In recent years, global energy has become increasingly scarce, and environmental pollution has become increasingly severe. The traditional energy industry has overloaded the ecological environment, causing huge troubles to human normal activities and economic life. At present, China’s untapped non renewable energy sources are far below the global average and face severe energy challenges. It is necessary to improve energy utilization efficiency to alleviate the pressure on China’s energy storage. Solar energy is undoubtedly the most ideal energy source. Due to its unique source, solar energy has enormous advantages such as green, wide applicability, abundant and sustainable energy. Power generation technology can achieve stable operation of solar power generation equipment and systems through intelligent control technology, reducing human resource costs. Photovoltaic power generation technology is considered a promising industry by developed countries and is one of the important measures for the sustainability of China’s economic development. Small independent photovoltaic power generation systems have outstanding advantages such as flexible location configuration, small environmental impact, low maintenance costs in the later stages of use, and strong adaptability, and have been widely used. Especially in remote areas and mountainous areas where it is difficult to ensure power supply, it plays an important role, such as supplying power to communication base stations; Small solar photovoltaic products are gradually becoming more abundant in daily life. Due to the influence of various factors on sunlight, it has a certain degree of randomness, and the electricity generated by photovoltaic power generation becomes unstable and discontinuous, leading to solar abandonment and grid disconnection.
At present, there are over 20 years of experience in using various materials and specifications of battery energy storage stations both domestically and internationally. The energy storage system is an important part of the photovoltaic system, which can control price fluctuations, reduce the difference between peak and low grid loads, and ensure the safe and stable operation of the system. The research on photovoltaic energy storage systems is crucial for solving practical operational problems, improving energy utilization and storage efficiency, and promoting the construction of intelligent photovoltaic power generation systems. As an extremely complex system, energy storage units are generally controlled comprehensively by a specific battery management system to ensure their normal operation. The functional module of the Battery Management System (BMS) consists of five parts.
From the basic framework of BMS, it can be seen that battery state estimation is the core of the BMS system and the foundation for the next step of system components such as safety protection. The main performance indicators of lithium-ion batteries during use are also the above two, which are closely related to the normal operation of the body. Therefore, accurate estimation of State of Charge (SOC) and State of Health (SOH) is a very important research topic. SOC and SOH, as battery state estimators for BMS, have significant implications for further guiding battery operation and protection. Only by constructing a battery state of charge detection system and health status management system that can accurately predict battery status, can overcharging and excessive discharge be prevented, ensuring the long-term safe and effective operation of the battery. However, SOC and SOH cannot be directly measured, and can only be estimated by detecting the external characterization parameters of the battery. These external parameters such as terminal voltage, current, temperature, and resistance generally do not have specific mathematical relationships, which makes it difficult to estimate SOC and SOH. The accurate estimation of SOC and SOH has practical significance, as follows. The estimated significance of SOC includes the following three points:
(1) With the use of battery packs, individual batteries may experience performance differences and require balancing treatment. Its important reference indicator is SOC. Accurately estimating the SOC value plays a guiding role in balancing the battery pack.
(2) Ensure the safety of the battery. For energy storage batteries, the performance of the battery itself is closely related to the way it is used. For example, excessive charging and discharging can lead to a shortened battery life. Therefore, improving the accuracy of SOC prediction can guide the correct use of the battery to a certain extent.
(3) In the entire photovoltaic power generation and energy storage system, if the selection of battery pack capacity is not appropriate, it will increase the economic cost of the project. Accurate estimation of SOC can help select the capacity that exactly meets the electrical energy requirements of the load. On the one hand, there is no need to choose battery packs with excessive capacity to prevent product surplus; On the other hand, the construction cost of photovoltaic power generation and energy storage systems can be controlled within a certain range, reducing operating costs.
The significance of SOH estimation includes the following two points:
(1) As the basis of battery state estimation, SOH not only affects the overall performance and safety of the system, but also affects battery and load management strategies. In addition, the estimation accuracy of high SOH can also serve as a correction for the prediction of SOC in energy storage devices.
(2) The performance of energy storage devices will also change with the number of times they are used. The Remaining Used Life (RUL) of a lithium battery is a performance evaluation standard that is predicted and judged based on the battery’s SOH value. Therefore, it also helps to predict potential catastrophic failures, thereby ensuring the reliable operation of the energy storage system and extending the overall system life.
In summary, studying the energy prediction model of energy storage systems requires effective research on SOC and SOH in order to better manage the energy prediction of energy storage systems.
1. Research status of photovoltaic power generation and energy storage systems
With the development of new energy, the advantages of photovoltaic power generation are becoming increasingly apparent. Due to the fact that the energy source of photovoltaic energy storage systems is not subjectively controlled, the energy output of the system may experience intermittency and fluctuations. Energy storage technology is divided into thermal energy storage and electrical energy storage. Photovoltaic power generation and energy storage systems belong to the category of energy storage, which stores energy through energy storage devices for use by loads. Different energy storage technologies have different characteristics and are used in different scenarios. There are two types of electric energy storage technologies, among which electrochemical energy storage has a great potential in the field of power applications due to its advantages of short construction time, convenient dispersion and placement, and high energy density. Energy storage technology can solve important problems in photovoltaic power generation systems. Currently, the key technology for energy research internationally is the development of new energy and energy storage systems. By developing energy storage systems, photovoltaic power generation systems can be optimized and high load supply quality and stable power input can be achieved. Photovoltaic power generation and energy storage systems have the following significance:
(1) It has the function of peak shaving and valley filling, and can regulate the fluctuation of electrical energy during the photovoltaic power generation process.
(2) To achieve power supply stability and reliability for loads in photovoltaic power generation systems. When encountering system faults or being affected by external conditions, the system cannot supply power normally. The energy storage system can provide backup energy, avoiding the problem of the system being unable to supply power normally due to various factors.
(3) The energy storage module is one of the important devices in the system, which can store excess electrical energy and timely supply energy to the load end when the system cannot provide sufficient energy, playing an important balancing role.
At present, various forms of complementary power generation systems have been proposed in relevant research fields both domestically and internationally. Among them, battery energy storage technology has the advantages of high storage capacity and high cost-effectiveness, gradually becoming an important energy storage device in photovoltaic power generation systems. The comparison of several common battery parameters in battery energy storage technology is shown in Table 1.
| Performance indicators | Sodium sulfur battery | Lead-acid battery | All vanadium battery | Lithium ion battery |
| Working voltage of single battery (V) | 2.0 | 2.0 | 1.4 | 3.6 |
| Working temperature (℃) | 300-360 | -5~40 | 0~45 | -30~60 |
| Self discharge rate (month) | 2% | 2.5% | 1% | 1% |
| Specific energy (Wh/kg) | 100~120 | 30~50 | 10~30 | 75~200 |
| Specific power (W/kg) | 150~230 | 75~300 | 120~150 | 150~315 |
| Cost expense ($/kWh) | 300~500 | 200~400 | 150~1000 | 600~2500 |
| Cost per cycle ($/kWh) | 8~20 | 20~100 | 5~80 | 15~100 |
| Environmental indicators | Environment protection | Containing toxic substances | Containing toxic substances | Environment protection |
| Security | Due to the ceramic material of the diaphragm, it is more susceptible to external impact and damage | Mature technology, relatively safe | Although the exchange membrane of the battery is damaged, the electrolyte can still be used, so all vanadium batteries are safer | Overcharging and discharging can damage the battery to varying degrees, creating potential hazards |
According to the above table, it can be seen that compared with other energy storage technologies, battery energy storage technology has a variety of types of batteries, a wide range of options, and can meet various needs. In terms of comprehensive comparison of the energy storage technology characteristics of all batteries in Table 1, it was found that lithium-ion batteries have outstanding advantages.
At present, lead-acid batteries are the type of battery most commonly used in energy storage units of photovoltaic power generation systems. However, research has found that lithium-ion batteries have more advantages, such as higher energy density, slow self-discharge efficiency, multiple cycles, and the ability to optimize the energy storage efficiency of photovoltaic power generation energy storage systems. With the update and development of lithium-ion battery technology, this battery is widely used in various electrical systems and is used as the main energy storage device for photovoltaic systems. Among them, the positive electrode of the battery uses lithium iron phosphate, which has gradually become the mainstream application direction of energy storage devices in energy storage systems.
The energy storage efficiency of lithium-ion batteries can reach about 90%, which is about 10% higher than traditional lead-acid batteries. Therefore, choosing lithium-ion batteries as energy storage units for energy storage systems can significantly improve the efficiency and overall service life of the energy storage system, and reduce losses.
2. Current status of research on SOC estimation strategies for state of charge
SOC can directly reflect the remaining capacity of the battery and is a key parameter to ensure battery reliability. SOC is defined as the ratio of the remaining discharge power (in Ah) to the fully charged state (in Ah) of a battery that has been in use for a period of time or has not been used for a long time. Percentage (value range: 0-100%). The definition formula is as follows.
In the formula, Q is the current remaining capacity; Qnow is the current available full capacity. SOC=100% indicates fully charged, SOC=0% indicates fully discharged.
Due to the inability to directly measure SOC using detectors, and factors affecting the estimation of SOC including capacitance attenuation, self discharge, consistency, temperature, and discharge current, it is difficult to model and accurately estimate energy storage systems.
Because batteries have complex internal reactions during operation and are influenced by various non battery factors, they are nonlinear systems. The existing estimation methods use measured characterization parameters and algorithms to estimate SOC values.
Here are several main estimation methods:
(1) Open circuit voltage method
The principle of the Open Circuit Voltage (OCV) method is to use the curve relationship of SOC-OCV to achieve estimation function. The functional relationship curve between the two is not only influenced by the structure and material of the battery, but also closely related to other factors, such as temperature, charge and discharge current, and battery capacity. And the SOC OCV curve of the battery varies depending on the conditions. However, the open circuit voltage OCV and SOC of the battery are monotonically increasing and in a relatively fixed state. OCV can be used to estimate battery level. This method is simple and suitable for estimating offline situations. However, when the battery is taken offline, the battery voltage will increase and it will take a long time to remain stationary to obtain a more accurate SOC-OCV curve relationship.
(2) Ampere hour integral method
The ampere hour integration method further estimates the SOC of the battery based on real-time calculation of the integration of current and charge discharge time. Although this method is very accurate and easy to implement, it also has many drawbacks. According to the ampere integration method of the experiment, the current drift generated over time will gradually increase the SOC error and be affected by the initial SOC uncertainty and changes in battery capacity. This does not apply to long-term or online estimation.
(3) Extended Kalman Filter Method
Extended Kalman Filter (EKF) can correct errors during model operation based on its effectiveness. It expands nonlinear systems through Taylor’s formula, which is an improvement of Extended Kalman Filter (KF). Overcoming the nonlinear characteristics of lithium-ion battery charging and discharging. Establishing an accurate battery model and treating the battery system as a nonlinear time-varying battery system can achieve better results using the EKF method.
(4) Neural Network Method
Most predictive models require the creation of complex electrochemical models, but neural network models skip this complex step. This method utilizes neural networks to learn and train sample data, which can generate models for predicting SOC. New sample data can be used for prediction and comparison. This method is mainly used to simulate how neurons handle problems and predict the output parameters of nonlinear systems. Compared to extended Kalman filters, neural network methods are not based on physical, electrical, chemical, or thermal models.
(5) Support Vector Machine
Support Vector Machine (SVM) has great advantages in handling regression and classification problems, and belongs to supervised learning. It can be used in nonlinear time-varying systems, such as lithium-ion batteries. Usually, temperature, measured current, and measured voltage are used as input variables, and SOC is used as output variable.
(6) Fuzzy logic method
The fuzzy logic method utilizes fuzzy rules, fuzzy logic, and membership functions to classify and refine complex problems that cannot use the same rule, in order to solve such problems. For nonlinear time-varying battery systems, the fuzzy logic control method does not require the establishment of a battery model, avoiding complex battery models. Only parameters such as temperature, measured current, and measured voltage are used as input variables. By establishing a fuzzy control rule table and using SOC as the control output variable, the higher the dimension of the fuzzy control, the higher the prediction accuracy when establishing the model.
(7) Internal resistance method
The internal resistance method is divided into AC impedance and DC internal resistance. Generally speaking, it is suitable for offline prediction of battery status, but not for online prediction. In addition, the relationship between the impedance spectrum, internal resistance, and charging state of the battery is extremely complex and influenced by various external factors. Due to the difficulty in determining the internal resistance of the battery, further prediction of SOC values will be even more difficult.
(8) Discharge experiment method
The discharge test method is set to conduct a fixed rate discharge test on the battery under constant current conditions (usually 0.3C or 1C), and after reaching the cut-off voltage, the experiment is stopped. Therefore, the amount of electricity released can be obtained by multiplying the discharge time of the battery and the experimental current. However, this method requires interrupting the operation of the battery, so it is not suitable for online measurement. It is only suitable for battery testing and maintenance, which is time-consuming and inefficient, so it is only suitable for laboratory use.
The advantages and disadvantages of the above methods are shown in Table 2.
| Method | Advantages | Disadvantages |
| Open circuit voltage method | Simple method, accurate prediction, and easy implementation | The experimental conditions require complex requirements, long standing time, and cannot be estimated online |
| The experimental conditions require complex requirements, long standing time, and cannot be estimated online | High precision, easy to implement, and can be measured online | Not applicable for long-term and online estimation |
| Extended Kalman Filter Method | High prediction accuracy and good convergence | Need to establish an accurate battery model, relying on the model and requiring a large amount of computation |
| Neural Network Method | No need to establish a battery model, good prediction effect | Requires a large amount of sample data |
| Support Vector Machine | Strong nonlinear approximation ability | High sensitivity to training data |
| Fuzzy logic method | Good ability in nonlinear approximation | High requirements for the selection of membership functions, high computational complexity, and large storage capacity |
| Internal resistance method | Simple and easy to implement | Internal resistance is affected by temperature and requires a large amount of calculation |
| Discharge experiment method | Simple and reliable method | Only applicable in the laboratory |
In summary, traditional SOC estimation methods cannot achieve online estimation of SOC, and their accuracy does not meet the requirements of online estimation. The discharge experiment method is subject to significant limitations and is only applicable to laboratory implementation. Although the fuzzy logic method has excellent nonlinear approximation ability, it has high requirements for membership functions and directly affects experimental results. However, the artificial neural network methods that are suitable for online estimation of SOC and high estimation accuracy in practical operations require a large amount of sample data. In order to improve the prediction effect, it is necessary to improve the method of estimating the SOC value of the state of charge.
3. Current Status of Research on SOH Estimation Strategies for Health Status
The SOH of lithium-ion batteries is a characteristic of the degree of battery aging and is related to the reduction of performance indicators such as capacity, performance, and internal resistance. With the increase of charging and discharging work and usage frequency of the battery, the degree of aging increases, which will affect performance, mainly manifested in a decrease in capacity and power, and an increase in internal resistance. During this process, SOH showed a downward trend. The SOH of batteries cannot be directly measured using equipment to obtain specific values, and can only be predicted using other means. By evaluating the battery’s SOH in real-time, it is possible to check the battery’s deterioration and remaining battery capacity at any time. For batteries that are about to malfunction, SOH can provide users with strong evaluation criteria, timely replacement, and achieve the effect of reducing system damage. At the same time, it can also provide a basis for correcting the SOC value of the battery.
The SOH of a battery is determined by multiple factors, not a single one, so it can be defined from multiple perspectives. SOH is typically expressed as a percentage value between 0 and 100%. For various characteristics, the SOH mathematical calculation formula can be obtained as follows:
(1) Capacity angle
SOH can directly reflect the capacity of lithium-ion batteries, that is, their ability to store charge. The ratio of the actual usable capacity to the rated capacity of lithium-ion batteries is used as the definition value of SOH through the capacity method. The formula is given by the equation:
In the formula, Cnew is the rated capacity of the battery; Cn wo is the current actual usable capacity of the battery.
This refers to the actual amount of charge released from the battery by SOC between 100% and 0 under standard discharge conditions.
According to IEEE standard 1188.1996, the standard for discarded batteries is to reduce the actual usable capacity Cnwo to less than 80% of the rated capacity Cnew, i.e., SOH<80%. At this point, it is necessary to replace the old batteries.
(2) Internal resistance angle
When using lithium-ion batteries, the internal resistance will change, and there is a certain correspondence between the internal resistance and the battery’s SOH. When SOH is defined by internal resistance, it is represented by the equation.
In the formula, Rnow is the internal resistance of the battery at the current moment; Rnew is the internal resistance of the battery when it leaves the factory; Error is the internal resistance when the battery capacity is reduced to 80% of the rated capacity.
A large amount of experimental data indicates that if the internal resistance of the battery increases by a quarter, the battery may have potential hazards. When the internal resistance of the battery increases by half, the battery has a serious fault. If the internal resistance of the battery doubles, the battery will be discarded. Therefore, the internal resistance of the battery can reflect the SOH of the battery.
Currently, there are two main methods for predicting the SOH value of lithium-ion batteries: model driven and data driven. On this basis, there is a hybrid method that combines more than two methods. In addition, there are some relatively rare methods, collectively referred to as other methods.
(1) Model driven SOH estimation for lithium-ion batteries
Model based SOH estimation for lithium-ion batteries requires prior knowledge to create accurate battery models, such as electrochemical and physical models. Internal state variables need to be considered, such as electrolyte concentration, charge concentration, aging coefficient, and battery diffusion coefficient. In this case, the selection of state variables must take into account all the effects of battery operation and aging mechanisms. The model-based approach is based on the operating mechanism of the battery and needs to further accurately reflect the changes in its operating mechanism. The SOH estimation value driven by the model can provide a deeper understanding of the properties of the battery and provide a more accurate SOH estimation value. However, in most cases, creating a model that accurately reflects the changes in battery operating mechanism is more difficult. If the degradation process of the battery is complex and unclear, modeling will be even more difficult. At present, there is no battery model that can fully explain the changes in the operating mechanism of batteries.
(2) Data driven SOH estimation for lithium-ion batteries
The SOH of data-driven lithium-ion batteries does not require prior battery knowledge, nor does it require the creation of accurate mathematical or physical models of the batteries. Working battery data (such as internal resistance and capacity) can be used to analyze battery data. Obtain SOH based on the corresponding relationship between these performance parameters and SOH. By classifying and predicting the sequence features and potential information of battery data, the variation patterns and estimated values of SOH can be obtained through analysis. Therefore, it is not necessary to have a complex and accurate battery model to estimate the SOH of lithium-ion batteries based on data. It is only necessary to use appropriate experimental methods to obtain data about the working battery and select appropriate prediction models based on the characteristics of the data. By continuously learning and training the prediction model, the accuracy of SOH estimation can be improved and the error of the prediction model can be reduced.
The advantages and disadvantages of the above two methods are shown in Table 3.
| Method Type | Advantages | Disadvantages |
| Data driven | Black box model; Suitable for offline monitoring | Requires a large amount of data; Need to establish appropriate prediction models; Unstable prediction performance |
| Model driven | No need for a large amount of data; Stable predictive performance | Need prior knowledge; Need to establish an accurate battery model; Not suitable for offline monitoring |
According to the table above, it can be seen that model-driven lithium-ion batteries do not require too much data to create predictive models, but it is indeed necessary to create accurate physical models to describe internal chemical reactions, with stable predictions and high accuracy. However, physical models have certain limitations and cannot be widely used. Model prediction has problems with multiple model parameters, multiple environmental factors, and sensitivity to external noise. The modeling process is relatively complex, and the established model usually has poor dynamic accuracy, robustness, and adaptability. Data driven lithium-ion battery health management is suitable for offline monitoring. It is more suitable than model driven methods, and does not require explaining the complex structure of internal chemical reactions in lithium-ion batteries or establishing an equivalent model for complex nonlinear time-varying lithium-ion batteries. The SOH value can be directly obtained from the algorithm. However, the rapid development of intelligent algorithms has greatly improved data-driven methods. In order to break away from the current situation of relying on the electrochemical principles inside the battery, methods based on machine learning prediction are becoming increasingly rich. By learning a large amount of data, the prediction model can be debugged, and the debugged model can directly predict SOH in an orderly manner.
In order to better manage the photovoltaic energy storage system and ensure that the energy storage system always provides sufficient energy to stabilize the entire photovoltaic power generation system in emergency situations. By obtaining real-time SOC and SOH of lithium-ion batteries, the energy of photovoltaic power generation systems can be more accurately controlled and the safety of energy storage systems can be ensured. To achieve this, EKF and Deep Reinforcement Learning (DRL) were used to predict the SOC of lithium-ion batteries. The method based on IMOCS-BP neural network was used to predict the SOH of lithium-ion batteries, ensuring the safety of the lithium-ion battery charging and discharging system.
Firstly, the research status of photovoltaic energy storage systems was investigated, and then it was pointed out that energy storage systems play a crucial role in photovoltaic power generation systems. On the basis of analyzing the advantages and disadvantages of different types of lithium-ion batteries, lithium iron phosphate batteries were selected as the main research focus.