Coordinated control method and system for integrated energy multi-agent collaborative group control and autonomous decision-making

Abstract

The coordination control method and system of comprehensive energy multi-agent collaborative group control and autonomous decision-making comprises: establishing a relaxation strategy of an agent corresponding to a distributed power supply; a main coordinator determines a global coordination signal by iteratively solving a global optimization target of relaxation according to global operation data of a comprehensive energy system and operation data of each agent, so as to minimize total cost, maximize system energy efficiency, minimize carbon emissions and minimize a global penalty coefficient under global relaxation constraints; each agent takes the minimum operation cost, the minimum response deviation to the global coordination signal and the minimum deviation between the actual state and the reference state as control targets; an electricity price autonomous decision-making model and a response power boundary autonomous decision-making model of the global coordination signal are established; and the control target values of each agent are predicted based on the control targets, the electricity price autonomous decision-making model and the response power boundary autonomous decision-making model of each agent, so as to realize the trade-off between the global optimization target and local autonomy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of integrated energy system collaborative control, and particularly relates to a kind of integrated energy multi-agent collaborative group control and autonomous decision coordination control method and system. BACKGROUND

[0002] With the rapid development of distributed energy system, a large number of distributed power, energy storage, flexible load and other intelligent agent devices with autonomous decision-making capability are widely connected to the integrated energy system. Multi-agent collaborative group control and autonomous decision-making in integrated energy system represents an important development direction of energy system intelligence. Collaborative group control achieves system global optimization through hierarchical architecture, while autonomous decision-making gives each intelligent unit the ability to independently respond to local uncertainty and optimize its own operation. In an integrated energy system with multiple agents, the key contradiction is that both the system's overall optimization goals of minimizing total cost, maximizing energy efficiency and minimizing carbon emissions must be achieved, and the local operation constraints and decision-making freedom of each intelligent unit must be fully respected. The key is to establish an efficient two-way regulation mechanism to achieve the coordination and unity of "centralized optimization guidance" and "distributed autonomous decision-making".

[0003] In the prior art, integrated energy multi-agent collaborative control methods are mainly divided into centralized, distributed and hybrid control, and all have some limitations. Centralized optimization relies on global scheduling by a central controller, which can achieve theoretical optimal solution, but requires full data sharing, resulting in heavy communication burden, and faces the risk of single point failure and privacy leakage, making it difficult to adapt to the plug-and-play needs of distributed energy. Completely distributed optimization uses Nash game and other models to coordinate the interests of multiple agents, but it is easy to fall into local optimum, and the convergence depends on the initial conditions, making it difficult to ensure global stability. Traditional hybrid control methods combine centralized and distributed architectures, but the collaboration between levels is insufficient; the global layer optimization lags behind local disturbances, and the intelligent agent role master-slave fault dynamic switching problem is not considered, resulting in insufficient robustness.

[0004] In summary, whether using centralized optimization control, distributed optimization control or traditional hybrid control method, it is difficult to effectively balance global optimization goals and local autonomy, resulting in high system total cost or low decision-making freedom of each intelligent unit. SUMMARY

[0005] To solve the problems in the prior art, the present application provides an integrated energy multi-agent collaborative group control and autonomous decision coordination control method and system, which can deeply integrate global collaborative optimization and local autonomous decision-making, and has high robustness and dynamic adaptability, achieving the lowest total cost of the system while ensuring the operation constraints and decision-making freedom of each intelligent unit, solving the problem of balancing global optimization goals and local autonomy.

[0006] The application adopts the technical solutions as follows.

[0007] The application provides a coordinated control method of comprehensive energy multi-agent collaborative group control and autonomous decision-making.

[0008] A master coordinator is determined from all agents based on a capability evaluation function of each agent.

[0009] A relaxation strategy of the agent corresponding to the distributed power source is established, including a relaxation boundary of electric power and a dynamic penalty coefficient of the relaxation boundary; a normalized value of an average value of the dynamic penalty coefficients of all agents corresponding to the distributed power sources is taken as a global penalty coefficient; the master coordinator takes the minimum total cost, the highest system energy efficiency, the minimum carbon emission and the minimum global penalty coefficient under the global relaxation constraint as the global optimization target of relaxation based on global operation data of the comprehensive energy system and operation data of each agent; and the master coordinator iteratively solves the global optimization target of relaxation to determine a global coordination signal, including a global virtual electricity price signal, a global power elasticity lower boundary and an elasticity upper boundary, and an upper limit of global carbon emission.

[0010] Each agent takes the minimum operation cost, the minimum response deviation to the global coordination signal and the minimum deviation between the actual state and the reference state as the control target; an electricity price autonomous decision-making model of each agent is established; a response power boundary autonomous decision-making model of each agent to the global coordination signal is established; each agent is subjected to model predictive control based on the control target, the electricity price autonomous decision-making model and the response power boundary autonomous decision-making model of each agent, and the control target value of each agent is predicted.

[0011] Each agent sends the control target value to the corresponding device according to a control period; each agent sends feedback data of the corresponding device after executing the control target value to the master coordinator, and the master coordinator evaluates the operation state of the comprehensive energy system in combination with external information; when it is determined that the control target value of each agent needs to be adjusted according to the evaluation result of the operation state of the comprehensive energy system, the master coordinator triggers rolling optimization to update the global coordination signal.

[0012] Capability evaluation function As follows:

[0013]

[0014] In the formula, , , are weight coefficients, CP is a calculation capability index of the agent, CR is a communication reliability index of the agent, and PA is a power supply stability index of the agent.

[0015] Establish a two-layer hybrid architecture for the integrated energy system, which includes a global coordination layer and an autonomous decision-making layer. The global coordination layer includes at least one master coordinator, and the autonomous decision-making layer includes multiple intelligent agents.

[0016] The monitoring module located in the global coordination layer monitors the current state of the master coordinator in real time; if the monitoring module detects that the master coordinator has failed, it triggers the master coordinator update mechanism to determine a new master coordinator from all agents based on the capability evaluation function.

[0017] If the monitoring module does not detect a failure of the main coordinator, the main coordinator acquires the global operating data of the integrated energy system and the operating data of each agent.

[0018] The smaller of the standard deviation of the output power of distributed power sources and the upper limit of the power of the integrated energy system is used as the standard deviation. Relaxation boundary of power during time period :

[0019]

[0020] In the formula, for The standard deviation of the output power of distributed power sources over a given period of time. for The upper limit of the power output of the integrated energy system during a given time period. It is a relaxation factor;

[0021] Using the rated power and output power fluctuation of the distributed power source, the dynamic penalty coefficient for the relaxation boundary is determined as follows:

[0022]

[0023] In the formula, for The dynamic penalty coefficient of the time-period relaxation boundary. This is the initial value for the dynamic penalty coefficient. The fluctuation sensitivity coefficient for Fluctuation in the output power of distributed power sources over time periods This refers to the rated power of the distributed power source.

[0024] Total cost function The following relationship must be satisfied:

[0025]

[0026] In the formula, and They are respectively The price of electricity purchased from the grid and the price of natural gas during the specified time period. and are the power purchased from the grid and the natural gas consumption in the time period, is a set of time periods, is the total cost of the day-ahead schedule;

[0027] the system energy efficiency function satisfies the following relation:

[0028]

[0029] where, is the energy efficiency coefficient of the energy conversion equipment in the system, is a set of energy conversion equipment in the system, , is the useful energy output and input of equipment in the time period;

[0030] the carbon emission function satisfies the following relation:

[0031]

[0032] where, is the carbon emission factor of the grid power in the time period, is the carbon emission factor of the natural gas, is the average value of the historical carbon emission of the system;

[0033] the global relaxation constraint function of the time period satisfies the following relation:

[0034]

[0035] where, , are the lower and upper bound relaxation variables corresponding to the global thermal power balance constraint in the time period, , are the lower and upper bound relaxation variables corresponding to the global electric power balance constraint in the time period, is the relaxation variable corresponding to the global carbon emission constraint in the time period; the relaxed global optimization objective

[0036] is as follows:

[0037] ​​​

[0038] wherein, , , are the weight coefficients of the total cost function, the system energy efficiency function and the carbon emission function, satisfying + + = 1, is the global penalty coefficient.

[0039] Global relaxation constraints include:

[0040] 1) Global electrical power balance constraint, as follows:

[0041]

[0042] wherein, is the output power of the distributed power source in the time period, is the set of energy storage devices, is the energy storage device in the time period, is the electrical load demand in the time period, and are the lower and upper limits of the electrical power of the integrated energy system in the time period;

[0043] 2) Global thermal power balance constraint, as follows:

[0044]

[0045] wherein, is the thermal conversion efficiency, is the thermal power in the time period, is the thermal load demand in the time period, , are the lower and upper limits of the thermal power of the integrated energy system in the time period;

[0046] 3) Global carbon emission constraint, as follows:

[0047]

[0048] wherein, is the upper limit of the total amount of carbon emissions allowed by the integrated energy system.

[0049] The global coordination signals satisfy the following relationships, respectively:

[0050]

[0051]

[0052]

[0053] In the formula, for Global virtual electricity price signal for the time period for Lagrange multipliers of power balance constraints during time periods , They are respectively The lower and upper boundaries of the global power elasticity for a given time period; , They are respectively The optimal relaxation variables corresponding to the lower and upper bounds of the power balance constraint during the time period. This represents the upper limit of global carbon emissions. for The optimal slack variable corresponding to the global carbon emission constraint for a given time period.

[0054] The control objective satisfies the following relationship:

[0055]

[0056] In the formula, For the agent to predict the steps Operating costs, including but not limited to equipment operating costs and depreciation costs; The benchmark value for the operating cost of the intelligent agent is the average operating cost of the intelligent agent during the same period on a typical day; For the agent to predict the steps Response deviation to global coordination signals; The baseline value for the response deviation is the maximum power deviation allowed by the agent; For the agent to predict the steps The deviation between the actual state and the reference state; , , These are the weighting coefficients for the agent's operating cost, responsiveness to global coordination signals, and deviation between the actual state and the reference state, respectively. To predict the time domain length, u is the control variable vector.

[0057]

[0058] In the formula, For the agent to predict the steps The equipment operating costs, For the agent to predict the steps Equipment depreciation costs; the global virtual price signal at the prediction time step ; the algebraic sum of the total power of all distributed power sources, energy storage discharges, and loads within the jurisdiction of the agent at the prediction time step , which is positive if the agent is delivering power to the integrated energy system network and negative if the agent is absorbing power from the network; the length of the optimization period; the dynamic penalty coefficient for violating the elasticity boundary, the deviation of the power autonomously decided by the agent at the prediction time step from the global power elasticity lower boundary or the global power elasticity upper boundary; , the actual state vector and the reference state vector of the agent at the prediction time step , respectively, the weighted Euclidean norm, and Q is a diagonal weight matrix, the elements on the diagonal of the matrix representing the importance of the tracking error of the state variables in the state vector in the objective function.

[0059] The price autonomous decision-making model of each agent is as follows:

[0060]

[0061] In the formula, the actual price signal (unit: yuan / kWh) of the agent at the prediction time step ; the fluctuation amount (kW) of the distributed power output at the prediction time step ; the price correction amplitude factor; the fluctuation threshold of the distributed power output, which in an embodiment is in the range of ; the sign function, outputting 1 when the fluctuation amount of the distributed power output is positive and outputting -1 when the fluctuation amount of the distributed power output is negative.

[0062] The response power boundary autonomous decision-making model of each agent to the global coordination signal is as follows:

[0063]

[0064]

[0065] In the formula, , the lower boundary and the upper boundary of the response power of the agent to the global coordination signal, respectively; , the global power elasticity lower boundary and the global power elasticity upper boundary at the prediction time step , respectively. is a prediction time step is a load prediction standard deviation; is a power adjustment coefficient.

[0066] The feedback data includes: the running state of the equipment, the actual running parameter of the equipment, and the disturbance data of the equipment;

[0067] The external information includes: grid dispatching instructions and market price updates;

[0068] The running state of the comprehensive energy system is evaluated, including:

[0069] 1) if the value of the system energy efficiency function is greater than 5%, it is determined that a major disturbance occurs;

[0070] 2) if the value of the total cost function is greater than 1.05 times of the total cost of the day-ahead plan running , and the value of the carbon emission function is greater than 1.05 times of the average value of the historical carbon emissions of the system , it is determined that a global target deviation risk occurs;

[0071] 3) if the value of the relaxed global optimization target is less than 90%, it is determined that the coordination signal does not match.

[0072] The application further provides a coordination control system for comprehensive energy multi-agent collaborative group control and autonomous decision-making, comprising:

[0073] An architecture building module is configured to set one agent corresponding to each device, determine a master coordinator from all agents based on a preset capability evaluation function of each agent;

[0074] A global coordination module is configured to establish a relaxed strategy of the agent corresponding to the distributed power source, including: a relaxed boundary of electric power and a dynamic penalty coefficient of the relaxed boundary; a normalized value of the average value of the dynamic penalty coefficients of all agents corresponding to the distributed power source is taken as a global penalty coefficient; the master coordinator takes the minimum total cost, the highest system energy efficiency, the minimum carbon emission, and the minimum global penalty coefficient under the global relaxed constraint as the relaxed global optimization target according to the global running data of the comprehensive energy system and the running data of each agent; and the master coordinator iteratively solves the relaxed global optimization target to determine a global coordination signal, including: a global virtual electricity price signal, a global power elasticity lower boundary and elasticity upper boundary, and a global upper limit of carbon emission;

[0075] An autonomous decision module is configured to take the minimum operation cost of each intelligent agent, the minimum response deviation of each intelligent agent to a global coordination signal, and the minimum deviation of an actual state from a reference state as control targets, to establish a price autonomous decision model of each intelligent agent, to establish a response power boundary autonomous decision model of each intelligent agent to the global coordination signal, and to perform model predictive control on each intelligent agent based on the control targets of each intelligent agent, the price autonomous decision model, and the response power boundary autonomous decision model of each intelligent agent to predict the control target values of each intelligent agent.

[0076] A feedback control module is configured to perform the following operations: each intelligent agent sends the control target values to corresponding devices according to a control period; each intelligent agent sends feedback data of the corresponding devices after performing the control target values to the main coordinator, which combines external information to evaluate the operation state of the comprehensive energy system; and when it is determined according to the evaluation result of the operation state of the comprehensive energy system that the control target values of each intelligent agent need to be adjusted, the main coordinator triggers rolling optimization to update the global coordination signal.

[0077] The application also provides a terminal, including a processor and a storage medium, the storage medium is used for storing instructions, and the processor is used for operating according to the instructions to perform the steps of the method.

[0078] The application also provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the steps of the method.

[0079] The application has the advantages that, compared with the prior art, the technical contradictions of single-point failure of centralized control, insufficient global optimization of distributed control, and response lag of hybrid control are solved by constructing a hierarchical-distributed hybrid architecture and a dynamic role division mechanism; an elastic boundary constraint model is adopted to generate a dynamic power boundary and a virtual price signal containing a slack variable, thereby providing a flexible decision space for lower intelligent agent proxies (APs); meanwhile, global signals and local real-time data are fused, model predictive control is performed through a dynamically corrected virtual price and an adaptive power boundary, the local decision freedom is significantly improved under the premise of ensuring the lowest global total cost and controlled low-carbon emissions; in addition, based on the deviation evaluation of AP feedback and MC targets, dynamic signal fine-tuning or architecture reconstruction is triggered, a two-way regulation mechanism of "elastic boundary optimization-autonomous decision execution-closed-loop feedback correction" is formed, and the bottleneck that global optimization and local autonomy are difficult to be compatible in the prior art is fundamentally broken through. BRIEF DESCRIPTION OF DRAWINGS

[0080] Figure 1 is a flow chart of a comprehensive energy multi-agent collaborative group control and autonomous decision coordination control method proposed by the application. DETAILED DESCRIPTION

[0081] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present application. The embodiments described in the present application are only a part of the embodiments of the present application, rather than all the embodiments. Based on the spirit of the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0082] The present application provides a coordinated control method for comprehensive energy multi-agent collaborative group control and autonomous decision-making, as shown in Figure 1 , comprising:

[0083] Step S1, obtaining various devices of a comprehensive energy system, each device corresponding to an agent, determining a master coordinator from all agents based on a preset ability evaluation function of each agent.

[0084] Specifically, step S1 comprises:

[0085] Step S11, obtaining various devices of a comprehensive energy system, each device corresponding to an agent;

[0086] The various devices of the comprehensive energy system include but are not limited to: distributed power supply, energy storage unit, flexible load, microgrid controller;

[0087] Step S12, determining a master coordinator from all agents based on a preset ability evaluation function of each agent.

[0088] The ability evaluation function is constructed based on the computing power, communication reliability, power supply stability and other indexes of the agent, and the agent corresponding to the maximum of the ability evaluation function is selected as the master coordinator, and the number of the determined master coordinator is at least one;

[0089] Specifically, the ability evaluation function is as follows:

[0090]

[0091] In the formula, , , are weight coefficients, and in the embodiments =0.4, =0.4, =0.2; CP is the computing power index of the agent; CR is the communication reliability index of the agent, and the reciprocal of the recent communication packet loss rate and delay is adopted; PA is the power supply stability index of the agent, and the remaining power or the Boolean value of whether to access a stable power grid is adopted.

[0092] Step S13, a double-layer hybrid architecture of the comprehensive energy system is established, including a global coordination layer and an autonomous decision layer, the global coordination layer includes at least one master coordinator, and the autonomous decision layer includes a plurality of agents; the double-layer hybrid architecture is specifically as follows:

[0093] 1) The global coordination layer (Global Coordinator Layer, GCL) is composed of one or more master coordinators (Master Coordinator, MC) with strong computing power; the MC is responsible for setting the system-level optimization target, maintaining the global constraint, generating the key coordination signal, and monitoring the overall operation state and the agent role state of the system;

[0094] 2) The autonomous decision layer (Autonomous Decision Layer, ADL) is composed of a plurality of agents (Agent Proxy, AP); each AP represents a physical entity, in the embodiment, AP1 represents a distributed power supply, AP2 represents an energy storage unit, AP3 represents a flexible load, AP4 represents a microgrid controller, and is embedded with a local controller.

[0095] Step S14, a monitoring module located in the global coordination layer monitors the current state of the MC in real time; if the monitoring module detects that the MC is invalid, the MC updating mechanism is triggered, that is, returning to step S12 to determine a new MC from all APs; if the monitoring module does not detect that the MC is invalid, step S15 is entered;

[0096] In the embodiment, all APs and MCs periodically broadcast their current state information to other nodes in the communication network, including but not limited to: self-computing load, communication state, power supply state; the monitoring module located in the global coordination layer is responsible for monitoring the current state of the MC, realizing role state monitoring and failure detection; when the monitoring module detects that the MC is invalid, the MC updating mechanism is triggered immediately, realizing dynamic switching between the master coordinator and the agent; wherein the MC failure includes but is not limited to: communication interruption, computing timeout; based on the ability evaluation function of the agent, a new master coordinator is determined from all agents.

[0097] Step S15, the master coordinator obtains global operation data of the comprehensive energy system and operation data of each agent;

[0098] In the embodiment, the MC extracts the global operation data of the comprehensive energy system from the global state snapshot, and requests each AP to report the current state and operation data; the global operation view of the comprehensive energy system is established based on the global operation data and the operation data of all APs. The global operation view includes but is not limited to: load prediction, distributed power supply output prediction, market price, and key AP state report.

[0099] Step S2: Establish relaxation strategies for agents corresponding to distributed power sources, including: relaxation boundaries of electrical power and dynamic penalty coefficients of relaxation boundaries; use the normalized value of the average dynamic penalty coefficients of all agents corresponding to distributed power sources as the global penalty coefficient; the main coordinator, based on the global operation data of the integrated energy system and the operation data of each agent, sets the global optimization objectives for relaxation as the lowest total cost, highest system energy efficiency, lowest carbon emissions, and the lowest global penalty coefficient under global relaxation constraints; the main coordinator iteratively solves the global optimization objectives for relaxation to determine the global coordination signals, including: global virtual electricity price signals, global power elasticity lower and upper boundaries, and the upper limit of global carbon emissions.

[0100] Specifically, step S2 includes:

[0101] Step S21: Establish the relaxation strategy for the agent corresponding to the distributed power source, including: the relaxation boundary of the power and the dynamic penalty coefficient of the relaxation boundary; use the normalized value of the average dynamic penalty coefficient of all agents corresponding to the distributed power source as the global penalty coefficient.

[0102] The smaller of the standard deviation of the output power of distributed power sources and the upper limit of the power of the integrated energy system is used as the standard deviation. Relaxation boundary of power during time period :

[0103]

[0104] In the formula, for The standard deviation of the output power of distributed power sources over a given period of time. for The upper limit of the power output of the integrated energy system during a given time period. As a relaxation factor, in the examples It is 2.0;

[0105] Using the rated power and output power fluctuation of the distributed power source, the dynamic penalty coefficient for the relaxation boundary is determined as follows:

[0106]

[0107] In the formula, for The dynamic penalty coefficient of the time-period relaxation boundary. This is the initial value for the dynamic penalty coefficient. The fluctuation sensitivity coefficient is used in the example. It is 1.5. for Fluctuation in the output power of distributed power sources over time periods This refers to the rated power of the distributed power source.

[0108] The normalized value of the average dynamic penalty coefficient of all agents corresponding to distributed power sources is used as the global penalty coefficient. The global penalty coefficient characterizes the uncertainty of new energy output.

[0109] The relaxation strategy of the agent corresponding to the distributed power source provides the AP corresponding to the distributed power source in the autonomous decision-making layer with flexible space for autonomous decision-making, which is also in line with the volatility and flexibility of the distributed power source. Moreover, the relaxation strategy introduced by the master coordinator when performing global optimization is the basis for realizing the coordinated control of multi-agent collaborative group control and flexible autonomous decision-making under the global optimum for the distributed power source.

[0110] Step S22: The main coordinator, based on the global operation data of the integrated energy system and the operation data of each agent, sets the global optimization objective as the lowest total cost, the highest system energy efficiency, the lowest carbon emissions, and the lowest global penalty coefficient under global relaxation constraints.

[0111] Specifically, the total cost function The following relationship must be satisfied:

[0112]

[0113] In the formula, and They are respectively Electricity price purchased from the grid during the specified time period (unit: yuan / kWh) and Natural gas price per hour (unit: yuan / ) ), and They are respectively Power purchased from the grid during the specified time period (unit: kW) and Natural gas consumption per period (unit: ), For time period sets, The total planned operating cost (in yuan) is used as the baseline value for the total cost.

[0114] Specifically, the system energy efficiency function The following relationship must be satisfied:

[0115]

[0116] In the formula, Energy conversion equipment in the system Energy efficiency coefficient, It is a collection of energy conversion devices in the system. For equipment exist Useful energy output during a given period (unit: kWh). for the device In the energy input of the period (unit: kWh);

[0117] Specifically, the carbon emission function satisfies the following relationship:

[0118]

[0119] In the formula, is the carbon emission factor of the grid power of the period (unit: kgCO2 / kWh), is the carbon emission factor of natural gas (unit: kgCO2 / ), is the average value of the historical carbon emission of the system (unit: kgCO2), as the carbon emission reference value;

[0120] Specifically, the global relaxation constraint function of the period satisfies the following relationship:

[0121]

[0122] In the formula, , are the lower and upper bounds of the relaxation variable corresponding to the global thermal power balance constraint of the period, , are the lower and upper bounds of the relaxation variable corresponding to the global electric power balance constraint of the period, is the relaxation variable corresponding to the global carbon emission constraint of the period;

[0123] The total cost function, the system energy efficiency function, and the carbon emission function are respectively assigned weight coefficients , , satisfying + + =1; is the weight coefficient of the global relaxation constraint function; the global optimization objective after relaxation is obtained as follows:

[0124]

[0125] ​The above formula reflects the goals of minimizing total cost, maximizing system energy efficiency, minimizing carbon emissions, and minimizing the global penalty coefficient under global relaxation constraints. Minimizing total cost, maximizing system energy efficiency, and minimizing carbon emissions are global optimization objectives that can significantly improve the performance of integrated energy systems. At the same time, minimizing the global penalty coefficient under global relaxation constraints fully demonstrates that when multi-agent collaborative group control is performed for distributed power sources, it is necessary not only to meet global constraints such as electrical power, thermal power, and carbon emissions, but also to make autonomous decisions that minimize individual penalties under the optimal global optimization objective, thus achieving a dynamic balance between global coordination and autonomous decision-making.

[0126] The global relaxation constraints include:

[0127] 1) Global power balance constraints, as follows:

[0128]

[0129] In the formula, for Output power of distributed power sources during specific time periods (unit: kW). A collection of energy storage devices, For energy storage devices exist Charging and discharging power during a given period (unit: kW). for Electricity load demand during specific time periods (unit: kW) and for Lower and upper limits of the power output of the integrated energy system during a given time period (unit: kW);

[0130] 2) Global thermal power balance constraints, as follows:

[0131]

[0132] In the formula, Thermal conversion efficiency (unit: kWh / ), for Thermal power over a period of time (unit: kWh) for Heat load demand during different time periods (unit: kWh) , They are respectively Lower and upper limits of thermal power of integrated energy system during a given time period (unit: kWh);

[0133] 3) Global carbon emission constraints, as follows:

[0134]

[0135] In the formula, The maximum allowable carbon emissions for a comprehensive energy system (unit: kgCO2).

[0136] In the embodiment, the slack variables obtained in the current iteration that satisfy the global power balance constraint are... and Relaxed variables that satisfy global thermal power balance constraints and Slack variables that satisfy global carbon emission constraints This feedback is then fed back to the relaxed global optimization objective for the next iteration of optimization.

[0137] Each slack variable satisfies the following constraints:

[0138]

[0139] In the embodiment, when MC has a period of 1 hour or when it detects a voltage deviation rate greater than 5% caused by a disturbance, it establishes a relaxed global optimization target based on the global operation data of the integrated energy system and the operation data of each intelligent agent.

[0140] Step S23: The master coordinator iteratively solves the relaxed global optimization objective to determine the global coordination signal, including: the global virtual electricity price signal. Global power elastic lower boundary and elastic upper boundary The upper limit of global carbon emissions ;

[0141] Specifically, the global coordination signals satisfy the following relationships:

[0142]

[0143]

[0144]

[0145] in, for Global virtual electricity price signal for the time period for Lagrange multipliers of power balance constraints during time periods , They are respectively The lower and upper boundaries of the global power elasticity for a given time period; , They are respectively The optimal relaxation variables corresponding to the lower and upper bounds of the power balance constraint during the time period. This represents the upper limit of global carbon emissions. for optimal relaxation variable corresponding to the period global carbon emission constraint;

[0146] The MC broadcasts a global coordination signal to all APs.

[0147] The application is based on a double-layer hybrid architecture, the main coordinator fuses the global information of the system and the local state information reported by each agent, performs rolling optimization and generates a global coordination signal containing elastic boundary constraints, global power marginal cost and carbon emissions.

[0148] Step S3, taking the minimum operating cost of each agent, the minimum response deviation to the global coordination signal, and the minimum deviation of the actual state from the reference state as the control target; establishing a price autonomous decision-making model of each agent; establishing a response power boundary autonomous decision-making model of each agent to the global coordination signal; based on the control target of each agent, the price autonomous decision-making model and the response power boundary autonomous decision-making model of each agent, performing model predictive control on each agent to predict the control target value of each agent.

[0149] Specifically, step S3 includes:

[0150] Step S31, taking the minimum operating cost of each agent, the minimum response deviation to the global coordination signal, and the minimum deviation of the actual state from the reference state as the control target, satisfying the following relationship:

[0151]

[0152] In the formula, is the operating cost of the agent at the prediction time step , including but not limited to equipment operation cost and loss cost; is the reference value of the operating cost of the agent, which is the average operating cost of the agent at the same period of a typical day; is the response deviation of the agent to the global coordination signal at the prediction time step ; is the reference value of the response deviation, which is the maximum power deviation allowed by the agent; is the deviation of the actual state from the reference state of the agent at the prediction time step ; , , are the weight coefficients of the operating cost, the response degree to the global coordination signal, and the deviation of the actual state from the reference state of the agent, respectively; is the prediction time domain length, and u is the control variable vector;

[0153] Specifically, the operating cost of the agent, the response degree to the global coordination signal, and the deviation of the actual state from the reference state satisfy the following relationship, respectively:

[0154]

[0155] wherein, is the equipment operation cost of the agent at the prediction time step , is the equipment wear cost of the agent at the prediction time step ; is the global virtual electricity price signal at the prediction time step ; is the total power algebra of all distributed power sources, energy storage discharges, and loads within the jurisdiction of the agent within the prediction time step , and the value is positive when the agent transmits power to the integrated energy system network and negative when the agent absorbs power from the network; is the length of the optimization period (unit: hour); is the dynamic penalty coefficient for violating the elastic boundary, is the deviation of the power autonomously decided by the agent at the prediction time step from the global power elasticity lower boundary or the global power elasticity upper boundary; , are the actual state vector and the reference state vector of the agent at the prediction time step , is the weighted Euclidean norm, and Q is the weight matrix;

[0156] Q is a diagonal weight matrix, and the elements on the diagonal are all greater than 0, representing the importance of the tracking error of the th state variable in the objective function in the state vector, and the greater the weight coefficient value, the greater the effort of the controller to reduce the tracking deviation of the state variable.

[0157] Step S32, an electricity price autonomous decision-making model of each agent is established, as follows:

[0158]

[0159] wherein, is the actual electricity price signal of the agent at the prediction time step (unit: yuan / kWh); is the fluctuation amount of the distributed power output at the prediction time step (kW); is the electricity price correction amplitude factor; is the fluctuation threshold of the distributed power output, is a sign function, and the output is 1 when the fluctuation amount of the distributed power output is positive and -1 when the fluctuation amount of the distributed power output is negative;

[0160] Different types of agents have different price constraints, especially considering the guiding influence of distributed power output power fluctuation on electricity price.

[0161] Step S33, establish the response power boundary autonomous decision model of each agent to the global coordination signal, as follows:

[0162]

[0163]

[0164] In the formula, , The lower boundary and the upper boundary of the response power of the agent to the global coordination signal (unit: kW) are respectively; , The global power elasticity lower boundary and the elasticity upper boundary of the predicted time step ; The load prediction standard deviation (unit: kW) of the predicted time step ; The power regulation coefficient.

[0165] Step S34, based on the control target of each agent, the autonomous decision model of electricity price and the autonomous decision model of the response power boundary, the model predictive control is carried out on each agent, and the control target value of each agent is predicted;

[0166] The control target value of each agent in the future several short periods predicted includes but is not limited to: the start-stop state of the device corresponding to each agent, the output power of the device corresponding to each agent;

[0167] Step S4, each agent sends the control target value to the corresponding device according to the control period; each agent sends the feedback data of the corresponding device after executing the control target value to the main coordinator, and the main coordinator combines the external information to evaluate the running state of the comprehensive energy system.

[0168] The feedback data includes but is not limited to: the running state of the device, the actual running parameter of the device, the disturbance data of the device;

[0169] The external information includes but is not limited to: power grid dispatching instruction, market price update;

[0170] Evaluating the running state of the comprehensive energy system includes:

[0171] 1) If the value of the system energy efficiency function is greater than 5%, it is determined that a major disturbance occurs;

[0172] 2) If the total cost function The value is greater than the total cost of the planned operation at the current date. 1.05 times that, and the carbon emission function The value is greater than the average historical carbon emissions of the system. When the value is 1.05 times the target, it is determined that there is a risk of deviation from the global target.

[0173] 3) If the global optimization objective is relaxed If the value is less than 90%, it is determined that the coordination signal is obviously mismatched.

[0174] When each agent periodically or based on event triggers reports critical status information to the master coordinator, the master coordinator triggers rolling optimization. This critical status information includes, but is not limited to: actual power, SOC, over-limit alarms, and local target achievement rate. When the operational status assessment results of the integrated energy system determine that adjustments to the control target values ​​of each agent are necessary, the master coordinator triggers rolling optimization in advance to update the global coordination signals, including:

[0175] 1) The master coordinator adjusts the global virtual electricity price signal, the lower and upper boundaries of global power elasticity, and the upper limit of global carbon emissions with a set fine-tuning step size, and broadcasts the adjustment to each agent.

[0176] 2) The master coordinator dynamically adjusts the weight coefficients in the relaxed global optimization objective, updates the global coordination signal, and broadcasts it to each agent.

[0177] This invention also proposes a coordinated control system for integrated energy multi-agent collaborative group control and autonomous decision-making, comprising:

[0178] The architecture building module is used to set up a corresponding agent for each type of device and to determine the main coordinator from all agents based on the preset capability evaluation function of each agent.

[0179] The global coordination module is used to establish relaxation strategies for agents corresponding to distributed power sources, including: relaxation boundaries of electrical power and dynamic penalty coefficients of the relaxation boundaries; the normalized value of the average dynamic penalty coefficients of all agents corresponding to distributed power sources is used as the global penalty coefficient; the main coordinator, based on the global operation data of the integrated energy system and the operation data of each agent, sets the global optimization objectives for relaxation as the lowest total cost, highest system energy efficiency, lowest carbon emissions, and the lowest global penalty coefficient under global relaxation constraints; the main coordinator iteratively solves the global optimization objectives for relaxation to determine the global coordination signals, including: global virtual electricity price signals, global power elasticity lower and upper boundaries, and global carbon emission upper limits;

[0180] The autonomous decision module is used for taking the minimum operation cost of each intelligent agent, the minimum response deviation of each intelligent agent to the global coordination signal, and the minimum deviation of the actual state from the reference state as control targets; an electricity price autonomous decision model of each intelligent agent is established; a response power boundary autonomous decision model of each intelligent agent to the global coordination signal is established; and model predictive control is performed on each intelligent agent based on the control targets of each intelligent agent, the electricity price autonomous decision model, and the response power boundary autonomous decision model of each intelligent agent, to predict the control target values of each intelligent agent.

[0181] The feedback control module is used for performing that each intelligent agent sends the control target values to the corresponding devices according to a control period; each intelligent agent sends feedback data of the corresponding devices after performing the control target values to the main coordinator, the main coordinator combines external information to evaluate the operation state of the comprehensive energy system; and when it is determined according to the evaluation result of the operation state of the comprehensive energy system that the control target values of each intelligent agent need to be adjusted, the main coordinator triggers rolling optimization to update the global coordination signal.

[0182] The present disclosure can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0183] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0184] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0185] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computing device, partly on the user's computing device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, for example, through the Internet using an Internet Service Provider. In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0186] Finally, it should be noted that the above-mentioned embodiments are merely intended for describing and illustrating, but not limiting the technical solutions of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A coordinated control method of integrated energy multi-agent collaborative group control and autonomous decision-making, each device corresponding to an agent, characterized in that, The method comprises the following steps: determining a main coordinator from all agents based on an ability evaluation function of each agent; the main coordinator solves the relaxed global optimization target iteratively to determine a global coordination signal, wherein the relaxed global optimization target is based on operation data of the integrated energy system and operation data of each agent, and the global optimization target is to minimize the total cost of the system, maximize energy efficiency, minimize carbon emissions, and minimize the global penalty coefficient under the global relaxation constraint; each agent establishes an electricity price autonomous decision-making model and a response power boundary autonomous decision-making model of the global coordination signal; 2. The method according to claim 1, wherein each agent sends the control target value to the corresponding device according to the control period, and sends feedback data of the corresponding device after executing the control target value to the main coordinator, and the main coordinator evaluates the operation state of the integrated energy system in combination with external information; when it is determined that the control target value of each agent needs to be adjusted according to the evaluation result of the operation state of the integrated energy system, the main coordinator triggers rolling optimization to update the global coordination signal. The various devices of the integrated energy system include: distributed power supply, energy storage unit, flexible load, micro-grid controller; each device corresponds to an agent; a capability evaluation function As follows: In the formula, , , are weight coefficients, CP is the computing power index of the agent, CR is the communication reliability index of the agent, and PA is the power supply stability index of the agent.

3. The method according to claim 2, wherein a double-layer hybrid architecture of the integrated energy system is established, which includes a global coordination layer and an autonomous decision-making layer, the global coordination layer includes at least one main coordinator, and the autonomous decision-making layer includes a plurality of agents; a monitoring module located in the global coordination layer monitors the current state of the main coordinator in real time, and triggers a main coordinator updating mechanism if the monitoring module detects that the main coordinator is invalid, and a new main coordinator is determined from all agents based on an ability evaluation function; if the monitoring module does not detect that the main coordinator is invalid, the main coordinator obtains global operation data of the integrated energy system and operation data of each agent.

4. The method according to claim 3, wherein Wherein, the smaller value between the standard deviation of the distributed power output power and the upper limit of the comprehensive energy system electric power is taken as Relaxation boundary of time period electric power : In the formula, is the standard deviation of the distributed power output power of the period, is the upper limit of the electric power of the integrated energy system of the period, is a relaxation factor; a relaxation boundary of electric power of the agent corresponding to the distributed power source and a dynamic penalty coefficient of the relaxation boundary are established, and a normalized value of an average value of the dynamic penalty coefficients of all agents corresponding to the distributed power source is taken as the global penalty coefficient; In the formula, is the dynamic penalty coefficient of the period relaxation boundary, is the initial value of the dynamic penalty coefficient, is the fluctuation sensitivity coefficient, is the fluctuation of the output power of the distributed power supply in the period, is the rated power of the distributed power supply. the dynamic penalty coefficient of the relaxation boundary is determined by using the rated power and output power fluctuation of the distributed power source, and is as follows: System total cost function satisfies the following relation: wherein and are respectively the price of electricity purchased from the grid and the price of natural gas for the period, and are respectively the power purchased from the grid and the natural gas consumption for the period, is a set of periods, is the total cost of the day-ahead schedule. System energy efficiency function satisfies the following relation: In the formula, Energy conversion equipment in the system Energy efficiency coefficient, It is a collection of energy conversion devices in the system. , For equipment exist Useful energy output and input during a given period; Carbon emission function satisfies the following relation: wherein is the carbon emission factor for grid electricity, is the carbon emission factor for grid electricity, is the carbon emission factor for natural gas, is the average value of the system historical carbon emissions; Global relaxation constraint function for time period satisfies the following relation: In the formula, , are respectively lower and upper bounds of the period global thermal power balance constraint corresponding to the relaxation variable, , are respectively lower and upper bounds of the period global electric power balance constraint corresponding to the relaxation variable, is a relaxation variable corresponding to the period global carbon emission constraint; Relaxed global optimization objective As follows: In the formula, , , are weight coefficients of the total cost function, the system energy efficiency function, and the carbon emission function, satisfying + + = 1, is a global penalty coefficient.

5. The method according to claim 4, wherein 6. The method according to claim 5, wherein the global relaxation constraint includes: wherein, is the output power of the distributed power source in the time period, is a set of energy storage devices, is an energy storage device in the charge and discharge power of the energy storage devices in the time period, is the electrical load demand in the time period, and is the lower and upper limits of the electrical power of the integrated energy system in the time period; 1) a global electric power balance constraint, as follows: In the formula, is the thermal conversion efficiency, is period thermal power, is period thermal load demand, , are lower and upper limits of the period comprehensive energy system thermal power; 2) a global thermal power balance constraint, as follows: 3) a global carbon emission constraint, as follows: In the formula, is the upper limit of the total amount of carbon emissions allowed for the integrated energy system.

7. The method of claim 6, wherein the global coordination signal comprises a global virtual electricity price signal, a global power elasticity lower bound and an elasticity upper bound, and a global upper limit of carbon emissions, and satisfies the following relationships:

8. The method of claim 1, wherein the control target satisfies the following relationship: In the formula, for Global virtual electricity price signal for the time period for Lagrange multipliers of power balance constraints during time periods , They are respectively The lower and upper boundaries of the global power elasticity for a given time period; , They are respectively The optimal relaxation variables corresponding to the lower and upper bounds of the power balance constraint during the time period. This represents the upper limit of global carbon emissions. for The optimal slack variable corresponding to the global carbon emission constraint for a given time period.

9. The method of claim 8, wherein the electricity price autonomous decision-making model of each agent is as follows:

10. The method of claim 9, wherein the response power boundary autonomous decision-making model of each agent to the global coordination signal is as follows: In the formula, is the operation cost of the agent at the prediction time step , including the device operation cost and the loss cost; is the benchmark value of the operation cost of the agent, which is the average operation cost of the agent at the same time period of a typical day; is the response deviation of the agent to the global coordination signal at the prediction time step ; is the benchmark value of the response deviation, which is the maximum power deviation allowed by the agent; is the deviation between the actual state and the reference state of the agent at the prediction time step ; , , are the weight coefficients of the operation cost, the response degree to the global coordination signal, and the deviation between the actual state and the reference state of the agent, respectively; is the prediction time domain length, and u is the control variable vector; wherein, is the device operation cost of the agent at the prediction time step is the device wear-out cost of the agent at the prediction time step is the global virtual price signal at the prediction time step is the total power algebra sum of all distributed power sources, energy storage discharges, and loads within the jurisdiction of the agent during the prediction time step , and its value is positive when the agent transmits power to the integrated energy system network and negative when the agent absorbs power from the network; is the length of the optimization period; is the dynamic penalty coefficient for violating the flexibility boundary, is the deviation of the power autonomously decided by the agent at the prediction time step from the global power flexibility lower boundary or the global power flexibility upper boundary; , are the actual state vector and the reference state vector of the agent at the prediction time step is the weighted Euclidean norm, and Q is a diagonal weight matrix, wherein the elements on the diagonal of the matrix represent the importance of the tracking error of the state variables in the state vector in the objective function.​​​​ 11. The method of claim 5, wherein the feedback data comprises an operating state of the device, an actual operating parameter of the device, and disturbance data of the device.

12. The method of claim 5, wherein the external information comprises a power grid dispatching instruction and market price updates. wherein, is the actual price signal at the prediction time step for the intelligent agent; is the fluctuation of the distributed generation output power at the prediction time step for the intelligent agent; is the price correction amplitude factor; is the fluctuation threshold of the distributed generation output power, is the sign function, outputting 1 when the fluctuation of the distributed generation output power is positive and -1 when it is negative.

13. The method of claim 5, wherein the evaluation of the operating state of the integrated energy system comprises:

14. The method of claim 5, wherein the architecture building module is configured to set one agent for each device, and determine a master coordinator from all agents based on a capability evaluation function of each agent. In the formula, , are the lower and upper bounds of the response power of the agent to the global coordination signal, respectively; , are the lower and upper bounds of the power elasticity of the global coordination signal at the prediction time step ; is the standard deviation of the load prediction at the prediction time step ; is the power adjustment coefficient.

15. The method of claim 5, wherein the global coordination module is configured to establish a relaxation strategy for the agents corresponding to the distributed power sources, including a relaxation boundary of electric power and a dynamic penalty coefficient of the relaxation boundary; take a normalized value of an average value of the dynamic penalty coefficients of all agents corresponding to the distributed power sources as a global penalty coefficient; and take the minimum total cost, the highest system energy efficiency, the minimum carbon emissions, and the minimum global penalty coefficient under the global relaxation constraint as the global optimization target of relaxation, and determine the global coordination signal, including the global virtual electricity price signal, the global power elasticity lower bound and the elasticity upper bound, and the global upper limit of carbon emissions, by iteratively solving the global optimization target of relaxation.

16. The method of claim 5, wherein the autonomous decision-making module is configured to take the minimum operating cost of each agent, the minimum response deviation to the global coordination signal, and the minimum deviation between the actual state and the reference state as the control target; establish an electricity price autonomous decision-making model of each agent; establish a response power boundary autonomous decision-making model of each agent to the global coordination signal; and perform model predictive control on each agent based on the control target, the electricity price autonomous decision-making model, and the response power boundary autonomous decision-making model of each agent to predict the control target value of each agent.

17. The method of claim 5, further comprising:

18. The method of claim 5, further comprising: 1) If the value of the system efficiency function is greater than 5%, then a major disturbance is determined to have occurred. 2) If the value of the total cost function is greater than 1.05 times the total cost of the day-ahead plan operation , and the value of the carbon emission function is greater than 1.05 times the average value of the historical carbon emissions of the system , then it is determined that there is a risk of deviation from the global target; 3) If the global optimization objective is relaxed If the value is less than 90%, the coordination signal is considered mismatched.

12. A coordinated control system of integrated energy multi-agent collaborative group control and autonomous decision-making using the method of any one of claims 1-11, characterized in that, 19. The method of claim 5, wherein each agent sends feedback data of the corresponding device after executing the control target value to the master coordinator, and the master coordinator evaluates the operating state of the integrated energy system in combination with the external information. ​ ​ ​ 13. The integrated energy multi-agent collaborative group control and autonomous decision coordination control system according to claim 12, characterized in that, ​ ​ ​ When it is determined according to the operation state evaluation result of the integrated energy system that it is necessary to adjust the control target value of each agent, the main coordinator triggers rolling optimization to update the global coordination signal. 14.A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is configured to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method of any one of claims 1-11.

15. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method of any one of claims 1-11.

Images