAB and NETLOGO, and the optimal power flow calculation is performed on the total power of each load agent in MATLAB to obtain the generator output And the electricity price corresponding to each network node, and return the electricity price of each network node to NETLOGO through the data exchange interface module between MATLAB and NETLOGO, and update the electricity price on the corresponding network node in the NETLOGO three-dimensional layer.

The signals transmitted by the interface between MATLAB and NETLOGO are P, f, V and C etc. among the present invention. Taking C as an example, the C obtained after MATLAB solves the electricity price is stored in the bus matrix. In the three-machine nine-node system, the electricity price information is stored in the 14th column of the Bus matrix. The Bus matrix is shown in Table 1.

Table 1

Continuing from the table

When NETLOGO calls the C of the load node in MATLAB, it must first determine the position of the node's electricity price in the bus matrix, and the Agent at the node gets the electricity price of the node through the interface command statement. On the other hand, Agents also need to call MATLAB with the total power of the load group through the command statement, and replace the load in the electricity price calculation with the existing value to recalculate the electricity price. Its code description is shown in Table 2.

Table 2

Step 9: Use the electricity price on each network node in the NETLOGO three-dimensional layer as a traction signal, and each load agent releases the electricity price on the corresponding network node to each load under its jurisdiction. Refers to:

Every time the power system dispatching platform runs in a fixed time period, at the end of each time period, it calculates the real-time electricity price, predicts the short-term electricity price, calculates the grid frequency and node voltage, and sends the electricity price of this period to each load agent and large load. Frequency, voltage, when necessary, send the historical and forecasted electricity price, frequency, and voltage before and after this period at the same time; the electricity price, frequency, and voltage are collectively called the traction signal, which guides each load to adjust its own electricity demand, and maximizes its own interests. At the same time serve the grid.

In the NETLOGO three-dimensional interface, the origin is located in the southwest corner, the horizontal direction represents the economic tendency, the vertical direction represents the comfort tendency, and the value range is 0-1. Each user in the user layer moves left and right, up and down on this layer Respectively represent the change of the economic tendency and the comfort tendency. While moving, the load is constantly changing until it finally reaches a point where the total target is the largest.

Step 10, according to the position of each load in the NETLOGO three-dimensional layer when the step 9 is completed, and the target inclination of each target of each load, update the initial strategy of each load, and update the initial strategy of each load according to the method in the step 5. Describe the strategy set corresponding to each load, and then according to the total objective function corresponding to each load, combined with the electricity price corresponding to each load, obtain the total objective function value of each load corresponding to each strategy in its strategy set;

The said obtaining the total objective function value of each load corresponding to each strategy in its policy set refers to: for each load, judge whether the total objective function value corresponding to the initial policy of the load is greater than the total target corresponding to other strategies in the policy set Function value, if greater than, the load will stop moving.

Step 11: For each load, judge whether the total objective function value corresponding to the initial policy of the load is greater than the total objective function value corresponding to other policies in its policy set, if yes, the load stops moving; otherwise, return to step 4.

To sum up, the system power balance control of the current multi-agent interactive grid has become a current research hotspot. The variability and uncertainty of the multi-agent interactive grid make it particularly difficult to control, and the method based on multi-agent multi-objective consistency can effectively deal with it. In the designed grid multi-agent control method, the modules composed of each power element have a certain degree of intelligence, can respond to external disturbances, and make positive responses, and at the same time realize self-adjustment through communication with surrounding modules to achieve a certain degree Autonomy, realize real-time scheduling and distributed scheduling, so as to improve the reliability and economy of power grid operation. Therefore, it is feasible to study the power components participating in the system power balance as a multi-agent system.

In order to realize the power balance control of the power system with multi-agent interaction, the present invention adopts a grid multi-agent modeling, simulation and control scheme. The simulation platform of this scheme is composed of NETLOGO and MATLAB. NETLOGO is responsible for the modeling of power system intelligent components and the work of multi-agent control of the power grid. MATLAB is responsible for various calculations of the power system. The entire system network is realized through the interface module between NETLOGO and MATLAB. Data interaction. In the simulation scheme, the power component agent and MATLAB transmit interactive information through the interface. The power component agent uploads relevant parameters to MATLAB for various calculations of the power system, and at the same time calls the latest signal in MATLAB to issue to the power component agent. Each power component agent Consider your own goals and respond positively.

The invention considers the characteristics of flexible loads. The loads are under different electricity prices, with the goal of pursuing economy and comfort, researching the distributed consistency optimization scheduling strategy, performing optimization and coordination operations, so that the system performance can be coordinated to achieve the optimal effect, and verifying the optimal operation. effectiveness of the strategy. By establishing the coordinated control model of smart grid multi-agent multi-objective system, indirect and distributed control is realized. Accurately describe the operating characteristics of power grid components and systems in the complex network theory system of smart grid, describe flexible loads in the form of agents, and obtain the multi-agent environment of grid interactive operation based on flexible load information interaction, and the interactive operation of grid components based on traction control. Coordinate the control strategy and the behavior specification of the autonomous operation of grid components based on the code of conduct, adopt appropriate algorithms and optimal operation strategies, and ensure the system has a good optimal operation effect on the basis of ensuring system reliability, and verify multi-agent multi-objective Coordinated control optimizes operating results.

Claims (8)

1. a kind of intelligent grid multiple agent multiple target uniformity optimization method, it is characterised in that how intelligent according to intelligent grid The characteristics of system of body Collaborative Control factor, analyze the different characteristic features of different force devices, and the mesh each proposed Mark is required, appropriate object function is chosen when target is various and obtains optimization operation control method and parameter, it is ensured that system operation When reliability and economy, and verify the validity of Optimal Operation Strategies；Implementation step includes：

Step 1, according to power system network structure, the union simulation platform based on MATLAB and NETLOGO is established, wherein, Power system component model is established in MATLAB, the intelligent body general module of power system component is represented defined in NETLOGO, Meanwhile build the data exchange interface module between MATLAB and NETLOGO and realize information exchange；

Step 2, for various load types, respectively according to the target of each target of load datum quantity, electricity price, and corresponding load Tendency degree, establish the load-Respondence to the Price of Electric Power characteristic model for corresponding respectively to various loads and power supply type；Described load includes Rigid load and flexible load, described power supply include distributed power source and energy-storage travelling wave tube；Wherein, rigid load refers to be not involved in The interactive load of power network, flexible load refer to participate in the interactive load of power network；

Step 3, according to the load for the corresponding to various load types respectively-Respondence to the Price of Electric Power characteristic model established in the step 2, divide The object function of each target of each load is not obtained；And each load is directed to respectively, by the mesh of each target of load Scalar functions are weighted processing, obtain the catalogue scalar functions of corresponding each load respectively；

Step 4, described each load is randomly dispersed in NETLOGO three-dimensional aspects, obtains the initial policy of each load； For the network node in NETLOGO three-dimensional aspects, electricity price is set at random, and establishes load agency；

Step 5, using the initial policy of each load as load datum quantity, respectively for each target of each load Goal orientation degree, obtain strategy corresponding to each load respectively by the way of+i or-i, and combine the initial plan of each load Slightly form the set of strategies of each load；Described i is each step iteration step length, and described step refers to that electricity price often changes once, is born The tactful respective change of lotus is once；

Step 6, using the intelligent grid uniformity optimized algorithm of the multi-objective coordinated control of multiple agent, respectively to each load Catalogue scalar functions optimize coordination computing, and selection obtains the plan that each load corresponds to its maximum general objective functional value respectively Slightly, the preference policy as each load；

Make x_iThe state of force device is represented, according to consistency protocol, the state value of and if only if network opens up all nodes of bowl spares When all equal, unanimously, i.e., the node of the network has all reached：

x₁=x₂=...=x_n

The intelligent grid uniformity optimized algorithm of the multi-objective coordinated control of described multiple agent, including procedure below：

Assuming that the cost of electricity-generating function of generating set and the electricity consumption benefit function of flexible load are quadratic function, generating set Cost of electricity-generating function is as follows：

C_i(P_Gi)=α_i+β_iP_Gi+γ_iP² _Gi, i ∈ S_G

The electricity consumption benefit function of flexible load is as follows：

B_i(P_Di)=a_i+b_iP_Di+c_iP² _Di,i∈S_D

Economic Dispatch Problem refers to that generator and flexible load under conditions of a series of operation constraints are met, make whole power train The optimization problem of the maximization of economic benefit of system operation, i.e.,：

<mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>B</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>C</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow>

<mrow> <mi>S</mi> <mo>.</mo> <mi>T</mi> <mo>.</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>=</mo> <mn>0</mn> </mrow>

P_Gi,min≤P_Gi≤P_Gi,max, i ∈ S_G

P_Dj,min≤P_Dj≤P_Dj,max, j ∈ S_D

Wherein, P_DjRepresent flexible load j demand power, P_GiRepresent generating set i power output；S_GRepresent generator collection Close, S_DRepresent flexible load set；Solved using the method for Lagrange multipliers of classics, make λ represent draw corresponding with equality constraint Ge Lang multipliers, do not consider to constrain, above-mentioned RegionAlgorithm for Equality Constrained Optimization can be converted into：

<mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> <mi>&amp;Gamma;</mi> <mo>=</mo> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>B</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>+</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>C</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>+</mo> <mi>&amp;lambda;</mi> <mrow> <mo>(</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow>

To variable P_Gi, P_DjLocal derviation is asked to obtain optimality condition with λ, i.e.,：

<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mfrac> <mrow> <mo>&amp;part;</mo> <mi>&amp;Gamma;</mi> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>C</mi> <mi>i</mi> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> </mrow> </mfrac> <mo>-</mo> <mi>&amp;lambda;</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mi>i</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mfrac> <mrow> <mo>&amp;part;</mo> <mi>&amp;Gamma;</mi> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>B</mi> <mi>j</mi> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> </mrow> </mfrac> <mo>+</mo> <mi>&amp;lambda;</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mfrac> <mrow> <mo>&amp;part;</mo> <mi>&amp;Gamma;</mi> </mrow> <mrow> <mo>&amp;part;</mo> <mi>&amp;lambda;</mi> </mrow> </mfrac> <mo>=</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> <mo>=</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced>

Above formula is the equation of comptability, can be obtained according to the equation of comptability：

<mrow> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>C</mi> <mn>1</mn> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mn>1</mn> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>C</mi> <mn>2</mn> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mn>2</mn> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mn>...</mn> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>C</mi> <mi>m</mi> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>m</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>B</mi> <mn>1</mn> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mn>1</mn> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>B</mi> <mn>2</mn> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mn>2</mn> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mn>...</mn> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>B</mi> <mi>k</mi> </msub> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>k</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <mi>&amp;lambda;</mi> </mrow>

That is the optimal solution of economic load dispatching is to make the incremental cost of generator equal with the increment benefit of flexible load, and wherein m is represented Generator number, k represent the number of flexible load；

Assuming that all flexible loads are run with generating set in its power constraints；In the consistency algorithm, hair The IC of group of motors and the IB of flexible load are defined as follows：

<mrow> <msub> <mi>IC</mi> <mi>i</mi> </msub> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>C</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <msub> <mi>&amp;lambda;</mi> <mi>i</mi> </msub> <mo>,</mo> <mi>i</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow>

<mrow> <msub> <mi>IB</mi> <mi>j</mi> </msub> <mo>=</mo> <mfrac> <mrow> <mo>&amp;part;</mo> <msub> <mi>B</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mo>&amp;part;</mo> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> </mrow> </mfrac> <mo>=</mo> <msub> <mi>&amp;lambda;</mi> <mi>j</mi> </msub> <mo>,</mo> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow>

IC and IB is selected as uniformity variable, using consistency algorithm, from generating set (Follower Generator) IC more new formula is：

<mrow> <msub> <mi>&amp;lambda;</mi> <mi>i</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>&amp;rsqb;</mo> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>I</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>a</mi> <mrow> <mi>i</mi> <mi>I</mi> </mrow> </msub> <msub> <mi>&amp;lambda;</mi> <mi>I</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>&amp;rsqb;</mo> <mo>,</mo> <mi>i</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow>

More new formula from the IB of load (Follower Load) is：

<mrow> <msub> <mi>&amp;lambda;</mi> <mi>j</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>&amp;rsqb;</mo> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>I</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>a</mi> <mrow> <mi>j</mi> <mi>I</mi> </mrow> </msub> <msub> <mi>&amp;lambda;</mi> <mi>I</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>&amp;rsqb;</mo> <mo>,</mo> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow>

In order to meet the power-balance constraint in power system, flexible load actual demand power and generating set are represented with Δ P Difference between real output：

<mrow> <mi>&amp;Delta;</mi> <mi>P</mi> <mo>=</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>D</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> </munder> <msub> <mi>P</mi> <mrow> <mi>G</mi> <mi>i</mi> </mrow> </msub> </mrow>

The IC of main generator group (Leader Generator) more new formula is：

<mrow> <msub> <mi>&amp;lambda;</mi> <mi>i</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>&amp;rsqb;</mo> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>f</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>a</mi> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </msub> <msub> <mi>&amp;lambda;</mi> <mi>f</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>&amp;rsqb;</mo> <mo>+</mo> <mi>&amp;epsiv;</mi> <mi>&amp;Delta;</mi> <mi>P</mi> <mo>,</mo> <mi>i</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>G</mi> </msub> </mrow> 2

The IB of main load (Leader Load) more new formula is：

<mrow> <msub> <mi>&amp;lambda;</mi> <mi>j</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>&amp;rsqb;</mo> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>a</mi> <mrow> <mi>j</mi> <mi>t</mi> </mrow> </msub> <msub> <mi>&amp;lambda;</mi> <mi>t</mi> </msub> <mo>&amp;lsqb;</mo> <mi>k</mi> <mo>&amp;rsqb;</mo> <mo>+</mo> <mi>&amp;epsiv;</mi> <mi>&amp;Delta;</mi> <mi>P</mi> <mo>,</mo> <mi>j</mi> <mo>&amp;Element;</mo> <msub> <mi>S</mi> <mi>D</mi> </msub> </mrow>

Wherein ε is convergence coefficient, is a positive scalar, it with main generator group and main load distributed optimization convergence rate It is relevant；

Step 7, the goal orientation degree of each target in the preference policy of described each load respectively, by each load Move to respectively in NETLOGO three-dimensional aspects on corresponding position, and update the goal orientation degree of each target of each load； Then load-Respondence to the Price of Electric Power characteristic model corresponding to, the power of now each load is obtained, and pin is acted on behalf of with reference to load Administration to corresponding load, the general power of each load agency is obtained respectively；

Step 8, the general power that described each load is acted on behalf of is sent into MATLAB by NETLOGO, obtained in MATLAB The electricity price of generator output and corresponding each network node, and be back in NETLOGO, it is right in NETLOGO three-dimensional aspects to update Answer the electricity price on network node；

Step 9, using the electricity price in described NETLOGO three-dimensional aspects on each network node as traction signal, and respectively by Electricity price on map network node is distributed to each load of its administration by described each load agency；

Step 10, the position of each load when being completed according to the step 9 in NETLOGO three-dimensional aspects, and each load Each target goal orientation degree, update the initial policy of each load, and according to the method in the step 5, update institute Set of strategies corresponding to each load is stated, it is electric with reference to corresponding to each load then according to the catalogue scalar functions of corresponding each load Valency, each load is obtained respectively and corresponds to each tactful general objective functional value in its set of strategies；

Step 11, judge whether general objective functional value corresponding to the initial policy of load is more than its plan for each load respectively The general objective functional value corresponding to other strategies is slightly concentrated, is the then load stop motion；Otherwise return to step 4.

A kind of 2. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that In the step 1, union simulation platform of the described foundation based on MATLAB and NETLOGO, refer to：

A kind of intelligent grid Multi-Agent simulation platform being made up of MATLAB and NETLOGO, wherein the calculating using MATLAB Function and programming technique, to establish the model of power system component and establish complicated electric power networks simulation model；And NETLOGO It is a programmable modeling environment emulated to nature and social phenomenon, suitable for the complication system progress to Temporal Evolution Modeling；Described NETLOGO completes building for power system component general module, and MATLAB carries out every meter of power system Calculate, the network parameter for solving to obtain realizes information exchange by the interface routine between MATLAB and NETLOGO.

A kind of 3. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that Described to obtain the catalogue scalar functions for corresponding to each flexible load respectively in the step 3, its process is：

If economic benefit κ₁=1 income as force device, it is defined as follows：

Wherein E_kFor the summation of net input and output, ρ_kThe price of electricity, D are bought for load_kFor load reference power, κ₁=1 is economic effect Benefit, μ_kFor the tendency degree of economy,For the tendency degree of comfort level, υ_kThe price of electricity, G are sold for distributed power source_kFor distributed electrical Source reference power；

It is as follows to define force device comfort level：

Wherein C_kFor force device comfort level；

The overall utility of force device is weighted to obtain general objective function representation by two object functions, and catalogue scalar functions define such as Under：

Wherein R_kFor the overall utility of force device.

A kind of 4. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that It is described that each load is randomly dispersed in NETLOGO three-dimensional aspects in the step 4, obtain the initial of each load Strategy；For the network node in NETLOGO three-dimensional aspects, electricity price is set at random, and establishes load agency, and its process is：

For the network node in described NETLOGO three-dimensional aspects, electricity price is set at random, and according to NETLOGO three-dimension layers Load bus in face, load agency is established, the quantity of described load agency is consistent with the quantity of load bus, and described is negative Lotus is acted on behalf of to be corresponded with load bus, the corresponding each load of described each load agency's administration, and described each negative Lotus agency is respectively used to the information transfer between each load and MATLAB of its administration.

A kind of 5. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that In the step 5, the initial policy using each load is as load datum quantity, respectively for each of each load The goal orientation degree of target, obtain strategy corresponding to each load respectively by the way of+i or-i, and combine each load Initial policy forms the set of strategies of each load：

Wherein, i=1, in NETLOGO three-dimensional aspects, eight points are included around each load, eight points are [μ respectively_k+ 1,φ_k], [μ_k-1,φ_k], [μ_k,φ_k+ 1], [μ_k,φ_k- 1], [μ_k+1,φ_k+ 1], [μ_k-1,φ_k- 1], Δ P, ε, i.e., each The corresponding eight different strategies of load, respectively constitute the set of strategies of each load.

A kind of 6. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that The implementation process of the step 8 is：

By the general power that described each load is acted on behalf of by the data exchange interface module between MATLAB and NETLOGO, by NETLOGO is sent into MATLAB, and optimal load flow calculating is carried out for the general power of each load agency respectively in MATLAB, Obtain the electricity price of generator output and corresponding each network node, and by the electricity price of each network node, by MATLAB with Data exchange interface module between NETLOGO is back in NETLOGO, updates map network section in NETLOGO three-dimensional aspects Electricity price on point.

A kind of 7. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that In the step 9, the electricity price in the three-dimensional aspect using NETLOGO on each network node is divided as traction signal Each load that the electricity price on map network node is distributed to its administration is not acted on behalf of by each load, is referred to：

Electric power system dispatching platform is often run once with a fixed time period, and electricity in real time is calculated at each period end Valency, prediction electricity price, calculating mains frequency and node voltage, and period electricity price, the frequency is issued to each load agency, big load in short-term Rate, voltage, it is necessary to when issue history and forecasted electricity market price, frequency, voltage before and after the period simultaneously；The electricity price, frequency, voltage Traction signal is referred to as, instructs to draw each load adjustment itself power demand, is served while number one is maximized Power network.

A kind of 8. intelligent grid multiple agent multiple target uniformity optimization method according to claim 1, it is characterised in that In the step 10, described each load that obtains respectively corresponds to each tactful general objective functional value in its set of strategies, is Refer to：

Each load is directed to respectively, judges whether general objective functional value corresponding to the initial policy of load is more than its in its set of strategies General objective functional value corresponding to its strategy, if greater than the then load stop motion.