A method of power optimization based on quadratic heuristic adaptive dynamic programming

Abstract

The present invention is applicable to the field of electric energy optimization technology, and provides an electric energy optimization method based on quadratic heuristic adaptive dynamic programming. The present invention can convert the potential energy of water flow into electric energy and supply the use of induction faucets and feed back to the power grid, thereby reducing energy consumption and electricity bills. At the same time, no external power supply is required, which improves the convenience of installation and use, and is particularly suitable for various occasions. The present invention makes full use of renewable energy, reduces dependence on traditional power supplies, enhances the safety and stability of the system, reduces the risk of leakage, and shows obvious advantages in intelligence, efficiency, energy saving, convenience, safety and environmental protection, and overcomes the problem that traditional induction switch faucets require external power supply and are inconvenient to install. For places with large water consumption such as hotels, hospitals, and bath centers, the present invention realizes the effective utilization of resources, saves electricity bills, alleviates the pressure on the power grid, and has the characteristics of water saving and energy saving, which is convenient for wide application and promotion.

Description

Technical Field

[0001] The invention belongs to the technical field of electric energy optimization, and in particular relates to an electric energy optimization method based on secondary heuristic adaptive dynamic programming. Background Art

[0002] For places with huge water consumption, such as hotels, hospitals, and bath centers, their water pipes often carry high pressure, which makes the tap water flowing through the pipes have large kinetic energy. However, this part of energy is mostly not effectively utilized in reality. Instead, the excessive kinetic energy of tap water causes a strong impact on water-using equipment, shortening the service life of the equipment. The field of tap water power generation and electric energy utilization has shown broad development prospects.

[0003] The currently widely used water control solutions have the following deficiencies and technical defects: (1) Power consumption: Many induction faucets rely on batteries or direct power supply. Although they consume less power in standby mode, long-term use will still accumulate a certain amount of power consumption, especially for models that require frequent battery replacement. (2) Inefficiency: Due to the lack of intelligent control, mechanical faucets may cause waste of water resources and energy during use, reducing overall efficiency. (3) Inconvenient installation and maintenance: Common induction faucets on the market usually require power cords or other power support, which has many problems such as inconvenient installation and non-energy-saving use. (4) Not adaptable to diverse needs: Ordinary faucets cannot be flexibly adjusted according to different usage scenarios or needs.

[0004] In view of the above problems, the present invention proposes an electric energy optimization method based on quadratic heuristic adaptive dynamic programming. Summary of the invention

[0005] The purpose of the present invention is to provide an electric energy optimization method based on quadratic heuristic adaptive dynamic programming, aiming to solve the problems raised in the above background technology.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] A method for optimizing electric energy based on secondary heuristic adaptive dynamic programming, the method is based on an electric energy optimization system, the electric energy optimization system includes a self-generating faucet, a secondary heuristic dynamic programming controller, a power grid and an electric energy optimization utilization unit; the self-generating faucet includes an induction faucet and a hydroelectric power generation device; the electric energy optimization utilization unit includes a rectifier, a charging controller, a battery pack, an inverter and an electric energy switching device; the induction faucet is connected to the electric energy switching device and the secondary heuristic dynamic programming controller; the hydroelectric power generation device is connected in series with a water supply pipe and the induction faucet; the rectifier connects the hydroelectric power generation device and the charging controller; the charging controller is connected to the battery pack and the secondary heuristic dynamic programming controller; the battery pack is connected to the inverter and the secondary heuristic dynamic programming controller; the inverter is connected to the electric energy switching device; the electric energy switching device is connected to the secondary heuristic dynamic programming controller and the power grid;

[0008] The electric energy optimization method comprises the following steps:

[0009] System operation: The induction faucet discharges water, driving the hydroelectric generator to generate electricity. The alternating current generated by the water flow generator is rectified and filtered by the rectifier to obtain smooth direct current. The direct current is stored in the battery pack under the control of the charge controller. The electric energy in the battery pack is converted from direct current to alternating current through the inverter device, and the electric energy is supplied to the induction faucet or incorporated into the power grid through the control of the electric energy switching device.

[0010] Power optimization: When the system is working, the quadratic heuristic dynamic programming controller uses the quadratic heuristic adaptive dynamic programming method to optimize the power. Under the condition that the fluctuation rate of the power grid and the state of charge of the battery group meet the limited range, the battery group discharge power and load power are adjusted to maximize the benefits.

[0011] Among them, the fluctuation rate of the electric energy fed back to the grid Discharge power of battery pack The rate of change, load power The rate of change and the system rated output power related: ;

[0012] Battery charge status The ratio of the current power of the battery pack to the capacity of the battery pack: ,in E 0 is the initial capacity of the battery pack, E total is the battery capacity, The power generated by the system, is the discharge power of the battery pack.

[0013] Furthermore, the quadratic heuristic dynamic programming controller includes a model network, two upper and lower evaluation networks and an execution network, and the model network, the evaluation network and the execution network are all constructed using a BP neural network;

[0014] In the execution network, the input is the state quantity of the system , that is, the fluctuation rate of the electric energy fed back to the grid and battery charge status ; Output is system Control strategy at all times , that is, the battery pack discharge power correction value And load power correction value ;

[0015] In the model network, the input is the control strategy and the state quantity of the system ; Output is the state quantity of the system in the next stage ;

[0016] The upper and lower evaluation networks have the same structure and parameters, but different inputs and outputs. The input of the upper evaluation network is the current state quantity. , the output is the cost function For the current state The partial derivative of ; The input of the next evaluation network is the state quantity of the controlled object in the next stage , the output is the cost function The state quantity of the next stage The partial derivative of .

[0017] Furthermore, the specific implementation steps of the quadratic heuristic dynamic programming controller using the quadratic heuristic adaptive dynamic programming method to perform power optimization are as follows:

[0018] Step 1: Determine the structure and parameters of each network in the quadratic heuristic dynamic programming controller;

[0019] Step 2: Use BP neural network to model the system, use the obtained model as the model network and train the model network;

[0020] Step 3: Calculate the current system rated output power, battery capacity, battery state of charge, sampling time, and power fluctuation rate of the fed-back grid.

[0021] Step 4: Initialize the execution network and evaluation network;

[0022] Step 5: Determine whether the power fluctuation rate of the current feedback grid and the state of charge of the battery group are within the limit range. If not, make the next correction to find the correction value of the battery group discharge power and the load power. If they are within the constraint range, do not correct the battery group discharge power and load power, that is, the battery group discharge power and load power remain unchanged, and directly output the control strategy, power fluctuation rate of the feedback grid, battery group state of charge and system benefits at each moment.

[0023] Step 6: Take the current power fluctuation rate of the grid and the battery charge state as the input of the execution network and execute the network output control strategy; the control strategy includes the battery discharge power correction value And load power correction value ;

[0024] Step 7: Save the current control strategy, input the power fluctuation rate of the current grid feedback, the battery charge state and the control strategy into the model network, and the model network outputs the state quantity of the controlled object in the next stage;

[0025] Step 8: Input the power fluctuation rate of the next stage of feedback grid and the state of charge of the battery pack into the lower evaluation network. The lower evaluation network outputs the partial derivative of the cost function of the next stage to the state of the controlled object and calculates the error of the execution network. , update the execution network weights;

[0026] Step 9: Input the power fluctuation rate of the current feedback grid and the state of charge of the battery pack into the upper evaluation network. The upper evaluation network outputs the partial derivative of the cost function at the current moment with respect to the state of the controlled object and calculates the error of the evaluation network. , update the evaluation network weights;

[0027] Step 10: Repeat steps 5 to 9 during the observation time until the control process is completed, and output the control strategy, feedback grid power fluctuation rate, battery charge state and system revenue at each moment.

[0028] Furthermore, in the feedback grid power energy fluctuation rate and battery charge status Under the conditions that all meet the limited range, the system benefits P for:

[0029] ;

[0030] in t represents the time variable, n is the upper limit of the number of time steps, is the sampling time, For each City electricity price, For each Feedback to the grid price, For conversion relationship .

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention can convert the potential energy of water flow into electrical energy and supply it for the use of induction faucets and feed it back to the power grid, reducing energy consumption and electricity expenses. At the same time, no external power supply is required, which greatly improves the convenience of installation and use, and is particularly suitable for various occasions. The present invention makes full use of renewable energy, reduces dependence on traditional power supplies, enhances the safety and stability of the system, reduces the risk of leakage, promotes the popularization of green energy-saving products, and shows obvious advantages in intelligence, efficiency, energy saving, convenience, safety and environmental protection, and overcomes the problem that traditional induction switch faucets require external power supply and are inconvenient to install. For places with large water consumption such as hotels, hospitals, and bath centers, the present invention realizes the effective use of resources, saves electricity expenses, alleviates the pressure on the power grid, and has the characteristics of water saving and energy saving, which is convenient for wide application and promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the structure of the power optimization system.

[0034] Figure 2 Schematic diagram of the structure of the quadratic heuristic dynamic programming controller.

[0035] Figure 3 Optimize flow chart for electrical energy.

[0036] Figure 4 This is the structural diagram of the BP neural network. DETAILED DESCRIPTION

[0037] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below, but it should not be construed as limiting the applicable scope of the present invention.

[0038] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0039] An embodiment of the present invention provides a power optimization method based on quadratic heuristic adaptive dynamic programming. The method is based on a power optimization system ( Figure 1), the electric energy optimization system includes a self-generating faucet, a DHP (Dual Heuristic Programming, quadratic heuristic dynamic programming) controller, a power grid and an electric energy optimization utilization unit; the self-generating faucet includes an induction faucet and a hydroelectric power generation device; the electric energy optimization utilization unit includes a rectifier, a charging controller, a battery pack, an inverter and an electric energy switching device; the induction faucet is connected to the electric energy switching device and the quadratic heuristic dynamic programming controller; the hydroelectric power generation device is connected in series with the water supply pipe and the induction faucet; the rectifier connects the hydroelectric power generation device and the charging controller; the charging controller is connected to the battery pack and the quadratic heuristic dynamic programming controller; the battery pack is connected to the inverter and the quadratic heuristic dynamic programming controller; the inverter is connected to the electric energy switching device; the electric energy switching device is connected to the quadratic heuristic dynamic programming controller and the power grid;

[0040] The electric energy optimization method comprises the following steps:

[0041] System operation: The induction faucet discharges water, driving the hydroelectric generator to generate electricity. The alternating current generated by the water flow generator is rectified and filtered by the rectifier to obtain smooth direct current. The direct current is stored in the battery pack under the control of the charge controller. The electric energy in the battery pack is converted from direct current to alternating current through the inverter device, and the electric energy is supplied to the induction faucet or incorporated into the power grid through the control of the electric energy switching device.

[0042] Power optimization: When the system is working, the quadratic heuristic dynamic programming controller uses the quadratic heuristic adaptive dynamic programming method to optimize the power. Under the condition that the fluctuation rate of the power grid and the state of charge of the battery group meet the limited range, the battery group discharge power and load power are adjusted to maximize the benefits.

[0043] Among them, the fluctuation rate of the electric energy fed back to the grid Discharge power of battery pack The rate of change, load power The rate of change and the system rated output power related: In order to improve the grid's ability to accept hydropower and reduce the impact of the power fed back to the grid on the grid, the fluctuation rate of the power fed back to the grid should be Control within the allowable range Within.

[0044] Battery charge status The ratio of the current power of the battery pack to the capacity of the battery pack: , affected by the system power generation And the battery discharge power The impact ofE 0 is the initial capacity of the battery pack, E total is the battery pack capacity. In order to extend the service life of the battery pack, the battery pack charge state should be Control within the rated range (determined by the actual battery pack used).

[0045] Fluctuation rate of power fed back to the grid and battery charge status Under the conditions that all meet the limited range, the system benefits P for:

[0046] ;

[0047] in t represents the time variable, n is the upper limit of the number of time steps, is the sampling time, For each City electricity price, For each Feedback to the grid price, For conversion relationship .

[0048] In the embodiment of the present invention, the hydroelectric power generation device can be a small hydroelectric generator, which generates electricity by connecting the generator in series in a water supply pipe. The output voltage is AC 220-400V, and the power generation power can be selected to be 600-3000W.

[0049] The rectifying device may be a bridge rectifier that matches the output voltage and current of the generator.

[0050] The charging controller can control the charging and discharging of each battery by connecting to the battery pack to prevent the battery from being overcharged or over-discharged.

[0051] The battery pack can be a lithium battery without memory effect. Lithium batteries are also safer than lead-acid batteries. The battery pack can be adapted to a corresponding number of lithium batteries according to the system power generation, and the DC voltage can be selected as 12V or 24V.

[0052] The inverter device may be a 12V / 24V to 220V inverter that matches the system power.

[0053] The power switching device is connected to the output end of the inverter device, and can control the alternating current output by the inverter to supply power to loads such as induction faucets or to be incorporated into the power grid.

[0054] The present invention uses a quadratic heuristic dynamic programming controller, which is widely used in complex nonlinear systems such as power systems and boiler combustion control systems. By using neural networks, the quadratic heuristic dynamic programming controller can achieve approximate optimal control of the system, effectively handle multiple constraints, and perform well under various load environments. The quadratic heuristic dynamic programming controller has the ability to monitor the state of the power grid in real time and can make rapid adjustments based on changes in the state of the power grid, thereby reducing system fluctuations and improving the overall stability of the power grid. This not only helps to manage power resources more effectively, but also optimizes the balance between power generation and load, thereby reducing energy waste.

[0055] As a preferred embodiment of the present invention, Figure 2 As shown in FIG. 1 , the structure of the quadratic heuristic dynamic programming controller includes a model network, two upper and lower evaluation networks, and an execution network, which is implemented using a neural network. γ is a discount factor, which is used to adjust the influence in the feedback signal to ensure that the control network is properly adjusted according to the evaluation results. In simple terms, it can be regarded as an adjustment factor to help the system better adapt to the changing environment.

[0056] 1) The model network, execution network and evaluation network in the present invention are all constructed using BP neural network. The role of the hidden layer nodes is to extract and store the internal rules from the samples. Each hidden node has several weights. , and each weight is a parameter that enhances the network mapping capability.

[0057] 2) The input of the model network is the control strategy and the state quantity of the system , the output is the state quantity of the system in the next stage .

[0058] 3) The input of the execution network is the state quantity of the system , that is, the fluctuation rate of the electric energy fed back to the grid and battery charge status , the output is the system Control strategy at all times , that is, the battery pack discharge power correction value And load power correction value The execution network of the present invention adopts a 2-6-2 structure, wherein the hidden layer adopts a bipolar function and the output layer adopts a linear function. The training of the execution network consists of a forward calculation process and a back propagation error process.

[0059] Forward calculation process:

[0060] ;

[0061] ;

[0062] ;

[0063] Back propagation error process:

[0064] ;

[0065] Weight update:

[0066] ;

[0067] in Represents the execution network hidden layer The input of each node; ( k ) represents the execution of the network input layer to the hidden layer The weight of each neuron; Represents the execution network hidden layer The output of each node; Indicates that the output layer of the execution network is at time k No. The final output value of each neuron; ( k ) represents the execution of the hidden layer of the network neurons to the output layer The weight of each neuron; Represents the error in performing network backpropagation; is the utility function, which is defined as the fluctuation rate of the power energy fed back to the grid according to the specific control target , Battery discharge power correction value and the current moment t Function of Indicates the execution of network weight update rules; is the learning rate; for Execute network weights at all times.

[0068] 4) The upper and lower evaluation networks have the same structure and parameters, but their inputs and outputs are different. The input of the upper evaluation network is the current state quantity , the output is the cost function J [ x ( k )] for the current state The partial derivative of ; The input of the next evaluation network is the state quantity of the controlled object in the next stage , the output is the cost function J [ x ( k+1)] for the next stage state quantity The partial derivative of In the present invention, the evaluation network adopts a 2-6-2 structure, wherein the hidden layer adopts a bipolar function and the output layer adopts a linear function. The training of the evaluation network consists of a forward calculation process and a back propagation error process.

[0069] Forward calculation process:

[0070] ;

[0071] ;

[0072] ;

[0073] Back propagation error process:

[0074] ;

[0075] Weight update:

[0076] ;

[0077] in represents the hidden layer of the evaluation network The input of each node; ( k ) represents the evaluation network input layer to the hidden layer The weight of each neuron; represents the hidden layer of the evaluation network The output of each node; Indicates that the output layer of the evaluation network is at time The final output value of the ith neuron, ( k ) represents the hidden layer of the evaluation network neurons to the output layer The weight of each neuron; Represents the error in the back propagation of the evaluation network; Represents the weight update rule of the evaluation network; is the learning rate; for Evaluate network weights at all times.

[0078] This design enables the two evaluation networks to evaluate the current state and the next state respectively, thereby helping the controller to optimize the control strategy. Compared with general dynamic programming methods, it shows significant advantages in dynamic response, convergence speed, control accuracy and stability.

[0079] 5) Utility Function U ( k) is defined as the fluctuation rate of the electric energy fed back to the grid , Battery discharge power correction value and the current moment t The function is defined according to the control objective.

[0080] Description of the controller: First, the current state of the system Input execution network and generate a control signal based on the current state The model network then receives the control signal and the current state , and predict the state quantity of the next stage Then the upper and lower evaluation networks evaluate the current state quantity and the next stage state Evaluate and determine the battery pack charge status and the fluctuation rate of the power fed back to the grid Finally, the evaluation network outputs the partial derivative of the cost function at the current moment with respect to the state of the controlled object and calculates the error of the evaluation network. , update the evaluation network weights, train the evaluation network; the next evaluation network outputs the partial derivative of the cost function of the next stage to the state of the controlled object and calculates the error of the execution network , update the execution network weights and train the execution network.

[0081] As a preferred embodiment of the present invention, Figure 3 As shown, the specific implementation steps of the quadratic heuristic dynamic programming controller using the quadratic heuristic adaptive dynamic programming method to perform power optimization are as follows:

[0082] Step 1: Determine the structure and parameters of each network in the quadratic heuristic dynamic programming controller; including the number of layers of the model network, evaluation network and execution network, the number of nodes in each layer, the type of transfer function and the learning rate.

[0083] Step 2: Train the model network; use BP neural network to model the system, use the obtained model as the model network and train the model network.

[0084] Step 3: Calculate the current system rated output power, battery capacity, battery state of charge, sampling time, and power fluctuation rate fed back to the grid.

[0085] Step 4: Initialize the execution network and evaluation network, including input variables and learning rate.

[0086] Step 5: Determine whether the fluctuation rate of the electric energy fed back to the grid and the state of charge of the battery pack are within the limit range at the current moment. If not, proceed to the next step of correction to find the correction value of the battery pack discharge power and the load power. If within the constraint range, do not correct the battery pack discharge power and load power, that is, the battery pack discharge power and load power remain unchanged, and directly output the control strategy, fluctuation rate of the electric energy fed back to the grid, state of charge of the battery pack and system benefits at each moment.

[0087] Step 6: Take the current power fluctuation rate of the grid and the battery charge state as the input of the execution network and execute the network output control strategy; the control strategy includes the battery discharge power correction value And load power correction value .

[0088] Step 7: Save the current control strategy, input the power fluctuation rate of the current grid feedback, the battery charge state and the control strategy into the model network, and the model network outputs the state quantity of the controlled object in the next stage;

[0089] Step 8: Input the power fluctuation rate of the next stage of feedback grid and the state of charge of the battery pack into the lower evaluation network. The lower evaluation network outputs the partial derivative of the cost function of the next stage to the state of the controlled object and calculates the error of the execution network. , update the execution network weights.

[0090] Step 9: Input the power fluctuation rate of the current feedback grid and the state of charge of the battery pack into the upper evaluation network. The upper evaluation network outputs the partial derivative of the cost function at the current moment with respect to the state of the controlled object and calculates the error of the evaluation network. , update the evaluation network weights.

[0091] Step 10: Repeat steps 5 to 9 during the observation time until the control process is completed, and output the control strategy, feedback grid power fluctuation rate, battery charge state and system revenue at each moment.

[0092] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A power optimization method based on quadratic heuristic adaptive dynamic programming, characterized in that: The method is based on an electric energy optimization system, which includes a self-generating faucet, a secondary heuristic dynamic programming controller, a power grid and an electric energy optimization utilization unit; the self-generating faucet includes an induction faucet and a hydroelectric power generation device; the electric energy optimization utilization unit includes a rectifier, a charging controller, a battery pack, an inverter and an electric energy switching device; the induction faucet is connected to the electric energy switching device and the secondary heuristic dynamic programming controller; the hydroelectric power generation device is connected in series with the water supply pipe and the induction faucet; the rectifier connects the hydroelectric power generation device and the charging controller; the charging controller is connected to the battery pack and the secondary heuristic dynamic programming controller; the battery pack is connected to the inverter and the secondary heuristic dynamic programming controller; the inverter is connected to the electric energy switching device; the electric energy switching device is connected to the secondary heuristic dynamic programming controller and the power grid; The electric energy optimization method comprises the following steps: System operation: The induction faucet discharges water, driving the hydroelectric generator to generate electricity. The alternating current generated by the water flow generator is rectified and filtered by the rectifier to obtain smooth direct current. The direct current is stored in the battery pack under the control of the charge controller. The electric energy in the battery pack is converted from direct current to alternating current through the inverter device, and the electric energy is supplied to the induction faucet or incorporated into the power grid through the control of the electric energy switching device. Power optimization: When the system is working, the quadratic heuristic dynamic programming controller uses the quadratic heuristic adaptive dynamic programming method to optimize the power. Under the condition that the fluctuation rate of the power grid and the state of charge of the battery group meet the limited range, the battery group discharge power and load power are adjusted to maximize the benefits. Among them, the fluctuation rate of the electric energy fed back to the grid Discharge power of battery pack The rate of change, load power The rate of change and the system rated output power related: ; Battery charge status The ratio of the current power of the battery pack to the capacity of the battery pack: ,in E 0 is the initial capacity of the battery pack, E total is the battery capacity, The power generated by the system, is the discharge power of the battery pack; The quadratic heuristic dynamic programming controller includes a model network, two upper and lower evaluation networks and an execution network, and the model network, the evaluation network and the execution network are all constructed using a BP neural network; In the execution network, the input is the state quantity of the system , that is, the fluctuation rate of the electric energy fed back to the grid and battery charge status ; Output is system Control strategy at all times , that is, the battery pack discharge power correction value And load power correction value ; In the model network, the input is the control strategy and the state quantity of the system ; Output is the state quantity of the system in the next stage ; The upper and lower evaluation networks have the same structure and parameters, but different inputs and outputs. The input of the upper evaluation network is the current state quantity. , the output is the cost function For the current state The partial derivative of ; The input of the next evaluation network is the state quantity of the controlled object in the next stage , the output is the cost function The state quantity of the next stage The partial derivative of ; The specific implementation steps of the quadratic heuristic dynamic programming controller using the quadratic heuristic adaptive dynamic programming method to perform power optimization are as follows: Step 1: Determine the structure and parameters of each network in the quadratic heuristic dynamic programming controller; Step 2: Use BP neural network to model the system, use the obtained model as the model network and train the model network; Step 3: Calculate the current system rated output power, battery capacity, battery state of charge, sampling time, and power fluctuation rate of the fed-back grid. Step 4: Initialize the execution network and evaluation network; Step 5: Determine whether the power fluctuation rate of the current feedback grid and the state of charge of the battery group are within the limit range. If not, make the next correction to find the correction value of the battery group discharge power and the load power. If they are within the constraint range, do not correct the battery group discharge power and load power, that is, the battery group discharge power and load power remain unchanged, and directly output the control strategy, power fluctuation rate of the feedback grid, battery group state of charge and system benefits at each moment. Step 6: Take the current power fluctuation rate of the grid and the battery charge state as the input of the execution network and execute the network output control strategy; the control strategy includes the battery discharge power correction value And load power correction value ; Step 7: Save the current control strategy, input the power fluctuation rate of the current grid feedback, the battery charge state and the control strategy into the model network, and the model network outputs the state quantity of the controlled object in the next stage; Step 8: Input the power fluctuation rate of the next stage of feedback grid and the state of charge of the battery pack into the lower evaluation network. The lower evaluation network outputs the partial derivative of the cost function of the next stage to the state of the controlled object and calculates the error of the execution network. , update the execution network weights; Step 9: Input the power fluctuation rate of the current feedback grid and the state of charge of the battery pack into the upper evaluation network. The upper evaluation network outputs the partial derivative of the cost function at the current moment with respect to the state of the controlled object and calculates the error of the evaluation network. , update the evaluation network weights; Step 10: Repeat steps 5 to 9 during the observation time until the control process is completed, and output the control strategy, feedback grid power fluctuation rate, battery charge state and system revenue at each moment.

2. The electric energy optimization method based on quadratic heuristic adaptive dynamic programming according to claim 1 is characterized in that: The fluctuation rate of the power energy fed back to the grid and battery charge status Under the conditions that all meet the limited range, the system benefits P for: ； in t represents the time variable, n is the upper limit of the number of time steps, is the sampling time, For each City electricity price, For each Feedback to the grid price, For conversion relationship .

Images