Power system regulation and control method based on distributed energy storage

Abstract

The invention relates to the field of power systems, in particular to a distributed energy storage-based power system regulation and control method, which comprises the following steps of: constructing a comprehensive communication network architecture fused with 5G communication, and processing communication and data security by using a block chain and edge computing; establishing a multi-subject benefit coordination model, and solving an optimal strategy according to a game theory algorithm; constructing a reliability and life management model based on reinforcement learning, and introducing a battery aging prediction algorithm; according to the real-time state and prediction information of a power system, the model is used for regulating and controlling the distributed energy storage charging and discharging power in real time, meanwhile, the power market price fluctuation influence is considered, the communication speed is remarkably improved, multi-subject benefits are more comprehensively coordinated, the service life and reliability of the energy storage system are prolonged, and the economic benefits of the power system are improved. And the ability of the system to deal with complex conditions is enhanced.

Description

Technical Field

[0001] This invention is a power system control method based on distributed energy storage, belonging to the field of power systems. Background Technology

[0002] With the increasing proportion of renewable energy in the power system, distributed energy storage, as a key supporting technology, plays a vital role in mitigating fluctuations in renewable energy generation and improving power system stability. However, there are currently many problems to be solved in power system regulation based on distributed energy storage.

[0003] Communication delay issues: Distributed energy storage devices are widely distributed, and significant delays occur when communicating with other components of the power system due to long transmission distances and complex network structures. Traditional communication network architectures are insufficient to meet the timeliness requirements of real-time control, resulting in untimely issuance of control commands and affecting the stable operation of the power system.

[0004] The challenge lies in coordinating the interests of multiple stakeholders: the power system encompasses grid companies, energy storage operators, and users, each with significantly differing interests. Grid companies prioritize ensuring the safe and stable operation of the grid and reducing operating costs; energy storage operators seek profit maximization, aiming to increase the utilization rate and revenue of energy storage devices; and users desire low electricity costs and reliable power supply. Currently, there is a lack of effective quantitative models and coordination mechanisms to balance the interests of all parties. For example, in pursuit of high profits, energy storage operators may over-discharge during peak grid load periods, which, while beneficial to themselves, could affect grid stability and increase the operational risks for grid companies. Information asymmetry among these stakeholders hinders the development of collaborative and optimized control strategies.

[0005] Reliability and lifespan management issues in large-scale energy storage systems: In large-scale energy storage systems, frequent charging and discharging of energy storage devices can easily lead to performance degradation and shortened lifespan. Traditional energy storage system management methods are mostly based on experience or simple rules, failing to fully consider the dynamic characteristics of energy storage devices and complex operating environments, making it difficult to balance reliability and lifespan. Taking lithium-ion batteries as an example, overcharging, over-discharging, or excessively high temperatures accelerate battery aging, reducing lifespan and reliability. Furthermore, existing methods cannot monitor and predict the health status of energy storage devices in real time, making it difficult to take timely maintenance measures, thus affecting the normal operation and economic benefits of the energy storage system.

[0006] To address the aforementioned issues, a power system regulation method based on distributed energy storage is proposed. Summary of the Invention

[0007] To address the problems in the existing technology, this invention provides a power system control method based on distributed energy storage.

[0008] The technical solution adopted by this invention to solve its technical problem is: a power system control method based on distributed energy storage, comprising the following steps:

[0009] Build an integrated communication network architecture that incorporates 5G communication, and use blockchain and edge computing to handle communication and data security;

[0010] Establish a multi-stakeholder interest coordination model and solve for the optimal strategy using game theory algorithms;

[0011] A reliability and lifespan management model based on reinforcement learning is constructed, and a battery aging prediction algorithm is introduced;

[0012] Based on the real-time status and forecast information of the power system, the above model is used to adjust the charging and discharging power of distributed energy storage in real time, while taking into account the impact of electricity market price fluctuations.

[0013] Furthermore, in the aforementioned communication network architecture, the blockchain consensus mechanism adopts the improved Practical Byzantine Fault Tolerance (PBFT) algorithm, the hash algorithm is used for data encryption and signing, and the edge computing nodes are equipped with a caching mechanism.

[0014] Furthermore, the multi-stakeholder interest coordination model incorporates multiple stakeholders, considers the costs, benefits, and contributions to system stability of each stakeholder, and sets demand response elasticity coefficients as game theory algorithm parameters for different types of users.

[0015] Furthermore, the aforementioned entities include power grid companies, energy storage operators, and users.

[0016] Furthermore, the multi-stakeholder interest coordination model also incorporates the additional costs and benefits of multiple stakeholders under extreme weather conditions, and adjusts the game theory algorithm strategy based on the failure probability.

[0017] Furthermore, the additional costs and benefits for the multiple entities under extreme weather conditions include grid repair costs, energy storage equipment damage repair costs, and user power outage losses and compensation.

[0018] Furthermore, the reliability and lifespan management model based on reinforcement learning uses a long short-term memory network (LSTM) combined with a particle filter algorithm to predict the battery health status, improves the deep Q network (DQN) algorithm, uses the prediction results as state space parameters, and optimizes the reward function to better balance reliability and lifespan.

[0019] Furthermore, the real-time control combines short-term power load and distributed power generation forecasts, uses time series analysis and machine learning algorithms to analyze historical data, formulates charging and discharging plans in advance, and arranges charging and discharging based on the coordination of the interests of multiple stakeholders and reliability and lifespan management models.

[0020] The beneficial effects of this invention are:

[0021] 1. Significantly improved communication and energy storage performance: The new architecture, which integrates 5G communication and digital energy storage technology, utilizes the low latency and high bandwidth characteristics of 5G to reduce data transmission delay, ensure that control commands are issued in real time, and meet the stringent requirements of the power system for timely communication.

[0022] 2. More Comprehensive Coordination of Multi-Stakeholder Interests: A comprehensive coordination model considering the interests of multiple stakeholders and the impact of extreme weather is introduced for the first time, incorporating extreme weather factors such as typhoons and torrential rains into the scope of multi-stakeholder interest coordination. By establishing a power system component failure probability model under extreme weather conditions, the actions of all parties can be coordinated, enabling power grid companies to arrange emergency repair resources in advance, and energy storage operators to adjust their energy storage strategies to assist in maintaining power supply in critical areas, thereby reducing power outage time and losses for users.

[0023] 3. By employing a Long Short-Term Memory (LSTM) network combined with a particle filter algorithm, the prediction error of battery capacity degradation is reduced compared to traditional methods, achieving accurate prediction of battery health status. The improved Deep Q-Network (DQN) algorithm further emphasizes the importance of battery health status maintenance by adjusting the state space and reward function, effectively extending the service life of energy storage equipment, reducing equipment replacement frequency, lowering the operating cost of the energy storage system, and ensuring the stable operation of the energy storage segment in the power system.

[0024] 4. Real-time regulation combines factors such as electricity market price fluctuations and employs machine learning ensemble algorithms to accurately predict electricity load, distributed power generation, and electricity market prices. Based on price trends, charging and discharging plans are formulated, charging during off-peak hours and discharging during peak hours. This maximizes the revenue of energy storage operators and minimizes electricity costs for users, fully tapping the price potential of the electricity market, optimizing power resource allocation, and improving the overall economic efficiency of the power system. Attached Figure Description

[0025] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0026] Figure 1 This is a flowchart illustrating a power system control method based on distributed energy storage according to the present invention. Detailed Implementation

[0027] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0028] Please see Figure 1 A power system control method based on distributed energy storage includes the following steps:

[0029] Step S1: Construct a comprehensive communication network architecture that integrates 5G communication, and use blockchain and edge computing to handle communication and data security;

[0030] 5G communication improves transmission efficiency: The introduction of 5G communication networks allows for collaborative work with edge computing. 5G base stations are deployed in areas with concentrated distributed energy storage devices, and edge computing nodes connect to these base stations. During data transmission, data from the distributed energy storage devices is rapidly transmitted to the edge computing nodes via the 5G network. The transmission latency of the 5G network is significantly reduced compared to traditional communication networks, meeting the stringent requirements of real-time control for timely communication.

[0031] Blockchain data interaction verification: Utilizing blockchain distributed ledger technology, transmitted data is encrypted and signed using a hash algorithm. The hash algorithm chosen is SHA-256, and each data block is processed as follows: Let the data block be M, its hash value H(M) = SHA-256(M), ensuring data integrity and immutability. When data is transmitted between nodes, the receiver recalculates the hash value and compares it with the hash value provided by the sender; if they match, the data has not been tampered with.

[0032] Edge computing reduces latency: Deploy edge computing nodes to preprocess local communication data. Edge computing nodes employ a caching mechanism, using the Least Recently Used (LRU) algorithm to manage the cache. With a cache capacity of C, when new data is requested, if the data is in the cache, it is returned directly; if the cache misses, data is retrieved from the source; if the cache is full, the least recently used data is evicted using the LRU algorithm, and the new data is stored in the cache. The blockchain consensus mechanism uses a modified Practical Byzantine Fault Tolerance (PBFT) algorithm. Given N nodes and F faulty nodes, N ≥ 3F + 1. The improvement lies in optimizing the node election process. By comparing node priorities (calculated by comprehensively considering node performance, network bandwidth, historical reliability, etc.), nodes with higher priority are selected as the master node, reducing the number of message passes required to reach consensus. Node fault detection combines heartbeat detection with data verification. The heartbeat detection period is set to T (e.g., 1 second). Every T, a node sends a heartbeat signal to its neighboring nodes. If a node does not receive a heartbeat signal from a node for K consecutive times (e.g., 3 times), data verification is initiated to verify the integrity of the node's data. If the data is abnormal, the node is determined to be faulty.

[0033] Step S2: Establish a multi-stakeholder interest coordination model and solve for the optimal strategy using game theory algorithms;

[0034] Inclusion of stakeholders and target setting: The model incorporates power grid companies, energy storage operators, and users, with the goal of maximizing the overall operational efficiency of the power system. Power grid company costs C_grid include transmission line maintenance costs C_line (related to line length L and unit length maintenance cost k_line, C_line = k_line × L), electricity purchase costs C_purchase (related to purchased volume Q_purchase and unit purchase price p_purchase, C_purchase = Q_purchase × p_purchase), and revenue R_grid represents electricity sales revenue (related to sold volume Q_sell and unit sales price p_sell, Rgrid = ...).

[0035] Q_sell×p_sell); Energy storage operator costs C_storage include energy storage equipment investment costs C_invest (related to equipment capacity E and unit capacity investment cost k_invest, C_invest=k_invest×E), operation and maintenance costs C_operation (related to equipment operating time t and unit time operation and maintenance cost k_operation, C_operation=k_operation×t), and revenue R_storage consists of compensation R_market for participating in power market dispatch (related to dispatch power P_market and compensation unit price p_market, R_market=P_market×p_market) and fees R_user for providing energy storage services to users (related to service power Q_user and service unit price p_user, R_user=

[0036] Q_user × p_user); User cost C_user is the electricity cost (related to electricity consumption Q_user_ele and electricity unit price p_user_ele, C_user = Q_user_ele × p_user_ele) and the potential cost of participating in demand response C_dr (related to the response level d and unit response cost k_dr, C_dr = d × k_dr), and revenue R_user is the demand response subsidy R_dr_subsidy (related to the response level d and subsidy unit price p_dr_subsidy, R_dr_subsidy = d × p_dr_subsidy).

[0037] Furthermore, extreme weather factors are incorporated, and the impact of extreme weather events such as typhoons and heavy rains on the power system is considered in the model. Taking typhoons as an example, a power system component failure probability model under extreme weather conditions is established using meteorological data and power system component parameters. When a typhoon is predicted to strike, the wind speed and direction on the distribution network lines are calculated based on parameters such as the typhoon's path and intensity, thereby determining the wind load acting on the distribution network components. The component failure probability is then calculated in conjunction with the line strength of the distribution network.

[0038] Optimizing the multi-agent strategy: This involves comprehensively considering the additional costs and benefits for each agent under extreme weather conditions. The power grid company incurs increased repair costs C_repair (related to the number of faulty components n and the average repair cost k_repair, C_repair = n × k_repair); the energy storage operator incurs increased equipment damage repair costs C_damage (related to the number of damaged devices m and the average repair cost k_damage, C_damage = m × k_damage); and users incur increased power outage losses C_loss (related to the outage time t and the unit time loss k_loss, C_loss = t × k_loss) and potential power outage compensation R_subsidy (related to the outage time t and the subsidy unit price p_subsidy, R_subsidy = t × p_subsidy). These factors are incorporated into the game theory algorithm's payoff function to re-solve for the optimal strategy for each agent. In extreme weather conditions, this model can coordinate the actions of all parties involved. For example, energy storage operators can adjust their energy storage strategies in advance to help the power grid maintain power supply in critical areas, reduce power outage losses for users, and at the same time, they can also receive corresponding compensation for participating in emergency response.

[0039] Step S3: Construct a reliability and lifespan management model based on reinforcement learning, learn charging and discharging strategies, and introduce a battery aging prediction algorithm;

[0040] Model Element Settings: Establish a reasonable state space S, action space A, and reward function R. The state space S includes parameters such as the energy storage device's state of charge (SOC, ranging from 0 to 1), temperature T (°C), charging / discharging power P (kW), and operating time t (h), i.e., S = {SOC, T, P, t}. The action space A represents the energy storage device's charging and discharging decisions, i.e., A = {charging, discharging, holding}. The reward function R comprehensively considers factors such as the energy storage device's health state H (assessed through indicators such as battery internal resistance and capacity decay), charging / discharging efficiency η (the ratio of actual charging / discharging energy to theoretical charging / discharging energy), and contribution to power system stability S_stability (e.g., its effect on voltage and frequency stability). Let H_norm, η_norm, and S_stability_norm be the normalized health state, charging / discharging efficiency, and contribution to power system stability, respectively. The reward function R = w1 × H_norm + w2 × η_norm + ...

[0041] w3×S_stability_norm, where w1, w2, and w3 are weighting coefficients set according to actual needs, such as w1=0.4, w2=0.3, and w3=0.3. Positive rewards are given if the health status of the energy storage device improves, its charging and discharging efficiency increases, or its contribution to system stability increases; conversely, negative rewards are given if these improvements are not achieved.

[0042] Furthermore, to accurately predict battery health status, a Long Short-Term Memory (LSTM) network combined with a particle filter algorithm is employed. The LSTM network learns from historical battery operating data (including charging / discharging current, voltage, temperature, and time) to capture battery aging trend characteristics. The particle filter algorithm is used to handle uncertainties in battery health status prediction, improving prediction accuracy. For example, in testing a lithium iron phosphate battery energy storage system, compared to traditional prediction methods, this method significantly reduced the error in predicting battery capacity degradation. The predicted battery health status is used as a new parameter in the state space, which, together with existing parameters such as state of charge, temperature, charging / discharging power, and operating time, constitutes a more comprehensive state space S = {SOC, T, P, t, SOH_prediction}.

[0043] Improved Reinforcement Learning Algorithm: Based on a new state space, an improved Deep Q-Network (DQN) algorithm is implemented. The reward function design further emphasizes the reward for maintaining battery health. Let H_norm, η_norm, and S_stability_norm represent the normalized health state, charge / discharge efficiency, and contribution to power system stability, respectively. The reward function R = w1 × H_norm + w2 × η_norm + ...

[0044] The weight of w1 is increased to 0.5 (from 0.4) to emphasize the importance of battery health. In the experience playback mechanism, a sampling strategy based on priority and time decay is adopted, assigning higher sampling probabilities to recent and important samples. Let the sample set be D, where each sample (d_i, r_i) corresponds to a priority p_i and a timestamp t_i. The sampling probability P(d_i) = p_i^α × e^(-β(t-t_i)) / Σ_j p_j^α × e^(-β(t-t_j)), where α is a parameter controlling the degree of priority influence (e.g., α = 0.6), β is a time decay parameter (e.g., β = 0.01), and t is the current time. Through continuous training, the agent learns better charging and discharging strategies, extending the lifespan of the energy storage device.

[0045] Step S4: Using the communication network architecture, multi-stakeholder interest coordination model, and reliability and lifespan management model established above, the distributed energy storage in the power system is regulated in real time, taking into account electricity market price factors.

[0046] Real-time Status Monitoring and Control: The system monitors the power system's operating status in real time, including parameters such as generation power (P_generation), load demand (P_load), grid voltage (V), and frequency (f). When the grid voltage (V) is detected to be below a set threshold (V_threshold, e.g., 0.95 per unit), a command is sent to the corresponding distributed energy storage system via the communication network. Based on a multi-stakeholder interest coordination model, and considering the interests of the grid company, energy storage operator, and users, the energy storage discharge power (P_discharge) is determined. Simultaneously, considering reliability and lifespan management models, and based on the current state of the energy storage equipment (e.g., state of charge, temperature), a reasonable charging and discharging strategy is employed to raise the grid voltage back to the normal range.

[0047] Combining forecasts with planning: Charging and discharging plans are formulated by combining short-term power load forecasts and distributed generation forecasts. Time series analysis (such as the ARIMA model) and machine learning (such as the LSTM model) algorithms are used to analyze historical power load data L_history and distributed generation data G_history. Taking the ARIMA(p,d,q) model for predicting power load as an example, let the time series be {y_t}, and the model expression be Φ(B)(1-B)^d y_t=Θ(B)ε_t, where Φ(B) and Θ(B) are lag operator polynomials, B is the shift operator, and ε_t is a white noise sequence. The model parameters p, d, and q are estimated using historical data to predict the load demand P_load_prediction for a future period. Taking the LSTM model for predicting distributed generation as an example, historical generation data, weather data, and other features are input, and the LSTM unit captures the time series features, outputting the future generation power G_prediction. Before predicting peak loads, distributed energy storage is scheduled to be charged in advance based on a multi-stakeholder interest coordination model and a reliability and lifespan management model. During periods of high distributed power generation, energy storage charging strategies are adjusted in advance to avoid energy waste and equipment damage caused by overcharging.

[0048] Furthermore, electricity market prices are simultaneously predicted, and price fluctuation trends are analyzed. For example, if it is predicted that electricity prices will rise at a certain point in the future, distributed energy storage discharge can be scheduled in advance by combining load and generation forecasts.

[0049] In real-time regulation, charging and discharging plans are adjusted based on electricity market price forecasts. When electricity prices are predicted to rise and load demand shows an increasing trend, distributed energy storage is pre-discharged, allowing the stored electricity to be sold during periods of high prices, thus increasing the revenue of energy storage operators, while ensuring the stability of the power system and the reliability of energy storage devices. When electricity prices are predicted to fall and distributed power generation is sufficient, energy storage is charged, reducing electricity costs. For example, in a microgrid system, implementing this regulation strategy has increased the monthly revenue of energy storage operators and significantly reduced the overall electricity costs for users.

[0050] The present invention will be further described in detail below with reference to specific embodiments.

[0051] Example 1: Communication Network Architecture Example: In a regional power system, distributed energy storage devices and edge computing nodes integrated with a 5G communication network are deployed. 5G base stations cover the project area, and edge computing nodes connect to the 5G base stations for rapid data processing from the distributed energy storage devices. A blockchain technology with an improved PBFT consensus mechanism is used to build the communication network, which has 15 edge computing nodes. Each node manages the communication of surrounding distributed energy storage devices. During data transmission, each data packet is encrypted and signed using the SHA-256 algorithm. The heartbeat detection period is set to 1 second; each node sends a heartbeat signal to its neighboring nodes every second. If a node fails to receive a heartbeat signal from a node for three consecutive times, data verification is initiated. Compared to traditional communication architectures, data transmission latency is significantly reduced.

[0052] Example 2: A coastal power system was hit by a typhoon. Using a comprehensive coordination model, the probability of power system component failures was predicted in advance, and the revenue functions of each entity were adjusted. The power grid company arranged repair resources in advance, and energy storage operators adjusted their energy storage strategies to help maintain power supply in critical areas. Post-typhoon statistics showed that compared to not using this model, user power outage time was significantly reduced, grid repair costs were lowered, and energy storage operators gained additional revenue from participating in emergency response. The interests of all parties were better balanced in extreme circumstances.

[0053] Example 3: For a reliability and lifespan management model based on reinforcement learning, a set of ternary lithium battery energy storage systems were selected for testing. LSTM combined with particle filter algorithm was used to predict battery health status, and the improved DQN algorithm was used to learn the charge and discharge strategy. After 15,000 training iterations, the average number of charge and discharge cycles of the battery was significantly increased compared to when the model was not used, the battery health status was better maintained, and the reliability of the energy storage system was significantly improved.

[0054] Although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A power system control method based on distributed energy storage, characterized in that, include: Build an integrated communication network architecture that incorporates 5G communication, and use blockchain and edge computing to handle communication and data security; Establish a multi-stakeholder interest coordination model and solve for the optimal strategy using game theory algorithms; A reliability and lifespan management model based on reinforcement learning is constructed, and a battery aging prediction algorithm is introduced; Based on the real-time status and forecast information of the power system, the above model is used to adjust the charging and discharging power of distributed energy storage in real time, while taking into account the impact of electricity market price fluctuations.

2. The power system control method based on distributed energy storage according to claim 1, characterized in that: In the aforementioned communication network architecture, the blockchain consensus mechanism adopts the improved Practical Byzantine Fault Tolerance (PBFT) algorithm, the hash algorithm is used for data encryption and signing, and the edge computing nodes are equipped with a caching mechanism.

3. The power system control method based on distributed energy storage according to claim 2, characterized in that: The multi-stakeholder interest coordination model incorporates multiple stakeholders, considering their costs, benefits, and contributions to system stability. Demand response elasticity coefficients are set as game theory algorithm parameters for different types of users.

4. The power system control method based on distributed energy storage according to claim 3, characterized in that: The various entities mentioned include power grid companies, energy storage operators, and users.

5. The power system control method based on distributed energy storage according to claim 4, characterized in that: The multi-stakeholder interest coordination model also incorporates the additional costs and benefits of multiple stakeholders under extreme weather conditions, and adjusts the game theory algorithm strategy based on the failure probability.

6. The power system control method based on distributed energy storage according to claim 5, characterized in that: The additional costs and benefits for the multiple stakeholders under extreme weather conditions include grid repair costs, repair costs for damaged energy storage equipment, and user power outage losses and compensation.

7. The power system control method based on distributed energy storage according to claim 1, characterized in that: The reliability and lifespan management model based on reinforcement learning uses a long short-term memory network (LSTM) combined with a particle filter algorithm to predict battery health status, improves the deep Q network (DQN) algorithm, uses the prediction results as state space parameters, and optimizes the reward function to better balance reliability and lifespan.

8. The power system control method based on distributed energy storage according to claim 1, characterized in that: The real-time control combines short-term power load and distributed power generation forecasts, uses time series analysis and machine learning algorithms to analyze historical data, formulates charging and discharging plans in advance, and arranges charging and discharging based on the coordination of the interests of multiple stakeholders and reliability and lifespan management models.

Images