Photovoltaic off-grid and grid-connected integrated control system based on big data technology

Abstract

The invention relates to the technical field of photovoltaic control systems, and discloses a photovoltaic off-grid and grid-connected integrated control system based on a big data technology, which integrates a photovoltaic data acquisition module and a photovoltaic off-grid and grid-connected control module, and constructs and maintains a photovoltaic off-grid and grid-connected control model by acquiring photovoltaic cell panel, environment and power grid state data in real time. According to the model, the whole process of photovoltaic power is modeled into a Markov decision process, and the model comprises a state space sub-module, an action space definition sub-module, a reward function design sub-module, a strategy optimization sub-module and the like. A deep reinforcement learning algorithm is adopted, historical data and real-time feedback are combined, the system continuously optimizes a control strategy, and long-term benefit maximization is pursued. The control instruction is generated and executed based on the optimization strategy, and intelligent control over the photovoltaic cell panel, the energy storage device, the power grid connection and the load device is achieved. The flexibility, the stability and the economical efficiency of the photovoltaic power generation system are improved, and large-scale application of photovoltaic power generation is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic control systems, and in particular to a photovoltaic off-grid and grid-connected integrated control system based on big data technology. Background Art

[0002] With the continuous growth of global energy demand and the increasing awareness of environmental protection, the development and utilization of renewable energy has become the focus of international attention. As one of the most promising renewable energy sources, photovoltaic power generation is of great importance. However, photovoltaic power generation systems face many challenges in practical applications, especially the flexible access and exit of photovoltaic power to the power grid (i.e., grid-connected and off-grid switching), the improvement of photovoltaic power generation efficiency, the maintenance of grid stability, the effective management of energy storage equipment, and the reliable guarantee of load power supply. These issues have become key factors restricting the large-scale application of photovoltaic power generation.

[0003] At present, photovoltaic control systems often focus on a single grid-connected or off-grid operation mode, making it difficult to achieve intelligent switching and efficient coordination between the two. At the same time, these systems lack comprehensive data collection and analysis capabilities, and are unable to accurately grasp multi-dimensional information such as photovoltaic power generation, grid status, energy storage conditions, and load demand in real time, resulting in greater blindness and lag in the formulation and implementation of control strategies. In addition, traditional control methods often find it difficult to achieve optimal control effects when dealing with complex and changeable operating scenarios, affecting the overall performance and economic benefits of photovoltaic power generation systems. Summary of the invention

[0004] The purpose of the present invention is to provide a photovoltaic off-grid and on-grid integrated control system based on big data technology to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a photovoltaic off-grid integrated control system based on big data technology, the system comprising: Photovoltaic data acquisition module: Integrates photovoltaic sensors, environmental monitors and power grid status monitors to collect real-time power generation data of photovoltaic panels, environmental factor data and real-time status data of the power grid, and perform preprocessing; Photovoltaic off-grid control module: Build and maintain a photovoltaic off-grid control model, which covers photovoltaic power generation nodes, grid connection nodes, energy storage nodes and load nodes, and models the entire process of photovoltaic power generation, storage, grid / off-grid switching and load power supply as a Markov decision process; the photovoltaic off-grid control decision module further includes: State space definition submodule: used to define the state space of the system, including the power generation power, power generation efficiency and light intensity dimensions of photovoltaic power generation nodes, the grid connection status, off-grid status and grid voltage dimensions of grid connection nodes, the remaining power, charging / discharging status and battery temperature dimensions of energy storage nodes, and the power demand, power type and priority dimensions of load nodes; Action space definition submodule: used to define the action space that the system can execute, including adjusting the tilt angle and orientation of photovoltaic panels, controlling the charging and discharging power of energy storage devices, performing grid-connected or off-grid operations, adjusting the power supply sequence and power distribution of loads, starting or shutting down backup energy equipment, performing equipment maintenance, and notifying operators; Reward function design submodule: Design reward functions for photovoltaic power generation efficiency, grid stability, energy storage equipment utilization efficiency, load power supply satisfaction, system operation and maintenance costs, and emergency response capabilities, conduct quantitative evaluation of the actions taken by the system, and give corresponding rewards or penalties; Strategy optimization submodule: uses deep reinforcement learning algorithms, combined with historical data and real-time feedback, to continuously learn and optimize control strategies in order to maximize long-term cumulative rewards; Control instruction generation and execution submodule: Based on the data provided by the photovoltaic data acquisition module and according to the optimized control strategy, specific control instructions are generated and sent to photovoltaic panels, energy storage devices, grid connection devices and load devices through the communication interface.

[0006] Preferably, the dimensions of power generation power and power generation efficiency of the photovoltaic power generation node are further refined into instantaneous power generation power, average daily power generation, power generation efficiency and power generation efficiency attenuation rate; the grid-connected status, off-grid status and grid voltage dimensions of the grid connection node include grid-connected / off-grid status identification, grid voltage value and voltage fluctuation range; the remaining power, charging / discharging status and battery temperature dimensions of the energy storage node are refined into current power, maximum charging / discharging power, battery temperature and battery health status; the power demand, power type and priority dimensions of the load node include total power consumption, peak power consumption period, power type classification and load power supply priority sorting.

[0007] Preferably, the reward function design submodule further includes: Photovoltaic power generation efficiency reward function: defined as ,in is the actual photovoltaic power generation power, is the rated power of the photovoltaic panel, The photovoltaic power generation efficiency bonus coefficient is adjusted according to the type, aging and environmental factors of the photovoltaic panels; Grid stability reward function: defined as

[0008] in and are the actual grid voltage and frequency, and are rated voltage and frequency respectively, The grid stability bonus factor is adjusted according to the specifications and requirements of the grid; Energy storage equipment efficiency reward function: defined as

[0009] in is the actual charge and discharge capacity of the energy storage device during the control cycle. is the total capacity of the energy storage device, and are the charge states of the energy storage device at the end and beginning of the control cycle, respectively. and are the maximum and minimum states of charge of the energy storage device, respectively. The energy storage equipment utilization efficiency reward coefficient is adjusted according to the type of energy storage equipment and usage strategy; Load power supply satisfaction reward function: defined as

[0010] in is the actual power supply, is the load demand power, For power supply delay time, is the maximum acceptable power supply delay time, The power supply satisfaction reward coefficient for the load is adjusted according to the type and importance of the load; System operation and maintenance cost penalty function: defined as ,in The actual operation and maintenance costs include equipment maintenance, troubleshooting, and labor costs. To estimate the operation and maintenance costs, It is the penalty coefficient of system operation and maintenance cost, which is adjusted according to the economy and operation and maintenance strategy of the system; Emergency response capability reward function: defined as ,in is the system's response time to a fault or abnormal situation, is the maximum response time threshold set, The reward coefficient for emergency response capability is adjusted according to the system's requirements for emergency response; The final reward function R output by the reward function design submodule is defined as:

[0011] This reward function is used to evaluate the control effect of the system in the current state and guide the strategy optimization and learning submodules to adjust and optimize the strategy.

[0012] Preferably, the photovoltaic off-grid and on-grid control model uses the Actor-Critic algorithm in the deep reinforcement learning algorithm to train the photovoltaic off-grid and on-grid control strategy.

[0013] Preferably, various sensors and monitoring devices are deployed in the photovoltaic system and the power grid to collect the actual data of the photovoltaic system's power generation, the power state of the energy storage device, the voltage and frequency of the power grid, and the power load in real time; combined with the preset off-grid switching rules and evaluation criteria, simulated control actions are performed on the collected actual data, and the corresponding immediate rewards or penalties are calculated; the actual data, simulated control actions, and immediate rewards or penalties are combined into a state-action-reward tuple as preliminary training data for the photovoltaic off-grid control model.

[0014] Preferably, the photovoltaic power generation records, power grid operation data and power load change data in the historical data are used to construct a historical status data set of the photovoltaic system and the power grid; for each data in the historical status data set, according to the preset off-grid switching rules and evaluation criteria, possible control actions are simulated and the corresponding immediate rewards or penalties are calculated; the historical status data, simulated control actions and immediate rewards or penalties are combined into additional state-action-reward tuples, which are supplemented into the preliminary training data to form a complete training data set for the photovoltaic off-grid control model.

[0015] Preferably, the training step of the photovoltaic off-grid and grid-connected control model includes: S1: Initialize the Actor network and the Critic network. The Actor network is used to generate control actions, and the Critic network is used to evaluate the value of the actions generated by the Actor network. At the same time, initialize the experience playback memory to store the state transition experience of photovoltaic off-grid and grid-connected. S2: Set the learning rate α, discount factor γ, exploration rate ε, and number of training rounds, where α controls the learning speed, γ controls the discounted value of future returns, ε controls the balance between exploring new actions and exploiting known actions, and the number of training rounds determines the total number of training iterations; S3: For each training round, perform the following steps: S301: Initialize the current state s of the photovoltaic system, energy storage device and power grid from the training data set; S302: The Actor network generates a control action a according to the current state s, which indicates the off-grid or on-grid operation of the photovoltaic system, or the charging and discharging operation of the energy storage device; S303: Execute action a, observe the next state s' and the immediate reward r or penalty p, where the immediate reward r or penalty p is calculated according to a preset evaluation standard; S304: storing the experience (s, a, r / p, s') into the experience playback memory; S305: Randomly sample a batch of experiences from the experience replay memory to update the Actor network and the Critic network; S306: Using the sampled experience and the Critic network to calculate the value of the action, and updating the parameters of the Actor network and the Critic network according to the loss function of the Actor-Critic algorithm; S307: Repeat steps S302 to S306 until the end condition of the training round is reached; S4: Repeat step S3 until the preset number of training rounds is reached or the output values ​​of the Actor network and the Critic network converge; S5: After the training is completed, the obtained Actor network is used as a photovoltaic off-grid control strategy model. For any given current state s of the photovoltaic system, energy storage equipment and power grid, the optimal control action a is generated through the forward propagation of the Actor network.

[0016] Preferably, the photovoltaic data acquisition module pre-processes the collected data by means of data cleaning and standardization.

[0017] Preferably, the method of cleaning the data includes: For missing values, linear interpolation algorithm is used to fill them; For outliers, the threshold judgment method is used to regard data points that exceed the preset threshold range as outliers and replace them with the mean of their neighboring points.

[0018] Preferably, the method of standardizing the data includes: using the Z-score standardization method to standardize the continuous features, specifically: subtracting the mean of each numerical data and dividing it by its standard deviation, so that the processed data conforms to the standard normal distribution, and the formula for standardization is: ,in, is the original data, is the mean of the original data, is the standard deviation of the original data, is the standardized data.

[0019] Compared with the prior art, the present invention has the following beneficial effects: By building and maintaining a photovoltaic off-grid control model, the system can automatically switch between grid connection and off-grid according to the grid status, photovoltaic power generation and load demand, ensuring the efficient use of photovoltaic power and the stable operation of the grid. This intelligent switching capability greatly improves the flexibility and adaptability of the photovoltaic power generation system. The photovoltaic data acquisition module integrated in the system can collect power generation data and environmental factor data of photovoltaic panels in real time, providing an accurate information basis for the photovoltaic off-grid control module. Combined with the deep reinforcement learning algorithm, the system can continuously optimize the control strategy and adjust the tilt angle and orientation of the photovoltaic panels to maximize the efficiency of photovoltaic power generation.

[0020] By monitoring the grid status data in real time and combining the reward function to design submodules to quantitatively evaluate the grid stability, the system can respond to grid fluctuations in a timely manner, adjust the output of photovoltaic power generation and energy storage equipment, and effectively maintain the stability and security of the grid. The system comprehensively monitors and manages the remaining power, charging / discharging status, and battery temperature of the energy storage nodes. By intelligently controlling the charging and discharging power of the energy storage equipment, it improves the efficiency of the energy storage equipment, extends its service life, and reduces the cost of energy storage. According to the power demand, power type, and priority dimension of the load node, the system intelligently adjusts the power supply sequence and power distribution of the load to ensure priority power supply to critical loads, improve the reliability of load power supply and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the photovoltaic off-grid and grid-connected integrated control system based on big data technology described in the present invention; Figure 2 Schematic diagram of the design of the reward function; Figure 3 This is a flow chart for building a photovoltaic off-grid control model. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1-3 The present invention provides a technical solution: a photovoltaic off-grid integrated control system based on big data technology, the system comprising: Photovoltaic data acquisition module: This module integrates photovoltaic sensors, environmental monitors and power grid status monitors. Photovoltaic sensors are responsible for real-time acquisition of photovoltaic panel power generation data, such as voltage, current and power; environmental monitors are used to monitor environmental factors data, such as light intensity, temperature, humidity, etc.; power grid status monitors obtain real-time power grid voltage, frequency, phase and other status data. After preprocessing, these data are formed into a standardized data format to provide a basis for subsequent control decisions.

[0024] Photovoltaic off-grid control module: further includes: State space definition submodule: This submodule defines the state space of the system, covering the power generation, power generation efficiency and light intensity of photovoltaic power generation nodes, the grid connection status, off-grid status and grid voltage of grid connection nodes, the remaining power, charging / discharging status and battery temperature of energy storage nodes, and the power demand, power type and priority of load nodes. These state variables together constitute a complete description of the current state of the system.

[0025] Action space definition submodule: This submodule defines the action space that the system can perform, including but not limited to adjusting the tilt angle and orientation of photovoltaic panels to optimize power generation efficiency, controlling the charging and discharging power of energy storage devices to balance supply and demand, performing grid-connected or off-grid operations to adapt to grid conditions, adjusting the power supply sequence and power distribution of loads to meet different needs, starting or shutting down backup energy equipment to deal with emergencies, performing equipment maintenance to maintain system performance, and notifying operators for manual intervention.

[0026] Reward function design submodule: Reward functions are designed based on factors such as photovoltaic power generation efficiency, grid stability, efficiency of energy storage equipment, satisfaction with load power supply, system operation and maintenance costs, and emergency response capabilities. For each action taken by the system, the reward function will be quantitatively evaluated and corresponding rewards or penalties will be given. This helps guide the system to adopt a better control strategy.

[0027] Strategy optimization submodule: It uses deep reinforcement learning algorithms, combined with historical data and real-time feedback, to continuously learn and optimize control strategies. The algorithm is updated iteratively to maximize the long-term cumulative rewards, thereby finding the optimal or near-optimal control strategy.

[0028] Control instruction generation and execution submodule: Based on the data provided by the photovoltaic data acquisition module and the optimization strategy output by the strategy optimization submodule, this submodule generates specific control instructions. These instructions are sent to photovoltaic panels, energy storage devices, grid connection devices and load devices through the communication interface to achieve real-time control of the system. The instructions include but are not limited to adjusting the angle and direction of photovoltaic panels, controlling the charging and discharging of energy storage devices, switching between grid-connected and off-grid states, and adjusting load power supply.

[0029] The present invention will be further described below in conjunction with Examples 1 to 3:

[0030] Embodiment 1: The reward function design submodule is used to calculate and output a comprehensive reward value according to the operating state and control effect of the system to guide the strategy optimization and learning submodule to adjust the strategy.

[0031] The reward function design submodule further includes: Photovoltaic power generation efficiency reward function: defined as ,in, is the actual photovoltaic power generation power, in kilowatts (kW); is the rated power of the photovoltaic panel, in kilowatts (kW); is the photovoltaic power generation efficiency bonus coefficient, and its value is set according to the type, aging degree and environmental factors of the photovoltaic panels. For example, for newly installed high-efficiency photovoltaic panels, It can be set to a higher value to encourage the system to prioritize the use of photovoltaic power generation.

[0032] Grid stability reward function: defined as

[0033] in, and are the actual grid voltage and frequency, in Volt (V) and Hertz (Hz), respectively; and They are rated voltage and frequency, also in volts (V) and hertz (Hz); It is the grid stability bonus coefficient, and its value is set according to the specifications and requirements of the grid.

[0034] Energy storage equipment efficiency reward function: defined as

[0035] in, The actual charge and discharge amount of the energy storage device during the control period, in kilowatt-hours (kWh); is the total capacity of the energy storage device, also in kilowatt-hours (kWh); and are the state of charge of the energy storage device at the end and beginning of the control cycle, expressed in percentage; and They are the maximum and minimum states of charge of the energy storage device, also expressed in percentages; It is the energy storage equipment utilization efficiency reward coefficient, and its value is set according to the type of energy storage equipment and its usage strategy.

[0036] Load power supply satisfaction reward function: defined as

[0037] in, is the actual power supply, in kilowatts (kW); is the load demand power, also in kilowatts (kW); is the power supply delay time, in seconds (s); is the maximum acceptable power supply delay time, also in seconds (s); The power supply satisfaction reward coefficient for the load is set according to the type and importance of the load.

[0038] System operation and maintenance cost penalty function: defined as ,in, is the actual operation and maintenance cost, including equipment maintenance, troubleshooting and labor costs, in Yuan (RMB); is the estimated operation and maintenance cost, also in RMB; It is the system operation and maintenance cost penalty coefficient, and its value is set according to the system's economy and operation and maintenance strategy.

[0039] Emergency response capability reward function: defined as ,in, The system's response time to a fault or abnormal situation, in seconds (s); The maximum response time threshold is also in seconds (s). It is the emergency response capability reward coefficient, and its value is set according to the system's requirements for emergency response.

[0040] Combining the above reward and penalty functions, the final reward function output by the reward function design submodule is Defined as:

[0041] This reward function is used to evaluate the control effect of the system in the current state and guide the strategy optimization and learning submodule to adjust and optimize the strategy. By continuously adjusting and optimizing the strategy, the system can improve the efficiency of photovoltaic power generation, grid stability, energy storage equipment utilization efficiency, load power supply satisfaction and emergency response capabilities while meeting the load power supply needs, and reduce the system operation and maintenance costs.

[0042] Embodiment 2: This embodiment aims to build a photovoltaic off-grid and grid-connected integrated control model that can intelligently decide whether the photovoltaic system is off-grid or on-grid, as well as the charging and discharging strategy of the energy storage device. The model is trained based on the Actor-Critic algorithm in deep reinforcement learning to achieve the optimal control strategy.

[0043] Step S1: Initialize network and storage Initialize the Actor network, whose input is the current status of the photovoltaic system, energy storage equipment and power grid (such as photovoltaic power generation, energy storage equipment power, grid voltage and frequency, etc.), and the output is the control action (such as off-grid, grid-connected, energy storage equipment charging or discharging).

[0044] Initialize the Critic network, whose input is the state-action pair and output is the value estimate of the action.

[0045] Initialize the experience replay memory to store state transition experience (s, a, r / p, s'), where s is the current state, a is the action performed, r / p is the immediate reward or penalty after the action is performed, and s' is the new state after the action is performed.

[0046] Step S2: Setting parameters The learning rate α is set to 0.001 to control the learning speed of the model.

[0047] Set the discount factor γ to 0.9 to calculate the discounted value of future returns.

[0048] The exploration rate ε is set to 0.1 to balance the ratio of exploring new actions and utilizing known actions.

[0049] Set the number of training rounds to 10000 as the total number of training iterations.

[0050] Step S3: Training process For each training round, the following steps are performed: S301: randomly extract a state s from the training data set as the current state, which includes information such as photovoltaic power generation power, energy storage device power, grid voltage and frequency, etc.

[0051] S302: The Actor network generates a control action a according to the current state s, for example, deciding whether the photovoltaic system is connected to the grid or whether the energy storage device is charged.

[0052] S303: Execute action a in the simulation environment, observe the next state s' and the immediate reward r or penalty p. For example, if the action is grid-connected and the grid is stable, a positive reward is obtained; if the action is off-grid but the energy storage device is insufficient, a penalty may be obtained.

[0053] S304: Store the experience (s, a, r / p, s') into the experience playback memory.

[0054] S305: Randomly sample a batch of experiences from the experience replay memory to update the Actor network and the Critic network.

[0055] S306: Using the sampled experience, combined with the Critic network, the value of the action is calculated, and the parameters of the Actor network and the Critic network are updated according to the loss function of the Actor-Critic algorithm. The loss function aims to minimize the difference between the predicted value and the actual value.

[0056] S307: Repeat steps S302 to S306 until the end condition of the training round is reached (such as reaching a preset number of steps).

[0057] Step S4: Training iterations Repeat step S3 until the preset number of training rounds, 10,000, is reached, or the output values ​​of the Actor network and the Critic network converge, that is, the control strategy tends to be stable.

[0058] Step S5: Model application After the training is completed, the obtained Actor network is used as a photovoltaic off-grid control strategy model. In practical applications, for any given current state s of the photovoltaic system, energy storage equipment and power grid, the optimal control action a can be generated through the forward propagation of the Actor network, realizing the intelligent off-grid control of the photovoltaic system and the intelligent charging and discharging management of the energy storage equipment.

[0059] Embodiment 3: This embodiment aims to construct a comprehensive and rich photovoltaic off-grid control strategy training data set by combining real-time data collection with historical data reuse, providing a solid foundation for subsequent control strategy model training based on machine learning or deep learning.

[0060] Step 1: Real-time data collection and initial training data generation Deployment of sensors and monitoring equipment: Light intensity sensors, temperature sensors, etc. are deployed in the photovoltaic system to monitor relevant parameters of photovoltaic power generation in real time.

[0061] Install a power monitor at the energy storage device (such as a battery) to obtain the power status of the energy storage device in real time.

[0062] Install voltage and frequency monitoring equipment at the grid access point to monitor the operating status of the grid in real time.

[0063] Install smart meters at power loads to monitor power consumption in real time.

[0064] Data collection and transmission: Through the Internet of Things technology, the data collected by the above sensors and monitoring equipment are transmitted to the data center in real time.

[0065] The data center pre-processes the received data, including data cleaning, format conversion, etc., to ensure the accuracy and consistency of the data.

[0066] Simulate control actions and reward calculation: According to the preset off-grid switching rules and evaluation criteria, simulated control actions are performed on the real-time collected data. For example, when the photovoltaic power generation is greater than the power load and the grid is stable, the grid-connected action is simulated; when the grid fails or the photovoltaic power generation is insufficient to meet the load, the off-grid action is simulated.

[0067] According to the system state after the simulated control action, the corresponding immediate reward or penalty is calculated. For example, a positive reward is given when the grid connection is successful and the grid is stable, and a penalty is given when the energy storage device runs out of power after leaving the grid.

[0068] Generate preliminary training data: The real-time collected data, simulated control actions, and calculated immediate rewards or penalties are combined into a state-action-reward tuple (s, a, r).

[0069] Store these tuples in the preliminary training dataset as the basis for subsequent training.

[0070] Step 2: Historical data reuse and training data supplement Historical data collection: Extract photovoltaic power generation data from the historical power generation records of the photovoltaic system.

[0071] Obtain historical data such as grid voltage and frequency from grid operation records.

[0072] Extract load power consumption data from the historical change records of power load.

[0073] Construct a historical status dataset: The above historical data are combined into a historical status data set in chronological order, and each status contains key information such as photovoltaic power generation power, grid voltage and frequency, and power load.

[0074] Simulate control actions and reward calculation (historical data): For each piece of data in the historical status data set, possible control actions are simulated according to the preset off-grid switching rules and evaluation criteria.

[0075] Calculate the immediate reward or penalty after the simulated control action in the same way as when collecting real-time data.

[0076] Generate additional training data: Combine historical state data, simulated control actions, and computed immediate rewards or penalties into additional state-action-reward tuples.

[0077] These tuples are added to the preliminary training data set to form a complete training data set for photovoltaic off-grid and grid-connected control strategies.

[0078] Assume that on a certain day, the real-time data collected is as follows: Photovoltaic power generation: 10kW Energy storage device power: 80% Grid voltage: 220V Grid frequency: 50Hz Power load: 8kW According to the preset rules, the grid-connected action is simulated and the instant reward is calculated as +1 (indicating that the grid-connected action is successful and the grid is stable).

[0079] At the same time, the status data extracted from the historical data to a certain day is as follows: Photovoltaic power generation: 5kW Grid voltage: 218V Grid frequency: 49.9Hz Power load: 6kW According to the preset rules, the off-grid action is simulated (because the photovoltaic power generation is not enough to meet the load), and the immediate penalty is calculated to be -0.5 (indicating that the energy storage device needs to discharge to meet the load, but the power is reduced).

[0080] The above real-time data and historical data are combined into state-action-reward tuples and added to the training dataset to enrich the diversity and representativeness of the dataset.

[0081] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0082] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Photovoltaic off-grid and grid-connected integrated control system based on big data technology, characterized in that: The system comprises: Photovoltaic data acquisition module: Integrates photovoltaic sensors, environmental monitors and power grid status monitors to collect real-time power generation data of photovoltaic panels, environmental factor data and real-time status data of the power grid, and perform preprocessing; Photovoltaic off-grid control module: Build and maintain a photovoltaic off-grid control model, which covers photovoltaic power generation nodes, grid connection nodes, energy storage nodes and load nodes, and models the entire process of photovoltaic power generation, storage, grid / off-grid switching and load power supply as a Markov decision process; the photovoltaic off-grid control decision module further includes: State space definition submodule: used to define the state space of the system, including the power generation power, power generation efficiency and light intensity dimensions of photovoltaic power generation nodes, the grid connection status, off-grid status and grid voltage dimensions of grid connection nodes, the remaining power, charging / discharging status and battery temperature dimensions of energy storage nodes, and the power demand, power type and priority dimensions of load nodes; Action space definition submodule: used to define the action space that the system can execute, including adjusting the tilt angle and orientation of photovoltaic panels, controlling the charging and discharging power of energy storage devices, performing grid-connected or off-grid operations, adjusting the power supply sequence and power distribution of loads, starting or shutting down backup energy equipment, performing equipment maintenance, and notifying operators; Reward function design submodule: Design reward functions for photovoltaic power generation efficiency, grid stability, energy storage equipment utilization efficiency, load power supply satisfaction, system operation and maintenance costs, and emergency response capabilities, conduct quantitative evaluation of the actions taken by the system, and give corresponding rewards or penalties; Strategy optimization submodule: uses deep reinforcement learning algorithms, combined with historical data and real-time feedback, to continuously learn and optimize control strategies in order to maximize long-term cumulative rewards; Control instruction generation and execution submodule: Based on the data provided by the photovoltaic data acquisition module and according to the optimized control strategy, specific control instructions are generated and sent to photovoltaic panels, energy storage devices, grid connection devices and load devices through the communication interface.

2. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 1 is characterized by: The power generation power and power generation efficiency dimensions of the photovoltaic power generation node are further refined into instantaneous power generation power, average daily power generation, power generation efficiency and power generation efficiency attenuation rate; the grid connection status, off-grid status and grid voltage dimensions of the grid connection node include grid connection / off-grid status identification, grid voltage value and voltage fluctuation range; the remaining power, charging / discharging status and battery temperature dimensions of the energy storage node are refined into current power, maximum charging / discharging power, battery temperature and battery health status; the power demand, power type and priority dimensions of the load node include total power consumption, peak power consumption period, power type classification and load power supply priority sorting.

3. The photovoltaic off-grid integrated control system based on big data technology according to claim 2 is characterized in that: The reward function design submodule further includes: Photovoltaic power generation efficiency reward function: defined as ,in is the actual photovoltaic power generation power, is the rated power of the photovoltaic panel, The photovoltaic power generation efficiency bonus coefficient is adjusted according to the type, aging and environmental factors of the photovoltaic panels; Grid stability reward function: defined as ； in and are the actual grid voltage and frequency, and are rated voltage and frequency respectively, The grid stability bonus factor is adjusted according to the specifications and requirements of the grid; Energy storage equipment efficiency reward function: defined as ； in is the actual charge and discharge capacity of the energy storage device during the control cycle. is the total capacity of the energy storage device, and are the charge states of the energy storage device at the end and beginning of the control cycle, respectively. and are the maximum and minimum states of charge of the energy storage device, respectively. The energy storage equipment utilization efficiency reward coefficient is adjusted according to the type of energy storage equipment and usage strategy; Load power supply satisfaction reward function: defined as ； in is the actual power supply, is the load demand power, For power supply delay time, is the maximum acceptable power supply delay time, The power supply satisfaction reward coefficient for the load is adjusted according to the type and importance of the load; System operation and maintenance cost penalty function: defined as ,in The actual operation and maintenance costs include equipment maintenance, troubleshooting, and labor costs. To estimate the operation and maintenance costs, It is the penalty coefficient of system operation and maintenance cost, which is adjusted according to the economy and operation and maintenance strategy of the system; Emergency response capability reward function: defined as ,in is the system's response time to a fault or abnormal situation, is the maximum response time threshold set, The reward coefficient for emergency response capability is adjusted according to the system's requirements for emergency response; The final reward function R output by the reward function design submodule is defined as: ； This reward function is used to evaluate the control effect of the system in the current state and guide the strategy optimization and learning submodules to adjust and optimize the strategy.

4. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 1 is characterized by: The photovoltaic off-grid and on-grid control model adopts the Actor-Critic algorithm in the deep reinforcement learning algorithm to train the photovoltaic off-grid and on-grid control strategy.

5. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 4 is characterized by: By deploying various sensors and monitoring equipment in the photovoltaic system and the power grid, the actual data of the photovoltaic system's power generation, the power status of the energy storage device, the voltage and frequency of the power grid, and the power load are collected in real time; combined with the preset off-grid switching rules and evaluation criteria, simulated control actions are performed on the collected actual data, and the corresponding immediate rewards or penalties are calculated; the actual data, simulated control actions, and immediate rewards or penalties are combined into a state-action-reward tuple as the preliminary training data for the photovoltaic off-grid control model.

6. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 5 is characterized by: The photovoltaic power generation records, grid operation data and power load change data in the historical data are used to construct a historical status data set of the photovoltaic system and the grid. For each data in the historical status data set, the possible control actions are simulated and the corresponding immediate rewards or penalties are calculated according to the preset off-grid switching rules and evaluation criteria. The historical status data, simulated control actions and immediate rewards or penalties are combined into additional state-action-reward tuples, which are supplemented into the preliminary training data to form a complete training data set for the photovoltaic off-grid control model.

7. The photovoltaic off-grid integrated control system based on big data technology according to claim 6 is characterized in that: The training steps of the photovoltaic off-grid and grid-connected control model include: S1: Initialize the Actor network and the Critic network. The Actor network is used to generate control actions, and the Critic network is used to evaluate the value of the actions generated by the Actor network. At the same time, initialize the experience playback memory to store the state transition experience of photovoltaic off-grid and grid-connected. S2: Set the learning rate α, discount factor γ, exploration rate ε, and number of training rounds, where α controls the learning speed, γ controls the discounted value of future returns, ε controls the balance between exploring new actions and exploiting known actions, and the number of training rounds determines the total number of training iterations; S3: For each training round, perform the following steps: S301: Initialize the current state s of the photovoltaic system, energy storage device and power grid from the training data set; S302: The Actor network generates a control action a according to the current state s, which indicates the off-grid or on-grid operation of the photovoltaic system, or the charging and discharging operation of the energy storage device; S303: Execute action a, observe the next state s' and the immediate reward r or penalty p, where the immediate reward r or penalty p is calculated according to a preset evaluation standard; S304: storing the experience (s, a, r / p, s') into the experience playback memory; S305: Randomly sample a batch of experiences from the experience replay memory to update the Actor network and the Critic network; S306: Using the sampled experience and the Critic network to calculate the value of the action, and updating the parameters of the Actor network and the Critic network according to the loss function of the Actor-Critic algorithm; S307: Repeat steps S302 to S306 until the end condition of the training round is reached; S4: Repeat step S3 until the preset number of training rounds is reached or the output values ​​of the Actor network and the Critic network converge; S5: After the training is completed, the obtained Actor network is used as a photovoltaic off-grid control strategy model. For any given current state s of the photovoltaic system, energy storage equipment and power grid, the optimal control action a is generated through the forward propagation of the Actor network.

8. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 1, characterized in that: The photovoltaic data acquisition module pre-processes the collected data in a manner including data cleaning and standardization.

9. The photovoltaic off-grid integrated control system based on big data technology according to claim 8 is characterized in that: Methods for cleaning data include: For missing values, linear interpolation algorithm is used to fill them; For outliers, the threshold judgment method is used to regard data points that exceed the preset threshold range as outliers and replace them with the mean of their neighboring points.

10. The photovoltaic off-grid and on-grid integrated control system based on big data technology according to claim 8, characterized in that: The method of standardizing the data includes: using the Z-score standardization method to standardize the continuous features, specifically: subtracting the mean of each numerical data and dividing it by its standard deviation, so that the processed data conforms to the standard normal distribution. The formula for standardization is: ,in, is the original data, is the mean of the original data, is the standard deviation of the original data, is the standardized data.

Images