Source network load storage comprehensive scheduling method based on multi-agent learning

Abstract

A source network load storage comprehensive scheduling method based on multi-agent learning comprises the following steps: acquiring energy flow parameters between an electric vehicle and a building, and constructing an energy interaction dynamic model containing bidirectional energy flow between the electric vehicle and the building based on a CFD simulation model; and taking the energy interaction dynamic model and the battery SOC and building energy consumption of the electric vehicle as the input of the adaptive neural fuzzy inference model, and predicting the nonlinear relationship between the battery SOC and the building energy consumption to output the change trend of the battery SOC and a charging and discharging optimization strategy. And two-way energy scheduling between the electric vehicle and the building is optimized based on the charging and discharging optimization strategy through multi-agent reinforcement learning, so that an optimized two-way energy flow scheduling scheme is output. According to the method, nonlinear characteristics can be effectively captured, the situation of energy transmission decision errors is reduced, multi-agent reinforcement learning is adopted to optimize energy scheduling, and the uncertainty of charging and discharging states can be dealt with when bidirectional energy flow is processed.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy dispatch optimization of intelligent buildings and intelligent electric vehicles, and more specifically, relates to a source-grid-load-storage integrated dispatching method based on multi-agent learning. Background Art

[0002] With the rapid development of smart cities and driverless technologies, the energy systems of future cities will face great challenges. In this context, millions of driverless electric vehicles (EVs) flow freely in cities, not only providing power for transportation, but also realizing two-way energy flow between their batteries and buildings. These vehicles will be linked with central air-conditioning systems and other building energy-consuming equipment in parking lots, garages, charging stations and other areas, becoming important energy dispatching units. The source-grid-load-storage system plays a vital role in this ecosystem. It not only needs to dispatch power supply, but also needs to coordinate multiple links such as energy storage, demand response, and load dispatch to achieve global energy optimization.

[0003] At present, the basic task of the existing source-grid-load-storage integrated dispatching system is to dynamically allocate between the power grid, energy storage system, and load demand (including building load and electric vehicle charging and discharging demand) through intelligent dispatching algorithms to ensure balanced and efficient energy supply. However, although certain dispatching models and algorithms have been applied in a single field, how to effectively coordinate dynamic energy flows in multiple dimensions remains a difficult problem. Especially in the energy interaction process between vehicle clusters and buildings, the dispatching system needs to capture a large number of complex nonlinear dynamic characteristics in real time, taking into account factors such as vehicle location, charging status, battery health, etc., as well as energy consumption fluctuations and air conditioning load changes in buildings.

[0004] However, the existing integrated dispatching system of sources, grids, loads and storage has real-time nonlinear control problems in the energy interaction between vehicle clusters and buildings. That is, in the process of two-way energy interaction between electric vehicles and buildings, the relationship between the battery state (SOC) and building energy consumption is nonlinear and relatively complex. Traditional dispatching algorithms find it difficult to effectively capture these nonlinear characteristics, which in turn leads to errors in energy transmission decisions.

[0005] In addition, electric vehicles not only need to be charged from buildings, but may also provide energy to buildings (such as reverse charging). The existing source-grid-load-storage scheduling methods have difficulty coping with the high uncertainty of charging and discharging states when dealing with bidirectional energy flows, which in turn greatly increases the complexity of scheduling. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the purpose of the present invention is to solve the above-mentioned defects and further propose a.

[0007] The present invention adopts the following technical solution.

[0008] The first aspect of the present invention discloses a source-grid-load-storage integrated scheduling method based on multi-agent learning, the method comprising: Obtain the energy flow parameters between electric vehicles and buildings, and build an energy interaction dynamic model including the two-way energy flow between electric vehicles and buildings based on the CFD simulation model; The energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy; Optimizing the bidirectional energy scheduling between the electric vehicle and the building based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow scheduling plan; Among them, the energy flow parameters include battery data of electric vehicles, building load data, vehicle location and charging rate. The adaptive neural fuzzy inference model is constructed based on the ANFIS algorithm and trained based on historical energy flow parameters. The bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

[0009] Further, the battery data includes battery capacity, battery SOC, battery voltage and battery current, and the building load data includes air conditioning load, lighting load and total building load; The energy flow parameters between the electric vehicle and the building are obtained, and an energy interaction dynamic model including bidirectional energy flow between the electric vehicle and the building is constructed based on the CFD simulation model, including: Building a basic data set based on the battery data, building load data, vehicle location and charging rate, and establishing an energy flow model between the electric vehicle battery and the building through a mathematical model, wherein the energy flow model includes the relationship between the battery SOC change and the building load data and the charging rate; Wherein, the expression of the battery SOC change is: In the formula, is the battery charging state of the electric vehicle at time t, is the battery charging state of the electric vehicle at time t-1, is the charging power, is the battery capacity, is the time step; The relationship between the building load data and the charging rate is expressed as follows: In the formula, is the charging power at time t, is the total load data of the building at time t, is the spatial distance between the electric vehicle and the building.

[0010] Furthermore, the method of obtaining energy flow parameters between the electric vehicle and the building and constructing an energy interaction dynamic model including bidirectional energy flow between the electric vehicle and the building based on the CFD simulation model also includes: The thermal effect generated during the battery charging and discharging process is simulated by the CFD thermal effect model computational fluid dynamics to obtain the thermal effect simulation results. The fluid dynamics simulation equation is: In the formula, is the temperature field, which represents the temperature distribution inside and outside the battery. is the thermal diffusion coefficient, It is the heat source generated during the battery charging and discharging process. is the Laplace operator of the temperature field, which represents the diffusion process of heat inside the battery; Based on the energy flow model and the thermal effect simulation results, an energy interaction dynamic model of bidirectional energy flow between electric vehicles and buildings is established. The expression of the energy interaction dynamic model is: In the formula, is the total energy flow at time t, representing the comprehensive energy exchange between electric vehicles and buildings, is the charging power at time t, is the thermal effect at time t, is the time step.

[0011] Furthermore, the energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy, including: Based on the energy interaction dynamic model, a battery SOC, building load data, charging rate and historical energy flow parameters of the electric vehicle are obtained to construct a training data set, where the historical energy flow parameters are the battery SOC, building load data and charging rate in a first historical time period; An ANFIS model is constructed by using an ANFIS algorithm, and the ANFIS model is trained based on the historical energy flow parameters to adjust the fuzzy rules of the ANFIS model and the weights in the neural network, and the error between the predicted value and the actual value of the ANFIS model is minimized by using an error function to optimize the ANFIS model to obtain a trained ANFIS model; The ANFIS model is composed of fuzzy rules and neural networks. The fuzzy rules are used to characterize the fuzzy relationship between input variables, and the neural network is used to optimize rule parameters by learning the pattern of input data.

[0012] Furthermore, the energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy, and further includes: Calling the trained ANFIS model to process the battery SOC of the electric vehicle and the building load data, so as to dynamically adjust the charging rate and time window based on the fluctuation of the building load data and the change trend of the battery SOC, and optimize the energy flow between the electric vehicle and the building; The ANFIS model includes an input layer, a fuzzification layer, a rule layer, a defuzzification layer and an output layer. The input layer is used to input battery SOC, building load data and charging rate. The fuzzification layer is used to fuzzify the input layer data through a membership function to obtain the membership value of each input layer data. The rule layer is used to generate and output rules according to fuzzy rules. The defuzzification layer is used to convert the fuzzified output into a predicted value through weighted averaging. The output layer is used to output the predicted charging rate and time window adjustment parameters. The expressions for prediction and adjustment of the trained ANFIS model are: In the formula, is the predicted battery charging state of the electric vehicle at time t+1, is the battery charging state of the electric vehicle at time t, is the charging power, is the battery capacity, is the time step.

[0013] Furthermore, the bidirectional energy scheduling between the electric vehicle and the building is optimized based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow scheduling scheme, including: Establishing a MARL model for describing the self-optimization adjustment behavior of each agent based on the energy flow parameters, and optimizing the behavior strategy of each agent through a Q-value update algorithm; The MARL model is trained by repeatedly updating the Q value to control the agent to learn the optimal behavior strategy, and the trained MARL model is called in conjunction with the Q value update algorithm to optimize resource allocation for multiple agents in a collaborative manner; The intelligent agent includes the electric vehicle and the building air-conditioning system, and each intelligent agent has a state, an action and a reward. The state includes the battery SOC of the electric vehicle, the building load data, the charging rate and the building system load. The action includes the charging and discharging rate of the electric vehicle and the power scheduling of the building air-conditioning system. The reward is the degree of optimization of the energy scheduling.

[0014] Further, the bidirectional energy flow scheduling scheme includes charging behavior of electric vehicles, scheduling of building air conditioning systems, and optimization of charging time windows, and the mathematical model of the bidirectional energy flow scheduling scheme includes a bidirectional energy flow model between electric vehicles and buildings and a building load balance equation; The expression of the bidirectional energy flow model between electric vehicles and buildings is: In the formula, is the change of battery SOC at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t, is the battery capacity, is the time step; The building load balance equation is expressed as: In the formula, is the total load of the building air conditioning system at time t, is the building load data at time t, is the power dispatch of the building air conditioning system at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t.

[0015] The second aspect of the present invention discloses a source-grid-load-storage integrated dispatching device, the device comprising: The dynamic model building module is used to obtain the energy flow parameters between electric vehicles and buildings, and to build an energy interaction dynamic model including the two-way energy flow between electric vehicles and buildings based on the CFD simulation model; An optimization strategy prediction module, used to use the energy interaction dynamic model and the battery SOC of the electric vehicle and the building energy consumption as inputs of an adaptive neuro-fuzzy inference model, predict the nonlinear relationship between the battery SOC and the building energy consumption, and output a battery SOC change trend and a charge and discharge optimization strategy; An energy flow scheduling module, used for optimizing the bidirectional energy scheduling between electric vehicles and buildings based on the charging and discharging optimization strategy through multi-agent reinforcement learning, so as to output an optimized bidirectional energy flow scheduling solution; Among them, the energy flow parameters include battery data of electric vehicles, building load data, vehicle location and charging rate. The adaptive neural fuzzy inference model is constructed based on the ANFIS algorithm and trained based on historical energy flow parameters. The bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

[0016] The third aspect of the present invention discloses a terminal, including a processor and a storage medium; the characteristics are: 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 described in the first aspect.

[0017] The fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0018] The beneficial effects of the present invention are that, compared with the prior art, the present invention has the following advantages: By obtaining the energy flow parameters between electric vehicles and buildings, and building an energy interaction dynamic model including the bidirectional energy flow between electric vehicles and buildings based on the CFD simulation model, the nonlinear relationship between the battery SOC and the building energy consumption is predicted by taking the energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption as the input of the adaptive neuro-fuzzy inference model, so as to output the battery SOC change trend and the charging and discharging optimization strategy. Finally, the bidirectional energy scheduling between electric vehicles and buildings is optimized based on the charging and discharging optimization strategy through multi-agent reinforcement learning, so as to output the optimized bidirectional energy flow scheduling scheme. This method predicts the nonlinear relationship between the battery SOC and the building energy consumption through the adaptive neuro-fuzzy inference model, and then outputs the battery SOC change trend and the charging and discharging optimization strategy, which can effectively capture the nonlinear characteristics and reduce the occurrence of energy transmission decision errors. In addition, the use of multi-agent reinforcement learning to optimize the bidirectional energy scheduling between electric vehicles and buildings can cope with the uncertainty of the charging and discharging state when processing the bidirectional energy flow, and reduce the complexity of its scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the source-grid-load-storage comprehensive scheduling method based on multi-agent learning provided by the present invention. DETAILED DESCRIPTION

[0020] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0021] like Figure 1 As shown, in one embodiment, a source-grid-load-storage integrated scheduling method based on multi-agent learning includes the following steps: Step S110, obtaining energy flow parameters between the electric vehicle and the building, and constructing an energy interaction dynamic model including bidirectional energy flow between the electric vehicle and the building based on the CFD simulation model.

[0022] Energy flow parameters include electric vehicle battery data, building load data, vehicle location, and charging rate.

[0023] In some embodiments, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention, the battery data includes battery capacity, battery SOC, battery voltage and battery current, and the building load data includes air conditioning load, lighting load and total building load. The energy flow parameters between the electric vehicle and the building are obtained, and an energy interaction dynamic model including the two-way energy flow between the electric vehicle and the building is constructed based on the CFD simulation model, specifically including the following steps: Step S111, construct a basic data set based on battery data, building load data, vehicle location and charging rate, and establish an energy flow model between electric vehicle batteries and buildings through a mathematical model. The energy flow model includes the relationship between battery SOC changes and building load data and charging rate.

[0024] Among them, the expression of battery SOC change is: In the formula, is the battery charging state of the electric vehicle at time t, is the battery charging state of the electric vehicle at time t-1, is the charging power, is the battery capacity, is the time step.

[0025] The relationship between building load data and charging rate is expressed as: In the formula, is the charging power at time t, is the total load data of the building at time t, is the spatial distance between the electric vehicle and the building.

[0026] In some embodiments, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention obtains energy flow parameters between electric vehicles and buildings, and constructs an energy interaction dynamic model including bidirectional energy flow between electric vehicles and buildings based on a CFD simulation model, specifically including the following steps: Step S112, using a CFD thermal effect model to calculate fluid dynamics to simulate the thermal effect generated during the battery charging and discharging process, to obtain a thermal effect simulation result, and the fluid dynamics simulation equation is: In the formula, is the temperature field, which represents the temperature distribution inside and outside the battery. is the thermal diffusion coefficient, It is the heat source generated during the battery charging and discharging process. is the Laplace operator of the temperature field, which represents the diffusion process of heat inside the battery.

[0027] Step S113, based on the energy flow model and the thermal effect simulation results, an energy interaction dynamic model of bidirectional energy flow between the electric vehicle and the building is established. The expression of the energy interaction dynamic model is: In the formula, is the total energy flow at time t, representing the comprehensive energy exchange between electric vehicles and buildings, is the charging power at time t, is the thermal effect at time t, is the time step.

[0028] Step S120, using the energy interaction dynamic model and the battery SOC of the electric vehicle and the building energy consumption as inputs of the adaptive neuro-fuzzy inference model, predicts the nonlinear relationship between the battery SOC and the building energy consumption, and outputs the battery SOC change trend and the charging and discharging optimization strategy.

[0029] Among them, the adaptive neuro-fuzzy inference model is constructed based on the ANFIS algorithm and trained based on historical energy flow parameters.

[0030] In some embodiments, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention uses the energy interaction dynamic model and the battery SOC of the electric vehicle and the building energy consumption as the input of the adaptive neural fuzzy reasoning model, predicts the nonlinear relationship between the battery SOC and the building energy consumption, and outputs the battery SOC change trend and the charging and discharging optimization strategy, which specifically includes the following steps: Step S121, based on the energy interaction dynamic model, the battery SOC, building load data, charging rate and historical energy flow parameters of the electric vehicle are obtained to construct a training data set, where the historical energy flow parameters are the battery SOC, building load data and charging rate in the first historical time period.

[0031] Step S122, constructing an ANFIS model through the ANFIS algorithm, and training the ANFIS model based on historical energy flow parameters to adjust the fuzzy rules of the ANFIS model and the weights in the neural network, and at the same time optimizing the ANFIS model by minimizing the error between the predicted value and the actual value of the ANFIS model through the error function to obtain a trained ANFIS model.

[0032] Among them, the ANFIS model is composed of fuzzy rules and neural networks. The fuzzy rules are used to characterize the fuzzy relationship between input variables, and the neural network is used to optimize the rule parameters by learning the pattern of input data.

[0033] In some embodiments, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention uses the energy interaction dynamic model and the battery SOC of the electric vehicle and the building energy consumption as the input of the adaptive neural fuzzy reasoning model, predicts the nonlinear relationship between the battery SOC and the building energy consumption, and outputs the battery SOC change trend and the charging and discharging optimization strategy, and specifically includes the following steps: Step S123, calling the trained ANFIS model to process the battery SOC of the electric vehicle and the building load data, so as to dynamically adjust the charging rate and time window based on the fluctuation of the building load data and the change trend of the battery SOC, and optimize the energy flow between the electric vehicle and the building.

[0034] Among them, the ANFIS model includes an input layer, a fuzzification layer, a rule layer, a defuzzification layer and an output layer. The input layer is used to input battery SOC, building load data and charging rate. The fuzzification layer is used to fuzzify the input layer data through the membership function to obtain the membership value of each input layer data. The rule layer is used to generate and output rules according to fuzzy rules. The defuzzification layer is used to convert the fuzzified output into a predicted value through weighted averaging. The output layer is used to output the predicted charging rate and time window adjustment parameters. The expression of prediction and adjustment of the trained ANFIS model is: In the formula, is the predicted battery charging state of the electric vehicle at time t+1, is the battery charging state of the electric vehicle at time t, is the charging power, is the battery capacity, is the time step.

[0035] Step S130, optimizing the bidirectional energy scheduling between electric vehicles and buildings based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow scheduling solution.

[0036] Among them, the bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

[0037] In some embodiments, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention optimizes the bidirectional energy scheduling between electric vehicles and buildings based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow scheduling scheme, specifically including the following steps: Step S131, establish a MARL model for describing the self-optimization adjustment behavior of each agent based on energy flow parameters, and optimize the behavior strategy of each agent through a Q value update algorithm.

[0038] Step S132, by repeatedly updating the Q value to control the agent to learn the optimal behavior strategy, the MARL model is trained, and the trained MARL model is called in conjunction with the Q value update algorithm to optimize resource allocation for multiple agents to collaborate.

[0039] Among them, the intelligent agents include electric vehicles and building air-conditioning systems, and each intelligent agent has state, action and reward. The state includes the battery SOC of the electric vehicle, the building load data, the charging rate and the building system load. The action includes the charging and discharging rate of the electric vehicle and the power scheduling of the building air-conditioning system. The reward is the optimization degree of energy scheduling.

[0040] In some embodiments, the present invention provides a source-grid-load-storage integrated scheduling method based on multi-agent learning, in which a bidirectional energy flow scheduling scheme includes the charging behavior of electric vehicles, the scheduling of building air-conditioning systems, and the optimization of charging time windows, and the mathematical model of the bidirectional energy flow scheduling scheme includes a bidirectional energy flow model between electric vehicles and buildings and a building load balance equation.

[0041] The expression of the bidirectional energy flow model between electric vehicles and buildings is: In the formula, is the change of battery SOC at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t, is the battery capacity, is the time step.

[0042] The building load balance equation is expressed as: In the formula, is the total load of the building air conditioning system at time t, is the building load data at time t, is the power dispatch of the building air conditioning system at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t.

[0043] The above-mentioned source-grid-load-storage integrated scheduling method based on multi-agent learning obtains the energy flow parameters between electric vehicles and buildings, and constructs an energy interaction dynamic model including the bidirectional energy flow between electric vehicles and buildings based on the CFD simulation model. Then, the energy interaction dynamic model, the battery SOC of electric vehicles and the building energy consumption are used as the input of the adaptive neural fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy. Finally, the bidirectional energy scheduling between electric vehicles and buildings is optimized based on the charging and discharging optimization strategy through multi-agent reinforcement learning, so as to output the optimized bidirectional energy flow scheduling scheme. This method predicts the nonlinear relationship between the battery SOC and the building energy consumption through the adaptive neural fuzzy inference model, and then outputs the battery SOC change trend and the charging and discharging optimization strategy, which can effectively capture the nonlinear characteristics and reduce the occurrence of energy transmission decision errors. In addition, the use of multi-agent reinforcement learning to optimize the bidirectional energy scheduling between electric vehicles and buildings can cope with the uncertainty of the charging and discharging state when processing the bidirectional energy flow, and reduce the complexity of its scheduling.

[0044] In a specific embodiment, the source-grid-load-storage integrated scheduling method based on multi-agent learning provided by the present invention includes steps 1 to 3: Step 1: Establish a system dynamics model.

[0045] Specifically, a system dynamics model is constructed that includes the bidirectional energy flow between electric vehicles (EVs) and buildings. The system dynamics model takes into account multiple factors such as battery SOC (state of charge), building load demand (such as air conditioning load, lighting load, etc.), vehicle location, charging rate, etc., and uses computational fluid dynamics (CFD) simulation to predict the thermal effects and energy flow of the battery charging and discharging process.

[0046] Input: Electric vehicle battery data, building load data, vehicle location, charging rate, CFD simulation model.

[0047] Output: Dynamic model of energy interaction between electric vehicles and buildings, including information such as SOC, load demand and charging rate.

[0048] More specifically, it includes steps 1.1 to 1.4: Step 1.1, preparation of input data.

[0049] First, it is necessary to collect input data related to electric vehicle batteries, building load data, vehicle location, charging rate, and CFD simulation models to establish a system model and perform subsequent calculations. The input data includes: electric vehicle battery data (battery capacity, battery SOC, battery voltage, battery current), building load data (air conditioning load, lighting load, total building load), vehicle location data, and charging rate, and then construct a basic data set containing electric vehicle battery data, building load data, vehicle location, and charging rate to prepare for subsequent calculations.

[0050] Step 1.2, establish the energy flow model.

[0051] Specifically, the energy flow relationship between electric vehicle batteries and buildings is established through a mathematical model. The model needs to consider the interaction between battery SOC changes, building load data, and charging rate.

[0052] In this embodiment, the battery SOC change expression in the energy flow model is: In the formula, is the battery charging state of the electric vehicle at time t, is the battery charging state of the electric vehicle at time t-1, is the charging power, is the battery capacity, is the time step. This expression describes the process of battery SOC changing over time. In each time step, the battery SOC changes according to the charging power and battery capacity.

[0053] The relationship between building load data and charging rate is expressed as: In the formula, is the charging power at time t, is the total load data of the building at time t, is the spatial distance between the electric vehicle and the building. This expression describes the relationship between the building load data and the charging rate. As the load demand of the building and the distance between the electric vehicle and the building change, the charging rate is adjusted accordingly.

[0054] Step 1.3, introduce CFD simulation to predict thermal effects.

[0055] Specifically, computational fluid dynamics (CFD) models are used to simulate the thermal effects generated during battery charging and discharging, and to analyze the heat transfer process of energy flow. CFD simulation helps to understand how batteries affect the overall energy flow during charging and discharging, especially the impact of thermal effects on charging and discharging efficiency.

[0056] In this embodiment, the computational fluid dynamics simulation equation of the CFD thermal effect model is: In the formula, is the temperature field, which represents the temperature distribution inside and outside the battery. is the thermal diffusion coefficient, It is the heat source generated during the battery charging and discharging process. is the Laplace operator of the temperature field, which represents the diffusion process of heat inside the battery. This equation describes the temperature change during the heat conduction process in the battery. Through CFD simulation, the temperature fluctuation caused by the thermal effect during the battery charging and discharging process can be predicted, which in turn affects the battery charging and discharging efficiency. The output of CFD simulation provides the temperature field and thermal effect during the battery charging and discharging process, and the influence of thermal effect on energy flow, providing data support for further optimizing energy management.

[0057] Step 1.4, establish a comprehensive energy interaction dynamic model.

[0058] Specifically, the energy flow model in step 1.2 is combined with the thermal effect simulation results in step 1.3 to establish a comprehensive system dynamics model of bidirectional energy flow between electric vehicles and buildings. This model can not only describe the relationship between battery SOC, building load data and charging rate, but also consider the thermal effect of the battery, providing a comprehensive basis for energy management decision-making.

[0059] In this embodiment, the expression of the comprehensive energy interaction dynamic model is: In the formula, is the total energy flow at time t, representing the comprehensive energy exchange between electric vehicles and buildings, is the charging power at time t, is the thermal effect at time t, This expression combines the battery charging power with the thermal effect and gives the comprehensive energy flow between electric vehicles and buildings, which helps to evaluate the energy loss and thermal effect in the system and optimize the energy scheduling strategy.

[0060] Step 2: Apply ANFIS to predict nonlinear relationships.

[0061] The complex nonlinear relationship between electric vehicle battery SOC and building energy consumption is predicted in real time using an adaptive neuro-fuzzy inference system (ANFIS). Through the combination of fuzzy rules and neural networks, ANFIS can dynamically adjust the charging rate and time window to respond to fluctuations in building load.

[0062] Input: The energy interaction dynamic model output from step 1, real-time battery SOC and building energy consumption data.

[0063] Output: predicted battery SOC change trend and charging / discharging optimization strategy.

[0064] Specifically, it includes steps 2.1 to 2.4: Step 2.1, preparation of input data.

[0065] The input data includes battery SOC, building load data, charging rate, and historical data. The historical data contains battery SOC, building load demand, and charging rate over a period of time in the past, which is used to train the ANFIS model.

[0066] Step 2.2, establishment of ANFIS model.

[0067] The ANFIS model consists of five layers, namely the input layer, fuzzification layer, rule layer, defuzzification layer and output layer. The input layer is used to input battery SOC, building load data and charging rate, and the fuzzification layer is used to fuzzify the input variables through membership functions to obtain the membership value of each input. The rule layer is used to generate outputs according to fuzzy rules. Each rule describes the relationship between the combination of input variables and output variables. The defuzzification layer is used to convert the fuzzified output into a clear prediction value through the weighted average method. The output layer is used to give the predicted charging rate and time window adjustment parameters.

[0068] Step 2.3, ANFIS model training and optimization.

[0069] The ANFIS model is trained using historical data to adjust the weights in the fuzzy rules and neural network. During the training process, the error minimization method is used to optimize the ANFIS model by minimizing the error between the model predicted value and the actual value.

[0070] During the training process, the error function expression is: In the formula, is the charging rate predicted by the model, is the actual charging rate, is the error value, is the sample size.

[0071] The training algorithm uses gradient descent or other optimization algorithms to adjust the parameters in the ANFIS model to minimize the error.

[0072] Step 2.4, real-time prediction and optimization strategy.

[0073] Use the trained ANFIS model for real-time prediction, dynamically adjust the charging rate and time window based on the fluctuation of building load data and the changing trend of battery SOC, and optimize the energy flow between electric vehicles and buildings.

[0074] Among them, the expressions of prediction and adjustment of the trained ANFIS model are: In the formula, is the predicted battery charging state of the electric vehicle at time t+1, is the battery charging state of the electric vehicle at time t, is the charging power, is the battery capacity, is the time step.

[0075] In this embodiment, the optimization strategy includes: Increased charging rate when building load data is high.

[0076] When the battery SOC is close to full charge, reduce the charging rate to avoid overcharging.

[0077] According to the load fluctuation trend, the charging time window is dynamically adjusted to ensure load balance.

[0078] Step 3: Multi-agent reinforcement learning (MARL) optimizes bidirectional energy scheduling.

[0079] Based on the charging and discharging optimization strategy provided in step 2, multi-agent reinforcement learning (MARL) is applied to optimize the bidirectional energy scheduling between vehicles and buildings. Each agent (such as electric vehicles, building air conditioning systems, etc.) self-optimizes its charging and discharging behavior through reinforcement learning, and collaborates with other agents to jointly optimize the energy flow of the entire system.

[0080] Input: charging and discharging optimization strategy output from step 2, real-time energy consumption data of buildings, real-time SOC of vehicles, and scheduling constraints.

[0081] Output: The optimized bidirectional energy flow scheduling plan, including the charging and discharging time, charging power and discharging amount of each vehicle.

[0082] Specifically, it includes steps 3.1 to 3.5: Step 3.1, build a multi-agent reinforcement learning model.

[0083] First, a MARL model is established to describe how each agent in the system (such as electric vehicles, building air conditioning systems, etc.) adjusts its behavior based on the current environmental state (including battery SOC, building load data, etc.). Among them, when the electric vehicle is an agent, it is a charging / discharging agent, and its behavior mainly involves how to determine the charging rate based on the battery SOC and building load data. When the building air conditioning system is an agent, it adjusts energy scheduling according to the load demand of the building, and may store or discharge energy to help balance the energy flow between electric vehicles and buildings.

[0084] In this embodiment, each intelligent agent has a state, an action, and a reward. The state includes the battery SOC of the electric vehicle, the building load data, the charging rate, and the building system load. The action includes the charging and discharging rate of the electric vehicle and the power scheduling of the building air-conditioning system. The reward is the degree of optimization of the energy scheduling.

[0085] Step 3.2, Q-learning and value updating of the agent.

[0086] Each agent continuously optimizes its behavior strategy through the Q learning algorithm. The core of the Q learning algorithm is to update the Q value (that is, the expected reward of taking a certain action in the current state).

[0087] In this embodiment, the Q value update formula is: In the formula, At time t, agent i is in state Take action The Q value, is the learning rate, which determines the importance the agent attaches to new information and its value range is [0,1]. is the discount factor, which indicates the discount degree of future rewards, and its value range is [0,1]. is the reward value, indicating that an action is taken at time t Instant rewards afterward. Indicates that in the next state Under the condition of , the agent selects the maximum Q value of the optimal action. By repeatedly updating the Q value, the agent can learn the optimal charging / discharging behavior and coordinate with the behavior of other agents to optimize the energy flow of the entire system.

[0088] Step 3.3, joint training and multi-agent collaboration.

[0089] Multiple agents coordinate and optimize the energy scheduling of the entire system through joint training. The reward of each agent is affected not only by its own behavior but also by the behavior of other agents. This requires considering the joint Q-learning algorithm strategy in MARL so that multiple agents can collaborate to optimize resource allocation.

[0090] In addition, each agent will consider the states and actions of other agents and obtain corresponding rewards through interaction to optimize and update. For example, the charging rate of electric vehicles may be affected by the load demand of the building's air conditioning system, and the scheduling of the air conditioning system is affected by the charging behavior of electric vehicles. Therefore, the behavior update of each agent depends not only on its own state and action, but also on the cooperation and response of other agents.

[0091] Step 3.4, optimization and implementation of energy scheduling.

[0092] After completing MARL training, the agent will be able to adjust its charging and discharging behavior in real time according to state changes to achieve overall system optimization. The optimized bidirectional energy flow scheduling scheme is based on three factors: the charging behavior of electric vehicles, the scheduling of building air conditioning systems, and the optimization of charging time windows. In addition, the optimized bidirectional energy flow scheduling scheme can be expressed as the following mathematical model: The expression of the bidirectional energy flow model between electric vehicles and buildings is: In the formula, is the change of battery SOC at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t, is the battery capacity, is the time step.

[0093] The building load balance equation is expressed as: In the formula, is the total load of the building air conditioning system at time t, is the building load data at time t, is the power dispatch of the building air conditioning system at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t.

[0094] Step 3.5, optimized bidirectional energy flow scheduling scheme.

[0095] The final output of the optimized bidirectional energy flow scheduling scheme includes: The behavior strategy of each agent (e.g., charging / discharging strategy of an electric vehicle, scheduling strategy of a building’s air conditioning system).

[0096] The charging rate, discharging rate and energy scheduling scheme of each agent in different states.

[0097] The source-grid-load-storage integrated dispatching device provided by the present invention is described below. The source-grid-load-storage integrated dispatching device described below and the source-grid-load-storage integrated dispatching method based on multi-agent learning described above can be referenced to each other.

[0098] In one embodiment, a source-grid-load-storage integrated scheduling device includes a dynamic model building module, an optimization strategy prediction module, and an energy flow scheduling module.

[0099] The dynamic model building module is used to obtain the energy flow parameters between electric vehicles and buildings, and to build an energy interaction dynamic model including the two-way energy flow between electric vehicles and buildings based on the CFD simulation model.

[0100] The optimization strategy prediction module is used to predict the nonlinear relationship between battery SOC and building energy consumption using the energy interaction dynamic model, the battery SOC of electric vehicles and the building energy consumption as inputs to the adaptive neuro-fuzzy inference model, so as to output the battery SOC change trend and the charging and discharging optimization strategy.

[0101] The energy flow scheduling module is used to optimize the bidirectional energy scheduling between electric vehicles and buildings based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output the optimized bidirectional energy flow scheduling plan.

[0102] Among them, the energy flow parameters include the battery data of electric vehicles, building load data, vehicle location and charging rate. The adaptive neuro-fuzzy inference model is built based on the ANFIS algorithm and trained based on historical energy flow parameters. The bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

[0103] In this embodiment, the source-grid-load-storage integrated dispatching device provided by the present invention, the battery data includes battery capacity, battery SOC, battery voltage and battery current, and the building load data includes air conditioning load, lighting load and total building load.

[0104] The dynamic model building module is specifically used for: A basic data set is constructed based on battery data, building load data, vehicle location and charging rate, and an energy flow model between electric vehicle batteries and buildings is established through a mathematical model. The energy flow model includes battery SOC changes and the relationship between building load data and charging rate.

[0105] Among them, the expression of battery SOC change is: In the formula, is the battery charging state of the electric vehicle at time t, is the battery charging state of the electric vehicle at time t-1, is the charging power, is the battery capacity, is the time step.

[0106] The relationship between building load data and charging rate is expressed as: In the formula, is the charging power at time t, is the total load data of the building at time t, is the spatial distance between the electric vehicle and the building.

[0107] In this embodiment, the source-grid-load-storage integrated dispatching device provided by the present invention, the dynamic model building module is specifically used for: The thermal effect generated during the battery charging and discharging process is simulated by the CFD thermal effect model computational fluid dynamics to obtain the thermal effect simulation results. The fluid dynamics simulation equation is: In the formula, is the temperature field, which represents the temperature distribution inside and outside the battery. is the thermal diffusion coefficient, It is the heat source generated during the battery charging and discharging process. is the Laplace operator of the temperature field, which represents the diffusion process of heat inside the battery.

[0108] Based on the energy flow model and thermal effect simulation results, an energy interaction dynamic model of bidirectional energy flow between electric vehicles and buildings is established. The expression of the energy interaction dynamic model is: In the formula, is the total energy flow at time t, representing the comprehensive energy exchange between electric vehicles and buildings, is the charging power at time t, is the thermal effect at time t, is the time step.

[0109] In this embodiment, the optimization strategy prediction module of the source-grid-load-storage integrated dispatching device provided by the present invention is specifically used for: Based on the energy interaction dynamic model, the battery SOC, building load data, charging rate and historical energy flow parameters of electric vehicles are obtained to construct a training data set. The historical energy flow parameters are the battery SOC, building load data and charging rate in the first historical time period.

[0110] The ANFIS model is constructed by the ANFIS algorithm, and is trained based on historical energy flow parameters to adjust the fuzzy rules of the ANFIS model and the weights in the neural network. At the same time, the error function is used to minimize the error between the predicted value and the actual value of the ANFIS model to optimize the ANFIS model and obtain the trained ANFIS model.

[0111] Among them, the ANFIS model is composed of fuzzy rules and neural networks. The fuzzy rules are used to characterize the fuzzy relationship between input variables, and the neural network is used to optimize the rule parameters by learning the pattern of input data.

[0112] In this embodiment, the source-grid-load-storage integrated dispatching device provided by the present invention, the optimization strategy prediction module is specifically used for: The trained ANFIS model is called to process the battery SOC of electric vehicles and building load data to dynamically adjust the charging rate and time window based on the fluctuation of building load data and the changing trend of battery SOC, so as to optimize the energy flow between electric vehicles and buildings.

[0113] Among them, the ANFIS model includes an input layer, a fuzzification layer, a rule layer, a defuzzification layer and an output layer. The input layer is used to input battery SOC, building load data and charging rate. The fuzzification layer is used to fuzzify the input layer data through the membership function to obtain the membership value of each input layer data. The rule layer is used to generate and output rules according to fuzzy rules. The defuzzification layer is used to convert the fuzzified output into a predicted value through weighted averaging. The output layer is used to output the predicted charging rate and time window adjustment parameters. The expression of prediction and adjustment of the trained ANFIS model is: In the formula, is the predicted battery charging state of the electric vehicle at time t+1, is the battery charging state of the electric vehicle at time t, is the charging power, is the battery capacity, is the time step.

[0114] In this embodiment, the source-grid-load-storage integrated dispatching device provided by the present invention, the energy flow dispatching module is specifically used for: A MARL model is established to describe the self-optimization and adjustment behavior of each agent based on energy flow parameters, and the behavior strategy of each agent is optimized through the Q-value updating algorithm.

[0115] The MARL model is trained by repeatedly updating the Q value to control the agent to learn the optimal behavior strategy, and the trained MARL model is called in conjunction with the Q value update algorithm to optimize resource allocation for multiple agents to collaborate.

[0116] Among them, the intelligent agents include electric vehicles and building air-conditioning systems, and each intelligent agent has state, action and reward. The state includes the battery SOC of the electric vehicle, the building load data, the charging rate and the building system load. The action includes the charging and discharging rate of the electric vehicle and the power scheduling of the building air-conditioning system. The reward is the optimization degree of energy scheduling.

[0117] In this embodiment, the source-grid-load-storage integrated scheduling device provided by the present invention, the bidirectional energy flow scheduling scheme includes the charging behavior of electric vehicles, the scheduling of building air-conditioning systems, and the optimization of charging time windows, and the mathematical model of the bidirectional energy flow scheduling scheme includes a bidirectional energy flow model between electric vehicles and buildings and a building load balance equation.

[0118] The expression of the bidirectional energy flow model between electric vehicles and buildings is: In the formula, is the change of battery SOC at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t, is the battery capacity, is the time step.

[0119] The building load balance equation is expressed as: In the formula, is the total load of the building air conditioning system at time t, is the building load data at time t, is the power dispatch of the building air conditioning system at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t.

[0120] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0121] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, 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 disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0122] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0123] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0124] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0125] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0126] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0127] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A source-grid-load-storage integrated scheduling method based on multi-agent learning, characterized in that: The method comprises: Obtain the energy flow parameters between electric vehicles and buildings, and build an energy interaction dynamic model including the two-way energy flow between electric vehicles and buildings based on the CFD simulation model; The energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy; Optimizing the bidirectional energy scheduling between the electric vehicle and the building based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow scheduling plan; Among them, the energy flow parameters include battery data of electric vehicles, building load data, vehicle location and charging rate. The adaptive neural fuzzy inference model is constructed based on the ANFIS algorithm and trained based on historical energy flow parameters. The bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

2. The source-grid-load-storage integrated scheduling method based on multi-agent learning according to claim 1 is characterized in that: The battery data includes battery capacity, battery SOC, battery voltage and battery current, and the building load data includes air conditioning load, lighting load and total building load; The energy flow parameters between the electric vehicle and the building are obtained, and an energy interaction dynamic model including bidirectional energy flow between the electric vehicle and the building is constructed based on the CFD simulation model, including: Building a basic data set based on the battery data, building load data, vehicle location and charging rate, and establishing an energy flow model between the electric vehicle battery and the building through a mathematical model, wherein the energy flow model includes the relationship between the battery SOC change and the building load data and the charging rate; Wherein, the expression of the battery SOC change is: ； In the formula, is the battery charging state of the electric vehicle at time t, is the battery charging state of the electric vehicle at time t-1, is the charging power, is the battery capacity, is the time step; The relationship between the building load data and the charging rate is expressed as follows: ； In the formula, is the charging power at time t, is the total load data of the building at time t, is the spatial distance between the electric vehicle and the building.

3. The source-grid-load-storage integrated scheduling method based on multi-agent learning according to claim 2 is characterized in that: The method of obtaining energy flow parameters between the electric vehicle and the building and constructing an energy interaction dynamic model including bidirectional energy flow between the electric vehicle and the building based on the CFD simulation model also includes: The thermal effect generated during the battery charging and discharging process is simulated by the CFD thermal effect model computational fluid dynamics to obtain the thermal effect simulation results. The fluid dynamics simulation equation is: ； In the formula, is the temperature field, which represents the temperature distribution inside and outside the battery. is the thermal diffusion coefficient, It is the heat source generated during the battery charging and discharging process. is the Laplace operator of the temperature field, which represents the diffusion process of heat inside the battery; Based on the energy flow model and the thermal effect simulation results, an energy interaction dynamic model of bidirectional energy flow between electric vehicles and buildings is established. The expression of the energy interaction dynamic model is: ； In the formula, is the total energy flow at time t, representing the comprehensive energy exchange between electric vehicles and buildings, is the charging power at time t, is the thermal effect at time t, is the time step.

4. The source-grid-load-storage integrated scheduling method based on multi-agent learning according to claim 3 is characterized in that: The energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy, including: Based on the energy interaction dynamic model, a battery SOC, building load data, charging rate and historical energy flow parameters of the electric vehicle are obtained to construct a training data set, where the historical energy flow parameters are the battery SOC, building load data and charging rate in a first historical time period; An ANFIS model is constructed by using an ANFIS algorithm, and the ANFIS model is trained based on the historical energy flow parameters to adjust the fuzzy rules of the ANFIS model and the weights in the neural network, and the error between the predicted value and the actual value of the ANFIS model is minimized by using an error function to optimize the ANFIS model to obtain a trained ANFIS model; The ANFIS model is composed of fuzzy rules and neural networks. The fuzzy rules are used to characterize the fuzzy relationship between input variables, and the neural network is used to optimize rule parameters by learning the pattern of input data.

5. The source-grid-load-storage integrated scheduling method based on multi-agent learning according to claim 4 is characterized in that: The energy interaction dynamic model, the battery SOC of the electric vehicle and the building energy consumption are used as inputs of the adaptive neuro-fuzzy inference model to predict the nonlinear relationship between the battery SOC and the building energy consumption, so as to output the battery SOC change trend and the charging and discharging optimization strategy, and also includes: Calling the trained ANFIS model to process the battery SOC of the electric vehicle and the building load data, so as to dynamically adjust the charging rate and time window based on the fluctuation of the building load data and the change trend of the battery SOC, and optimize the energy flow between the electric vehicle and the building; The ANFIS model includes an input layer, a fuzzification layer, a rule layer, a defuzzification layer and an output layer. The input layer is used to input battery SOC, building load data and charging rate. The fuzzification layer is used to fuzzify the input layer data through a membership function to obtain the membership value of each input layer data. The rule layer is used to generate and output rules according to fuzzy rules. The defuzzification layer is used to convert the fuzzified output into a predicted value through weighted averaging. The output layer is used to output the predicted charging rate and time window adjustment parameters. The expressions for prediction and adjustment of the trained ANFIS model are: ； In the formula, is the predicted battery charging state of the electric vehicle at time t+1, is the battery charging state of the electric vehicle at time t, is the charging power, is the battery capacity, is the time step.

6. The source-grid-load-storage integrated scheduling method based on multi-agent learning according to claim 5 is characterized in that: The method of optimizing the bidirectional energy dispatch between the electric vehicle and the building based on the charging and discharging optimization strategy through multi-agent reinforcement learning to output an optimized bidirectional energy flow dispatching scheme includes: Establishing a MARL model for describing the self-optimization adjustment behavior of each agent based on the energy flow parameters, and optimizing the behavior strategy of each agent through a Q-value update algorithm; The MARL model is trained by repeatedly updating the Q value to control the agent to learn the optimal behavior strategy, and the trained MARL model is called in conjunction with the Q value update algorithm to optimize resource allocation for multiple agents in a collaborative manner; The intelligent agent includes the electric vehicle and the building air-conditioning system, and each intelligent agent has a state, an action and a reward. The state includes the battery SOC of the electric vehicle, the building load data, the charging rate and the building system load. The action includes the charging and discharging rate of the electric vehicle and the power scheduling of the building air-conditioning system. The reward is the degree of optimization of the energy scheduling.

7. The source-grid-load-storage integrated dispatching method based on multi-agent learning according to claim 6 is characterized in that: The bidirectional energy flow scheduling scheme includes charging behavior of electric vehicles, scheduling of building air conditioning systems, and optimization of charging time windows, and the mathematical model of the bidirectional energy flow scheduling scheme includes a bidirectional energy flow model between electric vehicles and buildings and a building load balance equation; The expression of the bidirectional energy flow model between electric vehicles and buildings is: ； In the formula, is the change of battery SOC at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t, is the battery capacity, is the time step; The building load balance equation is expressed as: ； In the formula, is the total load of the building air conditioning system at time t, is the building load data at time t, is the power dispatch of the building air conditioning system at time t, is the charging rate of the electric vehicle at time t, is the discharge rate of the electric vehicle at time t.

8. A source-grid-load-storage integrated dispatching device, characterized in that: The device comprises: The dynamic model building module is used to obtain the energy flow parameters between electric vehicles and buildings, and to build an energy interaction dynamic model including the two-way energy flow between electric vehicles and buildings based on the CFD simulation model; An optimization strategy prediction module, used to use the energy interaction dynamic model and the battery SOC of the electric vehicle and the building energy consumption as inputs of an adaptive neuro-fuzzy inference model, predict the nonlinear relationship between the battery SOC and the building energy consumption, and output a battery SOC change trend and a charge and discharge optimization strategy; An energy flow scheduling module, used for optimizing the bidirectional energy scheduling between electric vehicles and buildings based on the charging and discharging optimization strategy through multi-agent reinforcement learning, so as to output an optimized bidirectional energy flow scheduling solution; Among them, the energy flow parameters include battery data of electric vehicles, building load data, vehicle location and charging rate. The adaptive neural fuzzy inference model is constructed based on the ANFIS algorithm and trained based on historical energy flow parameters. The bidirectional energy flow scheduling plan includes the charging and discharging time, charging power and discharge amount of electric vehicles.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images