A Method and Device for Optimizing Matching and Coordinating Switching Control of Distributed Controllable Resource Operating Modes under Different Power Deficits

Abstract

The present invention discloses a method for optimizing matching and coordinated switching control of operating modes of distributed controllable resources based on a MAS hierarchical architecture for different power deficit situations in a distribution network. First, for DERs with unique operating characteristics and complex working modes, based on hybrid systems and finite automata, combined with the power characteristics of photovoltaic, wind power, energy storage units, fuel cells, and loads, five hybrid automata models of DERs under hybrid systems are respectively constructed at the bottom layer. Secondly, in the face of the complex operating data generated during the operation of the distribution network, the selective pattern-based aggregated single-layer dependence classification algorithm is applied in the middle layer to process the data, and the best coordinated switching control command is selected from the control command set by performing classification operations, thereby effectively improving the intelligence of the overall system. Finally, at the upper layer, considering constraints such as system load demand, power balance, and DER power output limits, an optimal scheduling model considering system operating costs, environmental benefits, and power quality is proposed, and by implementing an improved fruit fly optimization algorithm with random walk, flexible scheduling and optimal operation of each unit of the system are realized.

Description

Technical Field

[0001] The present invention relates to a method and device for optimizing the matching and coordinating switching control of the operation modes of distributed controllable resources under different power deficits, and belongs to the technical field of smart grids. Background Technique

[0002] With the increasingly prominent global energy crisis and environmental problems, establishing a more secure, efficient and sustainable energy development and utilization model has become the main form and important development direction of the future energy system. The shortage of traditional fossil energy and the environmental pollution problems caused by combustion have become the main bottlenecks restricting the development of the world economy. The application of distributed generation technologies such as wind energy, solar energy, and energy storage provides an effective way to solve the above problems. At present, China is vigorously promoting the utilization of DERs (distributed energy resources), mainly including photovoltaic power generation, wind power generation, storage batteries, fuel cells, etc. The integration of a large number of DERs with different power output characteristics into the distribution network will cause a series of problems, such as power fluctuations and power quality degradation. Therefore, realizing the flexible control and coordinated control of DERs operating in different working modes is of great significance for improving the safe, stable and economic operation of the distribution network.

[0003] The optimal operation of the distribution network needs to be realized through the multi-modal switching of DERs. The operation characteristics of various types of distributed controllable resources are different and the working modes are complex. How to add a description of the dynamic operation characteristics of DERs to the control strategy to achieve discrete and continuous control during the switching process is crucial for realizing the multi-modal switching of DERs. Generally, a dynamic system composed of the interaction of continuous variables and discrete events is regarded as a hybrid system, and the distribution network is a typical hybrid system. To sum up, the management object of the traditional distribution network energy management system is single and the structure is simple, and its control means cannot realize the flexible control of various types of DERs. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method and device for optimizing the matching and coordinating switching control of the operation modes of distributed controllable resources under different power deficits, which can realize the flexible regulation of a variety of distributed controllable resources under different power deficits, with high intelligence and high accuracy.

[0005] To achieve the above object, the present invention is implemented by the following technical solutions:

[0006] In the first aspect, the present invention provides a method for optimizing the matching and coordinating switching control of the operation modes of distributed controllable resources under different power deficits, including the following steps:

[0007] Step 1: For the coordinated control objective of the distribution network, establish a hierarchical architecture of the distribution network based on MAS (multi-agent system, MAS), and divide the distribution network control structure into the upper layer, the middle layer, and the bottom layer;

[0008] Step 2: Considering that in the control process, there are both discrete control commands for switching the operation modes of different distributed energy resources (DERs) and load units, and continuous control instructions for the execution-end inverters. Therefore, based on the concept of finite automata, continuous variables are added to construct a hybrid automaton model represented by a 6-tuple array;

[0009] Step 3: For the operating characteristics of photovoltaic power generation units, wind power generation units, energy storage units, fuel cells, and loads, establish corresponding hybrid automaton models respectively;

[0010] Step 4: Divide the local control strategy at the bottom layer into an internal switching control strategy and a continuous dynamic management strategy. Among them, the internal switching control strategy is based on the coordinated control command of the middle layer, and the continuous dynamic management is designed as a dual-loop control method based on SPWM;

[0011] Step 5: Design a coordinated control strategy in the multi-agent structure of the middle layer, construct a system voltage security evaluation index and perform hierarchical processing. According to the system voltage security evaluation results, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units;

[0012] Step 6: To minimize the number of switching times of the DERs operation modes and improve the intelligence of the system and the rapidity and accuracy of the execution of control commands, use the aggregation single-layer dependence classification algorithm based on selective mode, use the coordinated control command set as the label set, and select the best coordinated switching control command in the control command set by performing classification operations;

[0013] Step 7: The upper-layer agent constructs a comprehensive objective function for reducing operating costs, reducing pollutant emissions, and improving power quality considering different power deficits, power balance, the output power of DERs, and the battery capacity constraint. Use the improved fruit fly optimization algorithm of random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy.

[0014] Furthermore, in Step 1, to achieve multi-modal switching, establish a hierarchical structure based on MAS, including:

[0015] Step 1.1 Upper-layer agent - negotiation agent. Achieve energy management through the optimization process to enable the system to obtain the maximum economic, environmental, and power quality benefits. The optimization model is designed based on the real-time data in the processing module. The time scale of the optimization process is once per hour.

[0016] Step 1.2 Intermediate layer agent - negotiation agent. Coordinate the switching of the working modes of DERs to ensure the safety of the system voltage. The time scale of the switching control is at the hour or minute level.

[0017] Step 1.3 Bottom layer agent - hybrid agent, including a reaction layer and a negotiation layer. The reaction layer is set to "perception - action" and has priority to quickly respond to emergencies. The negotiation layer is highly intelligent and can control or guide the behavior of the agent. The negotiation layer is set to "belief - desire - intention", which is highly intelligent and controls or plans the behavior of the agent to achieve its desires or intentions.

[0018] Furthermore, in Step 2, considering that there are both discrete values and continuous values in the process of issuing control commands, on the basis of the concept of finite automata, continuous variables are added to construct a hybrid automata model represented by a 6 - element array.

[0019] As one of the modeling methods for hybrid systems, automata have the advantage of intuitiveness. Ordinary automata models are used to process discrete events. If continuous dynamic characteristics are added to them, a hybrid automata model will be formed. A hybrid automata mainly consists of a discrete state space, enabled states, and continuous states. In each discrete state space, there are corresponding continuous state changes. When a certain enabled state is satisfied, the hybrid automata will migrate from one discrete state space to another discrete state space, and the corresponding continuous states will also change.

[0020] A typical hybrid automata model can be represented by a 6 - element array: H=(D, L, f, S, F, I). Among them, D = {δ1, δ2,...} represents a set of a series of discrete state spaces; L represents a set of a series of continuous state spaces; f = {f1(δ1), f2(δ2), f3(δ3),...} represents the change law of the continuous state space under each discrete state space; S = {S1, S2, S3,...} represents the mapping between the discrete state space and the continuous state space; F = {F1, F2, F3,...} represents the conditions for state space transition; I represents the initial state.

[0021] Furthermore, in Step 3, construct hybrid system models for different DERs, including:

[0022] Step 3.1 Construct a hybrid automata model for a photovoltaic power generation unit:

[0023]

[0024] In the formula, P pv is the output power of the photovoltaic power generation unit; P mpptP is the output power of the photovoltaic power generation unit operating in the maximum power point mode; L is the light intensity; M is the light intensity threshold, which is determined by the performance of the photovoltaic cell. According to the photovoltaic output characteristics, the photovoltaic power generation unit is set to two working modes: shutdown mode and MPPT mode, and the two working modes are set as the discrete states of the hybrid automaton model of the photovoltaic power generation unit; the power output value of the photovoltaic power generation unit is set as the continuous state; the change of the light intensity is set as the transition condition.

[0025] Step 3.2 Construct the hybrid automaton model of the wind power generation unit:

[0026]

[0027] In the formula, P w is the output power of the wind power generation unit; P e is the rated output power of the wind power generation unit; v is the real-time wind speed; v i is the cut-in wind speed; v r is the rated wind speed; v o is the cut-out wind speed. According to the operating characteristics of wind power generation, its working mode is divided into three types: shutdown mode, MPPT mode and P e mode. According to the hybrid automaton model, the three working modes are set as the discrete states of the hybrid automaton model of the wind power generation unit; the power output value of the wind power generation unit is set as the continuous state; the wind speed change is set as the transition condition.

[0028] Step 3.3 Construct the hybrid automaton model of the energy storage unit:

[0029] According to the characteristics of the battery itself, it is divided into five working modes: charging mode, discharging mode, shutdown mode (three). Among them, the three shutdown modes include normal shutdown, overcharge shutdown, and over-discharge shutdown.

[0030] According to the hybrid automaton model, when SOC ≤ SOC down , the storage battery switches from the discharging mode to the over-discharge shutdown mode, and when there is a relevant switching command, it will switch from the over-discharge shutdown mode to the charging mode again.

[0031] When SOC up ≤ SOC, the storage battery will switch from the charging mode to the overcharge shutdown mode, and when there is a relevant switching command, it will switch from the overcharge shutdown mode to the discharging mode again.

[0032] When SOC down <SOC<SOC up , the storage battery will switch between the discharging mode, the charging mode and the normal shutdown mode according to the system requirements.

[0033] Step 3.4 Construct the hybrid automaton model of the fuel cell:

[0034] According to the hybrid automata model, two operating modes of the fuel cell are designed: shutdown mode and rated output mode.

[0035] Step 3.5 Construct the load hybrid automata model:

[0036] Loads are divided into two categories: interruptible loads and non-interruptible loads. The power supply priority of interruptible loads is higher than that of non-interruptible loads. According to the hybrid automata model, two operating modes of the load unit are designed: normal operation mode and load shedding mode.

[0037] Step 3.6 Set the initial state of the hybrid automata model of all units, activate the discrete state space, set it to logical "1", and set the remaining discrete state spaces to logical "0". Thus, we can obtain

[0038] Photovoltaic power generation unit: D = {δ1, δ2} = [1, 0]; Wind turbine unit: D = {δ1, δ2, δ3} = [1, 0, 0]; Battery unit: D = {δ1, δ2, δ3, δ4, δ5} = [1, 0, 0, 0, 0]; Fuel cell unit: D = {δ1, δ2} = [1, 0]; Load unit: D = {δ1, δ2} = [1, 0].

[0039] Furthermore, in step 4, the local control strategy is divided into an internal switching control strategy and a continuous dynamic management strategy, which are specifically as follows:

[0040] For the photovoltaic power generation unit, the internal switching control commands are as follows:

[0041]

[0042]

[0043] For the wind turbine unit, the internal switching control commands are as follows:

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052] The control commands for the battery unit are as follows:

[0053]

[0054]

[0055] In step 4.2, a dual-loop control method based on SPWM is designed for the inverters of DERs, where the outer loop is designed as a controller using droop control, and the inner loop is designed as a controller in the dq rotating coordinate system.

[0056] Furthermore, in step 5, system voltage security evaluation indicators are constructed and classified. According to the system voltage security evaluation results, event-triggered hybrid control is applied to coordinate the switching of the working modes of DERs and load units, specifically as follows:

[0057] In step 5.1, the voltage sequence of the i-th bus of the distribution network system is extracted using the wide-area signal measurement method, denoted as After that, the average value of the instantaneous voltage of the i-th node at the j-th moment obtained from the voltage sequence is expressed as follows:

[0058]

[0059] The voltage deviation value of the i-th node at the j-th moment is expressed as:

[0060]

[0061] In the formula, is the actual voltage value of the i-th node at the j-th moment.

[0062] Therefore, the voltage security evaluation indicator of the i-th node at the j-th moment can be expressed as:

[0063]

[0064] In step 5.2, the voltage security evaluation indicators of each node are fused using the information fusion method based on the T-S fuzzy neural network.

[0065] Use to replace x = (x1, x2,..., x n ) as the input of the fuzzy neural network, and the output of the fuzzy neural network can be normalized to the required voltage security evaluation indicator. The output of the fuzzy neural network is expressed as:

[0066]

[0067] In the formula, is the weight of the fuzzy neural network; θj It is the product of membership degrees. The comprehensive voltage security evaluation index of each node obtained after information fusion is denoted as u.

[0068] Step 5.3 Classify the evaluation index u:

[0069] When 0.9 ≥ u or u ≥ 1.1, it is the voltage collapse state, and the distribution network system stops operating.

[0070] When 0.95 > u ≥ 0.9, the distribution network system has insufficient energy supply for the load, and the fuel cell operates or load shedding operations are performed.

[0071] When 0.98 > u ≥ 0.95, the distribution network system has insufficient energy supply for the load, and the battery discharges to balance the power.

[0072] When 1.02 > u ≥ 0.98, the system voltage is within the normal fluctuation range.

[0073] When 1.05 > u ≥ 1.02, the energy supply of the distribution network system is greater than the load demand, and the battery charges to balance the power.

[0074] When 1.1 > u ≥ 1.05, the energy supply of the distribution network system is much greater than the load demand, and the fuel cell switches to the shutdown mode or restores the interruptible load.

[0075] Step 5.4 Fully consider the trigger duration, switching sequence, and time interval, and use as the trigger event, and u and SOC as the trigger conditions. The specific design is as follows: The specific design is as follows:

[0076]

[0077] c2:H S ((δ 12 ,δ 22 ,δ 23 ,δ 33 ,δ 42 ),(F 12 ,F 25 ,F 26 ,F 33 ,F 42 ))

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091] where c n is the tag set, n ∈ [1, 2, …, 15]; is the trigger duration; Δt is the switching time interval. Using the above coordinated control commands, flexible regulation is carried out for the changes in the actions of DERs in the distribution network.

[0092] Furthermore, in step 6, the selective pattern-based aggregated single-layer dependence classification algorithm is used, and the coordinated control command set is used as the tag set to select the best coordinated switching control command in the control command set by performing classification operations, as follows:

[0093] Step 6.1 Attribute and attribute value mining: The selective pattern-based aggregated single-layer dependence classification algorithm is used to perform classification operations. The values of U and SOC during the operation of the distribution network system are used as continuous attributes, and the working modes of the battery, fuel cell, and load are used as discrete attributes, and attribute mining is carried out on this; the attribute set is: X = {(U, a1), (SOC, a2), (Battery, a3), (Fuel, a4), (Load, a5)}, and a1 to a5 are the attribute values corresponding to the five attributes.

[0094] Step 6.2 Process data of continuous attributes: The sample space (attribute values) of U is divided into seven continuous intervals, and referring to fuzzy control, the seven levels are defined as seven set values from low to high: "VL", "ML", "L", "Z", "H", "MH", "VH". Similarly, the sample space of SOC is divided into five continuous intervals, and the five levels are defined as five set values from low to high: "VL", "L", "Z", "H", "VH"; Process data of discrete attributes: For the battery unit, the sample space is set as {0, 1, 2}, where "0" represents the charging mode, "1" represents the discharging mode, and "2" represents the outage mode; For the fuel cell unit, the sample space is set as {0, 1}, where "0" represents the rated output mode and "1" represents the outage mode; For the load unit, the sample space is set as {0, 1}, "0" represents the normal operation mode, and "1" represents the load shedding mode.

[0095] Step 6.3 Assume that the dataset D has n attributes, and the training instance is represented as X = (a1, a2...a n ), where a i (1 ≤ i ≤ n) is the value of the instance X for the i-th attribute. The class of the training instance belongs to one of the classes C = {c1, c2...c m}, and below, c is used to represent the class of the instance. The probability P(c|X) that X belongs to class c can be expressed as:

[0096]

[0097] Let the set of attributes included in the set f be {a1, a2,..., a i}, and the Bayesian network is obtained:

[0098]

[0099] Apply the pattern classification ability to the Bayesian network, aggregate all situations corresponding to the patterns, and obtain the Bayesian probability prediction formula based on selectivity:

[0100]

[0101] Step 6.4 To weaken the dependence relationship between the attributes of the selective patterns and other attributes in the Bayesian algorithm, use the aggregated single-layer dependence classification algorithm based on selective patterns, and its final probability prediction formula is:

[0102]

[0103] In the formula, H = {j|i + 1 ≤ j ≤ n ∧ F(a j ) ≥ m}, F(a j ) is the attribute value containing a jThe number of training instances is restricted by the parameter m to achieve the support degree required for conditional probability estimation.

[0104] In step 6.5, for the aggregated single-layer dependence classification algorithm based on the selective pattern, after selecting the corresponding attributes, the attribute values are mined according to the data generated by the operation of the distribution network. At the same time, the coordinated control command set is used as the label set, and by performing the classification operation, the selection of the optimal coordinated switching control command is realized.

[0105] Furthermore, in step 7, a comprehensive objective function is established for the upper-layer intelligent agent, and a fruit fly optimization algorithm based on random walk is applied to achieve objective optimization, specifically as follows:

[0106] Step 7.1 Construct the optimal scheduling objective function:

[0107] Cost objective function:

[0108]

[0109] In the formula: i: the number of DERs; s: the working mode of DERs; α is : When the distribution network system is in the working mode, α is = 1, otherwise, α is = 0; r i : The fuel cost of the i-th DERs. For renewable energy, r i = 0; The active power output of the i-th DERs in the s mode; E is : Consumption characteristic function; M i : The maintenance cost of the i-th DERs, proportional to ; C i : The start-up cost of the i-th DERs; β is : When the i-th DERs works in the s mode, β is = 1, otherwise, β is = 0.

[0110] Carbon emission objective function:

[0111]

[0112] In the formula: F2: The actual carbon emission during the operation of the distribution network; π1: The carbon emission intensity per unit active power output of the thermal power unit; π2: The carbon emission intensity per unit active power output of the gas turbine; P in.t : The electricity purchased from the main grid per unit time period; P out.t : The power generation power of the gas turbine within the unit time period t; The carbon emission of clean energy is 0.

[0113] Power quality objective function:

[0114]

[0115] Where: σ is is the power quality coefficient of the i-th DERs in the s mode. σ is is determined by the voltage security assessment index.

[0116] Step 7.2 Construct the constraint conditions:

[0117] Power balance constraint:

[0118] P l = P pv + P w + P F + (-1) n P bat (14)

[0119] Where, P l is the load demand of the system; P pv is the output power of the photovoltaic power generation unit; P w is the output power of the wind power generation unit; P bat is the output / absorbed power of the battery; n ∈ [0, 1], when the battery is in the charging mode, n = 1, when the battery is in the discharging mode, n = 0; P F is the output power of the fuel cell.

[0120] Output power constraint of DERs:

[0121]

[0122] Where, P i is the output power of the i-th DERs; is the minimum output power of the i-th DERs; is the maximum output power of the i-th DERs.

[0123] Battery capacity constraint:

[0124] SOC down < SOC < SOC up (16)

[0125] Where, SOC is the state of charge of the battery; SOC down is the minimum capacity state value of the battery; SOC up is the maximum capacity state value of the battery.

[0126] Step 7.3 Based on an improved fruit fly optimization algorithm for the implementation of optimal scheduling.

[0127] Step 7.4 processes the multi-objective optimization problem in the form of weight coefficients. Energy management can be summarized as solving the following objective optimization function subject to constraints:

[0128]

[0129] In the formula, ω1, ω2, and ω3 represent the weight coefficients of each objective function, and ω1 + ω2 + ω3 = 1.

[0130] In a second aspect, the present invention provides an optimized matching and coordinated switching control device for the operating modes of distributed controllable resources under different power deficits. The device includes:

[0131] An architecture establishment module: used to establish a hierarchical architecture of the distribution network based on MAS for the coordinated control objectives of the distribution network, and divide the control structure of the distribution network into an upper layer, a middle layer, and a bottom layer;

[0132] A model construction module: used to construct a hybrid automaton model represented by a six-element array by adding continuous variables on the basis of the concept of a finite automaton;

[0133] A model establishment module: used to establish corresponding hybrid automaton models for the operating characteristics of photovoltaic power generation units, wind power generation units, energy storage units, fuel cells, and loads respectively;

[0134] A management strategy module: used to divide the local control strategy of the bottom layer into an internal switching control strategy and a continuous dynamic management strategy. Among them, the internal switching control strategy is based on the coordinated control command of the middle layer, and the continuous dynamic management is a dual-loop control method based on SPWM;

[0135] A coordinated control module: used to design a coordinated control strategy in the multi-agent structure of the middle layer, construct a system voltage security evaluation index and perform hierarchical processing, and apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units according to the system voltage security evaluation results;

[0136] A classification operation module: used to use the selective pattern-based aggregated single-layer dependence classification algorithm, use the coordinated control command set as a label set, and select the best coordinated switching control command in the control command set by performing classification operations;

[0137] A multi-objective optimization module: used for the upper-layer agent to construct a comprehensive objective function of reducing operating costs, reducing pollutant emissions, and improving power quality considering power balance, output power of DERs, and battery capacity constraints for different power deficits, and use an improved fruit fly optimization algorithm with random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy.

[0138] In a third aspect, the present invention provides a control device for optimizing the matching and coordinated switching of the operating modes of distributed controllable resources under different power deficits, characterized by comprising a processor and a storage medium;

[0139] The storage medium is used for storing instructions;

[0140] The processor is configured to operate according to the instructions to execute the steps of the method according to the first aspect.

[0141] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0142] The present invention constructs a hierarchical coordination architecture for the distribution network based on MAS. Under different power demands, according to the coordination optimization strategy formulated by the hierarchical architecture, it realizes the optimization of the matching and coordinated switching control of the operating modes of distributed controllable resources;

[0143] The hybrid automaton model constructed by the present invention based on the hybrid system and the automaton can match the operating characteristics of photovoltaic power generation, wind power generation, storage batteries, fuel cells, loads, etc. For the complex operating data generated during the operation of the distribution network, the machine learning algorithm can effectively process it, and realizes the optimal coordinated switching control of the operating modes of DERs through the aggregation single-layer dependence classification algorithm based on the selective mode. Considering the system's supply-demand balance constraint and the operating constraints of DERs, an optimization objective function for reducing operating costs, reducing pollutant emissions, and improving power quality is constructed, and is solved by an improved random walk fruit fly algorithm, thereby realizing the safe and economic optimal operation of the system. Description of the Drawings

[0144] Figure 1 is the hybrid automaton model of the photovoltaic power generation unit;

[0145] Figure 2 is the hybrid automaton model of the wind power generation unit;

[0146] Figure 3 is the hybrid automaton model of the storage battery unit;

[0147] Figure 4 is the hybrid automaton model of the fuel cell unit;

[0148] Figure 5 is the hybrid automaton model of the load unit;

[0149] Figure 6 is the hierarchical structure diagram of the distribution network based on MAS;

[0150] Figure 7 is the structure diagram of the outer-loop controller based on P-f and Q-U droop control;

[0151] Figure 8It is the structure diagram of the inner loop controller in the dq rotating coordinate system;

[0152] Figure 9 It is the method flow chart of the present invention. Specific implementation manners

[0153] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, and cannot be used to limit the protection scope of the present invention.

[0154] Embodiment 1:

[0155] This embodiment provides an optimization matching and coordinated switching control method for the operation mode of distributed controllable resources under different power deficits, and the applied scenario is the distribution network system. To solve the problems that the management object of the traditional distribution network energy management system is single and the management mode is simple, and it is impossible to achieve efficient regulation of existing DERs with different characteristics, a hierarchical structure of the distribution network based on MAS is designed as Figure 6 shown, and an optimization and coordination control strategy is formulated according to the hierarchical control structure. The hybrid automaton model of the photovoltaic power generation unit established based on the hybrid automaton model and its discrete states and transition conditions are as Figure 1 shown in, Table 1, and Table 2; the hybrid automaton model of the wind power unit and its discrete states and transition conditions are as Figure 2 shown in, Table 3, and Table 4; the hybrid automaton model of the battery unit and its discrete states and transition conditions are as Figure 3 shown in, Table 5, and Table 6; the hybrid automaton model of the fuel cell unit and its discrete states and transition conditions are as Figure 4 shown in, Table 7, and Table 8; the hybrid automaton model of the load unit and its discrete states and transition conditions are as Figure 5 shown in, Table 9, and Table 10. For the large amount of complex data generated during the operation of the distribution network, the selective mode-based aggregated single-layer dependence classification algorithm and the improved fruit fly optimization algorithm based on random walk are used for effective processing, and the working state of the distribution network system is identified through classification operations, improving the intelligence of the distribution network and the accuracy of control command execution.

[0156] Table 1 Discrete state description of the hybrid automaton model of the photovoltaic power generation unit

[0157] Discrete state Discrete state description <![CDATA[δ 11 > Outage mode of photovoltaic power generation unit <![CDATA[δ 12 > MPPT mode of photovoltaic power generation unit

[0158] Table 2 Transition condition description of the hybrid automaton model of the photovoltaic power generation unit

[0159] Transfer condition Transfer condition description <![CDATA[F 11 > Photovoltaic power generation unit switches from outage mode to MPPT mode <![CDATA[F 12 > Photovoltaic power generation unit switches from MPPT mode to outage mode

[0160] Table 3 Discrete state description of the hybrid automaton model of the wind power generation unit

[0161]

[0162]

[0163] Table 4 Description of Transition Conditions of Hybrid Automata Model for Wind Power Generation Unit

[0164] Transfer condition Transfer condition description <![CDATA[F 21 > As the wind speed increases, the wind power generation unit switches from outage mode to MPPT mode <![CDATA[F 22 > <![CDATA[As the wind speed increases, the wind power generation unit switches from the shutdown mode to the P e output mode]]> <![CDATA[F 23 > <![CDATA[The wind power generation unit switches from the MPPT mode to the P e output mode]]> <![CDATA[F 24 > <![CDATA[The wind power generation unit switches the output mode from P e to the MPPT mode]]> <![CDATA[F 25 > Wind power generation unit switches from MPPT mode to outage mode <![CDATA[F 26 > <![CDATA[The wind power generation unit switches the output mode from P e to the shutdown mode]]> <![CDATA[F 27 > As the wind speed decreases, the wind power generation unit switches from outage mode to MPPT mode <![CDATA[F 28 > <![CDATA[As the wind speed decreases, the wind power generation unit switches from the shutdown mode to the P e output mode]]>

[0165] Table 5 Description of Discrete States of Hybrid Automata Model for Battery Unit

[0166] Discrete state Discrete state description <![CDATA[δ 31 > Normal outage mode of battery unit <![CDATA[δ 32 > Discharge mode of battery unit <![CDATA[δ 33 > Charging mode of battery unit <![CDATA[δ 34 > Over-discharge outage mode of battery unit <![CDATA[δ 35 > Over-charge outage mode of battery unit

[0167] Table 6 Description of Transition Conditions of Hybrid Automata Model for Battery Unit

[0168]

[0169]

[0170] Table 7 Description of Discrete States of Hybrid Automata Model for Fuel Cell Unit

[0171] Discrete state Discrete state description <![CDATA[δ 41 > Outage mode of fuel cell unit <![CDATA[δ 42 > Rated output mode of fuel cell unit

[0172] Table 8 Description of Transition Conditions of Hybrid Automata Model for Fuel Cell Unit

[0173] Transfer condition Transfer condition description <![CDATA[F 41 > Fuel cell unit switches from outage mode to rated output mode <![CDATA[F 42 > Fuel cell unit switches from rated output mode to outage mode

[0174] Table 9 Description of Discrete States of Hybrid Automata Model for Load Unit

[0175] Discrete state Discrete state description <![CDATA[δ 51 > Normal operation mode of load unit <![CDATA[δ 52 > Load shedding mode of load unit

[0176] Table 10 Description of Transition Conditions of Hybrid Automata Model for Load Unit

[0177] Transfer condition Transfer condition description <![CDATA[F 51 > Load unit switches from normal operation mode to load shedding mode <![CDATA[F 52 > Load unit switches from load shedding mode to normal operation mode

[0178] The upper, middle, and bottom layers respectively involve energy management strategies, coordinated control strategies, internal switching control strategies, and continuous dynamic management strategies; the bottom layer designs local control as hybrid control, designs internal switching control strategies based on event-triggering mechanisms and hybrid automata models, and uses droop control methods of P-f and Q-U to achieve outer-loop control as Figure 7 shown, and the structure diagram of the inner-loop controller in the dq rotating coordinate system is designed as Figure 8As shown; the middle layer designs safety indicators for safety assessment, classifies the safety indicators based on T-S fuzzy control, as shown in Table 11, and designs a coordinated control command set; the upper layer sets up a main function, and uses an improved fruit fly optimization algorithm based on random walk to achieve a single-objective optimization process on the basis of multi-objective optimization. Through the hierarchical control structure of the distribution network and combined with the corresponding control strategies, the optimal matching and coordinated switching control of the operation modes of distributed controllable resources under different power deficits are realized.

[0179] Table 11 Classification of Voltage Safety Assessment Indicators

[0180] Level Level interval 1 0.9≥u 2 0.95＞u≥0.9 3 0.98＞u≥0.95 4 1.02＞u≥0.98 5 1.05＞u≥1.02 6 1.1＞u≥1.05 7 u≥1.1

[0181] Step 1, for the coordinated control objective of the distribution network, establish a hierarchical architecture of the distribution network based on MAS, and divide the control structure of the distribution network into the upper layer, the middle layer and the bottom layer:

[0182] (1) Upper layer agent - negotiation agent. Achieve energy management through the optimization process to enable the system to obtain the maximum economic, environmental and power quality benefits. The optimization model is designed based on the real-time data in the processing module. The time scale of the optimization process is once per hour.

[0183] (2) Middle layer agent - negotiation agent. Coordinate the switching of the working modes of DERs to ensure the voltage safety of the system. The time scale of the switching control is at the hour or minute level.

[0184] (3) Bottom layer agent - hybrid agent, including a reaction layer and a negotiation layer. The reaction layer is set to "perception-action" and has priority to quickly respond to emergencies. The negotiation layer is set to "belief-desire-intention", which is highly intelligent and controls or plans the behavior of the agent to achieve its desires or intentions.

[0185] Step 2, considering that in the control process, there are both discrete control commands for switching the operation modes of different distributed power sources and load units, and continuous control instructions for the execution-end inverter. Therefore, on the basis of the concept of finite automata, continuous variables are added to construct a hybrid automata model represented by a six-tuple:

[0186] As one of the modeling methods for hybrid systems, automata have the advantage of intuitiveness. Ordinary automata models are used to process discrete events. If continuous dynamic characteristics are added to them, a hybrid automata model will be formed. A hybrid automata mainly consists of a discrete state space, an enabled state, and a continuous state. In each discrete state space, there are corresponding continuous state changes. When a certain enabled state is satisfied, the hybrid automata will migrate from one discrete state space to another discrete state space, and the corresponding continuous state will also change.

[0187] A typical hybrid automaton model can be represented by a 6-tuple: H = (D, L, f, S, F, I). Where D = {δ1, δ2, …} represents a set of discrete state spaces; L represents a set of continuous state spaces; f = {f1(δ1), f2(δ2), f3(δ3), …} represents the variation law of the continuous state space under each discrete state space; S = {S1, S2, S3, …} represents the mapping between the discrete state space and the continuous state space; F = {F1, F2, F3, …} represents the conditions for state space transition; I represents the initial state.

[0188] Step 3, according to the operating characteristics of the photovoltaic power generation unit, wind power generation unit, energy storage unit, fuel cell, and load, establish corresponding hybrid automaton models respectively;

[0189] (1) Construct a hybrid automaton model for the photovoltaic power generation unit:

[0190]

[0191] In the formula, P pv is the output power of the photovoltaic power generation unit; P mppt is the output power of the photovoltaic power generation unit operating in the maximum power point mode; L is the light intensity; M is the light intensity threshold, which is determined by the performance of the photovoltaic cell. According to the photovoltaic output characteristics, the photovoltaic power generation unit is set to two operating modes: shutdown mode and MPPT mode, and the two operating modes are set as the discrete states of the hybrid automaton model of the photovoltaic power generation unit; the power output value of the photovoltaic power generation unit is set as the continuous state; the change of the light intensity is set as the transition condition.

[0192] (2) Construct a hybrid automaton model for the wind power generation unit:

[0193]

[0194] In the formula, P w is the output power of the wind power generation unit; P e is the rated output power of the wind power generation unit; v is the real-time wind speed; v i is the cut-in wind speed; v r is the rated wind speed; v o is the cut-out wind speed. According to the operating characteristics of wind power generation, its operating mode is divided into three types: shutdown mode, MPPT mode, and P e mode. According to the hybrid automaton model, the three operating modes are set as the discrete states of the hybrid automaton model of the wind power generation unit; the power output value of the wind power generation unit is set as the continuous state; the change of the wind speed is set as the transition condition.

[0195] (3) Construct a hybrid automaton model for the energy storage unit:

[0196] According to the characteristics of the battery itself, it is divided into five working modes: charging mode, discharging mode, and three shutdown modes. Among them, the three shutdown modes include normal shutdown, overcharge shutdown, and overdischarge shutdown.

[0197] According to the hybrid automaton model, when SOC ≤ SOC down the battery switches from the discharging mode to the overdischarge shutdown mode, and when there is a relevant switching command, it will switch from the overdischarge shutdown mode to the charging mode.

[0198] When SOC up ≤ SOC, the battery will switch from the charging mode to the overcharge shutdown mode, and when there is a relevant switching command, it will switch from the overcharge shutdown mode to the discharging mode.

[0199] When SOC down <SOC<SOC up the battery will switch between the discharging mode, the charging mode, and the normal shutdown mode according to the system requirements.

[0200] (4) Construct the fuel cell hybrid automaton model:

[0201] According to the hybrid automaton model, design two working modes for the fuel cell: shutdown mode and rated output mode.

[0202] (5) Construct the load hybrid automaton model:

[0203] The load is divided into two categories: interruptible load and non-interruptible load. The power supply priority of the interruptible load is higher than that of the non-interruptible load. According to the hybrid automaton model, design two working modes for the load unit: normal operation mode and load shedding mode.

[0204] (6) Set the initial state of the hybrid automaton model of all units, activate the discrete state space, set it to logical "1", and set the remaining discrete state spaces to logical "0". Thus, we can obtain

[0205] Photovoltaic power generation unit: D = {δ1, δ2} = [1, 0]; Wind turbine unit: D = {δ1, δ2, δ3} = [1, 0, 0]; Battery unit: D = {δ1, δ2, δ3, δ4, δ5} = [1, 0, 0, 0, 0]; Fuel cell unit: D = {δ1, δ2} = [1, 0]; Load unit: D = {δ1, δ2} = [1, 0].

[0206] Step 4: Divide the underlying local control strategy into an internal switching control strategy and a continuous dynamic management strategy. Among them, the internal switching control strategy is based on the coordinated control command of the middle layer, and the continuous dynamic management is designed as a dual-loop control method based on SPWM:

[0207] (1) For the photovoltaic power generation unit, the internal switching control commands are as follows:

[0208]

[0209]

[0210] For the wind power generation unit, the internal switching control commands are as follows:

[0211]

[0212]

[0213]

[0214]

[0215]

[0216]

[0217]

[0218]

[0219] The control commands for the battery unit are as follows:

[0220]

[0221]

[0222] (2) A dual-loop control method based on SPWM is designed for the inverter of DERs, where the outer loop is designed as a controller using droop control, and the inner loop is designed as a controller in the dq rotating coordinate system.

[0223] The method of the outer loop controller is determined as:

[0224]

[0225] where f and U are the frequency and amplitude of the output voltage respectively; f0 and U0 are the reference values of the voltage frequency and amplitude respectively; P0 and Q0 are the reference values of the active power and reactive power respectively; P and Q are the actual values of the active power and reactive power respectively; K p and K q are the droop coefficients.

[0226] The inner loop control uses the Park transformation to transform the three-phase control problem in the original abc natural coordinate system into a two-phase control problem in the dq rotating coordinate system.

[0227] Step 5: Design a coordinated control strategy in the middle-layer multi-agent structure, construct a system voltage security assessment index and perform hierarchical processing. According to the system voltage security assessment results, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units:

[0228] (1) Use the wide-area signal measurement method to extract the voltage sequence of the i-th bus in the distribution network system, denoted as After that, the average value of the instantaneous voltage of the i-th node at the j-th moment obtained from the voltage sequence is expressed as follows:

[0229]

[0230] The voltage deviation value of the i-th node at the j-th moment is expressed as:

[0231]

[0232] In the formula, is the actual voltage value of the i-th node at the j-th moment.

[0233] Thus, the voltage security assessment index of the i-th node at the j-th moment can be expressed as:

[0234]

[0235] (2) Use the information fusion method based on the T-S fuzzy neural network to fuse the voltage security assessment indexes of each node.

[0236] Use to replace x = (x1, x2,..., x n ) as the input of the fuzzy neural network, and the output of the fuzzy neural network can be normalized to the required voltage security assessment index. The output of the fuzzy neural network is expressed as:

[0237]

[0238] In the formula, is the weight of the fuzzy neural network; θ j is the membership degree product.

[0239] and θ j The expressions of are shown in formulas (26) and (27) respectively.

[0240]

[0241] In the formula, α is the learning rate of the fuzzy neural network; e = 0.5(y s -y m ) 2, which is the deviation between the actual output value and the target value, and is also a performance metric in the algorithm learning process; y s is the expected value of the output; y m is the actual output value.

[0242]

[0243] In the formula, is a fuzzy set; is the membership function for each input.

[0244] The expression of

[0245]

[0246] In the formula, and are the center and width of the membership function respectively.

[0247] and The expressions of

[0248]

[0249]

[0250] The comprehensive voltage security assessment index of each node obtained after information fusion is denoted as u.

[0251] (3) Perform hierarchical processing on the evaluation index u:

[0252] When 0.9 ≥ u or u ≥ 1.1, it is the voltage collapse state, and the distribution network system stops operating.

[0253] When 0.95 > u ≥ 0.9, the distribution network system has insufficient energy supply for the load, and the fuel cell works or load shedding operations are performed.

[0254] When 0.98 > u ≥ 0.95, the distribution network system has insufficient energy supply for the load, and the battery performs a discharging operation to balance the power.

[0255] When 1.02 > u ≥ 0.98, the system voltage is within the normal fluctuation range.

[0256] When 1.05 > u ≥ 1.02, the energy supply of the distribution network system is greater than the load demand, and the battery performs a charging operation to balance the power.

[0257] When 1.1 > u ≥ 1.05, the energy supply of the distribution network system is much greater than the load demand, and the fuel cell switches to the shutdown mode or restores the interruptible load.

[0258] (4) Fully consider the trigger duration, switching sequence, and time interval, and use as the trigger event, and u and SOC as the trigger conditions. For the specific design is as follows:

[0259]

[0260] c2:H S ((δ 12 , δ 22 , δ 23 , δ 33 , δ 42 ), (F 12 , F 25 , F 26 , F 33 , F 42 ))

[0261]

[0262]

[0263]

[0264]

[0265]

[0266]

[0267]

[0268]

[0269]

[0270]

[0271]

[0272]

[0273]

[0274] In the formula, c n is the label set, n ∈ [1, 2, …, 15]; is the trigger duration; Δt is the switching time interval. Using the above coordinated control commands, flexible regulation is carried out for the changes in the actions of DERs in the distribution network.

[0275] Step 6. To minimize the number of DERs operation mode switches and improve the intelligence of the system, as well as the rapidity and accuracy of control command execution, the selective mode-based aggregated single-layer dependence classification algorithm is utilized, with the coordinated control command set used as the tag set, and the best coordinated switching control command is selected from the control command set by performing classification operations:

[0276] (1) Attribute and attribute value mining: The selective mode-based aggregated single-layer dependence classification algorithm is adopted to perform classification operations. The values of U and SOC during the operation of the distribution network system are used as continuous attributes, and the working modes of the battery, fuel cell, and load are used as discrete attributes, and attribute mining is carried out for this; the attribute set is: X = {(U, a1), (SOC, a2), (Battery, a3), (Fuel, a4), (Load, a5)}, where a1 to a5 are the attribute values corresponding to the five attributes.

[0277] (2) Processing continuous attribute data: The sample space (attribute value) of U is divided into seven continuous intervals, and with reference to fuzzy control, the seven levels are defined as seven set values from low to high: "VL", "ML", "L", "Z", "H", "MH", "VH". Similarly, the sample space of SOC is divided into five continuous intervals, and the five levels are defined as five set values from low to high: "VL", "L", "Z", "H", "VH"; Processing discrete attribute data: For the battery unit, the sample space is set as {0, 1, 2}, where "0" represents the charging mode, "1" represents the discharging mode, and "2" represents the outage mode; for the fuel cell unit, the sample space is set as {0, 1}, where "0" represents the rated output mode and "1" represents the outage mode; for the load unit, the sample space is set as {0, 1}, "0" represents the normal operation mode, and "1" represents the load shedding mode.

[0278] (3) Assume that the data set D has n attributes, and the training instance is represented as X = (a1, a2...a n ), where a i (1 ≤ i ≤ n) is the value of the instance X on the i-th attribute. The category of the training instance belongs to one of the categories C = {c1, c2...c m}, and below, c is used to represent the category of the instance. The probability P(c|X) that the category of X is c can be expressed as:

[0279]

[0280] Let the attribute set contained in the set f be {a1, a2,..., a i}, and the Bayesian network is obtained:

[0281]

[0282] Apply the pattern classification ability to the Bayesian network, aggregate the situations corresponding to all patterns, and obtain the Bayesian probability prediction formula based on selectivity:

[0283]

[0284] (4) To weaken the dependence relationship between the attributes of the selective pattern and other attributes in the Bayesian algorithm, adopt the aggregated single-layer dependence classification algorithm based on the selective pattern, and its final probability prediction formula is:

[0285]

[0286] In the formula, H = {j|i + 1 ≤ j ≤ n ∧ F(a j ) ≥ m}, F(a j ) is the number of training instances whose attribute values contain a j , restricted by the parameter m to achieve the support degree required for conditional probability estimation.

[0287] (5) For the aggregated single-layer dependence classification algorithm based on the selective pattern, after selecting the corresponding attributes, mine the attribute values according to the data generated by the operation of the distribution network. At the same time, use the coordinated control command set as the label set, and through the classification operation, realize the selection of the optimal coordinated switching control command.

[0288] Step 7, for different power deficits, the upper-layer intelligent agent constructs a comprehensive objective function that reduces operating costs, reduces pollutant emissions, and improves power quality, considering power balance, the output power of DERs, and the battery capacity constraint, and uses a random-walk fruit fly optimization algorithm to solve the multi-objective optimization problem and realize the upper-layer control strategy:

[0289] (1) Construct the optimization scheduling objective function:

[0290] Cost objective function:

[0291]

[0292] In the formula: i: the number of DERs; s: the working mode of DERs; α is : when the distribution network system is in the working mode, α is = 1, otherwise, α is = 0; r i : the fuel cost of the i-th DERs, for renewable energy, r i = 0; The active power output of the i-th DERs in the s mode; E is : consumption characteristic function; M i : the maintenance cost of the i-th DERs, proportional to ; Ci : Startup cost of the \(i\) - th DERs; \(\beta\) is : When the \(i\) - th DERs operates in mode \(s\), \(\beta\) is \( = 1\); otherwise, \(\beta\) is \( = 0\).

[0293] Carbon emission objective function:

[0294]

[0295] In the formula: \(F_2\): Actual carbon emissions during the operation of the distribution network; \(\pi_1\): Carbon emission intensity per unit active power output of thermal power units; \(\pi_2\): Carbon emission intensity per unit active power output of gas turbines; \(P\) in.t : Electricity purchased from the main grid per unit time period; \(P\) out.t : Power generation power of the gas turbine within unit time period \(t\); The carbon emission of clean energy is 0.

[0296] Power quality objective function:

[0297]

[0298] In the formula: \(\sigma\) is is the power quality coefficient of the \(i\) - th DERs in mode \(s\). \(\sigma\) is is determined by the voltage security assessment index.

[0299] (2) Construct constraint conditions:

[0300] Power balance constraint:

[0301] \(P\) l \( = P\) pv \(+P\) w \(+P\) F \(+(-1)\) n \(P\) bat (20)

[0302] In the formula, \(P\) l is the load demand of the system; \(P\) pv is the output power of the photovoltaic power generation unit; \(P\) w is the output power of the wind power generation unit; \(P\) bat is the power output / absorption of the battery; \(n\in[0,1]\), when the battery is in the charging mode, \(n = 1\), when the battery is in the discharging mode, \(n = 0\); \(P\) F is the output power of the fuel cell.

[0303] Output power constraint of DERs:

[0304]

[0305] In the formula, \(P\) iis the output power of the $i$-th DERs; is the minimum output power of the $i$-th DERs; is the maximum output power of the $i$-th DERs.

[0306] Battery capacity constraint:

[0307] SOC down <SOC<SOC up (22)

[0308] In the formula, SOC is the state of charge of the battery; SOC down is the minimum capacity state value of the battery; SOC up is the maximum capacity state value of the battery.

[0309] (3) Based on an improved fruit fly optimization algorithm for implementing the optimal scheduling. The steps of the traditional fruit fly optimization algorithm are as follows:

[0310] Position initialization:

[0311]

[0312] In the formula: LR is used to set the position range of the fruit fly swarm; x start y start Determine the initial position of the fruit fly swarm.

[0313] Calculate the position of each fruit fly:

[0314]

[0315] In the formula: V is the range of the fruit fly swarm, that is, the radius of the current fruit fly swarm.

[0316] Distance calculation:

[0317]

[0318] Usually, D i is the distance from fly $i$ to the origin, and S i is the concentration of the food, which is related to the distance D i Smell i is the individual odor concentration of fly $i$, which can be calculated by any specific function Fitness().

[0319] Best fruit fly identification:

[0320] step4: [bestSmell bestIndex]=max(Smell i )(26)

[0321] The position and concentration value of the fly with the maximum concentration are recorded as bestIndex and bestSmell respectively.

[0322] Retention of the best fruit fly and replacement of the positions of the fly swarm:

[0323]

[0324] Given the position bestIndex of the fly with the maximum concentration value, the position of the bee swarm will be updated accordingly. x axis y axis represents the position of the best fruit fly.

[0325] Iterate the above steps until the accuracy or the number of iterations requirement is met.

[0326] During the iteration process, the distance between the current solution and the optimal solution cannot be predicted. Therefore, the characteristics of fruit fly random walk (semi-supervised method) are added to optimize the above algorithm steps 1 to 5, and the position weight calculation is added:

[0327]

[0328] In the formula: ω i,j is the position weight of two continuously iterated fly swarms, represents the Euclidean distance between two fly swarms, and the calculation formula of η is as follows:

[0329]

[0330] (4) Process the multi-objective optimization problem in the form of weight coefficients. Energy management can be summarized as solving the following objective optimization function that satisfies the constraint conditions:

[0331]

[0332] In the formula, ω1, ω2, and ω3 represent the weight coefficients of each objective function, and ω1 + ω2 + ω3 = 1.

[0333] Embodiment 2:

[0334] This embodiment provides an optimized matching and coordinated switching control device for the operation mode of distributed controllable resources under different power deficits. The device includes:

[0335] Architecture establishment module: used to establish a hierarchical architecture of the distribution network based on MAS for the coordinated control objective of the distribution network, and divide the distribution network control structure into an upper layer, a middle layer, and a bottom layer;

[0336] Model construction module: used to construct a hybrid automaton model represented by a 6-tuple by adding continuous variables on the basis of the concept of finite automata;

[0337] Model establishment module: used to respectively establish corresponding hybrid automata models for the operating characteristics of photovoltaic power generation units, wind power generation units, energy storage units, fuel cells, and loads;

[0338] Management strategy module: used to divide the underlying local control strategies into internal switching control strategies and continuous dynamic management strategies. Among them, the internal switching control strategy is based on the coordinated control commands of the middle layer, and the continuous dynamic management is a dual-loop control method based on SPWM;

[0339] Coordinated control module: used to design coordinated control strategies in the middle-layer multi-agent structure, construct system voltage security evaluation indicators and perform hierarchical processing. According to the system voltage security evaluation results, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units;

[0340] Classification operation module: used to use the aggregation single-layer dependence classification algorithm based on the selective mode, use the coordinated control command set as the label set, and select the best coordinated switching control command in the control command set by performing classification operations;

[0341] Multi-objective optimization module: used for the upper-layer agent to construct a comprehensive objective function for reducing operating costs, reducing pollutant emissions, and improving power quality considering power balance, the output power of DERs, and the battery capacity constraint for different power deficits, and use the improved fruit fly optimization algorithm with random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy.

[0342] The device of this embodiment can be used to implement the method described in Embodiment 1.

[0343] Embodiment 3:

[0344] This embodiment provides a device for optimizing the matching and coordinated switching control of the operating modes of distributed controllable resources under different power deficits, characterized by including a processor and a storage medium;

[0345] The storage medium is used to store instructions;

[0346] The processor is used to operate according to the instructions to execute the steps of the method described in Embodiment 1.

[0347] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0348] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0349] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0350] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0351] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for optimizing matching and coordinated switching control of distributed controllable resource operation modes under different power deficits, characterized in that: It includes the following steps: Step 1: For the coordinated control objective of the distribution network, establish a hierarchical architecture of the distribution network based on MAS, and divide the control structure of the distribution network into the upper layer, the middle layer, and the bottom layer; Step 2: On the basis of the concept of finite automata, add continuous variables to construct a hybrid automata model represented by a six-element array; Step 3: For the operating characteristics of photovoltaic power generation units, wind power generation units, energy storage units, fuel cells, and loads, establish corresponding hybrid automata models respectively; Step 4: Divide the local control strategy at the bottom layer into an internal switching control strategy and a continuous dynamic management strategy. Among them, the internal switching control strategy is based on the coordinated control command of the middle layer, and the continuous dynamic management is a double-loop control method based on SPWM; Step 5: Design a coordinated control strategy in the multi-agent structure of the middle layer, construct a system voltage security evaluation index and perform hierarchical processing. According to the system voltage security evaluation results, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units; Step 6: Use the aggregated single-layer dependence classification algorithm based on the selective mode, use the coordinated control command set as the label set, and select the best coordinated switching control command in the control command set by performing classification operations; Step 7: For different power shortages, the upper-layer agent constructs a comprehensive objective function that reduces operating costs, reduces pollutant emissions, and improves power quality considering power balance, the output power of DERs, and battery capacity constraints. Use the improved fruit fly optimization algorithm of random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy; Step 7. The upper-layer agent constructs a comprehensive objective function that reduces operating costs, reduces pollutant emissions, and improves power quality considering power balance, the output power of DERs, and battery capacity constraints for different power shortages. Use the improved fruit fly optimization algorithm of random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy, including: Step 7.1 Construct a comprehensive objective function: Cost objective function: Where: i is the number of DERs; s is the operating mode of DERs; α is is a parameter. When the distribution network system is in the operating mode, α is = 1; otherwise, α is = 0; r i is the fuel cost of the i-th DER. For renewable energy, r i = 0; is the active power output of the i-th DER in mode s; E is is the consumption characteristic function; M i is the maintenance cost of the i-th DER, proportional to ; C i is the start-up cost of the i-th DER; β is is a cost parameter. When the i-th DER operates in mode s, β is = 1; otherwise, β is = 0; Carbon emission objective function: Where: F2: the actual carbon emissions during the operation of the distribution network; π1: the carbon emission intensity per unit active power output of the thermal power unit; π2: the carbon emission intensity per unit active power output of the gas turbine; P in.t : the electricity purchased from the main grid in a unit time period; P out.t : the power generation power of the gas turbine within the unit time period t; the carbon emissions of clean energy are 0; Power quality objective function: Where: σ is is the power quality coefficient of the i-th DERs in the s mode; σ is is determined by the voltage security assessment index; Step 7.2 Construct constraint conditions: Power balance constraint: P l = P pv + P w + P F + (-1) n P bat (14) where P l is the load demand of the system; P pv is the output power of the photovoltaic power generation unit; P w is the output power of the wind power generation unit; P bat is the power output / absorbed by the battery; n ∈ [0, 1], when the battery is in the charging mode, n = 1, when the battery is in the discharging mode, n = 0; P F is the output power of the fuel cell; Output power constraint of DERs: P i down <P i <P i up (15) Wherein, P i is the output power of the i-th DERs; P i down is the minimum output power of the i-th DERs; P i up is the maximum output power of the i-th DERs; Battery capacity constraint: SOC down <SOC<SOC up (16) Where SOC is the state of charge of the battery; SOC down is the minimum capacity state value of the battery; SOC up is the maximum capacity state value of the battery; Step 7.3 Based on an improved fruit fly optimization algorithm for the implementation of optimal scheduling; Step 7.4 Process the multi-objective optimization problem in the form of weight coefficients. Energy management can be summarized as solving the following comprehensive objective function that satisfies the constraint conditions: In the formula, ω1, ω2, and ω3 represent the weight coefficients of each objective function, and ω1 + ω2 + ω3 = 1.

2. The method for optimizing matching and coordinated switching control of distributed controllable resource operation modes under different power deficits according to claim 1, characterized in that: Step 1. Establish a hierarchical structure based on MAS, including: Step 1.1 Upper-layer agent - negotiation agent; realize energy management through the optimization process to enable the system to obtain the maximum economic, environmental, and power quality benefits. The optimization model is designed based on the real-time data in the processing module; the time scale of the optimization process is once per hour; Step 1.2 Middle-layer agent - negotiation agent; coordinate the switching of the working modes of DERs to ensure system voltage security. The time scale of the switching control is at the hour or minute level; Step 1.3 The underlying agent - the hybrid agent, including a reactive layer and a negotiation layer; the reactive layer is set to "perception-action" and has priority to quickly respond to emergencies; the reactive layer has priority to quickly respond to emergencies, and the negotiation layer can control or guide the behavior of the agent.

3. The method for optimizing matching and coordinated switching control of distributed controllable resource operation modes under different power deficits according to claim 1, characterized in that: In Step 2, the hybrid automaton mainly consists of a discrete state space, an enabling state, and a continuous state; in each discrete state space, there is a corresponding change in the continuous state. When a certain enabling state is satisfied, the hybrid automaton migrates from one discrete state space to another discrete state space, and the corresponding continuous state also changes; The hybrid automaton model is represented by a 6-tuple: H = (D, L, f, S, F, I); where, D = {δ1, δ2,...} represents a set of a series of discrete state spaces; L represents a set of a series of continuous state spaces; f = {f1(δ1), f2(δ2), f3(δ3),...} represents the change rule of the continuous state space under each discrete state space; S = {S1, S2, S3,...} represents the mapping between the discrete state space and the continuous state space; F = {F1, F2, F3,...} represents the condition for state space transition; I represents the initial state.

4. The method for optimizing matching and coordinated switching control of the operating modes of distributed controllable resources under different power deficits according to claim 1, characterized in that: Step 3. Build a hybrid system model for different DERs, including: Step 3.1 Build a hybrid automaton model for the photovoltaic power generation unit: Wherein, P pv is the output power of the photovoltaic power generation unit; P mppt is the output power of the photovoltaic power generation unit operating in the maximum power point mode; L is the light intensity; M is the light intensity threshold, which is determined by the performance of the photovoltaic cell; for the photovoltaic output characteristics, the photovoltaic power generation unit is set to two working modes: the shutdown mode and the MPPT mode, and the two working modes are set as the discrete states of the hybrid automaton model of the photovoltaic power generation unit; the power output value of the photovoltaic power generation unit is set as the continuous state; the change of the light intensity is set as the transfer condition; Step 3.2 Build a hybrid automaton model for the wind power generation unit: Wherein, P w is the output power of the wind power generation unit; P e is the rated output power of the wind power generation unit; v is the real-time wind speed; v i is the cut-in wind speed; v r is the rated wind speed; v o is the cut-out wind speed; according to the operating characteristics of wind power generation, its working modes are divided into three types: shutdown mode, MPPT mode and P e mode; according to the hybrid automaton model, the three working modes are set as the discrete states of the hybrid automaton model of the wind power generation unit; the power output value of the wind power generation unit is set as the continuous state; the wind speed change is set as the transition condition; Step 3.3 Build a hybrid automaton model for the energy storage unit: According to the characteristics of the battery itself, it is divided into five working modes: charging mode, discharging mode, and three outage modes; among them, the outage modes include normal outage, overcharge outage, and overdischarge outage; According to the hybrid automaton model, when SOC ≤ SOC down the battery switches from the discharge mode to the over-discharge shutdown mode, and when there is a relevant switching command, it will switch from the over-discharge shutdown mode to the charging mode again; When the state of charge (SOC) up ≤ SOC, the battery will switch from the charging mode to the overcharge shutdown mode. When there is a relevant switching command, it will switch from the overcharge shutdown mode to the discharging mode again; When the state of charge (SOC) down <SOC<SOC up the battery will switch between the discharge mode, the charging mode and the normal shutdown mode according to the system requirements; Step 3.4 Build a hybrid automaton model for the fuel cell: According to the hybrid automaton model, design two working modes for the fuel cell: outage mode, rated output mode; Step 3.5 Build a hybrid automaton model for the load: The load is divided into two categories: interruptible load and non-interruptible load. The power supply priority of the interruptible load is higher than that of the non-interruptible load; according to the hybrid automaton model, design two working modes for the load unit: normal operation mode, load shedding mode; Step 3.6 Set the initial state of the hybrid automaton model of all units, activate the discrete state space, set it to logical "1", and set the remaining discrete state spaces to logical "0"; thus, we can get Photovoltaic power generation unit: D = {δ1, δ2} = [1, 0]; Wind turbine unit: D = {δ1, δ2, δ3} = [1, 0, 0]; Battery unit: D = {δ1, δ2, δ3, δ4, δ5} = [1, 0, 0, 0, 0]; Fuel cell unit: D = {δ1, δ2} = [1, 0]; Load unit: D = {δ1, δ2} = [1, 0].

5. The optimization matching and coordinated switching control method for the operating modes of distributed controllable resources under different power deficits according to claim 1, characterized in that: Divide the local control strategy into an internal switching control strategy and a continuous dynamic management strategy, including: Step 4.1 For the photovoltaic power generation unit, the internal switching control command is as follows: For the wind power generation unit, the internal switching control command is as follows: The control command for the battery unit is as follows: Step 4.2 designs a dual-loop control method based on SPWM for the inverters of DERs, where the outer loop is designed as a controller using droop control, and the inner loop is designed as a controller in the dq rotating coordinate system.

6. The method for optimizing matching and coordinated switching control of distributed controllable resource operation modes under different power deficit amounts according to claim 1, wherein: Step Five: Construct a system voltage security assessment index and perform hierarchical processing. According to the system voltage security assessment results, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units, including: Step 5.1 Use the wide-area signal measurement method to extract the voltage sequence of the i-th bus in the distribution network system, denoted as After that, the average value of the instantaneous voltage of the i-th node at the j-th moment obtained from the voltage sequence is expressed as follows: The voltage deviation value of the i-th node at the j-th moment is expressed as: Wherein, is the actual voltage value of the i-th node at the j-th moment; Therefore, the voltage security assessment index of the i-th node at the j-th moment can be expressed as: Step 5.2 Fuse the voltage security assessment indexes of each node using the information fusion method based on the T-S fuzzy neural network; Use Instead of x=(x1,x2,…,x n ) as the input of the fuzzy neural network, the output of the fuzzy neural network can be normalized to the required voltage security assessment index; the output of the fuzzy neural network is expressed as: In the formula, is the weight of the fuzzy neural network; θ j is the product of membership degrees; the comprehensive voltage security evaluation index of each node obtained after information fusion is denoted as u; Step 5.3 Perform hierarchical processing on the assessment index u: When 0.9 ≥ u or u ≥ 1.1, it is a voltage collapse state, and the distribution network system stops operating; When 0.95 > u ≥ 0.9, the distribution network system has insufficient energy supply for the load, and the fuel cell works or load shedding operations are performed; When 0.98 > u ≥ 0.95, the distribution network system has insufficient energy supply for the load, and the battery performs a discharging operation to balance the power; When 1.02 > u ≥ 0.98, the system voltage is within the normal fluctuation range; When 1.05 > u ≥ 1.02, the energy supply of the distribution network system is greater than the load demand, and the battery performs a charging operation to balance the power; When 1.1 > u ≥ 1.05, the energy supply of the distribution network system is much greater than the load demand, and the fuel cell switches to the shutdown mode or restores the interruptible load; Step 5.4 fully considers the trigger duration, switching sequence, and time interval, and uses as the trigger event, and u and SOC as the trigger conditions; for the design is as follows: c1: c2: c3: c4: c5: c6: c7: c8: c9: c 10 : c 11 : c 12 : c 13 : c 14 : c 15 : where c n is the label set, and n ∈ [1, 2, …, 15]; is the trigger duration; Δt is the switching time interval; With the above coordinated control commands, flexible regulation is performed for the changes in the actions of DERs in the distribution network.

7. The method for optimizing matching and coordinated switching control of distributed controllable resource operation modes under different power deficits according to claim 1, characterized in that: Step Six: Use the selective mode-based aggregated single-layer dependence classification algorithm, use the coordinated control command set as the label set, and select the best coordinated switching control command in the control command set by performing classification operations, including: Step 6.1 Attribute and attribute value mining: Use the selective mode-based aggregated single-layer dependence classification algorithm to perform classification operations. Use the values of U and SOC during the operation of the distribution network system as continuous attributes, and use the working modes of the battery, fuel cell, and load as discrete attributes, and perform attribute mining on this; the attribute set is: X = {(U, a1), (SOC, a2), (Battery, a3), (Fuel, a4), (Load, a5)}, and a1 to a5 are the attribute values corresponding to the five attributes; Step 6.2 Process data of continuous attributes: The sample space of U is divided into seven continuous intervals, and with reference to fuzzy control, seven levels are defined from low to high as seven set values: "VL", "ML", "L", "Z", "H", "MH", "VH"; Similarly, the sample space of SOC is divided into five continuous intervals, and five levels are defined from low to high as five set values: "VL", "L", "Z", "H", "VH"; Process data of discrete attributes: For the battery unit, the sample space is set as {0, 1, 2}, where "0" represents the charging mode, "1" represents the discharging mode, and "2" represents the outage mode; For the fuel cell unit, the sample space is set as {0, 1}, where "0" represents the rated output mode and "1" represents the outage mode; For the load unit, the sample space is set as {0, 1}, "0" represents the normal operation mode, and "1" represents the load shedding mode; Step 6.3 Assume that the dataset D has n attributes, and the training instance is represented as X = (a1, a2 … a n ), where a i , 1 ≤ i ≤ n, is the value of the instance X on the i-th attribute; the class of the training instance belongs to one of the classes C = {c1, c2 … c m}, and hereinafter c is used to represent the class of the instance; the probability P(c|X) that the class of X is c can be expressed as: Let the set of attributes included in set f be {a1, a2, …, a i}, and the Bayesian network is obtained: Apply the pattern classification ability to the Bayesian network, aggregate the situations corresponding to all patterns, and obtain the Bayesian probability prediction formula based on selectivity: Step 6.4 To weaken the dependence relationship between the attributes of the selective pattern and other attributes in the Bayesian algorithm, adopt the aggregated single-layer dependence classification algorithm based on the selective pattern, and its final probability prediction formula is: where \(H = \{j|i + 1\leq j\leq n^F(a j )\geq m\}\), \(F(a j )\) is the number of training instances whose attribute values contain \(a j , restricted by the parameter \(m\) to achieve the support degree required for conditional probability estimation; Step 6.5 Based on the aggregated single-layer dependence classification algorithm of the selective pattern, after selecting the corresponding attributes, mine the attribute values according to the data generated by the operation of the distribution network. At the same time, use the coordinated control command set as the label set, and through performing classification operations, realize the selection of the optimal coordinated switching control command.

8. A distributed controllable resource operation mode optimization matching and coordinated switching control device under different power deficits, characterized in that, The device includes: Architecture establishment module: Used to establish a distribution network hierarchical architecture based on MAS for the coordinated control objective of the distribution network, and divide the distribution network control structure into an upper layer, a middle layer, and a lower layer; Model construction module: Used to add continuous variables on the basis of the concept of finite automata to construct a hybrid automata model represented by a 6-tuple array; Model establishment module: Used to establish corresponding hybrid automata models respectively for the operating characteristics of photovoltaic power generation units, wind power generation units, energy storage units, fuel cells, and loads; Management strategy module: Used to divide the local control strategy of the lower layer into an internal switching control strategy and a continuous dynamic management strategy. Among them, the internal switching control strategy is based on the coordinated control command of the middle layer, and the continuous dynamic management is a dual-loop control method based on SPWM; Coordinated control module: Used to design a coordinated control strategy in the multi-agent structure of the middle layer, construct a system voltage security evaluation index and perform hierarchical processing, and according to the system voltage security evaluation result, apply event-triggered hybrid control to coordinate the switching of the working modes of DERs and load units; Classification operation module: Used to use the aggregated single-layer dependence classification algorithm based on the selective pattern, use the coordinated control command set as the label set, and select the optimal coordinated switching control command in the control command set through performing classification operations; Multi-objective optimization module: For the upper-layer agent to construct a comprehensive objective function of reducing operation cost, reducing pollutant emissions and improving power quality considering power balance, output power of DERs and battery capacity constraints for different power deficits, and use the improved fruit fly optimization algorithm based on random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy; Step 7: The upper-layer agent constructs a comprehensive objective function of reducing operation cost, reducing pollutant emissions and improving power quality considering power balance, output power of DERs and battery capacity constraints for different power deficits, and uses the improved fruit fly optimization algorithm based on random walk to solve the multi-objective optimization problem and implement the upper-layer control strategy, including: Step 7.1: Construct the optimization scheduling objective function: Cost objective function: Where: i is the number of DERs; s is the operating mode of the DERs; α is is a parameter. When the distribution network system is in the operating mode, α is = 1; otherwise, α is = 0; r i is the fuel cost of the i-th DER. For renewable energy, r i = 0; is the active power output of the i-th DER in the s mode; E is is the consumption characteristic function; M i is the maintenance cost of the i-th DER, proportional to ; C i is the start-up cost of the i-th DER; β is is a cost parameter. When the i-th DER is operating in the s mode, β is = 1; otherwise, β is = 0; Carbon emission objective function: Where: F2: actual carbon emissions during the operation of the distribution network; π1: carbon emission intensity per unit active power output of thermal power units; π2: carbon emission intensity per unit active power output of gas turbines; P in.t : electricity purchased from the main grid per unit time period; P out.t : power generation of gas turbines within unit time period t; carbon emissions of clean energy are 0; Power quality objective function: Where: σ is is the power quality coefficient of the i-th DERs in the s mode; σ is is determined by the voltage security assessment index; Step 7.2: Construct the constraint conditions: Power balance constraint: P l = P pv + P w + P F + (-1) n P bat (14) Wherein, P l is the load demand of the system; P pv is the output power of the photovoltaic power generation unit; P w is the output power of the wind power generation unit; P bat is the power output / absorbed by the battery; n ∈ [0, 1], when the battery is in the charging mode, n = 1, and when the battery is in the discharging mode, n = 0; P F is the output power of the fuel cell; Output power constraint of DERs: P i down <P i <P i up (15) where P i is the output power of the i-th DERs; P i down is the minimum output power of the i-th DERs; P i up is the maximum output power of the i-th DERs; Battery capacity constraint: SOC down <SOC<SOC up (16) Where SOC is the state of charge of the battery; SOC down is the minimum capacity state value of the battery; SOC up is the maximum capacity state value of the battery; Step 7.3: Based on an improved fruit fly optimization algorithm for the implementation of the optimization scheduling; Step 7.4: Deal with the multi-objective optimization problem in the form of weight coefficients. The energy management can be summarized as solving the following objective optimization function that satisfies the constraint conditions: In the formula, ω1, ω2, and ω3 represent the weight coefficients of each objective function, and ω1 + ω2 + ω3 = 1.

9. A distributed controllable resource operation mode optimization matching and coordinated switching control device under different power deficits, characterized in that, Including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 7.

Images