Charging system of omnibearing solar LED street lamp

Abstract

The invention relates to the technical field of LED lighting, in particular to a charging system of an omnibearing solar LED street lamp, which comprises a charging and discharging threshold value determining module, a charging and discharging threshold value determining module, a charging and discharging threshold value determining module, a charging and discharging threshold value determining module and a charging and discharging threshold value determining module, the historical data acquisition module collects historical illumination, energy consumption, environment, weather, illumination and power generation data of the street lamp and generates historical data of the solar LED street lamp. The energy consumption and generating capacity prediction module predicts the energy consumption demand and the solar generating capacity of the day according to the historical data of the solar LED street lamp. And the charging module obtains battery parameters and formulates a charging strategy in the charging time period, and the battery and the standby energy storage device are charged by using the strategy. And the discharging module obtains battery parameters in a discharging time period, formulates a discharging strategy, and supplies power to the street lamp through the battery and the standby energy storage device according to the strategy. The system realizes intelligent management of charging and discharging of the solar LED street lamp, and is helpful for prolonging the service life of a battery.

Description

Technical Field

[0001] The invention relates to the technical field of LED lighting, and in particular to a charging system for an omnidirectional solar LED street lamp. Background Art

[0002] With the development of modern cities, the construction of smart cities has gradually become the core of planning. As one of the infrastructures of smart cities, street lights play a key role in efficient lighting, intelligent management and green energy applications. In order to reduce energy consumption and improve management efficiency, more and more cities choose to install solar LED street lights. Solar LED street lights are an environmentally friendly and energy-saving outdoor lighting equipment, widely used in places such as roads, squares and highways. These street lights convert solar energy into electrical energy through solar panels and store it in batteries for nighttime lighting.

[0003] The illumination angle of the solar panels of traditional solar street lights is often fixed, which limits the utilization of light resources, especially in urban environments where the sunlight angle varies greatly or is affected by shadows from buildings, trees, etc., and solar energy cannot be effectively obtained. The fully designed solar LED street lights introduce adjustable solar panels or multi-directional solar panels, which can receive sunlight from different angles, not only improving the efficiency of light energy collection, but also providing more stable electricity under unstable sunlight conditions.

[0004] In the context of smart cities, although the existing charging and discharging systems have introduced some intelligent charging technologies, in actual applications, they still cannot be accurately adjusted according to the real-time status and energy consumption requirements of the battery. The dynamic adjustment strategy of the existing system is relatively simple, lacking overall planning and coordination of the charging and discharging process, resulting in a split control between the two, and failing to consider the interdependence of charging and discharging of solar LED street light batteries, resulting in frequent overcharging and over-discharging in battery management. Especially when weather conditions are unstable, sunshine is insufficient, or electricity demand fluctuates greatly at night, it is difficult for the system to flexibly respond to dynamic changes in the environment and demand. This not only affects the overall energy efficiency, but also accelerates battery aging and shortens the battery life.

[0005] Therefore, a charging system for an omnidirectional solar LED street lamp is proposed. Summary of the invention

[0006] The object of the present invention is to provide a charging system for an all-round solar LED street lamp, including a charging and discharging threshold determination module, a historical data acquisition module, an energy consumption and power generation prediction module, a charging module and a discharging module. The charging and discharging threshold determination module is used to obtain the current battery life cycle and the current battery health status of the solar LED street lamp, and determine the upper limit of charging and the lower limit of discharging based on this information. The historical data acquisition module is responsible for obtaining the historical street lamp lighting data, historical street lamp energy consumption data, historical environmental data, historical weather data, historical lighting data and historical solar panel power generation data of the street lamp, so as to generate the historical data of the solar LED street lamp. The energy consumption and power generation prediction module predicts the energy consumption demand in the discharge time period of the day and the solar power generation capacity in the charging time period of the day according to the historical data, and obtains the predicted consumption and predicted power generation. The charging module obtains the current parameters of the battery in the charging time period, obtains the first battery parameter set, and formulates a charging strategy based on the parameter set, the predicted power generation, the predicted consumption and the charging upper limit, and then charges the battery and the backup energy storage device. The discharge module obtains the current parameters of the battery during the discharge time period, obtains the second battery parameter set, and determines the discharge strategy based on the parameter set, the predicted consumption and the discharge lower limit, and finally provides power for the solar LED street lamp through the battery and the backup energy storage device. The present invention introduces a bidirectional energy flow management module to realize the bidirectional flow of energy between the battery and the backup energy storage device, so that the system can flexibly adjust the charging and discharging strategy according to the real-time battery parameters, predicted power generation and energy consumption requirements, thereby ensuring the safety and efficiency of the charging and discharging process. The present invention can dynamically adjust the charging upper limit and the discharging lower limit at different life cycle stages of the battery by determining the charging and discharging threshold according to the battery life cycle and health status. This method not only effectively avoids overcharging and over-discharging of the battery, ensures the safety and stability during the charging process, but also greatly extends the overall service life of the battery, and improves the energy management efficiency and sustainability of the system. The deep reinforcement learning algorithm is introduced to dynamically generate the charging and discharging strategies of the all-round solar LED street lamp, realizing more intelligent and accurate energy management. The system can flexibly adjust the charging and discharging strategies according to the real-time status of the battery and the predicted power generation and consumption, effectively avoiding overcharging, over-discharging and battery overheating, and ensuring the safety and service life of the battery.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A charging system for an all-round solar LED street light, comprising:

[0009] A charge and discharge threshold determination module, used to obtain the current battery life cycle and the current battery health status of the solar LED street lamp; determine the upper charge limit and the lower discharge limit according to the current battery life cycle and the current battery health status;

[0010] A historical data acquisition module is used to acquire the historical street lamp lighting data, historical street lamp energy consumption data, historical environmental data, historical weather data, historical lighting data and historical solar panel power generation data of the solar LED street lamp to obtain the historical data of the solar LED street lamp;

[0011] The energy consumption and power generation prediction module is used to predict the energy consumption demand in the discharge time period of the day and the solar power generation capacity in the charging time period of the day according to the historical data of the solar LED street lamp, and obtain the predicted consumption and predicted power generation;

[0012] A charging module, used to obtain current parameters of the battery within the charging time period, obtain a first battery parameter set, obtain a charging strategy according to the first battery parameter set, the predicted power generation, the predicted consumption and the charging upper limit; and use the charging strategy to charge the battery and the backup energy storage device;

[0013] A discharge module is used to obtain the current parameters of the battery during the discharge time period to obtain a second battery parameter set, and obtain a discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit; and power the solar LED street light through the battery and the backup energy storage device according to the discharge strategy.

[0014] Furthermore, the charging system also includes: an omnidirectional solar energy collection module for collecting sunlight at different angles and converting the sunlight into electrical energy; and includes a plurality of solar panels with adjustable angles.

[0015] Furthermore, the charging system further includes an early warning module, which includes:

[0016] An abnormality detection unit is used to generate a panel warning signal when a solar panel failure is detected; generate a life cycle warning signal when the current battery life cycle is detected to be in the late stage; generate a battery status warning signal when the current battery health state is lower than a preset threshold;

[0017] An information sending unit, used to send the battery panel warning signal, the life cycle warning signal and the battery status warning signal to a remote monitoring center through a communication network to obtain warning data;

[0018] A data storage unit, used to store the warning data in a specified format;

[0019] The early warning processing unit is used to analyze the early warning data and generate early warning maintenance suggestions.

[0020] Further, the charging system also includes a bidirectional energy flow management module, including: for managing the bidirectional flow of energy between the battery and the backup energy storage device; obtaining the remaining power of the first battery and the remaining power of the second battery according to the first battery parameter set and the second battery parameter set respectively; calculating the chargeable capacity according to the remaining power of the first battery and the charging upper limit; if the chargeable capacity exceeds the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery; if the chargeable capacity does not exceed the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery, and when the voltage of the battery reaches the charging upper limit, storing the electric energy generated by the omnidirectional solar energy collection module in the backup energy storage device; calculating the available power according to the remaining power of the first battery and the discharge lower limit; if the available power exceeds the predicted consumption, using the battery to power the omnidirectional solar LED street light; if the available power does not exceed the predicted consumption, using the battery to power the omnidirectional solar LED street light; if the voltage of the battery reaches the discharge lower limit, using the backup energy storage device to power the omnidirectional solar LED street light.

[0021] Further, determining the charging upper limit and the discharging lower limit according to the current battery life cycle and the current battery health state includes:

[0022] Obtaining the historical usage time, charge and discharge times, and rated service life of the battery; determining the current battery life cycle of the battery according to the historical usage time, charge and discharge times, and rated service life of the battery;

[0023] Acquire the charge and discharge data of multiple historical full-scale solar LED street lamps that have completed their life cycle from the historical charge and discharge database to obtain historical reference data; the historical reference data covers the full life cycle charge and discharge records of the batteries of the historical full-scale solar LED street lamps from initial use to end of life, and the full life cycle covers all life cycle stages, including the early, middle and late stages; select the historical reference data with a charge and discharge number greater than a preset threshold to obtain the long-life battery charge and discharge reference data;

[0024] Preprocessing the long-life battery charge and discharge reference data to obtain long-life standard battery charge and discharge reference data; the long-life standard battery charge and discharge reference data includes the battery health status during the charging process and the discharging process;

[0025] Extract the current charge and discharge times and current usage time of each of the omnidirectional solar LED street light batteries in the long-life standard battery charge and discharge reference data; determine the life cycle according to the current charge and discharge times and the current usage time;

[0026] Dividing the longevity standard battery charge and discharge reference data into N groups according to the life cycle and the battery health status to obtain N groups of battery status data, wherein the battery status data includes battery charging process data and battery discharging process data; the battery charging process data includes charging start voltage, charging end voltage, charging current, charging time and charging temperature change; the battery discharging process data includes discharging start voltage, discharging end voltage, discharging current, discharging time and discharging temperature change;

[0027] For each group of the battery status data, extract the maximum value of the charging end voltage in the battery charging process data and the minimum value of the discharging end voltage in the battery discharging process data to obtain the maximum charging voltage and the minimum discharging voltage;

[0028] According to the current battery life cycle and the current battery health status, filter out the groups in which the battery health status is the same as the current battery health status and the battery health status is the same as the current battery health status in the N groups of battery status data, and obtain the battery status data of the same stage; obtain the maximum charging voltage and the minimum discharging voltage of the battery status data of the same stage, and obtain the charging upper limit and the discharging lower limit.

[0029] Furthermore, the energy consumption and power generation prediction module includes a street lamp energy consumption prediction model, and the training process of the street lamp energy consumption prediction model includes the following steps:

[0030] Preprocessing the solar LED street light historical data to obtain solar LED street light standard historical data; the solar LED street light standard historical data includes historical standard street light lighting data, historical standard street light energy consumption data, historical standard environmental data and historical standard weather data; the historical standard environmental data includes image data and sound data of a designated lighting area;

[0031] The feature engineering technology is used to extract features from the standard historical data to obtain a feature set; the feature set includes time features, lighting intensity features, energy consumption features, street light control features, battery features, weather features, regional environment features and traffic flow features; the traffic flow features include the number of pedestrians and vehicles extracted based on the image data and the sound data; the regional environment features include the number of faults and fault distances of surrounding street lights in a specified area; interactive features are constructed based on the feature set, and the interactive features are integrated with the feature set into a first fused feature set; the interactive features include: time and traffic flow interactive features, weather and battery interactive features, regional environment and lighting intensity interactive features, and street light control and energy consumption interactive features;

[0032] The first fused feature set is used as input, and the historical street lamp energy consumption data in the solar LED street lamp historical data is used as a label to train the XGBoost model, and the street lamp energy consumption prediction model is obtained through hyperparameter optimization.

[0033] Furthermore, according to the historical data of the solar LED street lamp, the solar power generation capacity within the charging time period of the day is predicted to obtain the predicted power generation by: preprocessing the historical environmental data, the historical weather data and the historical solar panel power generation data in the historical data of the solar LED street lamp to obtain historical standard environmental data, historical standard weather data and historical standard solar panel power generation data; obtaining historical power generation data from the historical standard solar panel power generation data; using feature engineering technology to extract features of the historical standard environmental data, historical standard weather data, historical standard lighting data and historical standard solar panel power generation data and construct interactive features to obtain a second fused feature set; using the second fused feature set as input and the historical power generation data as labels, according to the XGBoost model, training a solar panel power generation prediction model; organizing the environmental data and weather forecast data of the day into the feature format used for training, and gradually inputting the trained solar panel power generation prediction model according to the time window to predict the power generation in each of the time windows; accumulating the power generation of all the time windows to obtain the predicted power generation.

[0034] Furthermore, the charging strategy is generated according to the deep reinforcement learning algorithm, and the generation process includes:

[0035] constructing a state set according to the first battery parameter set; the first battery parameter set comprising the voltage, current, temperature, state of charge and state of health of the battery during the charging time period;

[0036] defining an action set, including adjusting the charging current of the battery, selecting a charging mode, and switching to the backup energy storage device for charging;

[0037] By using the state transfer function, according to the changes of the current state set and the action set, the state transfer process under different actions is calculated to obtain the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery charge state, battery health state and backup energy storage device state;

[0038] The charging module evaluates the action according to the predicted power generation, the charging upper limit and the result of the state transfer, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overcharge penalty, and obtains the reward function;

[0039] The charging module generates the charging strategy according to the state set, the action set, the state transfer function and the reward function through the deep reinforcement learning algorithm.

[0040] Furthermore, the process of generating the charging strategy includes:

[0041] The policy network is used to select continuous actions according to the current state; the Q network is used to evaluate the long-term cumulative rewards of the continuous actions under the current state according to the reward function; the charging module repeatedly trains the policy network and the Q network through the experience replay mechanism, and optimizes and updates the weights of the policy network and the Q network by back propagation and gradient descent algorithms to maximize the long-term cumulative rewards. When the preset convergence conditions are met, the charging strategy is obtained.

[0042] Further, obtaining the discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit includes:

[0043] The discharge strategy is generated according to the deep reinforcement learning algorithm, and the generation process includes:

[0044] constructing a state set according to the second battery parameter set; the second battery parameter set comprising the voltage, current, temperature, state of charge and state of health of the battery during the discharge time period;

[0045] defining an action set, including adjusting the charging current of the battery, selecting a discharge mode, and switching to the backup energy storage device for discharge;

[0046] By using the state transfer function, according to the changes of the current state set and the action set, the state transfer process under different actions is calculated to obtain the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery charge state, battery health state and backup energy storage device state;

[0047] The discharge module evaluates the action according to the predicted consumption, the discharge lower limit and the result of the state transition, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overdischarge penalty, and obtains the reward function;

[0048] The discharge module generates the discharge strategy according to the state set, the action set, the state transfer function and the reward function through the deep reinforcement learning algorithm;

[0049] The strategy network is used to select continuous actions according to the current state; the Q network is used to evaluate the long-term cumulative rewards of the continuous actions under the current state according to the reward function; the discharge module repeatedly trains the strategy network and the Q network through the experience replay mechanism, and optimizes and updates the weights of the strategy network and the Q network by back propagation and gradient descent algorithms to maximize the long-term cumulative rewards. When the preset convergence conditions are met, the discharge strategy is obtained.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. The all-round solar LED street light charging system proposed in the present invention realizes the bidirectional flow of energy between the battery and the backup energy storage device through the bidirectional energy flow management module, so that the system can flexibly adjust the charging and discharging strategy according to the real-time battery parameters, predicted power generation and predicted consumption, thereby ensuring the safety and efficiency of the charging and discharging process. When the battery is in a low power state, the system will intelligently control the charging process according to the actual consumption demand to avoid battery loss caused by overcharging; when the battery voltage reaches the upper limit of charging, the system will transfer the excess power to the backup energy storage device to prevent overcharging and ensure that the battery works within a safe range. At the same time, during the discharge process, the system can monitor the remaining power and health status of the battery in real time, reasonably set the lower limit of discharge, and avoid battery loss caused by deep discharge. Through this dynamic management mechanism, not only the charging and discharging safety of the battery is greatly improved, but also the battery life can be effectively extended, and the risk of excessive use and excessive discharge of the battery can be reduced, thereby improving the long-term reliability and stability of the system.

[0052] 2. The present invention proposes a method for determining the charge and discharge thresholds according to the battery life cycle and health status, which can dynamically adjust the upper and lower limits of charging and discharging at different stages of the battery life cycle. By obtaining charge and discharge data covering the entire life cycle from the historical charge and discharge database, and combining the historical usage time, charge and discharge times and health status of the battery, the system can accurately identify the current battery life cycle stage of the battery. Selecting data with a lifespan greater than a preset threshold provides more representative and reference-worthy charge and discharge behaviors, ensuring that the system can extract the most suitable charge and discharge voltage range according to best practices during battery management and optimization. This method not only effectively avoids overcharging and over-discharging of the battery, ensuring safety and stability during the charging process, but also greatly extends the overall service life of the battery, and improves the energy management efficiency and sustainability of the system. This charge and discharge strategy based on big data analysis and historical experience enables the system to flexibly adapt to different environmental conditions and achieve efficient and stable operation.

[0053] 3. The present invention combines deep reinforcement learning algorithms to generate intelligent charging and discharging strategies, thereby achieving dynamic and precise management of solar LED street light batteries. During the charging process, the system can monitor the battery voltage, current, temperature, state of charge and health status in real time, and flexibly adjust the charging current and mode in combination with the predicted power generation and the set charging upper limit, effectively improving energy utilization efficiency, avoiding overcharging and energy waste, and extending battery life. During the discharge process, the system also optimizes the discharge strategy according to the battery status, predicted consumption and discharge lower limit, reasonably allocates the discharge tasks of the battery and the backup energy storage device, avoids over-discharge, and ensures stable operation of the system. The intelligent adjustment of the charging and discharging strategy greatly improves the overall energy efficiency of the street light system, extends the service life of the battery, improves the safety and stability of the charging and discharging process, and enables the system to flexibly respond to complex environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A structural diagram of a charging system for an omnidirectional solar LED street light provided by an embodiment of the present invention;

[0055] Figure 2 A structural diagram of a complete charging system for an all-round solar LED street light provided by an embodiment of the present invention;

[0056] Figure 3 A charging strategy decision framework diagram based on deep reinforcement learning provided by an embodiment of the present invention;

[0057] Figure 4 A discharge strategy decision framework diagram based on deep reinforcement learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] 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. According to 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.

[0059] Charging and discharging are two basic working processes of the battery. Charging refers to the omnidirectional solar panel converting the collected solar energy into electrical energy and storing it in the battery. Discharging refers to the battery supplying the previously stored electrical energy to the LED street lamp for lighting during the lighting period. The specific scheme of the present invention is described in detail through two embodiments below.

[0060] Embodiment 1

[0061] The same type of all-round solar LED street lights are installed on the highways in City A. The storage batteries of this batch of all-round solar LED street lights are all 3.2V lithium iron phosphate batteries (batteries), and their backup energy storage devices are grid backup power supply systems. The street lights have advanced adaptive light adjustment functions, and they will not remain on all night. They can flexibly adjust the brightness or on / off state of the lights according to the real-time ambient light intensity and traffic conditions. Because its lighting method is more flexible, it brings higher dynamic management requirements for charging and discharging control.

[0062] The omnidirectional solar LED street light B on the highway uses a charging system for omnidirectional solar LED street lights, such as Figure 1 As shown, including:

[0063] A charge and discharge threshold determination module is used to obtain the current battery life cycle and the current battery health state of the solar LED street lamp; determine the upper limit of charging and the lower limit of discharging according to the current battery life cycle and the current battery health state; specifically, obtain the current battery life cycle of the solar LED street lamp B - early stage, obtain the current battery health state of the solar LED street lamp B - 95%; determine the upper limit of charging and the lower limit of discharging according to the current battery life cycle and the current battery health state;

[0064] The all-round solar LED street light B includes a solar panel and a battery. The solar panel is responsible for obtaining energy from sunlight and converting light energy into electrical energy. The battery is responsible for storing the electrical energy generated by the solar panel and powering the LED street light at night or when there is insufficient sunlight. Together, they form the core of the all-round solar LED street light system. The batteries in this embodiment all refer to lithium iron phosphate batteries (rechargeable batteries).

[0065] A historical data acquisition module is used to acquire the historical street lamp lighting data, historical street lamp energy consumption data, historical environmental data, historical weather data, historical lighting data and historical solar panel power generation data of the solar LED street lamp to obtain the historical data of the solar LED street lamp;

[0066] The energy consumption and power generation prediction module is used to predict the energy consumption demand during the discharge time period of the day and the solar power generation capacity during the charging time period of the day based on the historical data of the solar LED street lamp, and obtain the predicted consumption and predicted power generation; in this embodiment, the prediction is performed before the start of the charging time period of the day.

[0067] A charging module is used to obtain the current parameters of the battery during the charging time period, obtain a first battery parameter set, and obtain a charging strategy based on the first battery parameter set, the predicted power generation, the predicted consumption, and the charging upper limit; use the charging strategy to charge the battery and the backup energy storage device; wherein the charging time period is obtained based on the sunrise time and sunset time of the location of the street lamp.

[0068] A discharge module is used to obtain the current parameters of the battery during the discharge time period to obtain a second battery parameter set, and obtain a discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit; and power the solar LED street light through the battery and the backup energy storage device according to the discharge strategy.

[0069] Furthermore, the charging system also includes: an omnidirectional solar energy collection module for collecting sunlight at different angles and converting the sunlight into electrical energy; the omnidirectional solar energy collection module includes a plurality of solar panels with adjustable angles.

[0070] The all-round solar energy collection module lays the foundation for the subsequent acquisition of more comprehensive historical solar panel power generation data. Since the all-round adjustable angle solar panels can dynamically adjust to the best angle to collect solar energy, the system can accumulate more dimensional and richer power generation data under different lighting conditions. With this data, the system can more accurately predict future power generation capacity, thereby further optimizing the charging strategy.

[0071] Furthermore, the charging system further includes an early warning module, which includes:

[0072] An abnormality detection unit is used to generate a panel warning signal when a solar panel failure is detected; generate a life cycle warning signal when the current battery life cycle is detected to be in the late stage; generate a battery status warning signal when the current battery health state is lower than a preset threshold;

[0073] An information sending unit, used to send the battery panel warning signal, the life cycle warning signal and the battery status warning signal to a remote monitoring center through a communication network to obtain warning data;

[0074] A data storage unit, used to store the warning data in a specified format, as shown in Table 1;

[0075] The early warning processing unit is used to analyze the early warning data and generate early warning maintenance suggestions.

[0076] Table 1. Example table of early warning data

[0077] Monitoring parameters Threshold / Standard Current data Warning signals generated Battery panel voltage 18—22V 12V Solar panel warning signal: Solar panel failure Current battery life cycle / Late stage Life cycle warning signs: Battery is approaching end of life Current battery health status 80％ 75％ Battery status warning signal: Health status is below the threshold

[0078] The present invention realizes intelligent monitoring and preventive maintenance of the solar LED street light system through the early warning module, which can improve the operation and maintenance efficiency of the system, extend the service life of the equipment, reduce the maintenance cost, and ensure the long-term stability and reliability of the solar LED street light system.

[0079] Furthermore, the charging system also includes a bidirectional energy flow management module, such as Figure 2 As shown, it includes: managing the bidirectional flow of energy between the battery and the backup energy storage device; obtaining the remaining power of the first battery and the remaining power of the second battery according to the first battery parameter set and the second battery parameter set respectively; calculating the chargeable capacity according to the remaining power of the first battery and the charging upper limit; if the chargeable capacity exceeds the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery; if the chargeable capacity does not exceed the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery, and when the voltage of the battery reaches the charging upper limit, storing the electric energy generated by the omnidirectional solar energy collection module in the backup energy storage device; calculating the available power according to the remaining power of the first battery and the discharge lower limit; if the available power exceeds the predicted consumption, using the battery to power the omnidirectional solar LED street light; if the available power does not exceed the predicted consumption, using the battery to power the omnidirectional solar LED street light; if the voltage of the battery reaches the discharge lower limit, using the backup energy storage device to power the omnidirectional solar LED street light.

[0080] The present invention improves the energy management efficiency and power supply stability of the solar LED street light system through a bidirectional energy flow management module. The module can dynamically manage the bidirectional energy flow between the battery and the backup energy storage device to ensure that the system reasonably distributes and stores electrical energy under different circumstances. According to the calculation of the remaining battery power and the charging upper limit, when the battery reaches the charging upper limit, the system will automatically transfer the excess solar power to the backup energy storage device to avoid battery overcharging and make full use of solar energy resources. At the same time, by calculating the available battery power and the lower discharge limit, when the battery power reaches the lower discharge limit, the system will intelligently switch to the backup energy storage device for power supply to ensure that the power supply to the street light is not interrupted. While ensuring the battery life, this module further improves the system's ability to cope with unstable power supply or poor environmental conditions, provides a more continuous and reliable power supply solution, and realizes efficient operation and safety assurance of the system.

[0081] Further, determining the charging upper limit and the discharging lower limit according to the current battery life cycle and the current battery health state includes:

[0082] Obtaining the historical usage time, charge and discharge times, and rated service life of the battery; determining the current battery life cycle of the battery according to the historical usage time, charge and discharge times, and rated service life of the battery;

[0083] Acquire the charging and discharging data of multiple historical full-scale solar LED street lamps that have completed their life cycles from the historical charging and discharging database to obtain historical reference data; the historical reference data covers the full life cycle charging and discharging records of the batteries of the historical full-scale solar LED street lamps from initial use to end of life, and the full life cycle covers all life cycle stages, including early, middle and late stages;

[0084] Table 2. Charging and discharging data of a historical full-scale solar LED street lamp that has completed its life cycle

[0085]

[0086] The charge and discharge data of a full range solar LED street light that has completed its life cycle can be found in Table 2. The charge and discharge times in Table 2 refer to the charge and discharge cycles, i.e. the battery completes a complete charge and discharge process. A total of 5,000 charge and discharge processes were performed during the entire life cycle of the full range solar LED street light.

[0087] The historical reference data with a charge and discharge frequency greater than a preset threshold is selected to obtain the long-life battery charge and discharge reference data; in this embodiment, the historical reference data with a charge and discharge frequency greater than a preset threshold (5000 times) in the historical omnidirectional solar LED street lamp is selected as the long-life battery charge and discharge reference data;

[0088] Preprocessing the long-life battery charge and discharge reference data to obtain long-life standard battery charge and discharge reference data; the long-life standard battery charge and discharge reference data includes the battery health status during the charging process and the discharging process;

[0089] Extract the current charge and discharge times and current usage time of each of the omnidirectional solar LED street light batteries in the long-life standard battery charge and discharge reference data; determine the life cycle according to the current charge and discharge times and the current usage time;

[0090] The charge and discharge reference data of the optimal life standard battery are divided into N groups according to the life cycle and the battery health status to obtain N groups of battery status data, wherein the battery status data includes battery charging process data and battery discharging process data; in this embodiment, the value of N is 8.

[0091] Specifically, the battery life cycle is divided into different stages according to the battery's service life and charge and discharge times. The typical life cycle of lithium iron phosphate batteries is divided into early (0-2 years, charge and discharge times: 0-1000 times), mid-term (2-6 years, charge and discharge times: 1000-3000 times) and late (6-10 years, charge and discharge times: more than 3000 times); each life cycle stage can be further subdivided according to the battery health status. Divided by the percentage of battery health status (SOC, State of Charge);

[0092] The battery charging process data includes charging start voltage, charging end voltage, charging current, charging time and charging temperature change; the battery discharging process data includes discharging start voltage, discharging end voltage, discharging current, discharging time and discharging temperature change;

[0093] Table 3. Lithium iron phosphate battery life cycle and health status grouped charge and discharge management table

[0094] Life cycle stage State of Health (SOH) Charging limit (V) Discharge lower limit (V) Early stage ≥90％ 3.6 2.5 Early stage 80％—90％ 3.5 2.6 Early stage 70％—80％ 3.4 2.7 Mid-term 80％—90％ 3.4 2.7 Mid-term 70％—80％ 3.3 2.8 Mid-term 60％—70％ 3.2 2.9 Late stage 60％—70％ 3.2 3.0 Late stage ≤60％ 3.0 3.1

[0095] For each group of the battery status data, the maximum value of the charging end voltage in the battery charging process data and the minimum value of the discharging end voltage in the battery discharging process data are extracted to obtain the maximum charging voltage and the minimum discharging voltage, as shown in Table 3.

[0096] According to the current battery life cycle and the current battery health status, filter out the groups in which the battery health status is the same as the current battery health status and the battery health status is the same as the current battery health status in the N groups of battery status data, and obtain the battery status data of the same stage; obtain the maximum charging voltage and the minimum discharging voltage of the battery status data of the same stage, and obtain the charging upper limit and the discharging lower limit.

[0097] The charging system of this all-round solar LED street light dynamically determines the upper and lower limits of charging and discharging by analyzing the historical usage data and current health status of the battery. At the same time, by combining the historical data of the excellent battery life, the system can extract reference data that matches the current battery health status to achieve targeted adjustments and avoid overcharging and deep discharge. This adaptive management method can effectively extend the service life of the battery and prevent premature aging of the battery due to improper charging and discharging. Secondly, the upper and lower limits of charging and discharging are dynamically adjusted to optimize the charging and discharging process according to different life cycle stages, improve the efficiency and safety of the battery, and ensure that a stable power supply can still be provided when the battery health declines, thereby improving the overall stability and reliability of the system.

[0098] Furthermore, the energy consumption and power generation prediction module includes a street lamp energy consumption prediction model, and the training process of the street lamp energy consumption prediction model includes the following steps:

[0099] Preprocessing the solar LED street light historical data to obtain solar LED street light standard historical data; the solar LED street light standard historical data includes historical standard street light lighting data, historical standard street light energy consumption data, historical standard environmental data and historical standard weather data; the historical standard environmental data includes image data and sound data of a designated lighting area;

[0100] The feature engineering technology is used to extract features from the standard historical data to obtain a feature set; the feature set includes time features, lighting intensity features, energy consumption features, street light control features, battery features, weather features, regional environmental features and traffic flow features; the traffic flow features include the number of pedestrians and vehicles extracted from the image data and the sound data; the regional environmental features include the number of faults and fault distances of surrounding street lights in a specified area;

[0101] Constructing interactive features according to the feature set, integrating the interactive features with the feature set into a first fusion feature set; the interactive features include: time and traffic flow interactive features, weather and battery interactive features, regional environment and lighting intensity interactive features, and street light control and energy consumption interactive features;

[0102] The first fused feature set is used as input, and the historical street lamp energy consumption data in the solar LED street lamp historical data is used as a label to train the XGBoost model, and the street lamp energy consumption prediction model is obtained through hyperparameter optimization.

[0103] By preprocessing and feature engineering multi-dimensional historical data, we comprehensively consider multi-dimensional factors such as time, lighting intensity, weather, traffic flow and regional environment, thereby improving the accuracy and stability of energy consumption prediction. The introduction of interactive features enables the street light energy consumption prediction model to more effectively capture the complex relationship between these features, especially in special cases such as high traffic flow or extreme weather, and accurately predict energy consumption fluctuations to achieve refined energy consumption management. At the same time, accurate daily energy consumption prediction can lay a data foundation for the formulation of daily charging and discharging strategies.

[0104] Further, according to the historical data of the solar LED street lamp, the solar power generation capacity within the charging time period of the day is predicted to obtain the predicted power generation by: preprocessing the historical environment data, the historical weather data and the historical solar panel power generation data in the historical data of the solar LED street lamp to obtain historical standard environment data, historical standard weather data and historical standard solar panel power generation data; obtaining historical power generation data from the historical standard solar panel power generation data; using feature engineering technology to extract features from the historical standard environment data, historical standard weather data, historical standard lighting data and historical standard solar panel power generation data and construct interactive features to obtain a second fused feature set; using the second fused feature set as input and the historical power generation data as labels, and training a solar panel power generation prediction model according to the XGBoost model;

[0105] The environmental data and weather forecast data of the day are organized into the characteristic format used for training, and the trained solar panel power generation prediction model is gradually input according to the time window to predict the power generation in each time window; the power generation in all the time windows is accumulated to obtain the predicted power generation.

[0106] By using feature engineering technology and the construction of interactive features, the system can capture the complex correlation between environmental, weather and power generation data, improving the accuracy of power generation forecasts. Secondly, by using the XGBoost model to train historical data, the system can make detailed predictions for each time period based on real-time environmental and weather forecast data to achieve more stable power generation planning. Finally, after accumulating the power generation in each time period, the system can obtain accurate full-day power generation forecasts, optimize the battery charging strategy, avoid energy waste, and ensure that solar LED street lights can still operate efficiently under changing weather conditions.

[0107] Furthermore, the charging strategy is generated according to the deep reinforcement learning algorithm, and the generation process includes:

[0108] constructing a state set according to the first battery parameter set; the first battery parameter set comprising the voltage, current, temperature, state of charge and state of health of the battery during the charging time period;

[0109] defining an action set, including adjusting the charging current of the battery, selecting a charging mode, and switching to the backup energy storage device for charging;

[0110] By using the state transfer function, according to the changes of the current state set and the action set, the state transfer process under different actions is calculated to obtain the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery charge state, battery health state and backup energy storage device state;

[0111] The charging module evaluates the action according to the predicted power generation, the charging upper limit and the result of the state transfer, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overcharge penalty, and obtains the reward function;

[0112] The charging module generates the charging strategy according to the state set, the action set, the state transfer function and the reward function through the deep reinforcement learning algorithm.

[0113] Furthermore, if Figure 3 As shown, the process of generating the charging strategy includes:

[0114] The policy network is used to select continuous actions according to the current state; the Q network is used to evaluate the long-term cumulative rewards of the continuous actions under the current state according to the reward function; the charging module repeatedly trains the policy network and the Q network through the experience replay mechanism, and optimizes and updates the weights of the policy network and the Q network by back propagation and gradient descent algorithms to maximize the long-term cumulative rewards. When the preset convergence conditions are met, the charging strategy is obtained.

[0115] In this embodiment, the following steps are specifically included:

[0116] S1: build state set S t ;

[0117] During the charging period, the current state set of the system is defined as:

[0118] S t = {V t , I t , T t , SOC t , SOH t , BSS t , V t};

[0119] Among them, V t Represents the voltage of the battery at the current time t; I t Represents the current of the battery at the current time t; T t Indicates the temperature of the battery at the current time t; SOC t Indicates the state of charge (State of Charge) of the battery at the current time t, that is, the current percentage of remaining power; SOH t Indicates the health status of the battery at the current time t; BSS trepresents the state of the backup energy storage device of the battery at the current time t, that is, the current charging status of the backup energy storage device; P represents the predicted consumption of the day; E represents the predicted power generation of the day; V t Represents the voltage of the battery at the current time t;

[0120] In the calculation process of the charging strategy, the predicted power generation P and the predicted consumption E are calculated in advance based on the environmental conditions of the day, historical data and battery health status, while the charging upper limit V max The parameters are set dynamically based on the current state of health and life cycle of the battery. Therefore, during a day's charging cycle, these parameters will not change further over time.

[0121] S2: Define the action set; in each state S t The system can take a series of optional actions, defining the action set A t :

[0122] A t ={a1, a2, a3};

[0123] Among them, a1 represents adjusting the battery charging current, such as selecting different charging rates; a2 represents selecting the charging mode, including constant current charging or constant voltage charging; a3 represents switching to the backup energy storage device for charging.

[0124] S3: Construct state transfer function; state transfer function T(S t , A t ) defines the system in performing action A t After that, how does the state transfer to the next state S? t+1 :

[0125] S t+1 =T(S t , A t );

[0126] In this embodiment, the state transfer function is derived through a physical model and an empirical formula.

[0127] S4: Define the reward function; the system uses the reward function R(S t , A t ) to evaluate the effect of each action and define the corresponding reward and penalty mechanism:

[0128] R(S t , A t )=α·η energy -β·P voltage -γ·P overheat -δ·D SOH -ζ·P overgeneration ;

[0129] P overgeneration =max(EQ charge ,0)

[0130] Among them, η energy It represents the energy utilization efficiency, that is, the ratio of the energy effectively stored in the battery to the total input energy; P voltage Battery voltage penalty. If the battery voltage exceeds the upper charging limit V max , the system is punished; P overheat Penalty for battery overheating. If the battery temperature exceeds the safe temperature range, the system will be penalized; D SOH Penalty for deterioration of battery health status. If SOH drops too fast or is too low, the system will be penalized. Q charge is the chargeable capacity calculated based on the remaining capacity of the battery and the charging upper limit; E represents the predicted power generation for the day; P overgeneration is the battery overcharge penalty, which means that when the predicted power generation exceeds the battery's chargeable capacity, the excess energy is still stored in the battery and the system is penalized; α, β, γ, δ and ζ are the weight coefficients of energy utilization efficiency, battery voltage penalty, battery overheat penalty, battery health deterioration penalty and battery overcharge penalty respectively; max() is the function for finding the maximum value;

[0131] Calculate the instant reward R at time t through the reward function t ;

[0132] S5: Deep reinforcement learning algorithm generates charging strategy;

[0133] Using the deep reinforcement learning algorithm, the system generates a charging strategy as follows:

[0134] Use the policy network to select actions: according to the current state S t , through the policy network π θ Choose an optimal action A t ,Right now:

[0135] A t =π θ (S t );

[0136] Evaluate the long-term cumulative reward of an action: through the Q network Q φ (S t , A t ) Evaluate the long-term cumulative reward of the selected action in the current state. The Q network update formula is:

[0137]

[0138] Among them, φ is the weight parameter of the Q network. With the training process, the weight of the network will be continuously optimized; Qφ (S t , A t ) means that at the current time t, given state S t and action A t , the Q value (i.e., the estimated value of the cumulative reward) evaluated by the Q network parameter φ; μ is the learning rate; σ is the discount factor, which controls the emphasis on future rewards; Indicates that at the next time t+1 state S t+1 Select the best action A t+1 The maximum Q value after

[0139] Experience replay mechanism: The system replays each state-action-reward-new state quadruple (S t , A t , R t， S t+1 ) is stored in the experience replay cache, and the policy network and Q network are repeatedly trained through the experience replay mechanism.

[0140] Backpropagation and Gradient Descent: Backpropagation and gradient descent algorithms are used to optimize and update the weights of the policy network and the Q network to maximize the long-term cumulative reward.

[0141] When any of the following convergence conditions is met, training stops:

[0142] The change in Q value is less than the preset threshold, or the change in cumulative reward is less than the preset threshold.

[0143] Therefore, the final charging strategy π θ (S t ) is the optimal action to choose in the current state

[0144]

[0145] in, It means that the choice can make the function Q φ (S t , A t )Action A that reaches the maximum value t .

[0146] S6: Execute charging strategy; according to the generated charging strategy, the system adjusts the charging current, charging mode or switches to the backup energy storage device for charging when necessary to optimize energy utilization and reduce adverse phenomena such as overcharging and overheating.

[0147] The charging system of this all-round solar LED street light generates a charging strategy through a deep reinforcement learning algorithm. It can flexibly adjust the charging mode and charging current according to multiple parameters such as battery voltage, current, temperature, state of charge and health status, and switch to a backup energy storage device when necessary. This charging strategy based on multi-dimensional state parameters and dynamic prediction greatly improves the energy utilization efficiency, reduces problems such as battery overcharging, overheating and over-discharging, and effectively extends the battery life. At the same time, the system ensures the stability and safety of the battery under long-term operation through mechanisms such as monitoring the battery health status and over-discharge punishment, and realizes intelligent and precise management of the battery charging process. At the same time, by combining the policy network with the Q network in deep reinforcement learning, it can select the optimal continuous charging action according to the current battery status, and continuously optimize the strategy through the experience replay mechanism. During the repeated training process, the system uses back propagation and gradient descent algorithms to optimize the weights of the policy network and Q network to ensure that the charging strategy can maximize the long-term cumulative reward under different conditions. This mechanism enables the system to achieve dynamic and adaptive charging control, greatly improve the energy utilization efficiency of the battery, reduce the risk of overcharging and battery aging, effectively extend the battery life, and maintain efficient and stable operation in complex environments.

[0148] Further, obtaining the discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit includes:

[0149] The discharge strategy is generated according to the deep reinforcement learning algorithm, such as Figure 4 As shown, the generation process includes:

[0150] constructing a state set according to the second battery parameter set; the second battery parameter set comprising the voltage, current, temperature, state of charge and state of health of the battery during the discharge time period;

[0151] defining an action set, including adjusting the charging current of the battery, selecting a discharge mode, and switching to the backup energy storage device for discharge;

[0152] By using the state transfer function, according to the changes of the current state set and the action set, the state transfer process under different actions is calculated to obtain the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery charge state, battery health state and backup energy storage device state;

[0153] The discharge module evaluates the action according to the predicted consumption, the discharge lower limit and the result of the state transition, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overdischarge penalty, and obtains the reward function;

[0154] The discharge module generates the discharge strategy according to the state set, the action set, the state transfer function and the reward function through the deep reinforcement learning algorithm;

[0155] The strategy network is used to select continuous actions according to the current state; the Q network is used to evaluate the long-term cumulative rewards of the continuous actions under the current state according to the reward function; the discharge module repeatedly trains the strategy network and the Q network through the experience replay mechanism, and optimizes and updates the weights of the strategy network and the Q network by back propagation and gradient descent algorithms to maximize the long-term cumulative rewards. When the preset convergence conditions are met, the discharge strategy is obtained.

[0156] The discharge module of this all-round solar LED street light uses a deep reinforcement learning algorithm, a policy network and a Q network to generate the optimal discharge strategy, and can dynamically adjust the discharge process according to real-time parameters such as battery voltage, current, temperature, state of charge and health status. Through the state transfer function and reward function, the system can not only efficiently manage energy output during the discharge process, but also avoid problems such as over-discharge and overheating of the battery. At the same time, the experience replay mechanism and back-propagation optimization enable the system to continuously update the discharge strategy to maximize the long-term cumulative reward. This mechanism effectively improves the energy utilization efficiency of the discharge process, extends the battery life, ensures that the system can operate efficiently and stably in complex and changeable environments, and further optimizes the overall intelligent level of power management.

[0157] The present invention realizes intelligent charging and discharging management of solar LED street light system by integrating charging and discharging threshold determination module, historical data acquisition module, energy consumption and power generation prediction module, and charging and discharging module. The system can accurately determine the upper limit of charging and the lower limit of discharging according to the current battery life cycle and health status of the battery, ensuring that the battery works in the best state and prolonging the service life. Through the acquisition and analysis of historical data, the system can accurately predict the energy consumption demand and power generation capacity of the day, so as to formulate a more reasonable charging and discharging strategy, adapt to the variability of discharge behavior, and avoid excessive discharge or overcharging. The charging and discharging process is interdependent, ensuring that stable power supply is maintained through the backup energy storage device when power generation is insufficient, improving the energy utilization efficiency and reliability of the system, and enabling the solar LED street light to continue to operate efficiently in the complex environment of the smart city. This system ensures the stability and efficient operation of the street light system in a complex and changeable environment through early prediction and dynamic adjustment.

[0158] Embodiment 2

[0159] Company C produced a batch of all-round solar LED street lights, all of which used an all-round solar LED street light charging system, including:

[0160] A charge and discharge threshold determination module, used to obtain the current battery life cycle and the current battery health status of the solar LED street lamp; determine the upper charge limit and the lower discharge limit according to the current battery life cycle and the current battery health status;

[0161] A historical data acquisition module is used to acquire the historical street lamp lighting data, historical street lamp energy consumption data, historical environmental data, historical weather data, historical lighting data and historical solar panel power generation data of the solar LED street lamp to obtain the historical data of the solar LED street lamp;

[0162] The energy consumption and power generation prediction module is used to predict the energy consumption demand in the discharge time period of the day and the solar power generation capacity in the charging time period of the day according to the historical data of the solar LED street lamp, and obtain the predicted consumption and predicted power generation;

[0163] A charging module, used to obtain current parameters of the battery within the charging time period, obtain a first battery parameter set, obtain a charging strategy according to the first battery parameter set, the predicted power generation, the predicted consumption and the charging upper limit; and use the charging strategy to charge the battery and the backup energy storage device;

[0164] A discharge module is used to obtain the current parameters of the battery during the discharge time period to obtain a second battery parameter set, and obtain a discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit; and power the solar LED street light through the battery and the backup energy storage device according to the discharge strategy.

[0165] Specifically, the discharge module evaluates the action according to the predicted consumption, the discharge lower limit and the result of the state transition, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overdischarge penalty, and obtains the reward function;

[0166] Further, obtaining the discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit includes:

[0167] S1: build state set S t ;

[0168] During the discharge period, the current state set of the system is defined as:

[0169] S t = {V t , I t , T t , SOC t , SOH t , BSS t , Vt};

[0170] Among them, V t Indicates the voltage of the battery at the current time t (during the discharge period); I t Represents the current of the battery at the current time t; T t Indicates the temperature of the battery at the current time t; SOC t Indicates the state of charge of the battery at the current time t, that is, the percentage of the current remaining power; SOH t Indicates the health status of the battery at the current time t; BSS t Indicates the state of the backup energy storage device of the battery at the current time t, that is, the discharge status of the current backup energy storage device; V t Represents the voltage of the battery at the current time t;

[0171] The predicted consumption P is a fixed value within the day. During the calculation of the discharge strategy, the predicted consumption has been calculated in advance based on the environmental conditions of the day, historical data and battery health status, and the discharge lower limit V min They are set dynamically based on the current state of health and life cycle of the battery. Therefore, during a one-day discharge cycle, these parameters will not change further over time.

[0172] S2: Define the action set; in each state S t The system can take a series of optional actions, defining the action set A t :

[0173] A t ={a1, a2, a3};

[0174] Among them, a1 represents adjusting the battery discharge current, such as selecting different discharge rates; a2 represents selecting the discharge mode, including constant current discharge or constant voltage discharge; a3 represents switching to the backup energy storage device for discharge.

[0175] S3: Construct state transfer function; state transfer function T(S t , A t ) defines the system in performing action A t After that, how does the state transfer to the next state S? t+1 :

[0176] S t+1 =T(S t , A t );

[0177] In this embodiment, the state transfer function is derived through a physical model and an empirical formula.

[0178] S4: Define the reward function; the system uses the reward function R(S t , A t ) to evaluate the effect of each action and define the corresponding reward and penalty mechanism:

[0179] R(S t , A t )=α·η energy -β·P voltage -γ·P overheat -δ·D SOH -ζ·P overuse ;

[0180] P overuse =max(PQ use ,0)

[0181] Among them, η energy It represents the energy utilization efficiency, that is, the ratio of the energy effectively stored in the battery to the total input energy; P voltage The battery voltage penalty is the battery voltage penalty. If the battery voltage is less than the discharge lower limit V min , the system is punished; P overheat Penalty for battery overheating. If the battery temperature exceeds the safe temperature range, the system will be penalized; D SOH Penalty for deterioration of battery health status. If SOH drops too fast or is too low, the system will be penalized. Q use is the available power calculated based on the remaining power of the battery and the lower discharge limit; P represents the predicted consumption for the day; P overuse为 Battery over-discharge penalty is used to evaluate whether the battery is overused (not switching to the backup charging device for discharge in time) when the predicted consumption exceeds the available power of the battery, and impose corresponding penalties; α, β, γ, δ and ζ are the weight coefficients of energy utilization efficiency, battery voltage penalty, battery overheat penalty, battery health deterioration penalty and battery over-discharge penalty respectively; max() is a function to find the maximum value;

[0182] Calculate the instant reward R at time t through the reward function t ;

[0183] S5: Deep reinforcement learning algorithm generates discharge strategies;

[0184] Using the deep reinforcement learning algorithm, the system generates a discharge strategy as follows:

[0185] Use the policy network to select actions: according to the current state S t , through the policy network π θ Choose an optimal action A t ,Right now:

[0186] At =π θ (S t );

[0187] Evaluate the long-term cumulative reward of an action: through the Q network Q φ (S t , A t ) Evaluate the long-term cumulative reward of the selected action in the current state. The Q network update formula is:

[0188]

[0189] Among them, φ is the weight parameter of the Q network. With the training process, the weight of the network will be continuously optimized; Q φ (S t , A t ) means that at the current time t, given state S t and action A t , the Q value (i.e., the estimated value of the cumulative reward) evaluated by the Q network parameter φ; μ is the learning rate; σ is the discount factor, which controls the emphasis on future rewards; Indicates that at the next time t+1 state S t+1 Select the best action A t+1 The maximum Q value after

[0190] Experience replay mechanism: The system replays each state-action-reward-new state quadruple (S t , A t , R t , S t+1 ) is stored in the experience replay cache, and the policy network and Q network are repeatedly trained through the experience replay mechanism.

[0191] Backpropagation and Gradient Descent: Backpropagation and gradient descent algorithms are used to optimize and update the weights of the policy network and the Q network to maximize the long-term cumulative reward.

[0192] When any of the following convergence conditions is met, training stops:

[0193] The change in Q value is less than the preset threshold or the change in cumulative reward is less than the preset threshold.

[0194] Therefore, the final discharge strategy π θ (S t ) is the optimal action to choose in the current state

[0195]

[0196] in, It means that the choice can make the function Q φ (S t , At )Action A that reaches the maximum value t .

[0197] S6: Execute the discharge strategy. According to the generated discharge strategy, the system adjusts the discharge current, discharge mode and switches to the backup energy storage device for discharge when necessary to optimize energy utilization and reduce adverse phenomena such as over-discharge and overheating.

[0198] 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. A charging system for an all-round solar LED street light, characterized in that: include: A charge and discharge threshold determination module is used to obtain the current battery life cycle and current battery health status of the solar LED street light; Determine a charging upper limit and a discharging lower limit according to the current battery life cycle and the current battery health state; A historical data acquisition module is used to acquire the historical street lamp lighting data, historical street lamp energy consumption data, historical environmental data, historical weather data, historical lighting data and historical solar panel power generation data of the solar LED street lamp to obtain the historical data of the solar LED street lamp; The energy consumption and power generation prediction module is used to predict the energy consumption demand in the discharge time period of the day and the solar power generation capacity in the charging time period of the day according to the historical data of the solar LED street lamp, and obtain the predicted consumption and predicted power generation; A charging module, configured to obtain current parameters of the battery within the charging time period, obtain a first battery parameter set, and obtain a charging strategy according to the first battery parameter set, the predicted power generation, the predicted consumption, and the charging upper limit; Charging the battery and the backup energy storage device using the charging strategy; A discharge module is used to obtain the current parameters of the battery during the discharge time period to obtain a second battery parameter set, and obtain a discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit; and power the solar LED street light through the battery and the backup energy storage device according to the discharge strategy.

2. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The charging system also includes: an omnidirectional solar energy collection module for collecting sunlight at different angles and converting the sunlight into electrical energy; and a plurality of solar panels with adjustable angles.

3. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The charging system also includes an early warning module, which includes: generating a solar panel early warning signal when a solar panel failure is detected; generating a life cycle early warning signal when the current battery life cycle is detected to be in the late stage; and generating a battery status early warning signal when the current battery health status is lower than a preset threshold.

4. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The charging system further includes a bidirectional energy flow management module, including: a module for managing the bidirectional flow of energy between the battery and the backup energy storage device; obtaining the remaining power of the first battery and the remaining power of the second battery according to the first battery parameter set and the second battery parameter set respectively; calculating the chargeable capacity according to the remaining power of the first battery and the charging upper limit; if the chargeable capacity exceeds the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery; if the chargeable capacity does not exceed the predicted power generation, storing the electric energy generated by the omnidirectional solar energy collection module in the battery, and when the voltage of the battery reaches the charging upper limit, storing the electric energy generated by the omnidirectional solar energy collection module in the backup energy storage device; The available power is calculated according to the remaining power of the first battery and the lower discharge limit; if the available power exceeds the predicted consumption, the battery is used to power the omnidirectional solar LED street light; if the available power does not exceed the predicted consumption, the battery is used to power the omnidirectional solar LED street light; if the voltage of the battery reaches the lower discharge limit, the backup energy storage device is used to power the omnidirectional solar LED street light.

5. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: Determining the charging upper limit and the discharging lower limit according to the current battery life cycle and the current battery health state includes: Obtaining the historical usage time, charge and discharge times, and rated service life of the battery; determining the current battery life cycle of the battery according to the historical usage time, charge and discharge times, and rated service life of the battery; Acquire the charge and discharge data of multiple historical full-scale solar LED street lamps that have completed their life cycles from the historical charge and discharge database to obtain historical reference data; the historical reference data covers the full life cycle charge and discharge records of the batteries of the historical full-scale solar LED street lamps from initial use to end of life, and the full life cycle covers all life cycle stages, including the early, middle and late stages; select the historical reference data with a charge and discharge number greater than a preset threshold to obtain the long-life battery charge and discharge reference data; pre-process the long-life battery charge and discharge reference data to obtain the long-life standard battery charge and discharge reference data; the long-life standard battery charge and discharge reference data includes the battery health status during the charging process and the discharging process; Extract the current charge and discharge times and current usage time of each of the omnidirectional solar LED street light batteries in the long-life standard battery charge and discharge reference data; determine the life cycle according to the current charge and discharge times and the current usage time; Dividing the longevity standard battery charge and discharge reference data into N groups according to the life cycle and the battery health status to obtain N groups of battery status data, wherein the battery status data includes battery charging process data and battery discharging process data; the battery charging process data includes a charging start voltage and a charging end voltage; the battery discharging process data includes a discharging start voltage and a discharging end voltage; For each group of the battery status data, extract the maximum value of the charging end voltage in the battery charging process data and the minimum value of the discharging end voltage in the battery discharging process data to obtain the maximum charging voltage and the minimum discharging voltage; According to the current battery life cycle and the current battery health status, groups that meet specified conditions are screened out from N groups of battery status data to obtain battery status data at the same stage; the maximum charging voltage and the minimum discharging voltage of the battery status data at the same stage are obtained to obtain the charging upper limit and the discharging lower limit; the specified conditions are: the battery health status is the same as the current battery health status and the battery health status is the same as the current battery health status.

6. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The energy consumption and power generation prediction module includes a street lamp energy consumption prediction model, and the training process of the street lamp energy consumption prediction model includes the following steps: Preprocessing the solar LED street light historical data to obtain solar LED street light standard historical data; the solar LED street light standard historical data includes historical standard street light lighting data, historical standard street light energy consumption data, historical standard environmental data and historical standard weather data; the historical standard environmental data includes image data and sound data of a designated lighting area; The feature engineering technology is used to extract features from the standard historical data to obtain a feature set; the feature set includes time features, lighting intensity features, energy consumption features, street light control features, battery features, weather features, regional environment features and traffic flow features; the traffic flow features include the number of pedestrians and vehicles extracted based on the image data and the sound data; the regional environment features include the number of faults and fault distances of surrounding street lights in a specified area; interactive features are constructed based on the feature set, and the interactive features are integrated with the feature set into a first fused feature set; the interactive features include: time and traffic flow interactive features, weather and battery interactive features, regional environment and lighting intensity interactive features, and street light control and energy consumption interactive features; The first fused feature set is used as input, and the historical street lamp energy consumption data in the solar LED street lamp historical data is used as a label to train the XGBoost model, and the street lamp energy consumption prediction model is obtained through hyperparameter optimization.

7. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: According to the historical data of the solar LED street lamp, the solar power generation capacity within the charging time period of the day is predicted to obtain the predicted power generation. Specifically, the historical environmental data, the historical weather data and the historical solar panel power generation data in the historical data of the solar LED street lamp are preprocessed to obtain historical standard environmental data, historical standard weather data and historical standard solar panel power generation data; historical power generation data is obtained from the historical standard solar panel power generation data; feature engineering technology is used to extract features of the historical standard environmental data, historical standard weather data, historical standard lighting data and historical standard solar panel power generation data and construct interactive features to obtain a second fused feature set; the second fused feature set is used as input, the historical power generation data is used as a label, and a solar panel power generation prediction model is trained according to the XGBoost model; the environmental data and weather forecast data of the day are organized into the feature format for training, and are gradually input into the trained solar panel power generation prediction model according to the time window to predict the power generation in each time window; the power generation of all the time windows is accumulated to obtain the predicted power generation.

8. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The charging strategy is generated according to the deep reinforcement learning algorithm, and the generation process includes: constructing a state set according to the first battery parameter set; the first battery parameter set comprising the voltage, current, temperature, state of charge and state of health of the battery during the charging time period; defining an action set, including adjusting the charging current of the battery, selecting a charging mode, and switching to the backup energy storage device for charging; By using the state transfer function, according to the changes of the current state set and the action set, the state transfer process under different actions is calculated to obtain the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery charge state, battery health state and backup energy storage device state; The charging module evaluates the action according to the predicted power generation, the charging upper limit and the result of the state transfer, defines the reward according to the energy utilization efficiency, the battery voltage penalty, the battery overheating penalty, the battery health state deterioration penalty and the battery overcharge penalty, and obtains the reward function; The charging module generates the charging strategy according to the state set, the action set, the state transfer function and the reward function through the deep reinforcement learning algorithm.

9. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: The process of generating the charging strategy includes: using the policy network to select continuous actions according to the current state; using the Q network to evaluate the long-term cumulative rewards of the continuous actions under the current state according to the reward function; through the experience replay mechanism, the charging module repeatedly trains the policy network and the Q network through the experience replay mechanism, and uses the back propagation and gradient descent algorithms to optimize and update the weights of the policy network and the Q network to maximize the long-term cumulative rewards. When the preset convergence conditions are met, the charging strategy is obtained.

10. The charging system for an omnidirectional solar LED street light according to claim 1, characterized in that: Obtaining the discharge strategy according to the second battery parameter set, the predicted consumption and the discharge lower limit includes: The discharge strategy is generated according to a deep reinforcement learning algorithm, and the generation process includes: constructing a state set according to the second battery parameter set; the second battery parameter set includes the voltage, current, temperature, state of charge and health state of the battery during the discharge time period; defining an action set, including adjusting the charging current of the battery, selecting a discharge mode, and switching to the backup energy storage device for discharge; through a state transfer function, according to the current state set and the changes in the action set, calculating the state transfer process under different actions, and obtaining the next moment state under each action; the next moment state includes battery voltage, battery current, battery temperature, battery state of charge, battery health state and backup energy storage device state; the discharge module performs an action according to the predicted consumption, the discharge lower limit and the result of the state transfer. An evaluation is performed, and rewards are defined according to energy utilization efficiency, battery voltage penalty, battery overheat penalty, battery health state deterioration penalty, and battery over-discharge penalty to obtain a reward function; the discharge module generates the discharge strategy according to the state set, the action set, the state transfer function, and the reward function through the deep reinforcement learning algorithm; the strategy network is used to select continuous actions according to the current state; according to the reward function, the Q network is used to evaluate the long-term cumulative rewards of the continuous actions under the current state; through the experience replay mechanism, the discharge module repeatedly trains the strategy network and the Q network through the experience replay mechanism, and optimizes and updates the weights of the strategy network and the Q network using back propagation and gradient descent algorithms to maximize the long-term cumulative rewards, and obtains the discharge strategy when the preset convergence conditions are met.

Images